Keysight Technologies Spectrum Analysis Basics

Home , Frequency counter, Preselector, Spectrum analyzer

Application Note 150 Keysight Technologies. Inc. dedicates this application note to Blake Peterson.

Blake’s outstanding service in technical support reached customers in all corners of the world during and after his 45-year career with Hewlett-Packard and Keysight. For many years, Blake trained new marketing and sales engineers in the “ABCs” of spectrum analyzer technology, which provided the basis for understanding more advanced technology. He is warmly regarded as a mentor and technical contributor in spectrum analysis.

Blake’s many accomplishments include: –Authored the original edition of the Spectrum Analysis Basics application note and contributed to subsequent editions –Helped launch the 8566/68 spectrum analyzers, marking the beginning of modern spectrum analysis, and the PSA Series spectrum analyzers that set new performance benchmarks in the industry when they were introduced –Inspired the creation of Blake Peterson University––required training for all engineering hires at Keysight

As a testament to his accomplishments and contributions, Blake was honored with Microwaves & RF magazine’s irst Living Legend Award in 2013.

2 Table of Contents

Chapter 1 – Introduction ...... 5 Frequency domain versus time domain ...... 5 What is a spectrum? ...... 6 Why measure spectra? ...... 6 Types of signal analyzers ...... 8

Chapter 2 – Spectrum Analyzer Fundamentals ...... 9 RF attenuator ...... 10 Low-pass filter or preselector ...... 10 Tuning the analyzer ...... 11 IF gain ...... 12 Resolving signals ...... 13 Residual FM ...... 15 Phase noise ...... 16 Sweep time ...... 18 Envelope detector ...... 20 Displays ...... 21 Detector types ...... 22 Sample detection ...... 23 Peak (positive) detection ...... 24 Negative peak detection ...... 24 Normal detection ...... 24 Average detection ...... 27 EMI detectors: average and quasi-peak detection ...... 27 Averaging processes ...... 28 Time gating ...... 31

Chapter 3 – Digital IF Overview ...... 36 Digital filters ...... 36 All-digital IF...... 37 Custom digital signal processing ...... 38 Additional video processing features ...... 38 Frequency counting ...... 38 More advantages of all-digital IF ...... 39

Chapter 4 – Amplitude and Frequency Accuracy ...... 40 Relative uncertainty ...... 42 Absolute amplitude accuracy ...... 42 Improving overall uncertainty ...... 43 Specifications, typical performance and nominal values ...... 43 Digital IF architecture and uncertainties ...... 43 Amplitude uncertainty examples ...... 44 Frequency accuracy ...... 44

3 Table of Contents continued

Chapter 5 – Sensitivity and Noise ...... 46 Sensitivity ...... 46 Noise floor extension ...... 48 Noise figure ...... 49 Preamplifiers ...... 50 Noise as a signal...... 53 Preamplifier for noise measurements ...... 54

Chapter 6 – Dynamic Range ...... 55 Dynamic range versus internal distortion ...... 55 Attenuator test ...... 56 Noise ...... 57 Dynamic range versus measurement uncertainty ...... 58 Gain compression ...... 60 Display range and measurement range ...... 60 Adjacent channel power measurements ...... 61

Chapter 7 – Extending the Frequency Range ...... 62 Internal harmonic mixing ...... 62 Preselection ...... 66 Amplitude calibration...... 68 Phase noise ...... 68 Improved dynamic range...... 69 Pluses and minuses of preselection...... 70 External harmonic mixing ...... 71 Signal identification ...... 73

Chapter 8 – Modern Signal Analyzers ...... 76 Application-specific measurements ...... 76 The need for phase information ...... 77 Digital modulation analysis ...... 79 Real-time spectrum analysis ...... 80

Chapter 9 – Control and Data Transfer ...... 81 Saving and printing data ...... 81 Data transfer and remote instrument control ...... 81 Firmware updates ...... 82 Calibration, troubleshooting, diagnostics and repair ...... 82

Summary ...... 82

Glossary of Terms ...... 83

4 Chapter 1. Introduction

This application note explains the Fourier1 theory tells us any time- Some measurements require that we fundamentals of swept-tuned, super- domain electrical phenomenon is preserve complete information about heterodyne spectrum analyzers and made up of one or more sine waves the signal frequency, amplitude and discusses the latest advances in spec- of appropriate frequency, amplitude, phase. However, another large group trum analyzer capabilities. and phase. In other words, we can of measurements can be made without transform a time-domain signal into its knowing the phase relationships among At the most basic level, a spec- frequency-domain equivalent. Mea- the sinusoidal components. This type trum analyzer can be described as a surements in the frequency domain of signal analysis is called spectrum frequency-selective, peak-responding tell us how much energy is present at analysis. Because spectrum analysis is voltmeter calibrated to display the rms each particular frequency. With proper simpler to understand, yet extremely value of a sine wave. It is important to iltering, a waveform such as the one useful, we begin by looking irst at how understand that the spectrum analyzer shown in Figure 1-1 can be decom- spectrum analyzers perform spectrum is not a power meter, even though it can posed into separate sinusoidal waves, analysis measurements, starting in be used to display power directly. As or spectral components, which we can Chapter 2. long as we know some value of a sine then evaluate independently. Each sine wave (for example, peak or average) and wave is characterized by its amplitude Theoretically, to make the transfor- know the resistance across which we and phase. If the signal we wish to mation from the time domain to the measure this value, we can calibrate our analyze is periodic, as in our case here, frequency domain, the signal must voltmeter to indicate power. With the Fourier says that the constituent sine be evaluated over all time, that is, advent of digital technology, modern waves are separated in the frequency over ininity. However, in practice, we spectrum analyzers have been given domain by 1/T, where T is the period of always use a inite time period when many more capabilities. In this note, we the signal2. making a measurement. You also can describe the basic spectrum analyzer make Fourier transformations from as well as additional capabilities made possible using digital technology and digital signal processing.

Frequency domain versus time domain Before we get into the details of describing a spectrum analyzer, we might irst ask ourselves: “Just what is a spectrum and why would we want to analyze it?” Our normal frame of reference is time. We note when certain events occur. This includes electrical events. We can use an oscilloscope to view the instantaneous value of a particular electrical event (or some other event converted to volts through an appropriate transducer) as a function of time. In other words, we use the oscilloscope to view the waveform of a signal in the time domain.

Figure 1-1. Complex time-domain signal

1. Jean Baptiste Joseph Fourier, 1768-1830. A French mathematician and physicist who discovered that periodic functions can be expanded into a series of sines and cosines. 2. If the time signal occurs only once, then T is ininite, and the frequency representation is a continuum of sine waves.

5 the frequency to the time domain. This is not a pure sinusoid, it does not give Why measure spectra? case also theoretically requires the us a deinitive indication of the reason evaluation of all spectral components why. Figure 1-2 shows our complex sig- The frequency domain also has its over frequencies to ininity. In reality, nal in both the time and frequency do- measurement strengths. We have making measurements in a inite band- mains. The frequency-domain display already seen in Figures 1-1 and 1-2 width that captures most of the signal plots the amplitude versus the frequen- that the frequency domain is better for energy produces acceptable results. cy of each sine wave in the spectrum. determining the harmonic content of a When you perform a Fourier transfor- As shown, the spectrum in this case signal. People involved in wireless com- mation on frequency domain data, the comprises just two sine waves. We now munications are extremely interested phase of the individual components is know why our original waveform was in out-of-band and spurious emissions. indeed critical. For example, a square not a pure sine wave. It contained a For example, cellular radio systems wave transformed to the frequency second sine wave, the second harmonic must be checked for harmonics of the domain and back again could turn into in this case. Does this mean we have no carrier signal that might interfere with a sawtooth wave if you do not preserve need to perform time-domain mea- other systems operating at the same phase. surements? Not at all. The time domain frequencies as the harmonics. Engi- is better for many measurements, and neers and technicians are also very concerned about distortion of the mes- What is a spectrum? some can be made only in the time domain. For example, pure time-domain sage modulated onto a carrier. So what is a spectrum in the context of measurements include pulse rise and this discussion? A spectrum is a collec- fall times, overshoot and ringing. Third-order intermodulation (two tones tion of sine waves that, when combined of a complex signal modulating each properly, produce the time-domain other) can be particularly troublesome signal under examination. Figure 1-1 because the distortion components can shows the waveform of a complex fall within the band of interest, which signal. Suppose that we were hoping to means they cannot be iltered away. see a sine wave. Although the waveform certainly shows us that the signal Spectrum monitoring is another important frequency-domain measurement activity. Government regulatory agencies allocate different frequencies for various radio services, such as broadcast television and radio, mobile phone systems, police and emergency communications, and a host of other applications. It is critical that each of these services operates at the assigned frequency and stays within the allocated channel bandwidth. Transmitters and other intentional radiators often must operate at closely spaced adjacent Time domain Frequency domain frequencies. A key performance measurements measurements measure for the power ampliiers and other components used in these Figure 1-2. Relationship between time and frequency domain systems is the amount of signal energy that spills over into adjacent channels and causes interference.

Electromagnetic interference (EMI) is a term applied to unwanted emissions from both intentional and unintentional radiators. These unwanted emissions, either radiated or conducted (through the power lines or other interconnecting wires), might impair the operation of other systems. Almost anyone designing or manufacturing

6 Figure 1-3. Harmonic distortion test of a transmitter Figure 1-4. GSM radio signal and spectral mask showing limits of unwanted emissions

Figure 1- 5. Two-tone test on an RF power ampliier Figure 1-6. Radiated emissions plotted against CISPR11 limits as part of an EMI test

7 electrical or electronic products in situations involving brief stops must test for emission levels versus for quick measurements. Due to More information frequency according to regulations advanced calibration techniques, For additional information on set by various government agencies or ield measurements made with these vector measurements, see Vector industry-standard bodies. handheld analyzers correlate with lab- Signal Analysis Basics–Application Note, grade bench-top spectrum analyzers literature number 5989-1121EN. For Noise is often the signal you want to within 10ths of a dB. information on FFT analyzers that measure. Any active circuit or device tune to 0 Hz, see the Web page for will generate excess noise. Tests such In this application note, we concentrate the Keysight 35670A at as noise igure and signal-to-noise ratio on swept amplitude measurements, www.keysight.com/ind/35670A. (SNR) are important for characterizing only briely touching on measurements the performance of a device and involving phase–see Chapter 8. its contribution to overall system performance. Note: When computers became Hewlett-Packard’s dominant business, Figures 1-3 through 1-6 show some of it created and spun off Keysight Tech- these measurements on an X-Series nologies in the late 1990’s to continue signal analyzer. the test and measurement business. Many older spectrum analyzers carry Types of signal analyzers the Hewlett-Packard name but are supported by Keysight. The irst swept-tuned superheterodyne analyzers measured only amplitude. This application note will give you However, as technology advanced insight into your particular spectrum or and communication systems grew signal analyzer and help you use this more complex, phase became a more versatile instrument to its maximum important part of the measurement. potential. Spectrum analyzers, now often labeled signal analyzers, have kept pace. By digitizing the signal, after one or more stages of frequency conversion, phase as well as amplitude is preserved and can be included as part of the information displayed. So today’s signal analyzers such as the Keysight X-Series combine the attributes of analog, vector and FFT (fast Fourier transform) analyzers. To further improve capabilities, Keysight’s X-Series signal analyzers incorporate a computer, complete with a removable disk drive that allows sensitive data to remain in a controlled area should the analyzer be removed.

Advanced technology also has allowed circuits to be miniaturized. As a result, rugged portable spectrum analyzers such as the Keysight FieldFox simplify tasks such as characterizing sites for transmitters or antenna farms. Zero warm-up time eliminates delays

8 Chapter 2. Spectrum Analyzer Fundamentals

This chapter focuses on the fundamen- will see why the ilter is here) to a mixer, and the block diagram shown in Figure tal theory of how a spectrum analyzer where it mixes with a signal from the 2-1. The differences are that the output works. While today’s technology makes local oscillator (LO). Because the mixer of a spectrum analyzer is a display it possible to replace many analog cir- is a non-linear device, its output in- instead of a speaker, and the local cuits with modern digital implementa- cludes not only the two original signals, oscillator is tuned electronically rather tions, it is useful to understand classic but also their harmonics and the sums than by a front-panel knob. spectrum analyzer architecture as a and differences of the original frequen- starting point in our discussion. cies and their harmonics. If any of the The output of a spectrum analyzer is mixed signals falls within the pass band an X-Y trace on a display, so let’s see In later chapters, we will look at the of the intermediate-frequency (IF) ilter, what information we get from it. The capabilities and advantages that digital it is further processed (ampliied and display is mapped on a grid (graticule) circuitry brings to spectrum analysis. perhaps compressed on a logarithmic with 10 major horizontal divisions and Chapter 3 discusses digital architec- scale). It is essentially rectiied by the generally 10 major vertical divisions. tures used in spectrum analyzers avail- envelope detector, iltered through The horizontal axis is linearly calibrated able today. the low-pass ilter and displayed. A in frequency that increases from left ramp generator creates the horizontal to right. Setting the frequency is a Figure 2-1 is a simpliied block diagram movement across the display from left two-step process. First we adjust the of a superheterodyne spectrum analyz- to right. The ramp also tunes the LO so frequency at the centerline of the grati- er. Heterodyne means to mix; that is, to its frequency change is in proportion to cule with the center frequency control. translate frequency. And super refers to the ramp voltage. Then we adjust the frequency range superaudio frequencies, or frequencies (span) across the full 10 divisions with above the audio range. In the Figure If you are familiar with superhetero- the frequency span control. These con- 2-1 block diagram, we see that an input dyne AM radios, the type that receive trols are independent, so if we change signal passes through an attenuator, ordinary AM broadcast signals, you will the center frequency, we do not alter then through a low-pass ilter (later we note a strong similarity between them the frequency span. Alternatively, we

RF input attenuator Log Envelope Mixer IF gain IF ﬁlter amp detector

Input signal

Pre-selector, or Video low-pass ﬁlter ﬁlter Local oscillator

Reference oscillator

Sweep generator Display

Figure 2-1. Block diagram of a classic superheterodyne spectrum analyzer

9 can set the start and stop frequencies with a maximum attenuation of 70 dB Low-pass ilter or preselector instead of setting center frequency and in increments of 2 dB. The blocking ca- span. In either case, we can determine pacitor is used to prevent the analyzer The low-pass ilter blocks high-fre- the absolute frequency of any signal from being damaged by a DC signal or quency signals from reaching the mixer. displayed and the relative frequency a DC offset of the signal being viewed. This iltering prevents out-of-band difference between any two signals. Unfortunately, it also attenuates low- signals from mixing with the local oscil- frequency signals and increases the lator and creating unwanted responses The vertical axis is calibrated in am- minimum useable start frequency of on the display. Microwave spectrum plitude. You can choose a linear scale the analyzer to 9 kHz, 100 kHz or 10 analyzers replace the low-pass ilter calibrated in volts or a logarithmic MHz, depending on the analyzer. with a preselector, which is a tun- scale calibrated in dB. The log scale able ilter that rejects all frequencies is used far more often than the linear In some analyzers, an amplitude refer- except those we currently wish to view. scale because it has a much wider us- ence signal can be connected as shown In Chapter 7, we go into more detail able range. The log scale allows signals in Figure 2-3. It provides a precise about the operation and purpose of the as far apart in amplitude as 70 to 100 frequency and amplitude signal, used preselector. dB (voltage ratios of 3200 to 100,000 by the analyzer to periodically self- and power ratios of 10,000,000 to calibrate. 10,000,000,000) to be displayed simultaneously. On the other hand, the linear scale is usable for signals differing by no more than 20 to 30 dB (voltage ratios of 10 to 32). In either case, we give the top line of the graticule, the reference level, an absolute value through calibration techniques1 and use the scaling per division to assign values to other locations on the graticule. Therefore, we can measure either the absolute value of a signal or the relative amplitude difference between any two signals.

Scale calibration, both frequency and amplitude, is shown by annotations written onto the display. Figure 2-2 shows the display of a typical analyzer.

Now, let’s turn our attention back to the spectrum analyzer components diagramed in Figure 2-1.

RF attenuator Figure 2-2. Typical spectrum analyzer display with control settings The irst part of our analyzer is the RF input attenuator. Its purpose is to ensure the signal enters the mixer at the 0 to 70 dB, 2 dB steps optimum level to prevent overload, gain RF input compression and distortion. Because attenuation is a protective circuit for the analyzer, it is usually set automatically, based on the reference level. Amplitude However, manual selection of attenua- reference signal tion is also available in steps of 10, 5, 2, or even 1 dB. The diagram in Figure 2-3 is an example of an attenuator circuit Figure 2-3. RF input attenuator circuitry

1. See Chapter 4, “Amplitude and Frequency Accuracy.” 10 Tuning the analyzer tune from 0 Hz to 3.6 GHz (actually Figure 2-4 illustrates analyzer tuning. from some low frequency because we In this igure, fLO is not quite high We need to know how to tune our cannot view a 0-Hz signal with this enough to cause the fLO – fsig mixing spectrum analyzer to the desired architecture). product to fall in the IF pass band, so frequency range. Tuning is a function there is no response on the display. If of the center frequency of the IF ilter, If we start the LO at the IF (LO minus IF we adjust the ramp generator to tune the frequency range of the LO and = 0 Hz) and tune it upward from there the LO higher, however, this mixing the range of frequencies allowed to to 3.6 GHz above the IF, we can cover product will fall in the IF pass band at reach the mixer from the outside world the tuning range with the LO minus IF some point on the ramp (sweep), and (allowed to pass through the low- mixing product. Using this information, we will see a response on the display. pass ilter). Of all the mixing products we can generate a tuning equation: emerging from the mixer, the two with The ramp generator controls both the the greatest amplitudes, and therefore fsig = fLO - fIF horizontal position of the trace on the the most desirable, are those created display and the LO frequency, so we from the sum of the LO and input signal where fsig = signal frequency can now calibrate the horizontal axis of and from the difference between the fLO = local oscillator frequency, the display in terms of the input signal LO and input signal. If we can arrange and frequency. things so that the signal we wish to fIF = intermediate frequency (IF) examine is either above or below the We are not quite through with the tun- LO frequency by the IF, then only one If we wanted to determine the LO fre- ing yet. What happens if the frequency of the desired mixing products will fall quency needed to tune the analyzer to of the input signal is 9.0 GHz? As the within the pass-band of the IF ilter and a low-, mid-, or high-frequency signal LO tunes through its 3.8- to 8.7-GHz be detected to create an amplitude (say, 1 kHz, 1.5 GHz, or 3 GHz), we range, it reaches a frequency (3.9 GHz) response on the display. would irst restate the tuning equation at which it is the IF away from the 9.0-

in terms of fLO: GHz input signal. At this frequency we We need to pick an LO frequency and have a mixing product that is equal to an IF that will create an analyzer with fLO = fsig + fIF the IF, creating a response on the dis- the desired tuning range. Let’s assume play. In other words, the tuning equa- that we want a tuning range from 0 to Then we would apply the numbers for tion could just as easily have been: 3.6 GHz. We then need to choose the the signal and IF in the tuning equa- IF. Let’s try a 1-GHz IF. Because this 2 tion : fsig = fLO + fIF frequency is within our desired tuning range, we could have an input signal f = 1 kHz + 5.1 GHz = 5.100001 GHz This equation says that the architec- at 1 GHz. The output of a mixer also LO f = 1.5 GHz + 5.1 GHz = 6.6 GHz or ture of Figure 2-1 could also result in a includes the original input signals, so LO tuning range from 8.9 to 13.8 GHz, but an input signal at 1 GHz would give fLO = 3 GHz + 5.1 GHz = 8.1 GHz. us a constant output from the mixer at the IF. The 1-GHz signal would thus Freq range pass through the system and give us A of analyzer IF a constant amplitude response on the display regardless of the tuning of the LO. The result would be a hole in the frequency range at which we could not properly examine signals because the fsig f f f amplitude response would be indepen- LO fLO – fsig fLO + f sig dent of the LO frequency. Therefore, a 1-GHz IF will not work. Freq range of analyzer Freq range of LO A Instead, we choose an IF that is above the highest frequency to which we wish to tune. In the Keysight X-Series signal analyzers that can tune to 3.6 GHz, the irst LO frequency range is 3.8 to f f 8.7 GHz, and the IF chosen is about LO 5.1 GHz. Remember that we want to Figure 2-4. The LO must be tuned to fIF + fsig to produce a response on the display

2. In the text, we round off some of the frequency values for simplicity, although the exact values are shown in the igures.

11 only if we allow signals in that range analyzer. The full tuning equation for LO feedthrough, can mask very low- to reach the mixer. The job of the input this analyzer is: frequency signals, so not all analyzers low-pass ilter in Figure 2-1 is to pre- allow the display range to include 0 Hz. vent these higher frequencies from get- fsig = fLO1 – (fLO2 + fLO3 + final IF) ting to the mixer. We also want to keep IF gain signals at the intermediate frequency However, itself from reaching the mixer, as previ- Referring back to Figure 2-1, we see the next component of the block dia- ously described, so the low-pass ilter fLO2 + fLO3 + final IF gram is a variable gain ampliier. It is must do a good job of attenuating sig- = 4.8 GHz + 300 MHz + 22.5 MHz nals at 5.1 GHz as well as in the range used to adjust the vertical position of = 5.1225 GHz, the irst IF. from 8.9 to 13.8 GHz. signals on the display without affecting the signal level at the input mixer. Simplifying the tuning equation by us- In summary, we can say that for a When the IF gain is changed, the ing just the irst IF leads us to the same single-band RF spectrum analyzer, we value of the reference level is changed answers. Although only passive ilters would choose an IF above the high- accordingly to retain the correct indi- are shown in igure 2-5, the actual est frequency of the tuning range. We cated value for the displayed signals. implementation includes ampliication would make the LO tunable from the Generally, we do not want the refer- in the narrower IF stages. The inal IF IF to the IF plus the upper limit of the ence level to change when we change section contains additional compo- tuning range and include a low-pass the input attenuator, so the settings nents, such as logarithmic ampliiers or ilter in front of the mixer that cuts off of the input attenuator and the IF gain analog -to-digital converters, depend- below the IF. are coupled together. ing on the design of the particular analyzer. To separate closely spaced signals (see A change in input attenuation will “Resolving signals” later in this chap- automatically change the IF gain to Most RF spectrum analyzers allow ter), some spectrum analyzers have IF offset the effect of the change in input an LO frequency as low as, and even bandwidths as narrow as 1 kHz; others, attenuation, thereby keeping the signal below, the irst IF. Because there is 10 Hz; still others, 1 Hz. Such narrow at a constant position on the display. inite isolation between the LO and IF ilters are dificult to achieve at a center ports of the mixer, the LO appears at frequency of 5.1 GHz, so we must add the mixer output. When the LO equals additional mixing stages, typically the IF, the LO signal itself is processed two to four stages, to down-convert by the system and appears as a re- from the irst to the inal IF. Figure 2-5 sponse on the display, as if it were an shows a possible IF chain based on input signal at 0 Hz. This response, the the architecture of a typical spectrum

3.6 GHz 5.1225 GHz 322.5 MHz 22.5 MHz Envelope detector

3.8 to 8.7 GHz

4.8 GHz 300 MHz

Sweep generator Display

Figure 2-5. Most spectrum analyzers use two to four mixing steps to reach the inal IF.

12 Resolving signals a deinite width on the display. The Two signals must be far enough apart output of a mixer includes the sum or the traces they make will fall on top After the IF gain ampliier, we ind the and difference products plus the two of each other and look like only one IF section, which consists of the analog original signals (input and LO). A band- response. Fortunately, spectrum ana- or digital resolution bandwidth (RBW) pass ilter determines the intermediate lyzers have selectable resolution (IF) ilters, or both. frequency, and this ilter selects the ilters, so it is usually possible to select desired mixing product and rejects all one narrow enough to resolve closely Analog ilters other signals. Because the input signal spaced signals. is ixed and the local oscillator is swept, Frequency resolution is the ability of the products from the mixer are also Keysight data sheets describe the a spectrum analyzer to separate two swept. If a mixing product happens to ability to resolve signals by listing the input sinusoids into distinct responses. sweep past the IF, the characteristic 3-dB bandwidths of the available IF Fourier tells us that a sine-wave signal shape of the bandpass ilter is traced ilters. This number tells us how close only has energy at one frequency, so on the display. See Figure 2-6. The nar- together equal-amplitude sinusoids can we should not have any resolution rowest ilter in the chain determines the be and still be resolved. In this case, problems. Two signals, no matter how overall displayed bandwidth, and in the there will be about a 3-dB dip between close in frequency, should appear as architecture of Figure 2-5, this ilter is the two peaks traced out by these two lines on the display. But a closer in the 22.5- MHz IF. signals. See Figure 2-7. The signals look at our superheterodyne receiver can be closer together before their shows why signal responses have

Figure 2-6. As a mixing product sweeps past the IF ilter, the ilter shape is traced on the display

3. If you experiment with resolution on a spectrum analyzer using the normal (rosenfell) detector mode (See “Detector types” later in this chapter) use enough video iltering to create a smooth trace. Otherwise, you will see smearing as the two signals interact. While the smeared trace certainly indicates the presence of more than one signal, it is dificult to determine the amplitudes of the individual signals. Spectrum analyzers with positive peak as their default detector mode may not show the smearing effect. You can observe the smearing by selecting the sample detector mode. Figure 2-7. Two equal-amplitude sinusoids separated by the 3-dB BW of the selected IF ilter can be resolved.

13 traces merge completely, but the 3-dB bandwidth is a good rule of thumb for resolution of equal-amplitude signals3.

More often than not, we are dealing with sinusoids that are not equal in amplitude. The smaller sinusoid can actually be lost under the skirt of the response traced out by the larger. This effect is illustrated in Figure 2-8. The top trace looks like a single signal, but in fact represents two signals: one at 300 MHz (0 dBm) and another at 300.005 MHz (–30 dBm). The lower trace shows the display after the 300- MHz signal is removed.

Another speciication is listed for the resolution ilters: bandwidth selectivity (or selectivity or shape factor). Band- width selectivity helps determine the Figure 2-8. A low-level signal can be lost under the skirt of the response to a larger signal resolving power for unequal sinusoids. For Keysight analyzers, bandwidth selectivity is generally speciied as the ratio of the 60-dB bandwidth to the 3-dB bandwidth, as shown in Figure 2-9. The analog ilters in Keysight analyzers are a four-pole, synchronously tuned design, with a nearly Gaussian shape4. This type of ilter exhibits a bandwidth selectivity of about 12.7:1.

For example, what resolution bandwidth must we choose to resolve

Figure 2-9. Bandwidth selectivity, ratio of 60-dB to 3-dB bandwidths

4. Some older spectrum analyzer models used ive-pole ilters for the narrowest resolution bandwidths to provide improved selectivity of about 10:1. Modern designs achieve even better bandwidth selectivity using digital IF ilters. 14 signals that differ by 4 kHz and 30 dB, assuming 12.7:1 bandwidth selectivity? Because we are concerned with rejection of the larger signal when the analyzer is tuned to the smaller signal, we need to consider not the full bandwidth, but the frequency difference from the ilter center frequency to the skirt. To determine how far down the ilter skirt is at a given offset, we use the following equation:

2 H(∆f) = –10(N) log10 [(∆f/f0) + 1]

Where H(∆f) is the ilter skirt rejection in dB, N is the number of ilter poles, ∆f is the frequency offset from the center in Hz, and

RBW f is given by 0 2 √ 21/N –1 Figure 2-10. The 3-kHz ilter (top trace) does not resolve the smaller signal; reducing the resolution bandwidth to 1 kHz (bottom trace) does For our example, N=4 and ∆f = 4000. Let’s begin by trying the 3-kHz RBW ilter. Digital ilters mixing product involving the LO, there was no point in having resolution band- First, we compute f0: Some spectrum analyzers use digital widths narrower than 1 kHz because it 3000 techniques to realize their resolu- was impossible to determine the cause f0 = 2 √ 2¼ –1 = 3448.44 tion bandwidth ilters. Digital ilters of any instability on the display. can provide important beneits, such Now we can determine the ilter rejec - as dramatically improved bandwidth tion at a 4-kHz offset: However, modern analyzers have selectivity. The Keysight PSA and X- dramatically improved residual FM. For Series signal analyzers implement all H(4000) = –10(4) log [(4000/3448.44)2 + 1] example, residual FM in Keysight PXA 10 resolution bandwidths digitally. Other = −14.8 dB Series analyzers is nominally 0.25 Hz; analyzers, such as the Keysight ESA-E in PSA Series analyzers, 1 to 4 Hz; and Series, take a hybrid approach, using in ESA Series analyzers, 2 to 8 Hz. This This is not enough to allow us to see analog ilters for the wider bandwidths allows bandwidths as low as 1 Hz in the smaller signal. Let’s determine and digital ilters for bandwidths of 300 many analyzers, and any instability we H(∆f) again using a 1-kHz ilter: Hz and below. Refer to Chapter 3 for see on a spectrum analyzer today is 1000 more information on digital ilters. due to the incoming signal. f = = 1149.48 0 2 √ 2¼ –1 Residual FM This allows us to calculate the ilter rejection: The instability and residual FM of the LOs in an analyzer, particularly the 2 H(4000) = –10(4) log10[(4000/1149.48) + 1] irst LO, often determine the minimum = −44.7 dB usable resolution bandwidth. The un- stable YIG (yttrium iron garnet) oscilla- Thus, the 1-kHz resolution bandwidth tor used in early analyzers typically had ilter does resolve the smaller signal, as a residual FM of about 1 kHz. Because illustrated in Figure 2-10. this instability was transferred to any

15 Phase noise No oscillator is perfectly stable. Even though we may not be able to see the actual frequency jitter of a spectrum analyzer LO system, there is still a manifestation of the LO frequency or phase instability that can be observed. This is known as phase noise (sometimes called sideband noise).

All are frequency or phase modulated by random noise to some extent. As previously noted, any instability in the LO is transferred to any mixing products resulting from the LO and input signals. So the LO phase noise modulation sidebands appear around any spectral component on the display that is far enough above the broadband noise loor of the system (Figure 2-11). The amplitude difference between a displayed spectral component and the Figure 2-11. Phase noise is displayed only when a signal is displayed far enough above the system noise loor phase noise is a function of the stability of the LO. The more stable the LO, the Generally, we can see the inherent – Optimize phase noise for fre- lower the phase noise. The amplitude phase noise of a spectrum analyzer quency offsets > 160 kHz from the difference is also a function of the only in the narrower resolution ilters, carrier This mode optimizes phase resolution bandwidth. If we reduce the when it obscures the lower skirts of noise for offsets above 160 kHz resolution bandwidth by a factor of 10, these ilters. The use of the digital away from the carrier. the level of the displayed phase noise ilters previously described does not decreases by 10 dB5. change this effect. For wider ilters, the – Optimize LO for fast tuning phase noise is hidden under the ilter When this mode is selected, LO The shape of the phase noise spec- skirt, just as in the case of two unequal behavior compromises phase noise trum is a function of analyzer design, sinusoids discussed earlier. at all offsets from the carrier below in particular, the sophistication of the approximately 2 MHz. This mode phase-lock loops employed to stabilize Today’s spectrum or signal analyz- minimizes measurement time and the LO. In some analyzers, the phase ers, such as Keysight’s X-Series, allow allows the maximum measurement noise is a relatively lat pedestal out to you to select different LO stabilization throughput when changing the the bandwidth of the stabilizing loop. In modes to optimize the phase noise for center frequency or span. others, the phase noise may fall away different measurement conditions. For as a function of frequency offset from example, the PXA signal analyzer offers the signal. Phase noise is speciied in three different modes: terms of dBc (dB relative to a carrier) and normalized to a 1-Hz noise power – Optimize phase noise for fre- bandwidth. It is sometimes speci- quency offsets < 140 kHz from ied at speciic frequency offsets. At the carrier In this mode, the LO other times, a curve is given to show phase noise is optimized for the the phase noise characteristics over a area close in to the carrier at the range of offsets. expense of phase noise beyond 140-kHz offset.

5. The effect is the same for the broadband noise loor (or any broadband noise signal). See Chapter 5, “Sensitivity and Noise.” 16 Figure 2-12a. Phase noise performance can be optimized for different mea- Figure 2-12b. Detail of the 140-kHz carrier offset region surement conditions

The PXA signal analyzers phase noise optimization can also be set to auto mode, which automatically sets the instrument’s behavior to optimize speed or dynamic range for various operating conditions. When the span is > 44.44 MHz or the RBW is > 1.9 MHz, or the source mode is set to “Tracking,” the PXA selects Fast Tuning mode. Other- wise, the PXA automatically chooses Best Close-In Phase Noise when center frequency < 195 kHz, or when center frequency ≥ 1 MHz and span ≤ 1.3 MHz and RBW ≤ 75 kHz. If these conditions are not met, the PXA automatically chooses Best Wide-Offset Phase Figure 2-13. Phase noise can prevent resolution of unequal signals Noise. These rules apply when you use swept spans, zero span or FFT spans.

In any case, phase noise becomes the ultimate limitation in an analyzer’s ability to resolve signals of unequal amplitude. As shown in Figure 2-13, we may have determined that we can resolve two signals based on the 3-dB bandwidth and selectivity, only to ind that the phase noise covers up the smaller signal.

17 Sweep time Analog resolution ilters If resolution were the only criterion on which we judged a spectrum analyzer, we might design our analyzer with the narrowest possible resolution (IF) ilter and let it go at that. But resolution affects sweep time, and we care very much about sweep time. Sweep time directly affects how long it takes to complete a measurement.

Resolution comes into play because the IF ilters are band-limited circuits that require inite times to charge and discharge. If the mixing products are swept through them too quickly, there will be a loss of displayed amplitude, as shown in Figure 2-14. (See “Envelope detector,” later in this chapter, for another ap- Figure 2-14. Sweeping an analyzer too fast causes a drop in displayed amplitude and a shift in indicated frequency proach to IF response time.) If we think about how long a mixing product stays in the pass band of the IF ilter, that time is directly proportional to bandwidth and inversely proportional to the sweep in Hz The important message here is that per unit time, or: a change in resolution has a dramatic effect on sweep time. Older analog ana- Time in pass band = lyzers typically provided values in a 1, 3, 10 sequence or in ratios roughly equal- RBW = (RBW)(ST) ing the square root of 10. So sweep time Span/ST Span was affected by a factor of about 10 Where with each step in resolution. Keysight X- RBW = resolution bandwidth and Series signal analyzers offer bandwidth ST = sweep time. steps of just 10% for an even better compromise among span, resolution and On the other hand, the rise time of sweep time. a ilter is inversely proportional to its bandwidth, and if we include a constant Spectrum analyzers automatically couple of proportionality, k, then: sweep time to the span and resolution bandwidth settings. Sweep time is k Rise time = adjusted to maintain a calibrated display. RBW If the need arises, we can override If we make the terms equal and solve for the automatic setting and set sweep sweep time, we have: time manually. If you set a sweep time longer than the maximum available, the k (RBW)(ST) analyzer indicates that the display is RBW = Span uncalibrated with a k (Span) “Meas Uncal” message in the upper-right or ST = 2 RBW part of the graticule. For the synchronously-tuned, near- Gaussian ilters used in many analog analyzers, the value of k is in the 2 to 3 range .

18 Digital resolution ilters The digital resolution ilters used in Keysight spectrum analyzers have an effect on sweep time that is different from the effects we’ve just discussed for analog ilters. For swept analysis, the speed of digitally implemented ilters, with no further processing, can show a two to four times improvement.

However, the X-Series signal analyzers with Option FS1 are programmed to correct for the effect of sweeping too fast for resolution bandwidths between about 3 kHz and 100 kHz. As a result, sweep times that would otherwise be many seconds may be reduced to milliseconds, depending upon the particular settings. See Figure 2-14a. The sweep time without the correction would be 79.8 seconds. Figure 2-14b shows a sweep time of 1.506 s with Figure 2-14a. Full span sweep speed, RBW of 20 kHz, without Option FS1 Option FS1 installed. For the widest resolution bandwidths, sweep times are already very short. For example, using the formula with k = 2 on a span of 1 GHz and a RBW of 1 MHz, the sweep time calculates to just 2 msec.

For narrower resolution bandwidths, analyzers such as the Keysight X-Series use fast Fourier transforms (FFTs) to process the data, also producing shorter sweep times than the formula pre- dicts. The difference occurs because the signal being analyzed is processed in frequency blocks, depending upon the particular analyzer. For example, if the frequency block was 1 kHz, then when we select a 10-Hz resolution bandwidth, the analyzer is in effect simultaneously processing the data in each 1-kHz block through 100 contigu- ous 10-Hz ilters. If the digital processing were instantaneous, we would expect sweep time to be reduced by a Figure 2-14b. Full span sweep speed, RBW of 20 kHz, with Option FS1 factor of 100. In practice, the reduction factor is less, but is still signiicant. For more information on the advantages of More information digital processing, refer to Chapter 3. A more detailed discussion about fast sweep measurements can be found in Using Fast-Sweep Techniques to Accelerate Spur Searches – Application Note, literature number 5991-3739EN

19 Envelope detector6 Older analyzers typically converted the IF signal to video with an enve- tt lope detector7. In its simplest form, an envelope detector consists of a diode, IF signal resistive load and low-pass ilter, as shown in Figure 2-15. The output of the

IF chain in this example, an amplitude Figure 2-15. Envelope detector modulated sine wave, is applied to the detector. The response of the detector follows the changes in the envelope of the IF signal, but not the instantaneous value of the IF sine wave itself.

For most measurements, we choose a resolution bandwidth narrow enough to resolve the individual spectral components of the input signal. If we ix the frequency of the LO so that our analyzer is tuned to one of the spectral components of the signal, the output of the IF is a steady sine wave with a constant peak value. The output of Figure 2-16. Output of the envelope detector follows the peaks of the IF signal the envelope detector will then be a constant (DC) voltage, and there is no The width of the resolution (IF) il- gible attenuation due to the roll-off of variation for the detector to follow. ter determines the maximum rate at the ilter8. The analyzer display will vary which the envelope of the IF signal can between a value that is twice the volt- However, there are times when we change. This bandwidth determines age of either (6 dB greater) and zero deliberately choose a resolution band- how far apart two input sinusoids can (minus ininity on the log scale). We width wide enough to include two or be so that after the mixing process they must remember that the two signals more spectral components. At other will both be within the ilter at the same are sine waves (vectors) at different times, we have no choice. The spectral time. Let’s assume a 22.5-MHz inal IF frequencies, and so they continually components are closer in frequency and a 100-kHz bandwidth. Two input change in phase with respect to each than our narrowest bandwidth. As- signals separated by 100 kHz would other. At some time they add exactly in suming only two spectral components produce mixing products of 22.45 and phase; at another, exactly out of phase. within the pass band, we have two sine 22.55 MHz and would meet the criteri- waves interacting to create a beat note, on. See Figure 2-16. The detector must So the envelope detector follows the and the envelope of the IF signal varies, be able to follow the changes in the changing amplitude values of the peaks as shown in Figure 2-16, as the phase envelope created by these two signals of the signal from the IF chain but not between the two sine waves varies. but not the 22.5-MHz IF signal itself. the instantaneous values, resulting in the loss of phase information. This The envelope detector is what makes gives the analyzer its voltmeter charac- More information the spectrum analyzer a voltmeter. teristics. Let’s duplicate the situation above and Additional information on envelope have two equal-amplitude signals in Digitally implemented resolution band- detectors can be found in Spectrum the pass band of the IF at the same widths do not have an analog envelope and Signal Analyzer Measurements and time. A power meter would indicate a detector. Instead, the digital processing Noise–Application Note, literature power level 3 dB above either signal, computes the root sum of the squares number 5966-4008E. that is, the total power of the two. of the I and Q data, which is mathemat- Assume that the two signals are close ically equivalent to an envelope detec- enough so that, with the analyzer tuned tor. For more information on digital half-way between them, there is negli- architecture, refer to Chapter 3.

6. The envelope detector should not be confused with the display detectors. See “Detector types” later in this chapter. 7. A signal whose frequency range extends from zero (DC) to some upper frequency determined by the circuit elements. Historically, spectrum analyzers with analog displays used this signal to drive the vertical delection plates of the CRT directly. Hence it was known as the video signal. 8. For this discussion, we assume the ilter is perfectly rectangular. 20 Displays Keysight (part of Hewlett-Packard to use in spectrum analyzers. Once a Up until the mid-1970s, spectrum at the time) pioneered a variable- trace had been digitized and put into analyzers were purely analog. The persistence storage CRT in which memory, it was permanently available displayed trace presented a continu- we could adjust the fade rate of the for display. It became an easy matter ous indication of the signal envelope, display. When properly adjusted, to update the display at a licker-free and no information was lost. However, the old trace would just fade out at rate without blooming or fading. The analog displays had drawbacks. The the point where the new trace was data in memory was updated at the major problem was in handling the updating the display. This display was sweep rate, and since the contents of long sweep times required for narrow continuous, had no flicker and avoided memory were written to the display at resolution bandwidths. In the extreme confusing overwrites. It worked quite a licker-free rate, we could follow the case, the display became a spot that well, but the intensity and the fade updating as the analyzer swept through moved slowly across the cathode ray rate had to be readjusted for each its selected frequency span just as we tube (CRT), with no real trace on the new measurement situation. When could with analog systems. display. So a meaningful display was digital circuitry became affordable not possible with the longer sweep in the mid-1970s, it was quickly put times.

Figure 2-17. When digitizing an analog signal, what value should be displayed at each point?

21 Detector types With digital displays, we had to decide what value should be displayed for each display data point. No matter how many data points we use across the display, each point must represent what has occurred over some frequency range and, although we usually do not think in terms of time when dealing with a spectrum analyzer, over some time interval.

It is as if the data for each interval is thrown into a bucket and we apply whatever math is necessary to extract the desired bit of information from our input signal. This datum is put into memory and written to the display. This process provides great lexibility.

Here we will discuss six different detector types. Figure 2-18. Each of the 1001 trace points (buckets) covers a 100-kHz frequency span and a 0.01-mil- lisecond time span In Figure 2-18, each bucket contains data from a span and timeframe that is determined by these equations: One bucket

Frequency: Positive peak bucket width = span/(trace points – 1) Time: bucket width = sweep time/(trace points – 1)

The sampling rates are different for various instruments, but greater ac- Sample curacy is obtained from decreasing the span or increasing the sweep time because the number of samples per bucket will increase in either case. Even in analyzers with digital IFs, sample Negative peak rates and interpolation behaviors are designed to be the equivalent of continuous-time processing. Figure 2-19. The trace point saved in memory is based on the detector type algorithm The “bucket” concept is important, as it will help us differentiate the six detec- peak, and negative peak are easy to display an analog system as faithfully tor types: understand and are visually represented as possible using digital techniques. – Sample in Figure 2-19. Normal, average, and Let’s imagine the situation illustrated in –Positive peak (also simply called quasipeak are more complex and will be Figure 2-17. We have a display that con- peak) discussed later. tains only noise and a single CW signal. – Negative peak – Normal Let’s return to the question of how to – Average – Quasipeak

The irst three detectors, sample,

22 Sample detection As a irst method, let us simply select the data point as the instantaneous level at the center of each bucket (see Figure 2-19). This is the sample detection mode. To give the trace a continuous look, we design a system that draws vectors between the points. Figure 2-20. Sample display mode using 10 points Figure 2-21. More points produce a display closer Comparing Figure 2-17 with 2-20, it to display the signal shown in Figure 2-17 to an analog display appears that we get a fairly reasonable display. Of course, the more points there are in the trace, the better the replication of the analog signal will be. The number of available display points can vary for different analyzers. On X-Series signal analyzers, the number of display points for frequency domain traces can be set from a minimum of 1 point to a maximum of 40,001 points. As shown in Figure 2-21, more points do indeed get us closer to the analog signal.

While the sample detection mode does a good job of indicating the randomness of noise, it is not a good mode for analyzing sinusoidal signals. If we were to look at a 100-MHz comb on a Keysight PXA, we might set it to span from 0 to 26.5 GHz. Even with 1,001 display points, each display point represents a span (bucket) of Figure 2-22a. A 10-MHz span of a 250-kHz comb in the sample display mode 26.5 MHz. This is far wider than the maximum 8-MHz resolution bandwidth.

As a result, the true amplitude of a comb tooth is shown only if its mixing product happens to fall at the center of the IF when the sample is taken. Figure 2-22a shows a 10-MHz span with a 750-Hz bandwidth using sample detection. The comb teeth should be relatively equal in amplitude, as shown in Figure 2-22b (using peak detection). Therefore, sample detection does not catch all the signals, nor does it necessarily relect the true peak values of the displayed signals. When resolution bandwidth is narrower than the sample interval (the bucket width), sample mode can give erroneous results.

Figure 2-22b. The actual comb over a 10-MHz span using peak (positive) detection

23 Peak (positive) detection Negative peak detection Normal detection One way to insure that all sinusoids Negative peak detection displays the To provide a better visual display of are reported at their true amplitudes is minimum value encountered in each random noise than offered by peak to display the maximum value en- bucket. It is generally available in most mode and yet avoid the missed-signal countered in each bucket. This is the spectrum analyzers, though it is not problem of the sample mode, the nor- positive peak detection mode, or peak. used as often as other types of detec- mal detection mode (informally known This mode is illustrated in Figure 2-22b. tion. Differentiating CW from impulsive as rosenfell9 mode) is offered on many Peak is the default mode offered on signals in EMC testing is one applica- spectrum analyzers. Should the signal many spectrum analyzers because it tion where negative peak detection is both rise and fall, as determined by the ensures that no sinusoid is missed, valuable. Later in this application note, positive peak and negative peak detec- regardless of the ratio between resolu- we will see how negative peak detectors, the algorithm classiies the signal tion bandwidth and bucket width. tion is also used in signal identiication as noise. However, unlike sample mode, peak routines when you use external mixers does not give a good representation of for high-frequency measurements. In that case, an odd-numbered data random noise because it only displays point displays the maximum value the maximum value in each bucket and encountered during its bucket. And an ignores the true randomness of the even-numbered data point displays the noise. So spectrum analyzers that use minimum value encountered during its peak detection as their primary mode bucket. See Figure 2-25. Normal and generally also offer sample mode as an sample modes are compared in Figures alternative. 2-23a and 2-23b.10

Figure 2-23a. Normal mode Figure 2-23b. Sample mode

9. Rosenfell is not a person’s name but rather a description of the algorithm that tests to see if the signal rose and fell within the bucket represented by a given data point. It is also sometimes written as “rose’n’fell.” 10. Because of its usefulness in measuring noise, the sample detector is usually used in “noise marker” applications. Similarly, the measurement of channel power and adjacent-channel power requires a detector type that gives results unbiased by peak detection. For analyzers without averaging detectors, sample detection is the best choice.

24 What happens when a sinusoidal signal is encountered? We know that as a mixing product is swept past the IF ilter, an analyzer traces out the shape of the ilter on the display. If the ilter shape is spread over many display points, we encounter a situation in which the displayed signal only rises as the mixing product approaches the center frequency of the ilter and only falls as the mixing product moves away from the ilter center frequency. In either of these cases, the positive-peak and negative-peak detectors sense an amplitude change in only one direction, and, according to the normal detection algorithm, the maximum value in each bucket is displayed. See Figure 2-24.

What happens when the resolution bandwidth is narrow, relative to a bucket? The signal will both rise and Figure 2-24. Normal detection displays maximum values in buckets where the signal only rises or only falls fall during the bucket. If the bucket happens to be an odd-numbered one, all is well. The maximum value encountered in the bucket is simply plotted as the next data point. How- ever, if the bucket is even-numbered, then the minimum value in the bucket is plotted. Depending on the ratio of resolution bandwidth to bucket width, the minimum value can differ from the true peak value (the one we want displayed) by a little or a lot. In the extreme, when the bucket is much wider than the resolution bandwidth, the difference between the maximum and minimum values encountered in the bucket is the full difference between the peak signal value and the noise. This is true for the example in Figure 2-25. See bucket 6. The peak value of the previous bucket is always compared to that of the current bucket. The greater of the two values is displayed if the bucket number is odd, as depicted in bucket 7. The signal peak actually occurs in bucket 6 but is not displayed until bucket 7.

25 The normal detection algorithm: If the signal rises and falls within a bucket: Even-numbered buckets display the minimum (negative peak) value in the bucket. The maximum is remembered. Odd-numbered buckets display the maxi(positive peak) value determined by comparing the current bucket peak with the previous (remembered) bucket peak. If the signal only rises or only falls within a bucket, the peak is displayed. See Figure 2-25.

This process may cause a maximum value to be displayed one data point too far to the right, but the offset is usually only a small percentage of the span. Some spectrum analyzers, such as the Keysight PXA signal analyzer, compensate for this potential effect by moving the LO start and stop frequencies. Figure 2-25. Trace points selected by the normal detection algorithm

Another type of error occurs when two peaks are displayed when only one actually exists. Figure 2-26 shows this error. The outline of the two peaks is displayed using peak detection with a wider RBW.

So peak detection is best for locat- ing CW signals well out of the noise. Sample is best for looking at noise, and normal is best for viewing signals and noise.

Figure 2-26. Normal detection can show two peaks when only one peak actually exists

26 Average detection Voltage averaging averages the linear detector. As a result, the average voltage data of the envelope signal detector has a lower variance result for Although modern digital modulation measured during the bucket interval. It the same measurement time. In swept schemes have noise-like character- is often used in EMI testing for measur- analysis, it also allows the convenience istics, sample detection does not ing narrowband signals (this topic will of reducing variance simply by extend- always provide us with the information be discussed further in the next sec- ing the sweep time. we need. For instance, when taking tion). Voltage averaging is also useful a channel power measurement on a for observing rise and fall behavior of W-CDMA signal, integration of the EMI detectors: average and AM or pulse-modulated signals such as quasipeak detection rms values is required. This measure- radar and TDMA transmitters. ment involves summing power across An important application of average a range of analyzer frequency buckets. Log-power (video) averaging averages detection is for characterizing devices Sample detection does not provide this the logarithmic amplitude values (dB) for electromagnetic interference (EMI). capability. of the envelope signal measured during In this case, voltage averaging, as de- the bucket interval. Log power aver- scribed in the previous section, is used While spectrum analyzers typically col- aging is best for observing sinusoidal for measuring narrowband signals that lect amplitude data many times in each signals, especially those near noise.11 might be masked by the presence of bucket, sample detection keeps only broadband impulsive noise. The aver- one of those values and throws away the Thus, using the average detector with age detection used in EMI instruments rest. On the other hand, an averaging the averaging type set to power pro- takes an envelope-detected signal and detector uses all the data values col- vides true average power based upon passes it through a low-pass ilter with lected within the time (and frequency) in- rms voltage, while the average detector a bandwidth much less than the RBW. terval of a bucket. Once we have digitized with the averaging type set to voltage The ilter integrates (averages) the the data, and knowing the circumstances acts as a general-purpose average higher-frequency components such as under which they were digitized, we can detector. The average detector with the noise. To perform this type of detec- manipulate the data in a variety of ways averaging type set to log has no other tion in an older spectrum analyzer that to achieve the desired results. equivalent. doesn’t have a built-in voltage averaging detector function, set the analyzer Some spectrum analyzers refer to the Average detection is an improvement in linear mode and select a video ilter averaging detector as an rms detector over using sample detection for the with a cut-off frequency below the low- when it averages power (based on the determination of power. Sample detec- est PRF of the measured signal. root mean square of voltage). Keysight tion requires multiple sweeps to collect X-Series signal analyzers have an enough data points to give us accurate Quasipeak detectors (QPD) are also average detector that can average the average power information. Average used in EMI testing. QPD is a weighted power, voltage or log of the signal by detection changes channel power form of peak detection. The mea- including a separate control to select measurements from being a summation sured value of the QPD drops as the the averaging type: over a range of buckets into integra- repetition rate of the measured signal tion over the time interval represent- decreases. Thus, an impulsive signal Power (rms) averaging averages rms ing a range of frequencies in a swept with a given peak amplitude and a levels by taking the square root of the analyzer. In a fast Fourier transfer (FFT) 10-Hz pulse repetition rate will have sum of the squares of the voltage data analyzer12, the summation used for a lower quasipeak value than a signal measured during the bucket interval, channel power measurements changes with the same peak amplitude but divided by the characteristic input from being a summation over display having a 1-kHz repetition rate. This impedance of the spectrum analyzer, buckets to being a summation over signal weighting is accomplished by normally 50 ohms. Power averaging FFT bins. In both swept and FFT cases, circuitry with speciic charge, discharge calculates the true average power and the integration captures all the power and display time constants deined by is best for measuring the power of com- information available, rather than just CISPR13. plex signals. that which is sampled by the sample QPD is a way of measuring and quan- 11. See Chapter 5, “Sensitivity and Noise.” tifying the “annoyance factor” of a sig- 12. Refer to Chapter 3 for more information on the FFT analyzers. They perform math computa- nal. Imagine listening to a radio station tions on many buckets simultaneously, which improves measurement speed. suffering from interference. If you hear an 13. CISPR, the International Special Committee on Radio Interference, was established in 1934 by a group of international organizations to address radio interference. CISPR is a non- occasional “pop” caused by noise once governmental group composed of National Committees of the International Electrotechnical Commission (IEC), as well as numerous international organizations. CISPR’s recommended standards generally form the basis for statutory EMC requirements adopted by governmental regulatory agencies around the world.

27 every few seconds, you can still listen to the program without too much trouble. However, if that same amplitude pop occurs 60 times per second, it becomes extremely annoying, making the radio program intolerable to listen to.

Averaging processes There are several processes in a spectrum analyzer that smooth the variations in envelope-detected amplitude. The irst method, average detection, was discussed previously. Two other methods, video iltering and trace averaging, are discussed next.14

Video iltering Figure 2-27. Spectrum analyzers display signal plus noise Discerning signals close to the noise is not just a problem when performing EMC tests. Spectrum analyzers display signals plus their own internal noise, as shown in Figure 2-27. To reduce the effect of noise on the displayed signal amplitude, we often smooth or average the display, as shown in Figure 2-28. Spectrum analyzers include a variable video ilter for this purpose. The video ilter is a low-pass ilter that comes after the envelope detector and determines the bandwidth of the video signal that will later be digitized to yield amplitude data. The cutoff frequency of the video ilter can be reduced to the point where it becomes smaller than the bandwidth of the selected resolution bandwidth (IF) ilter. When this occurs, the video system can no longer follow the more Figure 2-28. Display of Figure 2-27 after full smoothing rapid variations of the envelope of the signal(s) passing through the IF chain. The result is an averaging or smoothing

More information A more detailed discussion about noise markers can be found in Spectrum and Signal Analyzer Measure- ments and Noise – Application Note, literature number 5966-4008E

14. A fourth method, called a noise marker, is discussed in Chapter 5, “Sensitivity and Noise.” Figure 2-29. Smoothing effect of VBW-to-RBW ratios of 3:1, 1:10, and 1:100 28 of the displayed signal.

The effect is most noticeable in measuring noise, particularly when you use a wide-resolution bandwidth. As we reduce the video bandwidth, the peak-to-peak variations of the noise are reduced. As Figure 2-29 shows, the degree of reduction (degree of averaging or smoothing) is a function of the ratio of the video to resolution bandwidths. At ratios of 0.01 or less, the smoothing is very good. At higher ratios, the smoothing is not as good. The video ilter does not affect any part of the trace that is already smooth (for example, a sinusoid displayed well out of the noise).

If we set the analyzer to positive peak detection mode, we notice two things: First, if VBW > RBW, then changing the Figure 2-30a. Positive peak detection mode: reducing video bandwidth lowers peak noise but not resolution bandwidth does not make average noise much difference in the peak-to-peak fluctuations of the noise. Second, if VBW < RBW, changing the video bandwidth seems to affect the noise level. The fluctuations do not change much because the analyzer is displaying only the peak values of the noise. However, the noise level appears to change with video bandwidth because the averaging (smoothing) changes, thereby changing the peak values of the smoothed noise envelope. See Figure 2-30a. When we select average detection, we see the average noise level remains constant. See Figure 2-30b.

Because the video ilter has its own response time, the sweep time increases approximately inversely with video bandwidth when the VBW is less than the resolution bandwidth. The sweep time (ST) can therefore be described by this equation: Figure 2-30b. Average detection mode: noise level remains constant, regardless of VBW-to-RBW ratios k(Span) (3:1, 1:10 and 1:100) ST ≈ (RBW)(VBW)

The analyzer sets the sweep time automatically to account for video bandwidth as well as span and resolution bandwidth.

29 Trace averaging Digital displays offer another choice for smoothing the display: trace averaging. Trace averaging uses a completely different process from the smoothing performed using the average detector. In this case, averaging is accomplished over two or more sweeps on a point- by-point basis. At each display point, the new value is averaged in with the previously averaged data: n– 1 1 A Aavg = ((n ))prior avg + n An where Aavg = new average value Aprior avg = average from prior sweep An= measured value on current sweep n = number of current sweep

Thus, the display gradually converges to an average over a number of sweeps. Figure 2-31. Trace averaging for 1, 5, 20 and 100 sweeps, top to bottom (trace position offset for each set of sweeps As with video iltering, we can select the degree of averaging or smoothing. We do this by setting the number of ing or trace averaging. However, there As a result, we can get signiicantly sweeps over which the averaging oc- is a distinct difference between the different results from the two averag- curs. Figure 2-31 shows trace averaging two. Video iltering performs averaging methods on certain signals. For for different numbers of sweeps. While ing in real time. That is, we see the full example, a signal with a spectrum that trace averaging has no effect on sweep effect of the averaging or smoothing at changes with time can yield a different time, the time to reach a given degree each point on the display as the sweep average on each sweep when we use of averaging is about the same as with progresses. Each point is averaged only video iltering. However, if we choose video iltering because of the number of once, for a time of about 1/VBW on trace averaging over many sweeps, we sweeps required. each sweep. Trace averaging, on the will get a value much closer to the true other hand, requires multiple sweeps average. See Figures 2-32a and 2-32b. In many cases, it does not matter which to achieve the full degree of averaging, form of display smoothing we pick. If and the averaging at each point takes Figures 2-32a and 2-32b show how the signal is noise or a low-level sinu- place over the full time period needed video iltering and trace averaging yield soid very close to the noise, we get the to complete the multiple sweeps. different results on an FM broadcast same results with either video ilter- signal.

Figure 2-32a. Video iltering Figure 2-32b. Trace averaging 30 Time gating Measuring time division signal shown, as in Figure 2-33c. duplex signals Time gating can be achieved using Time-gated spectrum analysis allows three different methods we will discuss you to obtain spectral information To illustrate using time-gating capabil- below. However, there are certain basic about signals occupying the same part ity to perform dificult measurements, concepts of time gating that apply to of the frequency spectrum that are consider Figure 2-33a, which shows a any implementation. In particular, you separated in the time domain. Using simpliied digital mobile-radio signal must have, or be able to set, the fol- an external trigger signal to coordinate in which two radios, #1 and #2, are lowing four items: the separation of these signals, you can time-sharing a single frequency chan- perform the following operations: nel. Each radio transmits a single 1-ms –An externally supplied gate trigger burst, then shuts off while the other signal –Measure any one of several signals radio transmits for 1 ms. The challenge –The gate control or trigger mode separated in time (For example, is to measure the unique frequency (edge or level) (The X-Series signal you can separate the spectra of spectrum of each transmitter. analyzers can be set to gate-trig- two radios time-sharing a single ger holdoff to ignore potential false frequency.) Unfortunately, a traditional spectrum triggers.) –Measure the spectrum of a signal analyzer cannot do that. It simply –The gate delay setting, which de- in one time slot of a TDMA system shows the combined spectrum, as seen termines how long after the trigger –Exclude the spectrum of interfer- in Figure 2-33b. Using the time-gating signal the gate actually becomes ing signals, such as periodic pulse capability and an external trigger active and the signal is observed edge transients that exist for only a signal, you can see the spectrum of just –The gate length setting, which limited time radio #1 (or radio #2 if you wish) and determines how long the gate is on identify it as the source of the spurious and the signal is observed Why time gating is needed Traditional frequency-domain spectrum analysis provides only limited information for certain dificult-to-analyze signals. Examples include the following signal types:

– Pulsed RF – Time multiplexed – Time domain multiple access (TDMA) – Interleaved or intermittent –Burst modulated

In some cases, time-gating capability enables you to perform measurements that would otherwise be very dificult, Figure 2-33a. Simpliied digital mobile-radio signal in the time domain if not impossible to make.

Figure 2-33b. Frequency spectrum of combined Figure 2-33c. The time-gated spectrum of signal Figure 2-33d. The time-gated spectrum of signal signals. Which radio produces the spurious emis- #1 identiies it as the source of spurious emission #2 shows it is free of spurious emissions sions?

31 Controlling these parameters will allow us to look at the spectrum of the signal during a desired portion of the time. If you are fortunate enough to have a gating signal that is only true during the period of interest, you can use level gating, as shown in Figure 2-34. How- ever, in many cases the gating signal will not perfectly coincide with the time we want to measure the spectrum. Therefore, a more lexible approach is to use edge triggering in conjunction with a speciied gate delay and gate length to precisely deine the time period in which to measure the signal. Figure 2-34. Level triggering: the spectrum analyzer only measures the frequency spectrum when the Consider the GSM signal with eight gate trigger signal is above a certain level time slots in Figure 2-35. Each burst is 0.577 ms and the full frame is 4.615 ms. We may be interested in the spectrum of the signal during a speciic time slot. For the purposes of this example, let’s assume we are using only two of the eight available time slots (time slots 1 and 3), as shown in Figure 2-36. When we look at this signal in the frequency domain in Figure 2-37, we observe an unwanted spurious signal present in the spectrum. In order to troubleshoot the problem and ind the source of this interfering signal, we need to determine the time slot in which it is occurring. If we wish to look at time slot 3, we set up the gate to trigger on the rising edge of the burst in time slot 3, and, then specify a gate delay of 1.4577 ms and a gate length of 461.60 µs, as shown in Figure 2-38. The gate delay assures that we only measure the spectrum of time slot 3 while the burst is fully on. Note that the Figure 2-35. A TDMA format signal (in this case, GSM) with 8 time slots, time slot zero is “off”. gate start and stop value is carefully selected to avoid the rising and falling edge of the burst, as we want to allow time for the RBW iltered signal to settle out before we make a measurement. Figure 2-39 shows the spectrum of time slot 3, which reveals that the spurious signal is not caused by this burst.

Three methods are commonly used to perform time gating: –Gated FFT – Gated LO – Gated video

32 Figure 2-36. A zero span (time domain) view of the GSM signal with only time Figure 2-37. Frequency domain view of the GSM signal with 2 time slots “on” slots 1 and 3 “on”. showing an unwanted spurious signal present in the spectrum.

Figure 2-38. Time gating is used to look at the spectrum of the GSM time Figure 2-39. Spectrum of time slot 3 reveals that the spurious signal is not slot 3. caused by this burst.

33 RF IF resolution Envelope Video step bandwidth IF log detector bandwidth Peak/sample Analog-digital attenuator Mixer filter amplifier (IF to video) filter detector converter RF input

Display logic

Local Scan generator oscillator Gate control Display

Figure 2-40. In gated LO mode, the LO sweeps only during gate interval

Gated FFT Gated LO in this chapter. Using an X-Series signal analyzer, a standard, non-gated, spec- The Keysight X-Series signal analyzers LO gating, sometimes referred to as trum sweep over a 1-MHz span takes have built-in FFT capabilities. In this gated sweep, is another technique 14.6 ms, as shown in Figure 2-41. With mode, the data is acquired for an FFT for performing time gating. With this a gate length of 0.3 ms, the spectrum starting at a chosen delay following a method, we control the voltage ramp analyzer sweep must be built up in 49 trigger. The IF signal is digitized and produced by the scan generator to gate intervals (14.6 divided by 0.3). captured for a time period of 1.83 di- sweep the LO, as shown in Figure 2-40. Or, if the full frame of the GSM signal vided by resolution bandwidth. An FFT When the gate is active, the LO ramps is 4.615 ms, the total measurement is computed based on this data acqui- up in frequency like any spectrum time is 49 intervals times 4.615 ms = sition and the results are displayed as analyzer. When the gate is blocked, 226 ms. This represents a signiicant the spectrum. Thus, the spectrum is the voltage out of the scan genera- improvement in speed compared to the that which existed at a particular time tor is frozen, and the LO stops rising gated video technique, which will be of known duration. This is the fastest in frequency. This technique can be described in the following section. LO gating technique when the span is not much faster than gated video because gating is available on X-Series signal wider than the FFT maximum width. multiple buckets can be measured dur- analyzers and PSA Series spectrum ing each burst. As an example, let’s use analyzers. To get the maximum possible fre- the same GSM signal described earlier quency resolution, choose the narrowest available RBW with a capture time that its within the time period of interest. You may not always need that much resolution, however, and you could choose a wider RBW with a corresponding narrower gate length. The minimum usable RBW in gated FFT applications is always lower than the minimum usable RBW in other gating techniques, because the IF must fully settle during the burst in other techniques, which takes longer than 1.83 divided by RBW.

Figure 2-41. Spectrum of the GSM signal

34 Gated video display point, or bucket, so the peak 1.85 s. Some TDMA formats have cycle detector is able to see real data during times as large as 90 ms, resulting in Gated video is the analysis technique that time interval. Otherwise, there will long sweep times using the gated video used in a number of spectrum analyz- be trace points with no data, resulting technique. ers, including the Keysight 8560, 8590 in an incomplete spectrum. Therefore, and ESA Series. In this case, the video the minimum sweep time is N display Now that you’ve seen how a classic voltage is switched off, or to “negative buckets times burst cycle time. For ex- analog spectrum analyzer works and ininity decibels,” during the time the ample, in GSM measurements, the full how to use some of the important fea- gate is supposed to be in its “blocked” frame lasts 4.615 ms. For an ESA spec- tures and capabilities, let’s take a look mode. The detector is set to peak de- trum analyzer set to its default value of at how replacing some analog circuits tection. The sweep time must be set so 401 display points, the minimum sweep with digital technology improves spec- that the gates occur at least once per time for GSM gated video measure- trum analyzer performance. ments would be 401 times 4.615 ms or

RF IF resolution Envelope Video step bandwidth IF log detector bandwidth Peak/sample Analog-digital attenuator Mixer filter amplifier (IF to video) filter detector converter

RF input – dB ∞ Reset Gate control

Display logic

Local Scan generator oscillator

Display

Figure 2-42. Block diagram of a spectrum analyzer with gated video

35 Chapter 3. Digital IF Overview

Since the 1980s, one of the most techniques. As shown in Figure 3-1, A key beneit of the digital processing profound changes in spectrum analy- the linear analog signal is mixed down done in these analyzers is a bandwidth sis has been the application of digital to an 8.5-kHz IF and passed through a selectivity of about 4:1. This selectivity technology to replace portions of bandpass ilter only 1 kHz wide. This IF is available on the narrowest ilters, the spectrum analyzers that had been signal is ampliied, then sampled at an ones we would choose to separate the implemented previously as analog cir- 11.3-kHz rate and digitized. most closely spaced signals. cuits. With the availability of high-performance analog-to-digital converters, Once in digital form, the signal is put In Chapter 2, we did a ilter skirt selec- the latest spectrum analyzers digitize through a fast Fourier transform algo- tivity calculation for two signals spaced incoming signals much earlier in the rithm. To transform the appropriate sig- 4 kHz apart, using a 3-kHz analog signal path compared to spectrum ana- nal, the analyzer must be ixed-tuned ilter. Let’s repeat that calculation us- lyzer designs of just a few years ago. (not sweeping). That is, the transform ing digital ilters. A good model of the The change has been most dramatic in must be done on a time-domain signal. selectivity of digital ilters is a near- the IF section of the spectrum analyzer. Thus the ESA-E Series analyzers step in Gaussian model: Digital IFs1 have had a great impact on 900-Hz increments, instead of sweep- ∆f α H(∆f) = –3.01 dB x spectrum analyzer performance, with ing continuously, when we select one []RBW/2 signiicant improvements in speed, of the digital resolution bandwidths. accuracy and the ability to measure This stepped tuning can be seen on the where H(∆f) is the ilter skirt rejection in dB. complex signals using advanced DSP display, which is updated in 900-Hz techniques. increments as the digital processing is ∆f is the frequency offset from the completed. center in Hz, and α is a parameter that Digital ilters controls selectivity. α = 2 for an ideal As you will see in a moment, other Gaussian ilter. The swept RBW ilters You will ind a partial implementation of spectrum and signal analyzers, such as used in Keysight spectrum analyzers digital IF circuitry in the Keysight ESA- the Keysight X-Series analyzers, use an are based on a near-Gaussian model E Series spectrum analyzers. While the all-digital IF, implementing all resolu- with an α value equal to 2.12, resulting 1-kHz and wider RBWs are implement- tion bandwidth ilters digitally. in a selectivity ratio of 4.1:1. ed with traditional analog LC and crystal ilters, the narrowest bandwidths (1 Hz to 300 Hz) are realized using digital

Log 21.4 MHz

Video ADC µC

Linear

3rd LO Sample and hold at 11.3 kHz

8.5 kHz CF 1 kHz BW

Figure 3-1. Digital implementation of 1-, 3-, 10-, 30-, 100- and 300-Hz resolution ilters in ESA-E Series spectrum analyzers

1. Strictly speaking, once a signal has been digitized, it is no longer at an intermediate frequency, or IF. At that point, the signal is represented by digital data values. However, we use the term “digital IF” to describe the digital processing that replaces the analog IF processing found in traditional spectrum analyzers.

36 Entering the values from our example IF in the X-Series signal analyzer, as of the ADC clock (30 MHz) through into the equation, we get: shown in Figure 3-2. the anti-alias ilter. The delay allows 4000 2.12 time for an impending large signal to H(4 kHz) = –3.01 dB x []3000/2 In this case, all 160 resolution band- be recognized before it overloads the = –24.1 dB widths are digitally implemented. ADC. The logic circuitry controlling At an offset of 4 kHz, the 3-kHz digital However, there is some analog circuitry the autorange detector will decrease ilter is down –24.1 dB compared to the prior to the ADC, starting with several the gain in front of the ADC before a analog ilter which was only down –14.8 stages of down conversion, followed large signal reaches it, thus preventing dB. Because of its superior selectiv- by a pair of single-pole preilters (one clipping. If the signal envelope remains ity, the digital ilter can resolve more an LC ilter, the other crystal-based). small for a long time, the autoranging closely spaced signals. A preilter helps prevent succeeding circuit increases the gain, reducing the stages from contributing third-order effective noise at the input. The digital All-digital IF distortion in the same way a preilter gain after the ADC is also changed to would in an analog IF. In addition, it compensate for the analog gain in front Analyzers such as the Keysight X- enables dynamic range extension via of it. The result is a “loating point” Series combine several digital tech- autoranging. The output of the single- ADC with very wide dynamic range niques to achieve the all-digital IF. The pole preilter is routed to the autorange when autoranging is enabled in swept all-digital IF offers users a wealth of detector and the anti-alias ilter. mode. advantages. The combination of FFT analysis for narrow spans and swept As with any FFT-based IF architec- analysis for wider spans optimizes ture, the anti-alias ilter is required to sweeps for the fastest possible mea- prevent aliasing (the folding of out- surements. Architecturally, the ADC is of-band signals into the ADC sampled moved closer to the input port, a move data). This ilter has many poles and made possible by improvements to the thus has substantial group delay. Even A-to-D converters and other digital a very fast-rising RF burst, downcon- hardware. Let’s begin by taking a look verted to the IF frequency, will experi- at the block diagram of the all-digital ence a delay of more than three cycles

Custom IC

Anti-alias Counter ﬁlter I I, Q Display ADC VBW det Q r, log (r) log pwr pwr log Hilbert log v v log Ranging –1 transform Preﬁlter rules log log log log Autoranging ADC system

RISC processor

Display FFT Processing log/lin dB/div Display

Figure 3-2. Block diagram of the all-digital IF in the Keysight X-Series signal analyzers

37 Figure 3-3 illustrates the sweeping Amplitude behavior of the X-Series analyzers. The (log) single-pole preilter allows the gain to be turned up high when the analyzer is tuned far from the carrier. As the car- ADC rier gets closer, the gain falls and the clipping threshold ADC quantization noise rises. The noise Preﬁlter gain level will depend on the signal level Typical analog IF Digital IF RBW response frequency separation from the carrier, response Noise ﬂoor after autoranging so it looks like a step-shaped phase Typical LO phase noise noise. However, phase noise is different from this autoranging noise. Phase noise cannot be avoided in a spectrum analyzer. However, reducing the preil- Frequency or time ter width can reduce autoranging noise at most frequency offsets from the carrier. Since the preilter width is approxi- Figure 3-3. Autoranging keeps ADC noise close to the carrier and lower than LO noise or RBW ilter response mately 2.5 times the RBW, reducing the RBW reduces the autoranging noise. For swept analysis, the iltered I and power during the time the LO sweeps Q pairs are converted to magnitude through that point. The VBW ilter can Custom digital signal and phase pairs. For traditional swept be reconigured into an accumulator to processing analysis, the magnitude signal is video- perform averaging on either a log, volt- bandwidth (VBW) iltered and samples age or power scale. Turning back to the block diagram of are taken through the display detector the digital IF (Figure 3-2), after the ADC circuit. The log/linear display selection Frequency counting gain has been set with analog gain and and dB/division scaling occur in the corrected with digital gain, a custom processor, so a trace can be displayed Swept spectrum analyzers usually IC begins processing the samples. on any scale without remeasuring. have a frequency counter. This counter First, it splits the 30-MHz IF samples counts the zero crossings in the IF sig- into I and Q pairs at half the rate (15 Additional video processing nal and offsets that count by the known Mpairs/s). The I and Q pairs are given frequency offsets from LOs in the rest a high-frequency boost with a single- features of the conversion chain. If the count is stage digital ilter that has gain and The VBW ilter normally smoothes the allowed to run for a second, you can phase approximately opposite to that of log of the magnitude of the signal, but achieve a resolution of 1 Hz. the single-pole analog preilter. Next, I it has many additional features. It can and Q signals are low-pass iltered with convert the log magnitude to a voltage Because of its digitally synthesized a linear-phase ilter with nearly ideal envelope before iltering and convert LOs and all-digital RBWs, the native Gaussian response. Gaussian ilters it back for consistent behavior before frequency accuracy of the X-Series have always been used for swept spec- display detection. signal analyzer is very good (0.1% of trum analysis, because of their optimum span). In addition, the X-Series signal compromise between frequency domain Filtering the magnitude on a linear analyzer includes a frequency counter performance (shape factor) and time- voltage scale is desirable for observ- that observes not just zero crossings, domain performance (response to rapid ing pulsed-RF envelope shapes in zero but also the change in phase. Thus, it sweeps). With the signal bandwidth span. The log-magnitude signal also can resolve frequency to the tens-of- now reduced, the I and Q pairs may be can be converted to a power (magni- millihertz level in 0.1 second. With this decimated and sent to the processor tude squared) signal before iltering, design, the ability to resolve frequency for FFT processing or demodulation. Al- and then it can be converted back. changes is not limited by the spectrum though FFTs can be performed to cover Filtering the power allows the analyzer analyzer, but rather is determined a segment of frequency span up to the to give the same average response to by the noisiness of the signal being 10-MHz bandwidth of the anti-alias signals with noise-like characteristics, counted. ilter, even a narrower FFT span, such such as digital communications signals, as 1 kHz, with a narrow RBW, such as 1 as to CW signals with the same rms Hz, would require FFTs with 20 million voltage. An increasingly common mea- data points. Using decimation for nar- surement need is total power in a chan- rower spans, the number of data points nel or across a frequency range. In a needed to compute the FFT is greatly measurement such as this, the display reduced, speeding up computations. points might represent the average

38 More advantages of all-digital IF all digital IF implementations is speciied preilter. Again, the preilter is highly at ± 0.07 dB for any level up to –20 dBm stable and contributes only 20 percent We have already discussed a number of at the input mixer of the analyzer. The of the error that would exist with an advantages of signal analyzers with all- range of the log amp does not limit the RBW made of ive such stages. As a re- digital IF: power/voltage/log video ilter- log idelity at low levels, as it would be sult, most RBWs are within 2 percent of ing, high-resolution frequency counting, in an analog IF; the range is only limited their stated bandwidth, compared to 10 log/linear switching of stored traces, by noise around –155 dBm at the input to 20 percent speciications in analog-IF excellent shape factors, an average- mixer. Because of single-tone compres- analyzers. across-the display-point detector mode, sion in upstream circuits at higher pow- 160 RBWs, and of course, FFT or swept ers, the idelity speciication degrades Bandwidth accuracy is important for processing. In spectrum analysis, the to ± 0.13 dB for signal levels up to –10 minimizing the inaccuracy of chan- iltering action of RBW ilters causes dBm at the input mixer. By comparison, nel power measurements and similar errors in frequency and amplitude analog log amps are usually speciied measurements. The noise bandwidth measurements that are a function of the with tolerances in the ± 1 dB region. of the RBW ilters is known to much sweep rate. For a ixed level of these better speciications than the 2 percent errors, the all-digital IF’s linear phase Other IF-related accuracies are im- setting tolerance, and noise markers RBW ilters allow faster sweep rates proved as well. The IF preilter is analog and channel-power measurements are than analog ilters permit. The digital and must be aligned like an analog ilter, corrected to a tolerance of ± 0.5 per- implementation also allows well-known so it is subject to alignment errors, but it cent. Therefore, bandwidth uncertain- compensations to frequency and am- is much better than most analog ilters. ties contribute only ± 0.022 dB to the plitude readout, permitting sweep rates With only one stage to manufacture, amplitude error of noise density and typically twice as fast as older analyz- that stage can be made much more channel-power measurements. ers and excellent performance at even stable than the 4- and 5-stage ilters of four times the sweep speed. Keysight analog IF-based spectrum analyzers. Finally, with no analog reference-level- X-Series signal analyzers can achieve As a result, the gain variations between dependent gain stages, there is no “IF over 50 times faster sweep speeds (see RBW ilters is held to a speciication of gain” error at all. The sum of all these Chapter 2 - Digital resolution ilters). ± 0.03 dB for general digital IF imple- improvements means that the all-digital mentations, which is ten times better IF makes a quantum improvement in Digitally implemented logarithmic than all-analog designs. spectrum analyzer accuracy. It also ampliication is very accurate. Typical allows you to change analyzer settings errors of the entire analyzer are much The accuracy of the IF bandwidth is without signiicantly impacting mea- smaller than the measurement uncer- determined by settability limitations surement uncertainty. We will cover this tainty with which the manufacturer in the digital part of the iltering and topic in more detail in the next chapter. proves the log idelity. The log idelity on calibration uncertainties in the analog

39 Chapter 4. Amplitude and Frequency Accuracy

Now let’s look at amplitude accuracy, Impedance mismatch is an important Spectrum analyzer data sheets typi- or perhaps better, amplitude uncer- factor in measurement uncertainty that cally specify the input voltage stand- tainty. Most spectrum analyzers are is often overlooked. Analyzers do not ing wave ratio (VSWR). Knowing the speciied in terms of both absolute and have perfect input impedances, and VSWR, we can calculate ρ with the relative accuracy. However, relative signal sources do not have ideal output following equation: performance affects both, so let’s look impedances. When a mismatch exists, (VSWR–1) irst at factors affecting relative mea- the incident and relected signal vec- ρ = (VSWR+1) surement uncertainty. tors may add constructively or destruc- tively. Thus the signal received by the As an example, consider a spectrum Before we discuss these uncertainties, analyzer can be larger or smaller than analyzer with an input VSWR of 1.2 and let’s look again at the block diagram the original signal. In most cases, un- a device under test (DUT) with a VSWR of an analog swept-tuned spectrum certainty due to mismatch is relatively of 1.4 at its output port. The resulting analyzer, shown in Figure 4-1, and see small. However, as spectrum analyzer mismatch error would be ±0.13 dB. which components contribute to the amplitude accuracy has improved uncertainties. Later in this chapter, we dramatically in recent years, mismatch will see how a digital IF and various uncertainty now constitutes a more correction and calibration techniques signiicant part of total measurement can substantially reduce measurement uncertainty. In any case, improving the More information uncertainty. match of either the source or analyzer For more information about how reduces uncertainty. improving the match of either Components that contribute to uncer- the source or analyzer reduces tainty: The general expression used to calculate uncertainty, see the Keysight –Input connector (mismatch) the maximum mismatch error in dB is: PSA Performance Spectrum Analyzer –RF input attenuator Series Amplitude Accuracy – Technical – Mixer and input ilter (latness) Error (dB) = –20 log[1 ± |(ρanalyzer)(ρsource)|] Overview literature number 5980- – IF gain/attenuation (reference level ) 3080EN. – RBW ilters where ρ is the relection coeficient. – Display scale idelity – Calibrator (not shown)

RF input Log Envelope attenuator Mixer IF gain IF ﬁlter amp detector

Input signal

Pre-selector, or Video low-pass ﬁlter ﬁlter Local oscillator

Reference oscillator

Sweep generator Display

Figure 4-1. Spectrum analyzer block diagram

40 Since the analyzer’s worst-case match like in one frequency band. Frequency contribute to frequency response. occurs when its input attenuator is response is usually speciied as ± x dB However, some amplitude uncertainty set to 0 dB, we should avoid the 0 dB relative to the midpoint between the is always introduced and it depends on setting if we can. Alternatively, we can extremes. The frequency response of how accurately the IF ampliier and at- attach a well-matched pad (attenua- a spectrum analyzer represents the tenuator can be set to a desired value. tor) to the analyzer input and greatly overall system performance resulting This uncertainty is known as reference reduce mismatch as a factor. Adding from the latness characteristics and level accuracy. attenuation is a technique that works interactions of individual components well to reduce measurement uncertain- in the signal path up to and including Another parameter we might change ty when the signal we wish to measure the irst mixer. Microwave spectrum during the course of a measurement is well above the noise. However, in analyzers use more than one frequency is resolution bandwidth. Different cases where the signal-to-noise ratio is band to go above 3.6 GHz. This is done ilters have different insertion losses. small (typically ≤ 7 dB), adding attenu- by using a higher harmonic of the local Generally, we see the greatest differ- ation will increase measurement error oscillator, which will be discussed in ence when switching between LC ilters because the noise power adds to the detail in Chapter 7. When making rela- (typically used for the wider resolution signal power, resulting in an errone- tive measurements between signals in bandwidths) and crystal ilters (used ously high reading. different frequency bands, you must for narrow bandwidths). This results add the frequency response of each in resolution bandwidth switching Let’s turn our attention to the input at- band to determine the overall frequency uncertainty. tenuator. Some relative measurements response uncertainty. In addition, some are made with different attenuator set- spectrum analyzers have a band switch- The most common way to display sig- tings. In these cases, we must consider ing uncertainty which must be added to nals on a spectrum analyzer is to use the input attenuation switching uncer- the overall measurement uncertainty. a logarithmic amplitude scale, such as tainty. Because an RF input attenuator 10 dB per div or 1 dB per div. Therefore, must operate over the entire frequency After the input signal is converted to the IF signal usually passes through a range of the analyzer, its step accuracy an IF, it passes through the IF gain log ampliier. The gain characteristic of varies with frequency. The attenuator ampliier and IF attenuator, which are the log ampliier approximates a loga- also contributes to the overall fre- adjusted to compensate for changes rithmic curve. So any deviation from a quency response. At 1 GHz, we expect in the RF attenuator setting and mixer perfect logarithmic response adds to the attenuator performance to be quite conversion loss. Input signal ampli- the amplitude uncertainty. Similarly, good; at 26 GHz, not as good. tudes are thus referenced to the top when the spectrum analyzer is in linear line of the graticule on the display, mode, the linear ampliiers do not have The next component in the signal path known as the reference level. The IF a perfect linear response. This type is the input ilter. Spectrum analyzers ampliier and attenuator work only at of uncertainty is called display scale use a ixed low-pass ilter in the low one frequency and, therefore, do not idelity. band and a tunable bandpass ilter called a preselector (we will discuss the preselector in more detail in Chapter Frequency response 7) in the higher frequency bands. The Signals in the same harmonic band low-pass ilter has a better frequency response than the preselector and adds +0.5 dB a small amount of uncertainty to the frequency response error. A preselector, usually a YIG-tuned ilter, has a 0 larger frequency response variation, ranging from 1.5 dB to 3 dB at millimeter-wave frequencies. - 0.5 dB BAND 1 Following the input ilter are the mixer Speciﬁcation: 0.5 dB and the local oscillator, both of which add to the frequency response un- Figure 4-2. Relative frequency response in a single band certainty. Figure 4-2 illustrates what the frequency response might look

41 Relative uncertainty Absolute amplitude accuracy It is best to consider all known uncertainties and then determine which ones When we make relative measure- Almost all spectrum analyzers have a can be ignored when making a certain ments on an incoming signal, we use built-in calibration source that provides type of measurement. The range of val- either some part of the same signal a known reference signal of speciied ues shown in Table 4-1 represents the or a different signal as a reference. amplitude and frequency. We rely on speciications of a variety of spectrum For example, when we make second the relative accuracy of the analyzer analyzers. harmonic distortion measurements, we to translate the absolute calibration of use the fundamental of the signal as the reference to other frequencies and Some of the speciications, such as our reference. Absolute values do not amplitudes. Spectrum analyzers often frequency response, are frequency- come into play; we are interested only have an absolute frequency response range dependent. A 3-GHz RF analyzer in how the second harmonic differs in speciication, where the zero point might have a frequency response of amplitude from the fundamental. on the latness curve is referenced to ± 0.38 dB, while a microwave spectrum this calibration signal. Many Keysight analyzer tuning in the 26-GHz range In a worst-case relative measurement spectrum analyzers use a 50-MHz could have a frequency response of scenario, the fundamental of the signal reference signal. At this frequency, the ± 2.5 dB or higher. On the other hand, may occur at a point where the fre- speciied absolute amplitude accuracy other sources of uncertainty, such as quency response is highest, while the is extremely good: ± 0.28 dB for the changing resolution bandwidths, apply harmonic we wish to measure occurs at X-Series signal analyzers. equally to all frequencies. the point where the frequency response is the lowest. The opposite scenario is equally likely. Therefore, if our relative frequency response speciication is Table 4-1. Representative values of amplitude uncertainty for common spectrum analyzers ± 0.5 dB, as shown in Figure 4-2, then the total uncertainty would be twice Amplitude uncertainties (± dB) that value, or ± 1.0 dB. Relative RF attenuator switching uncertainty 0.18 to 0.7 Perhaps the two signals under test Frequency response 0.38 to 2.5 are in different frequency bands of the spectrum analyzer. In that case, a Reference level accuracy (IF attenuator/gain change) 0.0 to 0.7 rigorous analysis of the overall uncer- Resolution bandwidth switching uncertainty 0.03 to 1.0 tainty must include the sum of the lat- Display scale idelity 0.07 to 1.15 ness uncertainties of the two frequency Absolute bands. Calibrator accuracy 0.24 to 0.34 Other uncertainties might be irrelevant in a relative measurement, like RBW switching uncertainty or reference level accuracy, which apply to both signals at the same time.

42 Improving overall uncertainty ers available today have self-calibration Nominal values indicate expected per- routines. These routines generate error formance or describe product perfor- When we look at total measurement coeficients (for example, amplitude mance that is useful in the application uncertainty for the irst time, we may changes versus resolution bandwidth) of the product, but is not covered by well be concerned as we add up the un- that the analyzer later uses to correct the product warranty. Nominal param- certainty igures. The worst-case view measured data. As a result, these self- eters generally are not tested during assumes each source of uncertainty calibration routines allow us to make the manufacturing process. for your spectrum analyzer is at the good amplitude measurements with a maximum speciied value, and all are spectrum analyzer and give us more biased in the same direction at the same Digital IF architecture and freedom to change controls during the uncertainties time. The sources of uncertainty can be course of a measurement. considered independent variables, so it As described in the previous chapter, is likely that some errors will be positive a digital IF architecture eliminates or while others will be negative. Therefore, Speciications, typical performance and nominal values minimizes many of the uncertainties a common practice is to calculate the experienced in analog spectrum ana- root sum of squares (RSS) error. When evaluating spectrum analyzer lyzers. These include: accuracy, it is important to have a clear Regardless of whether we calculate the understanding of the many different worst-case or RSS error, we can take Reference level accuracy (IF values found on an analyzer data sheet. gain uncertainty) steps to improve the situation. First of Keysight deines three classes of instru- all, we should know the speciications ment performance data: Spectrum analyzers with an all-digital for our particular spectrum analyzer. IF, such as the Keysight X-Series, do These speciications may be good Speciications describe the perfor- not have IF gain that changes with enough over the range in which we are mance of parameters covered by the reference level. Therefore, there is no making our measurement. If not, Table product warranty over a temperature IF gain uncertainty. 4-1 suggests some opportunities to range of 0 to 55 °C (unless otherwise improve accuracy. noted). Each instrument is tested to Display scale idelity verify it meets the speciication and Before taking any data, we can step takes into account the measurement A digital IF architecture does not through a measurement to see if any uncertainty of the equipment used to include a log ampliier. Instead, the log controls can be left unchanged. We test the instrument. All of the units function is performed mathematically, might ind that the measurement can tested will meet the speciication. and traditional log idelity uncertainty be made without changing the RF at- does not exist. However, other factors, tenuator setting, resolution bandwidth Some test equipment manufacturers such as RF compression (especially or reference level. If so, all uncertain- use a “2 sigma” or 95% conidence for input signals above –20 dBm), ADC ties associated with changing these value for certain instrument specii- range gain alignment accuracy and controls drop out. We may be able cations. When evaluating data sheet ADC linearity (or quantization error) to trade off reference level accuracy speciications for instruments from contribute to display scale uncertainty. against display idelity, using whichever different manufacturers, it is important The quantization error can be im- is more accurate and eliminating the to make sure you are comparing like proved by the addition of noise, which other as an uncertainty factor. We can numbers in order to make an accurate smoothes the average of the ADC even get around frequency response comparison. transfer function. This added noise is if we are willing to go to the trouble of called dither. While the dither improves 2 characterizing our particular analyzer . Typical performance describes ad- linearity, it does slightly degrade the You can accomplish this by using a ditional product performance informa- displayed average noise level. In the power meter and comparing the read- tion that is not covered by the product PSA Series analyzers, we generally ing of the spectrum analyzer at the warranty. It is performance beyond recommend you use dither when the desired frequencies with the reading of speciication that 80% of the units ex- measured signal has a signal-to-noise the power meter. hibit with a 95% conidence level over ratio of greater than or equal to 10 the temperature range 20 to 30 °C. dB. When the signal-to-noise ratio is The same applies to the calibrator. If Typical performance does not include less than 10 dB, the degradations to we have a more accurate calibrator, or measurement uncertainty. During accuracy of any single measurement one closer to the frequency of interest, manufacture, all instruments are tested (in other words, without averaging) we may wish to use that in lieu of the for typical performance parameters. that come from a higher noise floor built-in calibrator. Finally, many analyz- are worse than the linearity problems solved by adding dither, so dither is 2. Should we do so, then mismatch may become a more signiicant error. best turned off.

43 RBW switching uncertainty Table 4-2. Amplitude uncertainties when measuring a 1-GHz signal

The digital IF in the X-Series signal Source of uncertainty Absolute uncertainty of 1-GHz, –20-dBm signal analyzers includes an analog preilter N9030A PXA N9020A MXA N9010A EXA set to 2.5 times the desired resolution bandwidth. This preilter has some un- Absolute amplitude accuracy ± 0.24 dB ± 0.33 dB ± 0.40 dB certainty in bandwidth, gain and center Frequency response ± 0.35 dB ± 0.45 dB ± 0.60 dB frequency as a function of the RBW Total worst-case uncertainty ± 0.59 dB ± 0.78 dB ± 1.00 dB setting. The rest of the RBW iltering is Total RSS uncertainty ± 0.42 dB ± 0.56 dB ± 0.72 dB done digitally in an ASIC in the digital IF section. Though the digital ilters are not perfect, they are very repeatable, and some compensation is applied to Table 4-3. Absolute and relative amplitude accuracy comparison (8563EC and N9030A PXA) minimize the error. This results in a Source of uncertainty Measurement of a 10-GHz signal at –10 dBm tremendous overall improvement to the RBW switching uncertainty compared Absolute uncertainty of Relative uncertainty of second to analog implementations. fundamental at 10 GHz harmonic at 20 GHz 8563EC N9030A PXA 8563EC N9030A PXA Amplitude uncertainty Calibrator ± 0.3 dB N/A N/A N/A examples Absolute amplitude N/A ± 0.24 dB N/A N/A accuracy Let’s look at some amplitude uncertainty examples for various measure- Attenuator N/A N/A N/A N/A ments. Suppose we want to measure a Frequency response ± 2.9 dB ± 2.0 dB ± (2.2 + 2.5) dB ± (2.0 + 2.0) dB 1-GHz RF signal with an amplitude of Band switching uncer- N/A N/A ± 1.0 dB N/A –20 dBm. If we use a Keysight PXA X- tainty Series signal analyzer with Atten = 10 IF gain N/A N/A N/A N/A dB, RBW = 1 kHz, VBW = 1 kHz, Span = 20 kHz, Ref level = –20 dBm, log scale, RBW switching N/A ± 0.03 dB N/A N/A and coupled sweep time, and an ambi- Display scale idelity N/A ± 0.07 dB ± 0.85 dB ± 0.07 dB ent temperature of 20 to 30 °C, the Total worst-case ± 3.20 dB ± 2.34 dB ± 6.55 dB ± 4.07 dB speciications tell us that the absolute uncertainty uncertainty equals ± 0.24 dB plus the Total RSS uncertainty ± 2.91 dB ± 2.02 dB ± 3.17 dB ± 2.83 dB absolute frequency response. The MXA X-Series signal analyzer measuring the same signal using the same settings would have a speciied uncertainty of ± 0.33 plus the absolute frequency re- Frequency accuracy sponse. These values are summarized Up until the late 1970s, absolute in Table 4-2. So far, we have focused almost exclu- frequency uncertainty was measured sively on amplitude measurements. in megahertz because the irst LO was At higher frequencies, the uncertainties What about frequency measurements? a high-frequency oscillator operating get larger. In this example, we want to Again, we can classify two broad above the RF range of the analyzer, and measure a 10-GHz signal with an am- categories, absolute and relative there was no attempt to tie the LO to plitude of –10 dBm. In addition, we also frequency measurements. Absolute a more accurate reference oscillator. want to measure its second harmonic measurements are used to measure Today’s LOs are synthesized to provide at 20 GHz. Assume the following mea- the frequencies of speciic signals. For better accuracy. Absolute frequency surement conditions: 0 to 55 °C, RBW example, we might want to measure uncertainty is often described under = 300 kHz, Atten = 10 dB, Ref level = a radio broadcast signal to verify it is the frequency readout accuracy speci- –10 dBm. In Table 4-3, we compare operating at its assigned frequency. ication and refers to center frequency, the absolute and relative amplitude Absolute measurements are also used start, stop and marker frequencies. uncertainty of two different Keysight to analyze undesired signals, such as spectrum and signal analyzers, an when you search for spurs. Relative With the introduction of the Keysight 8563EC (with analog IF) and N9030A measurements, on the other hand, 8568A in 1977, counter-like frequency PXA (with digital IF). are useful for discovering the distance accuracy became available in a gen- between spectral components or the eral-purpose spectrum analyzer, and modulation frequency. ovenized oscillators were used to reduce

44 drift. Over the years, crystal reference In a factory setting, there is often an careful to place the marker exactly at oscillators with various forms of indirect in-house frequency standard available the peak of the response to a spectral synthesis have been added to analyz- that is traceable to a national standard. component. If we place the marker at ers in all cost ranges. The broadest Most analyzers with internal reference some other point on the response, we deinition of indirect synthesis is that the oscillators allow you to use an external will get a different frequency reading. frequency of the oscillator in question is reference. The frequency reference For the best accuracy, we may narrow in some way determined by a reference error in the foregoing expression then the span and resolution bandwidth to oscillator. This includes techniques such becomes the error of the in-house minimize their effects and to make it as phase lock, frequency discrimination standard. easier to place the marker at the peak and counter lock. of the response. When you make relative measure- What we care about is the effect ments, span accuracy comes into play. Many analyzers have marker modes these changes have had on frequency For Keysight analyzers, span accuracy that include internal counter schemes accuracy (and drift). A typical readout generally means the uncertainty in the to eliminate the effects of span and accuracy might be stated: indicated separation of any two spectral resolution bandwidth on frequency components on the display. For exam- accuracy. The counter does not count ± [(freq readout x freq ref error) + A% of ple, suppose span accuracy is 0.5% of the input signal directly, but instead span + B% of RBW + C Hz] span and we have two signals separated counts the IF signal and perhaps one by two divisions in a 1-MHz span (100 or more of the LOs, and the processor Note that we cannot determine an kHz per division). The uncertainty of the computes the frequency of the input exact frequency error unless we know signal separation would be 5 kHz. The signal. A minimum signal-to-noise ratio something about the frequency refer- uncertainty would be the same if we is required to eliminate noise as a fac- ence. In most cases, we are given an used delta markers and the delta read- tor in the count. Counting the signal in annual aging rate, such as ± 1 x 10–7 ing was 200 kHz. So we would measure the IF also eliminates the need to place per year, though sometimes aging is 200 kHz ± 5 kHz. the marker at the exact peak of the given over a shorter period (for ex- signal response on the display. If you ample, ± 5 x 10–10 per day). In addition, When making measurements in the are using this marker counter function, we need to know when the oscillator ield, we typically want to turn our ana- placement anywhere near the peak of was last adjusted and how close it was lyzer on, complete our task, and move the signal suficiently out of the noise set to its nominal frequency (usually 10 on as quickly as possible. It is helpful to will do. Marker count accuracy might MHz). Other factors that we often over- know how the reference in our analyzer be stated as: look when we think about frequency behaves under short warm-up condi- accuracy include how long the refer- tions. For example, the Keysight ESA-E ± [(marker freq x freq ref error) ence oscillator has been operating. Series portable spectrum analyzers will + counter resolution] Many oscillators take 24 to 72 hours meet published speciications after a to reach their speciied drift rate. To 5-minute warm up. We must still deal with the frequency minimize this effect, some spectrum reference error, as we previously analyzers continue to provide power Most analyzers offer markers you can discussed. Counter resolution refers to to the reference oscillator as long as put on a signal to see amplitude and the least-signiicant digit in the counter the instrument is plugged into the AC absolute frequency. However, the readout, a factor here just as with any power line. In this case, the instrument indicated frequency of the marker is a simple digital counter. Some analyzers is not really turned “off.” It is more function of the frequency calibration of allow you to use the counter mode with accurate to say it is on “standby.” We the display, the location of the marker delta markers. In that case, the effects also need to consider the temperature on the display and the number of of counter resolution and the ixed stability, as it can be worse than the display points selected. Also, to get the frequency would be doubled. drift rate. In short, there are a number best frequency accuracy, we must be of factors to consider before we can determine frequency uncertainty.

45 Chapter 5. Sensitivity and Noise

Sensitivity high enough in amplitude that the noise it is well above the effective (displayed) generated in subsequent gain stages noise loor. The effective input noise One of the primary ways engineers use adds only a small amount to the total loor includes the losses caused by the spectrum analyzers is for searching out noise power. The input attenuator and input attenuator, mixer conversion loss, and measuring low-level signals. The one or more mixers may be between and other circuit elements prior to the limitation in these measurements is the the input connector of a spectrum ana- irst gain stage. We cannot do anything noise generated within the spectrum lyzer and the irst stage of gain, and all about the conversion loss of the mix- analyzer itself. This noise, generated by of these components generate noise. ers, but we can change the RF input the random electron motion in various However, the noise they generate is attenuator. This enables us to control circuit elements, is ampliied by multi- at or near the absolute minimum of the input signal power to the irst mixer ple gain stages in the analyzer and ap- –174 dBm/Hz, so they do not signii- and thus change the displayed signal- pears on the display as a noise signal. cantly affect the noise level input to the to-noise loor ratio. Clearly, we get the On a spectrum analyzer, this noise is irst gain stage, and its ampliication is lowest DANL by selecting minimum commonly referred to as the displayed typically insigniicant. (zero) RF attenuation. average noise level, or DANL1. The noise power observed in the DANL is a While the input attenuator, mixer and Because the input attenuator has no combination of thermal noise and the other circuit elements between the effect on the actual noise generated in noise igure of the spectrum analyzer. input connector and irst gain stage the system, some early spectrum ana- While there are techniques to measure have little effect on the actual system lyzers simply left the displayed noise signals slightly below the DANL, this noise, they do have a marked effect at the same position on the display noise power ultimately limits our ability on the ability of an analyzer to display regardless of the input attenuator set- to make measurements of low-level low-level signals because they at- ting. That is, the IF gain remained con- signals. tenuate the input signal. That is, they stant. In this case, the input attenuator reduce the signal-to-noise ratio and so affected the location of a true input Let’s assume a 50-ohm termination degrade sensitivity. signal on the display. As input attenu- is attached to the spectrum analyzer ation was increased, further attenuat- input to prevent any unwanted signals We can determine the DANL simply by ing the input signal, the location of the from entering the analyzer. This passive noting the noise level indicated on the signal on the display went down while termination generates a small amount display when the spectrum analyzer the noise remained stationary. of noise energy equal to kTB, where: input is terminated with a 50-ohm load. This level is the spectrum analyzer’s k = Boltzmann’s constant own noise loor. Signals below this level –23 (1.38 x 10 joule/K) are masked by the noise and cannot be T = temperature, in Kelvin seen. However, the DANL is not the ac- B = bandwidth in which the noise is tual noise level at the input, but rather measured, in Hertz the effective noise level. An analyzer display is calibrated to relect the level The total noise power is a function of of a signal at the analyzer input, so measurement bandwidth, so the value the displayed noise loor represents a is typically normalized to a 1-Hz band- ictitious or effective noise loor at the width. Therefore, at room temperature, input. the noise power density is –174 dBm/ Hz. When this noise reaches the irst The actual noise level at the input is gain stage in the analyzer, the ampli- a function of the input signal. Indeed, ier boosts the noise, plus adds some noise is sometimes the signal of inter- of its own. As the noise signal passes est. Like any discrete signal, a noise on through the system, it is typically signal is much easier to measure when

1. Displayed average noise level is sometimes confused with the term “sensitivity.” While related, these terms have different meanings. Sensitivity is a measure of the minimum signal level that yields a deined signal-to-noise ratio (SNR) or bit error rate (BER). It is a common metric of radio receiver performance. Spectrum analyzer speciications are always given in terms of the DANL.

46 Beginning in the late 1970s, spectrum analyzer designers took a different approach. In newer analyzers, an internal microprocessor changes the IF gain to offset changes in the input attenuator. Thus, signals present at the analyzer’s input remain stationary on the display as we change the input attenuator, while the displayed noise moves up and down. In this case, the reference level remains unchanged, as shown in Figure 5-1. As the attenuation increases from 5 to 15 to 25 dB, the displayed noise rises while the –30-dBm signal remains constant. In either case, we get the best signal-to-noise ratio by selecting minimum input attenuation.

Resolution bandwidth also affects signal-to-noise ratio, or sensitivity. The noise generated in the analyzer is random and has a constant amplitude Figure 5-1. In modern signal analyzers, reference levels remain constant when you change input attenuation over a wide frequency range. Since the resolution, or IF, bandwidth ilters come after the irst gain stage, the total noise power that passes through the ilters is determined by the width of the ilters. This noise signal is detected and ultimately reaches the display. The random nature of the noise signal causes the displayed level to vary as:

10 log (BW2/BW1) where BW = starting resolution bandwidth 1 BW = ending resolution bandwidth 2

So if we change the resolution bandwidth by a factor of 10, the displayed noise level changes by 10 dB, as shown in Figure 5-2. For continuous wave (CW) signals, we get best signal-to- noise ratio, or best sensitivity, using the minimum resolution bandwidth available in our spectrum analyzer2. Figure 5-2. Displayed noise level changes as 10 log (BW2 /BW1 )

2. Broadband, pulsed signals can exhibit the opposite behavior, where the SNR increases as the bandwidth gets larger.

47 A spectrum analyzer displays signal plus noise, and a low signal-to-noise ratio makes the signal dificult to distinguish. We noted previously that the video ilter can be used to reduce the amplitude fluctuations of noisy signals without affecting constant signals. Figure 5-3 shows how the video ilter can improve our ability to discern low- level signals. The video ilter does not affect the average noise level and so does not, by this deinition, affect the sensitivity of an analyzer.

In summary, we get best sensitivity for narrowband signals by selecting the minimum resolution bandwidth and minimum input attenuation. These settings give us the best signal-to-noise ratio. We can also select minimum video bandwidth to help us see a signal at or close to the noise level3. Of course, selecting narrow resolution and video bandwidths does lengthen the sweep time.

Noise loor extension While lowering an analyzer’s inherent noise loor through hardware design and component choices is obviously beneicial for dynamic range, there are practical limits, and another approach offers signiicant improvement. With suficient processing and other technical innovations, the noise power in a signal analyzer can be modeled and subtracted from measurement results to reduce the effective noise level. In the Keysight PXA signal analyzer this operation is called noise loor extension (NFE).

Generally, if you can accurately identify the noise power contribution of an analyzer, you can subtract this power from Figure 5-3. Video iltering makes low-level signals more discernible various kinds of spectrum measurements. Examples include signal power Keysight has been demonstrating somewhat inconvenient. It involves dis- or band power, ACPR, spurious, phase noise subtraction capability for some connecting the signal from the analyzer, noise, harmonic and intermodulation time, using trace math in vector signal measuring analyzer noise level with a distortion. Noise subtraction tech- analyzers to remove analyzer noise from large amount of averaging, reconnect- niques do not improve the performance spectrum and band power measure- ing the signal and using trace math to of vector analysis operations such as ments. (Similar trace math is available display a corrected result. It is neces- demodulation or time-domain displays in the Keysight X-Series signal analyz- sary to re-measure the analyzer noise of signals. ers.) This capability is effective, though power every time the analyzer conigu-

3. For the effect of noise on accuracy, see “Dynamic range versus measurement uncertainty” in Chapter 6.

48 ration (frequency center/span, attenuator/input range, resolution bandwidth) changed.

The Keysight PXA analyzers dramatically improve this measurement technique for many measurement situations. Critical parameters that determine the analyzer’s noise loor are measured when it is calibrated, and these parameters are used (with current measurement information such as analyzer temperature) to fully model the analyzer’s noise loor, including changes in analyzer coniguration and operating conditions. The analyzer’s noise power contribution is then automatically subtracted from spectrum and power measurements. This process in the PXA is called noise loor extension and is enabled with a keystroke in the Mode Setup menu. An example is Figure 5-4. Noise loor extension view of harmonics shown in Figure 5-4.

The effectiveness of NFE can be ex- Noise igure the signal level at the output (indicated pressed in several ways. Average noise on the display) is the same as the level power in the display (DANL) is usually Many receiver manufacturers specify at the input (input connector). So our reduced by 10 to 12 dB in the analyzer’s the performance of their receivers in expression, after substitution, cancel- low band (below 3.6 GHz) and about terms of noise igure, rather than sen- lation and rearrangement, becomes: 8 dB in its high band (above 3.6 GHz). sitivity. We will show you how the two can be equated. A spectrum analyzer While the apparent noise level will F = No/Ni be reduced, only the analyzer’s noise is a receiver, and we will examine noise power is being subtracted. Therefore, igure on the basis of a sinusoidal input. This expression tells us that all we need the apparent power of signals in the to do to determine the noise igure is display will be reduced if the analyzer’s Noise igure can be deined as the compare the noise level as read on noise power is a signiicant part of their degradation of signal-to-noise ratio the display to the true (not the effec- power, and not otherwise. as a signal passes through a device, a tive) noise level at the input connector. spectrum analyzer in our case. We can Noise igure is usually expressed in Thus measurements of both discrete express noise igure as: terms of dB, or: signals and the noise loor of signal Si/Ni sources connected to the PXA are more F = NF = 10 log(F) = 10 log(N ) – 10 log(N ). So/No o i accurately measured with NFE enabled. NFE works with all spectrum measure- where We use the true noise level at the input, ments regardless of RBW or VBW, and F = noise igure as power ratio (also rather than the effective noise level, it also works with any type of detector known as noise factor) because our input signal-to-noise ratio or averaging. Si = input signal power was based on the true noise. As we saw Ni = true input noise power earlier, when the input is terminated in So = output signal power 50 ohms, the kTB noise level at room No = output noise power temperature in a 1-Hz bandwidth is More information –174 dBm. We can simplify this expression for For more information on using our spectrum analyzer. First of all, the noise loor extension, please refer output signal is the input signal times to, Using Noise Floor Extension in the the gain of the analyzer. Second, the PXA Signal Analyzer – Application gain of our analyzer is unity because Note, literature number 5990- 5340EN.

49 We know the displayed level of noise on analyzer, we can obtain a system Testing by experiment means we must the analyzer changes with bandwidth. (preampliier/spectrum analyzer) noise have the equipment at hand. We do So all we need to do to determine the igure lower than that of the spectrum not need to worry about numbers. We noise igure of our spectrum analyzer analyzer alone. To the extent that we simply connect the preampliier to the is to measure the noise power in some lower the noise igure, we also improve analyzer, note the average displayed bandwidth, calculate the noise power the system sensitivity. noise level and subtract the gain of the that we would have measured in a 1-Hz preampliier. Then we have the sensitiv- bandwidth using 10 log(BW2/BW1), and When we introduced noise igure in the ity of the system. compare that to –174 dBm. previous discussion, we did so on the basis of a sinusoidal input signal. We However, we really want to know ahead For example, if we measured –110 dBm can examine the beneits of a pream- of time what a preampliier will do for in a 10-kHz resolution bandwidth, we pliier on the same basis. However, a us. We can state the two cases above would get: preampliier also ampliies noise, and as follows: this output noise can be higher than the If NF + G ≥ NF + 15 dB, NF = [measured noise in dBm] – effective input noise of the analyzer. In pre pre SA Then NFsys = NFpre – 2.5 dB 10 log(RBW/1) – kTBB=1 Hz the “Noise as a signal” section later in –110 dBm –10 log(10,000/1) – (–174 dBm) this chapter, you will see how a spec- And –110 – 40 + 174 trum analyzer using log power averaging = 24 dB displays a random noise signal 2.5 dB If NF + G ≤ NF – 10 dB, below its actual value. As we explore pre pre SA Then NF = NF – G Noise igure is independent of band- preampliiers, we shall account for this sys SA pre width4. Had we selected a different 2.5 dB factor where appropriate. resolution bandwidth, our results Using these expressions, we’ll see how would have been exactly the same. Rather than develop a lot of formu- a preampliier affects our sensitivity. For example, had we chosen a 1-kHz las to see what beneit we get from a Assume that our spectrum analyzer resolution bandwidth, the measured preampliier, let us look at two extreme has a noise igure of 24 dB and the noise would have been –120 dBm and cases and see when each might ap- preampliier has a gain of 36 dB and a 10 log(RBW/1) would have been 30. ply. First, if the noise power out of the noise igure of 8 dB. All we need to do Combining all terms would have given preampliier (in a bandwidth equal to is to compare the gain plus noise igure –120 – 30 + 174 = 24 dB, the same that of the spectrum analyzer) is at of the preampliier to the noise igure of the spectrum analyzer. The gain noise igure as above. least 15 dB higher than the DANL of the spectrum analyzer, the noise igure of plus noise igure of the preampliier is 44 dB, more than 15 dB higher than the The 24- dB noise igure in our example the system is approximately that of the tells us that a sinusoidal signal must preampliier less 2.5 dB. How can we noise igure of the spectrum analyzer, so the noise igure of the preampliier/ be 24 dB above kTB to be equal to the tell if this is the case? Simply connect spectrum-analyzer combination is that displayed average noise level on this the preampliier to the analyzer and particular analyzer. Thus we can use note what happens to the noise on the of the preampliier less 2.5 dB, or 5.5 dB. In a 10-kHz resolution bandwidth, noise igure to determine the DANL display. If it goes up 15 dB or more, we for a given bandwidth or to compare have fulilled this requirement. our preampliier/analyzer system has a sensitivity of: DANLs of different analyzers with the On the other hand, if the noise power kTB + 10 log(RBW/1) + NF same bandwidth.5 out of the preampliier (again, in the B=1 sys same bandwidth as that of the spec- = –174 + 40 + 5.5 = –128.5 dBm Preampliiers trum analyzer) is 10 dB or more lower than the displayed average noise level This is an improvement of 18.5 dB One reason for introducing noise igure on the analyzer, the noise igure of the over the –110 dBm noise loor without is that it helps us determine how much system is that of the spectrum analyzer the preampliier. beneit we can derive from the use of less the gain of the preampliier. Again a preampliier. A 24-dB noise igure, we can test by inspection. Connect However, there might be a drawback to while good for a spectrum analyzer, is the preampliier to the analyzer; if the using this preampliier, depending upon not so good for a dedicated receiver. displayed noise does not change, we our ultimate measurement objective. If However, by placing an appropriate have fulilled the requirement. preampliier in front of the spectrum we want the best sensitivity but no loss

4. This may not always be precisely true for a given analyzer because of the way resolution bandwidth ilter sections and gain are distributed in the IF chain. 5. The noise igure computed in this manner cannot be directly compared to that of a receiver because the “measured noise” term in the equation understates the actual noise by 2.5 dB. See the section titled “Noise as a signal” later in this chapter. 50 of measurement range, this preampliier ier in place, our measurement range To choose the correct preampliier, we is not the right choice. Figure 5-5 il- is 92.5 dB, 17.5 dB less than without must look at our measurement needs. lustrates this point. A spectrum analyzer the preampliier. The loss in measure- If we want absolutely the best sensitivity with a 24-dB noise igure will have an ment range equals the change in the and are not concerned about measure- average displayed noise level of –110 displayed noise when the preampliier is ment range, we would choose a high- dBm in a 10-kHz resolution bandwidth. connected. gain, low-noise-igure preampliier so If the 1-dB compression point6 for that that our system would take on the noise analyzer is 0 dBm, the measurement Finding a preampliier that will give us igure of the preampliier, less 2.5 dB. range is 110 dB. When we connect the better sensitivity without costing us If we want better sensitivity but cannot preampliier, we must reduce the maxi- measurement range dictates that we afford to give up any measurement mum input to the system by the gain of must meet the second of the above crite- range, we must choose a lower-gain the preampliier to –36 dBm. However, ria; that is, the sum of its gain and noise preampliier. when we connect the preampliier, the igure must be at least 10 dB less than Interestingly enough, we can use the displayed average noise level will rise by the noise igure of the spectrum analyz- input attenuator of the spectrum analyzer about 17.5 dB because the noise power er. In this case, the displayed noise loor to effectively degrade the noise igure out of the preampliier is that much will not change noticeably when we (or reduce the gain of the preampliier, higher than the analyzer’s own noise connect the preampliier, so although if you prefer). For example, if we need loor, even after accounting for the 2.5 we shift the whole measurement range slightly better sensitivity but cannot dB factor. It is from this higher noise down by the gain of the preampliier, we afford to give up any measurement level that we now subtract the gain of end up with the same overall range we range, we can use the above preampli- the preampliier. With the preampli- started with. ier with 30 dB of RF input attenuation

Spectrum analyzer Spectrum analyzer and preampliﬁer

1 dB compression 0 dBm

Gpre System 1 dB compression –36 dBm

110 dB spectrum analyzer range 92.5 dB system range DANL –92.5 dBm DANL –110 dBm Gpre System sensitivity –128.5 dBm

Figure 5-5. If displayed noise goes up when a preampliier is connected, measurement range is diminished by the amount the noise changes

6. See the section titled “Mixer compression” in Chapter 6.

51 on the spectrum analyzer. This attenuation increases the noise igure of the analyzer from 24 to 54 dB. Now the NF + 3 dB gain plus noise igure of the preampli - NFSA – Gpre + 3 dB pre ier (36 + 8) is 10 dB less than the NF – G + 2 dB NF + 2 dB noise igure of the analyzer, and we SA pre pre have met the conditions of the second System Noise NF – G + 1 dB NF + 1 dB criterion above. Figure (dB) SA pre pre

NF – G NF The noise igure of the system is now: SA pre pre

NF – 1 dB NFsys = NFSA – GPRE pre = 54 dB – 36 dB NF – 2 dB = 18 dB pre NF – 2.5 dB –10 –5 0 +5 +10 pre This represents a 6-dB improvement NFpre + Gpre – NFSA (dB) over the noise igure of the analyzer alone with 0 dB of input attenuation. So we have improved sensitivity by 6 dB Figure 5-6. System noise igure for sinusoidal signals and given up virtually no measurement range. Next, let’s try two numerical examples. with a built-in preampliier, because Above, we determined that the noise the preampliier/spectrum analyzer Of course, there are preampliiers that igure of our analyzer is 24 dB. What combination is calibrated as a system, fall in between the extremes. Figure would the system noise igure be if we and amplitude values displayed on 5-6 enables us to determine system add a Keysight 8447D ampliier, a pre- screen are already corrected for proper noise igure from a knowledge of the ampliier with a noise igure of about readout. With an external preampli- noise igures of the spectrum analyzer 8 dB and a gain of 26 dB? First, NFPRE ier, you must correct the spectrum and preampliier and the gain of the + GPRE – NFSA is +10 dB. From the analyzer reading with a reference level ampliier. We enter the graph of Figure graph of Figure 5-6 we ind a system offset equal to the preamp gain. Most 5-6 by determining NF + G – PRE PRE noise igure of about NFPRE – 1.8 dB, modern spectrum analyzers allow you NFSA. If the value is less than zero, we or about 8 – 1.8 = 6.2 dB. The graph to enter the gain value of the external ind the corresponding point on the accounts for the 2.5-dB factor. On the preampliier from the front panel. The dashed curve and read system noise other hand, if the gain of the preampli- analyzer then applies this gain offset igure as the left ordinate in terms of ier is just 10 dB, then NFPRE + GPRE to the displayed reference level value, dB above NF – G . If NF + G SA PRE PRE PRE – NFSA is –6 dB. This time the graph so you can directly view corrected – NF is a positive value, we ind the SA indicates a system noise igure of NFSA measurements on the display. corresponding point on the solid curve – GPRE + 0.6 dB, or 24 – 10 + 0.6 = 14.6 and read system noise dB. (We did not introduce the 2.5-dB igure as the right ordinate in terms of factor previously when we determined More information dB above NFPRE. the noise igure of the analyzer alone because we read the measured noise For more details on noise Let’s irst test the two previous extreme directly from the display. The displayed igure, see Fundamentals of cases. noise included the 2.5-dB factor.) RF and Microwave Noise Figure Measurements – Application Note,

As NFPRE + GPRE – NFSA becomes less Many modern spectrum analyzers have literature number than –10 dB, we ind that system noise optional built-in preampliiers available. 5952-8255E. igure asymptotically approaches Compared to external preampliiers, NFSA – GPRE. built-in preampliiers simplify measure- As the value becomes greater than ment setups and eliminate the need for +15 dB, system noise igure asymptoti- additional cabling. Measuring signal cally approaches NFPRE less 2.5 dB. amplitude is much more convenient

52 Noise as a signal Figure 5-7. Random noise has So far, we have focused on the noise a Gaussian amplitude distribution generated within the measurement system (analyzer or analyzer/preampliier). We described how the measurement system’s displayed average noise level limits the overall sensitivity. How- ever, random noise is sometimes the signal we want to measure. Because of the nature of noise, the superheterodyne spectrum analyzer indicates Figure 5-8. The a value that is lower than the actual envelope of band- value of the noise. Let’s see why this is limited Gaussian noise has a Ray- so and how we can correct for it. leigh distribution

By random noise, we mean a signal whose instantaneous amplitude has a Gaussian distribution versus time, as shown in Figure 5-7. For example, thermal or Johnson noise has this characteristic. Such a signal has no discrete spectral components, so we cannot select some particular component averaging. The mean value of a Rayleigh lower values. As a result, the output and measure it to get an indication of distribution is 1.253 σ. of the envelope detector is a skewed signal strength. In fact, we must deine Rayleigh distribution, and the mean what we mean by signal strength. If However, our analyzer is a peak- value that we get from video iltering or we sample the signal at an arbitrary responding voltmeter calibrated to in- averaging is another 1.45 dB lower. In instant, we could theoretically get any dicate the rms value of a sine wave. To the log mode, then, the mean or aver- amplitude value. We need some mea- convert from peak to rms, our analyzer age noise is displayed 2.5 dB too low. sure that expresses the noise level scales its readout by 0.707 (–3 dB). The Again, this error is not an ambiguity, averaged over time. Power, which is of mean value of the Rayleigh-distributed and we can correct for it7. course proportionate to rms voltage, noise is scaled by the same factor, satisies that requirement. giving us a reading of 0.886 σ (l.05 dB This is the 2.5-dB factor we accounted below σ). To equate the mean value dis- for in the previous preampliier discus- We have already seen that both video played by the analyzer to the rms volt- sion, when the noise power out of the iltering and video averaging reduce age of the input noise signal, we must preampliier was approximately equal to the peak-to-peak luctuations of a account for the error in the displayed or greater than the analyzer’s own noise. signal and can give us a steady value. value. Note, however, that the error is We must equate this value to either not an ambiguity; it is a constant error Another factor that affects noise power or rms voltage. The rms value that we can correct for by adding 1.05 measurements is the bandwidth in of a Gaussian distribution equals its dB to the displayed value. which the measurement is made. We standard deviation, σ. have seen how changing resolution In most spectrum analyzers, the display bandwidth affects the displayed level Let’s start with our analyzer in the linear scale (log or linear in voltage) controls of the analyzer’s internally generated display mode. The Gaussian noise at the the scale on which the noise distribu- noise. Bandwidth affects external noise input is band limited as it passes through tion is averaged with either the VBW signals in the same way. To compare the IF chain, and its envelope takes on ilter or with trace averaging. Normally, measurements made on different ana- a Rayleigh distribution (Figure 5-8). The we use our analyzer in the log display lyzers, we must know the bandwidths noise we see on our analyzer display, mode, and this mode adds to the error used in each case. the output of the envelope detector, is in our noise measurement. the Rayleigh-distributed envelope of the The gain of a log ampliier is a function Not only does the 3-dB (or 6-dB) input noise signal. To get a steady value, of signal amplitude, so the higher noise bandwidth of the analyzer affect the the mean value, we use video iltering or values are not ampliied as much as the measured noise level, the shape of the

7. In X-Series analyzers, the averaging can be set to video, voltage or power (rms), independent of display scale. When using power averaging, no correction is needed, since the average rms level is determined by the square of the magnitude of the signal, not by the log or envelope of the voltage.

53 resolution ilter also plays a role. To make comparisons possible, we deine a standard noise-power bandwidth: the NF – G + 3 dB NF + 3 dB width of a rectangular ilter that passes SA pre pre the same noise power as our analyzer’s System Noise NF – G + 2 dB ilter. For the near-Gaussian ilters Figure (dB) SA pre NFpre + 2 dB in Keysight analyzers, the equivalent noise-power bandwidth is about 1.05 NFSA – Gpre + 1 dB NFpre + 1 dB to 1.13 times the 3-dB bandwidth, depending on bandwidth selectivity. For NFSA – Gpre NFpre example, a 10-kHz resolution band- –10 –5 0 +5 +10 width ilter has a noise-power band- NFpre + G pre – NF SA (dB) width in the range of 10.5 to 11.3 kHz.

Figure 5-9. System noise igure for noise signals If we use 10 log(BW2/BW1) to adjust the displayed noise level to what we would have measured in a noise-power Many of today’s microprocessor-con- at the same level by the time it reaches bandwidth of the same numeric value trolled analyzers allow us to activate the display. The input and internal noise as our 3-dB bandwidth, we ind that a noise marker. When we do so, the signals add to raise the displayed noise the adjustment varies from: microprocessor switches the analyzer by 3 dB, a factor of two in power. So into the power (rms) averaging mode, we can deine the noise igure of our 10 log(10,000/10,500) = –0.21 dB computes the mean value of a number analyzer for a noise signal as: to of display points about the marker9, 10 log(10,000/11,300) = –0.53 dB normalizes and corrects the value to NFSA(N) = (noise loor)dBm/RBW – a 1-Hz noise-power bandwidth and 10 log(RBW/1) – kTBB=1 + 2.5 dB In other words, if we subtract some- displays the normalized value. thing between 0.21 and 0.53 dB from If we use the same noise loor we used the indicated noise level, we have the The analyzer does the hard part. It is previously, –110 dBm in a 10-kHz reso- noise level in a noise-power bandwidth easy to convert the noise-marker value lution bandwidth, we get: that is convenient for computations. to other bandwidths. For example, if we For the following examples, we will use want to know the total noise in a 4-MHz NFSA(N) = –110 dBm – 10 log(10,000/1) 0.5 dB as a reasonable compromise for communication channel, we add 10 – (–174 dBm) + 2.5 dB = 26.5 dB the bandwidth correction8. log(4,000,000/1), or 66 dB to the noise- marker value10. As was the case for a sinusoidal signal,

Let’s consider the various correction NFSA(N) is independent of resolution factors to calculate the total correction Preampliier for noise bandwidth and tells us how far above for each averaging mode: measurements kTB a noise signal must be to be equal to the noise loor of our analyzer. Linear (voltage) averaging: Noise signals are typically low-level Rayleigh distribution (linear mode): 1.05 dB signals, so we often need a preampliier When we add a preampliier to our 3-dB/noise power bandwidths: –0.50 dB to have suficient sensitivity to measure analyzer, the system noise igure and Total correction: 0.55 dB them. However, we must recalculate sensitivity improve. However, we have sensitivity of our analyzer irst. We pre- accounted for the 2.5-dB factor in our

Log averaging: viously deined sensitivity as the level of deinition of NFSA(N), so the graph of Logged Rayleigh distribution: 2.50 dB a sinusoidal signal that is equal to the system noise igure becomes that of 3-dB/noise power bandwidths: –0.50 dB displayed average noise loor. Since the Figure 5-9. We determine system noise Total correction: 2.00 dB analyzer is calibrated to show the prop- igure for noise the same way that we er amplitude of a sinusoid, no correction did previously for a sinusoidal signal. Power (rms voltage) averaging: for the signal was needed. But noise is Power distribution: 0.00 dB displayed 2.5 dB too low, so an input 3-dB/noise power bandwidths: –0.50 dB noise signal must be 2.5 dB above the Total correction: –0.50 dB analyzer’s displayed noise loor to be

8. The X-Series analyzers specify noise power bandwidth accuracy to within 0.5% (± 0.022 dB). 9. For example, the X-Series analyzers compute the mean over half a division, regardless of the number of display points. 10. Most modern spectrum analyzers make this calculation even easier with the channel power function. You enter the integration bandwidth of the channel and center the signal on the analyzer display. The channel power function then calculates the total signal power in the channel.

54 Chapter 6. Dynamic Range

Dynamic range is generally thought of We can expand this expression into a that for every 1 dB we drop the level of as the ability of an analyzer to mea- power series: the fundamental at the input mixer, the sure harmonically related signals and internally generated second harmonic 2 3 the interaction of two or more signals, i = IS(k1v + k2v + k3v +...) drops by 2 dB. See Figure 6-1. The for example, to measure second- or second term includes 3ω , the third where k = q/kT 1 third-harmonic distortion or third-order 1 harmonic, and the cube of the input- k = k12/2! intermodulation. In dealing with such 2 signal voltage, V 3. So a 1-dB change k = k13/3!, etc. 1 measurements, remember that the 3 in the fundamental at the input mixer input mixer of a spectrum analyzer is a Let’s now apply two signals to the changes the internally generated third nonlinear device, so it always gener- mixer. One will be the input signal we harmonic by 3 dB. ates distortion of its own. The mixer is wish to analyze; the other, the local nonlinear for a reason. It must be non- oscillator signal necessary to create Distortion is often described by its linear to translate an input signal to the the IF: order. The order can be determined by desired IF. But the unwanted distortion noting the coeficient associated with products generated in the mixer fall at v = V sin(ω t) + V sin(ω t) the signal frequency or the exponent the same frequencies as the distortion LO LO 1 1 associated with the signal amplitude. products we wish to measure on the If we go through the mathematics, we Thus second-harmonic distortion input signal. arrive at the desired mixing product is second order and third harmonic that, with the correct LO frequency, distortion is third order. The order also So we might deine dynamic range in equals the IF: indicates the change in internally gen- this way: it is the ratio, expressed in dB, erated distortion relative to the change of the largest to the smallest signals k2VLOV1 cos[(ωLO – ω1)t] in the fundamental tone that created it. simultaneously present at the input of the spectrum analyzer that allows A k V V cos[(ω + ω )t] term is also Now let us add a second input signal: measurement of the smaller signal to a 2 LO 1 LO 1 generated, but in our discussion of the given degree of uncertainty. tuning equation, we found that we want v = VLO sin(ωLO t) + V1 sin(ω1t) + V2 sin(ω2t) the LO to be above the IF, so (ω + ω ) Notice that accuracy of the LO 1 is also always above the IF. This time, when we go through the measurement is part of the deinition. math In the following examples, you will see With a constant LO level, the mixer to ind internally generated distortion, how both internally generated noise output is linearly related to the input in addition to harmonic distortion, we and distortion affect accuracy. signal level. For all practical purposes, get: this is true as long as the input signal is (k4/8)VLOV12V2cos[ωLO – (2ω1 – ω2)]t, Dynamic range versus more than 15 to 20 dB below the level (k4/8)VLOV1V22 cos[ωLO – (2ω2 – ω1)]t, etc. internal distortion of the LO. There are also terms involving harmonics of the input signal: To determine dynamic range versus These equations represent intermodu- distortion, we must irst determine just (3k3/4)VLOV12 sin(ωLO – 2 ω1)t, lation distortion, the interaction of the how our input mixer behaves. Most (k4/8)VLOV13 sin(ωLO – 3ω1)t, etc. two input signals with each other. The analyzers, particularly those using lower distortion product, 2ω1 – ω2, falls harmonic mixing to extend their tuning These terms tell us that dynamic range below ω1 by a frequency equal to the range1, use diode mixers. (Other types due to internal distortion is a function of difference between the two fundamen- of mixers would behave similarly.) The ω ω the input signal level at the input mixer. tal tones, 2 – 1. The higher distortion current through an ideal diode can be ω ω ω Let’s see how this works, using as our product, 2 2 – 1, falls above 2 by the expressed as: deinition of dynamic range, the difference same frequency. See Figure 6-1. in dB between the fundamental tone and i = I (eqv/kT–1) s the internally generated distortion. Once again, dynamic range is a func- The argument of the sine in the irst tion of the level at the input mixer. The where IS = the diode’s saturation current –19 term includes 2ω , so it represents the internally generated distortion changes q = electron charge (1.60 x 10 C) 1 2 second harmonic of the input signal. as the product of V1 and V2 in the irst v = instantaneous voltage 2 k = Boltzmann’s constant The level of this second harmonic is a case, of V1 and V2 in the second. If V1 (1.38 x 10–23 joule/K) function of the square of the voltage of and V2 have the same amplitude, the T= temperature in Kelvin 2 usual case when testing for distortion, the fundamental, V1 . This fact tells us

1. See Chapter 7, “Extending the Frequency Range.”

55 we can treat their products as cubed 3 3 D dB D dB D dB terms (V1 or V2 ). Thus, for every dB that we simultaneously change the level of the two input signals, there is a 2D dB 3D dB 3-dB change in the distortion compo- 3D dB 3D dB nents, as shown in Figure 6-1.

This is the same degree of change that we see for third harmonic distortion w 2 w 3 w 2w – w w w 2w – w in Figure 6-1. And in fact, this too, is 1 2 1 2 2 1 third-order distortion. In this case, we can determine the degree of distortion Figure 6-1. Changing the level of fundamental tones at the mixer by summing the coeficients of ω1 and second-harmonic example, then, when tion. However, from a mathematical ω2 (e.g., 2ω1 – 1ω2 yields 2 + 1 = 3) or the level at the mixer changes from –40 standpoint, TOI is a perfectly good data the exponents of V1 and V2. to –50 dBm, the internal distortion, and point because we know the slope of All this says that dynamic range thus our measurement range, changes the line. So even with TOI as a starting depends upon the signal level at the from –75 to –85 dBc. In fact, these point, we can still determine the degree mixer. How do we know what level we points fall on a line with a slope of 1 of internally generated distortion at a need at the mixer for a particular mea- that describes the dynamic range for given mixer level. surement? Most analyzer data sheets any input level at the mixer. include graphs to tell us how dynamic We can calculate TOI from data sheet range varies. However, if no graph is We can construct a similar line for information. Because third-order provided, we can draw our own2. third-order distortion. For example, dynamic range changes 2 dB for every a data sheet might say third-order 1 dB change in the level of the fun- We do need a starting point, and this distortion is –85 dBc for a level of damental tone(s) at the mixer, we get we must get from the data sheet. –30 dBm at this mixer. Again, this is TOI by subtracting half of the speciied Let’s look at second-order distortion our starting point, and we would plot dynamic range in dBc from the level of irst. Let’s assume the data sheet says the point shown in Figure 6-2. If we the fundamental(s): second-harmonic distortion is 75 dB now drop the level at the mixer to –40 down for a signal –40 dBm at the mixer. dBm, what happens? Referring again TOI = Afund – d/2 Because distortion is a relative mea- to Figure 6-1, we see that both third- surement, and, at least for the moment, harmonic distortion and third-order in- where Afund = level of the fundamental we are calling our dynamic range the termodulation distortion fall by 3 dB for in dBm difference in dB between fundamental every 1 dB that the fundamental tone d = difference in dBc (a negative tone or tones and the internally gener- or tones fall. Again, it is the difference value) between fundamental and ated distortion, we have our starting that is important. If the level at the distortion point. Internally generated second-or- mixer changes from –30 to –40 dBm, der distortion is 75 dB down, so we can the difference between fundamental Using the values from the previous measure distortion down 75 dB. We tone or tones and internally generated discussion: plot that point on a graph whose axes distortion changes by 20 dB. So the in- are labeled distortion (dBc) versus level ternal distortion is –105 dBc. These two TOI = –30 dBm – (–85 dBc)/2 = +12.5 dBm at the mixer (level at the input connec- points fall on a line with a slope of 2, tor minus the input-attenuator setting). giving us the third-order performance Attenuator test See Figure 6-2. What happens if the for any level at the mixer. Understanding the distortion graph is level at the mixer drops to –50 dBm? important, but we can use a simple test As noted in Figure 6-1, for every 1 dB Sometimes third-order performance to determine whether displayed distor- change in the level of the fundamental is given as TOI (third-order inter- tion components are true input signals at the mixer there is a 2 dB change in cept). This is the mixer level at which or internally generated signals. Change the internally generated second har- the internally generated third-order the input attenuator. If the displayed monic. But for measurement purposes, distortion would be equal to the value of the distortion components we are interested only in the relative fundamental(s), or 0 dBc. This situation remains the same, the components are change, that is, in what happened cannot be realized in practice because part of the input signal. If the displayed to our measurement range. In this the mixer would be well into satura- case, for every 1 dB the fundamental 2. For more information on how to construct a dynamic range chart, see Optimizing Dynamic changes at the mixer, our measurement Range for Distortion Measurements – Keysight PSA Performance Spectrum Analyzer Series range also changes by 1 dB. In our Product Note, literature number 5980-3079EN.

56 value changes, the distortion com- Figure 6-2. Dynamic range ponents are generated internally or TOI SHI 0 versus distortion are the sum of external and internally and noise generated signals. We continue chang- –10 ing the attenuator until the displayed –20 distortion does not change and then complete the measurement. –30 3rd order2nd order Noise –40 –50 Noise (10 kHz BW) Another constraint on dynamic range is the noise loor of our spectrum (dBc) –60 analyzer. Going back to our deinition of –70 dynamic range as the ratio of the larg- Maximum 2nd order est to the smallest signal we can mea- dynamic range –80 sure, the average noise of our spectrum Maximum 3rd order dynamic range analyzer puts the limit on the smaller –90 Optimum signal. So dynamic range versus noise mixer levels –100 becomes signal-to-noise ratio in which –60 –50 –40–30 –20 –10+0 10 the signal is the fundamental whose Mixer level (dBm) distortion we wish to measure.

We can easily plot noise on our dy- Figure 6-3. Reducing resolu- namic range chart. For example, sup- 0 TOI SHI tion bandwidth pose the data sheet for our spectrum improves dy- analyzer speciies a displayed average –10 namic range noise level of –110 dBm in a 10-kHz –20 resolution bandwidth. If our signal –30 fundamental has a level of –40 dBm at 3rd order the mixer, it is 70 dB above the average 2nd order –40 noise, so we have a 70-dB signal-to- noise ratio. For every 1 dB we reduce –50 Noise (10 kHz BW) the signal level at the mixer, we lose 1 (dBc) Noise (1 kHz BW) dB of signal-to-noise ratio. Our noise –60 curve is a straight line having a slope of –70 –1, as shown in Figure 6-2. 2nd order dynamic range improvement –80 3rd order If we ignore measurement accuracy dynamic range improvement considerations for a moment, the –90 best dynamic range will occur at the intersection of the appropriate distor- –60 –50 –40 –30 –20 –100 +10 tion curve and the noise curve. Figure Mixer level (dBm) 6-2 tells us that our maximum dynamic range for second-order distortion is certainly can improve dynamic range The inal factor in dynamic range is the 72.5 dB; for third-order distortion, 81.7 by narrowing the resolution bandwidth, phase noise on our spectrum analyzer dB. In practice, the intersection of the but there is not a one-to-one corre- LO, and this affects only third-order noise and distortion graphs is not a spondence between the lowered noise distortion measurements. For example, sharply deined point, because noise loor and the improvement in dynamic suppose we are making a two-tone, adds to the CW-like distortion prod- range. For second-order distortion, the third-order distortion measurement ucts, reducing dynamic range by 2 dB improvement is one half the change in on an ampliier, and our test tones are when you use the log power scale with the noise loor; for third-order distor- separated by 10 kHz. The third-order log scale averaging. tion, two-thirds the change in the noise distortion components will also be loor. See Figure 6-3. separated from the test tones by 10 Figure 6-2 shows the dynamic range kHz. For this measurement, we might for one resolution bandwidth. We

57 ind ourselves using a 1-kHz resolution bandwidth. Referring to Figure 6-3 and allowing for a 10-dB decrease in the –10 noise curve, we would ind a maximum dynamic range of about 88 dB. Sup- pose however, that our phase noise at –20 a 10-kHz offset is only –80 dBc. Then 80 dB becomes the ultimate limit of –30 dynamic range for this measurement, as shown in Figure 6-4. –40 In summary, the dynamic range of a spectrum analyzer is limited by three –50 factors: the distortion performance of the input mixer, the broadband noise loor (sensitivity) of the system and the –60 phase noise of the local oscillator. (dBc) –70 Dynamic range versus measurement uncertainty –80 Phase noise Dynamic range (10 kHz offset) In our previous discussion of amplitude reduction due accuracy, we included only those items –90 to phase noise listed in Table 4-1, plus mismatch. We did not cover the possibility of an internally generated distortion product (a –100 sinusoid) being at the same frequency as an external signal we wished to –110 measure. However, internally generated –60 –50 –40 –30 –20 –10+0 10 distortion components fall at exactly Mixer level (dBm) the same frequencies as the distortion components we wish to measure on Figure 6-4. Phase noise can limit third-order intermodulation tests external signals. The problem is that we have no way of knowing the phase relationship between the external and internal signals. So we can determine 6 5 only a potential range of uncertainty: 4 3 Uncertainty (in dB) = 20 log(l ± 10d/20) 2 1 Maximum where d = difference in dB between 0 –30 –25 –20 –15 –10–5 –1 error (dB) the larger and smaller sinusoid –2 (a negative number) –3 –4 See Figure 6-5. For example, if we set –5 up conditions such that the internally –6 generated distortion is equal in ampli- –7 Delta (dBc) –8 tude to the distortion on the incoming signal, the error in the measurement could range from +6 dB (the two Figure 6-5. Uncertainty versus difference in amplitude between two sinusoids at the same frequency signals exactly in phase) to ininity (the two signals exactly out of phase

58 and so canceling). Such uncertainty is unacceptable in most cases. If we put a limit of ±1 dB on the measure- 0 ment uncertainty, Figure 6-5 shows us that the internally generated distortion –10 product must be about 18 dB below the distortion product we wish to measure. To draw dynamic range curves for –20 second- and third-order measurements with no more than 1 dB of measure- –30 ment error, we must then offset the curves of Figure 6-2 by 18 dB as shown in Figure 6-6. –40 2nd order Noise 3rd order Next, let’s look at uncertainty due to –50 low signal-to-noise ratio. The distortion 5 dB

components we wish to measure are, (dBc) we hope, low-level signals, and often –60 they are at–or very close to–the noise level of our spectrum analyzer. In such –70 cases, we often use the video ilter to make these low-level signals more 18 dB discernible. Figure 6-7 shows the error –80 in displayed signal level as a function of displayed signal-to-noise for a typical –90 18 dB spectrum analyzer. The error is only in one direction, so we could correct for –100 it. However, we usually do not. So for –60 –50 –40 –30 –20–100 +10 our dynamic range measurement, let’s Mixer level (dBm) accept a 0.3-dB error due to noise and offset the noise curve in our dynamic range chart by 5 dB, as shown in Figure Figure 6-6. Dynamic range for 1.3-dB maximum error 6-6. Where the distortion and noise curves intersect, the maximum error possible would be less than 1.3 dB. 7

Let’s see what happened to our dynamic range as a result of our concern 6 with measurement error. As Figure 6-6 shows, second-order-distortion 5 dynamic range changes from 72.5 to 61 dB, a change of 11.5 dB. This is one 4 half the total offsets for the two curves (18 dB for distortion; 5 dB for noise). 3 Third-order distortion changes from 81.7 dB to about 72.7 dB for a change of about 9 dB. In this case, the change 2 Error in displayed signal level (dB) is one third of the 18-dB offset for the distortion curve plus two thirds of the 1 5-dB offset for the noise curve. 0 01 2 34567 8 Displayed S/N (dB)

Figure 6-7. Error in displayed signal amplitude due to noise

59 Gain compression A third method, called pulse compres- analyzers with analog IF circuitry. sion, measures the change in system For example, ESA-L Series spectrum In our discussion of dynamic range, we gain to a narrow (broadband) RF pulse analyzers use an 85-dB log ampli- did not concern ourselves with how while the power of the pulse is swept ier. Thus, only measurements that are accurately the larger tone is displayed, upward. When measuring pulses, within 85 dB below the reference level even on a relative basis. As we raise we often use a resolution bandwidth are calibrated. the level of a sinusoidal input signal, much narrower than the bandwidth of eventually the level at the input mixer the pulse, so our analyzer displays the The question is, can the full display becomes so high that the desired out- signal level well below the peak pulse range be used? From the previous put mixing product no longer changes power. As a result, we could be un- discussion of dynamic range, we know linearly with respect to the input signal. aware of the fact that the total signal the answer is generally yes. In fact, The mixer is in saturation, and the dis- power is above the mixer compression dynamic range often exceeds display played signal amplitude is too low. Satu- threshold. A high threshold improves range or log ampliier range. To bring ration is gradual rather than sudden. To signal-to-noise ratio for high-power, the smaller signals into the calibrated help us stay away from the saturation ultranarrow or widely-chirped pulses. area of the display, we must increase condition, the 1-dB compression point The threshold is about 12 dB higher IF gain. But in so doing, we may move is normally speciied. Typically, this gain than for two-tone compression in the the larger signals off the top of the compression occurs at a mixer level in Keysight X-Series signal analyzers. display, above the reference level. the range of –5 to +5 dBm. Thus we can Nevertheless, because different com- Some Keysight analyzers, such as the determine what input attenuator setting pression mechanisms affect CW, two- X-Series, allow measurements of sig- to use for accurate measurement of tone and pulse compression differently, nals above the reference level without 3 high-level signals . Spectrum analyzers any of the compression thresholds can affecting the accuracy with which the with a digital IF will indicate that ADC is be lower than any other. smaller signals are displayed, as shown over-ranged. in Figure 6-8 (see page 61). So we can indeed take advantage of the full dy- Actually, there are three different Display range and measurement range namic range of an analyzer even when methods of evaluating compres- the dynamic range exceeds the display sion. A traditional method, called CW Two additional ranges are often range. In Figure 6-8, even though the compression, measures the change confused with dynamic range: display reference level has changed from –20 in gain of a device (ampliier or mixer range and measurement range. Display dBm to –50 dBm, driving the signal far or system) as the input signal power range, often called display dynamic above the top of the screen, the marker is swept upward. This method is the range, refers to the calibrated ampli- readout remains unchanged. one just described. Note that the CW tude range of the spectrum analyzer compression point is considerably display. For example, a display with ten Measurement range is the ratio of the higher than the levels for the funda- divisions would seem to have a 100-dB largest to the smallest signal that can mentals indicated previously for even display range when we select 10 dB be measured under any circumstances. moderate dynamic range. So we were per division. This is certainly true for to- The maximum safe input level, typi- correct in not concerning ourselves day’s analyzers with digital IF circuitry, cally +30 dBm (1 watt) for most analyz- with the possibility of compression of such as the Keysight X-Series. It is also ers, determines the upper limit. These the larger signal(s). true for the Keysight ESA-E Series ana- analyzers have input attenuators you can lyzers when you use the narrow (10- to set to 60 or 70 dB, so you can reduce A second method, called two-tone 300-Hz) digital resolution bandwidths. +30 dBm signals to levels well below the compression, measures the change However, spectrum analyzers with ana- compression point of the input mixer and in system gain for a small signal while log IF sections typically are calibrated measure them accurately. The displayed the power of a larger signal is swept only for the irst 85 or 90 dB below the average noise level sets the other end of upward. Two-tone compression ap- reference level. In this case, the bottom the range. Depending on the minimum plies to the measurement of multiple line of the graticule represents resolution bandwidth of the particular CW signals, such as sidebands and signal amplitudes of zero, so the bot- analyzer and whether or not you are us- independent signals. The threshold of tom portion of the display covers the ing a preampliier, DANL typically ranges compression of this method is usually range from –85 or –90 dB to ininity, from –115 to –170 dBm. Measurement a few dB lower than that of the CW relative to the reference level. range, then, can vary from 145 to 200 method. This is the method used by dB. Of course, we cannot view a –170- Keysight Technologies to specify spec- The range of the log ampliier can dBm signal while a +30-dBm signal is trum analyzer gain compression. be another limitation for spectrum also present at the input.

3. Many analyzers internally control the combined settings of the input attenuator and IF gain so that a CW signal as high as the compression level at the input mixer creates a delection above the top line of the graticule. This feature keeps us from making incorrect measurements on CW signals inadvertently. 60 Adjacent channel power measurements TOI, SOI, 1-dB gain compression, and DANL are all classic measures of spectrum analyzer performance. However, with the tremendous growth of digital communication systems, other measures of dynamic range have become increasingly important. For example, adjacent channel power (ACP) measurements are often done in CDMA-based communication systems to determine how much signal energy leaks or “spills over” into adjacent or alternate channels located above and below a carrier. An example ACP measurement is shown in Figure 6-9.

Note the relative amplitude difference between the carrier power and the adjacent and alternate channels. You can measure up to six channels on either side of the carrier at a time.

Typically, we are most interested in the relative difference between the signal power in the main channel and the signal power in the adjacent or alternate channel. Depending on the particular communication standard, these measurements are often described as “adjacent channel power ratio” (ACPR) or “adjacent channel leakage ratio” (ACLR) tests. Because digitally modulated signals and the distortion they Figure 6-8. Display range and measurement range on the PXA spectrum analyzer generate are very noise-like in nature, the industry standards typically deine a channel bandwidth over which the signal power is integrated.

To accurately measure ACP performance of a device under test such as a power ampliier, the spectrum analyzer must have better ACP performance than the device being tested. There- fore, spectrum analyzer ACPR dynamic range has become a key performance measure for digital communication systems.

Figure 6-9. Adjacent channel power measurement using a PXA spectrum analyzer

61 Chapter 7. Extending the Frequency Range

As more wireless services continue to be But let us take one step at a time. In de- In summary, for the low band, up to 3.6 introduced and deployed, the available veloping our tuning equation in Chapter GHz, our irst IF is 5.1 GHz. For the upper spectrum has become more and more 2, we found that we needed the low-pass frequency bands, we switch to a irst IF crowded. As a result, there has been an ilter shown in Figure 2-1 to prevent of 322.5 MHz. In Figure 7-1, the second ongoing trend toward developing new higher-frequency signals from reach- IF is already 322.5 MHz, so all we need products and services at higher frequen- ing the mixer. The result was a uniquely to do when we want to tune to the higher cies. In addition, new microwave tech- responding, single-band analyzer that ranges is bypass the irst IF. nologies continue to evolve, driving the tuned to 3.6 GHz. To observe and mea- need for more measurement capability in sure higher-frequency signals, we must In Chapter 2, we used a mathematical the microwave bands. Spectrum analyzer remove the low-pass ilter. approach to conclude that we needed a designers have responded by developing low-pass ilter. The math becomes more instruments capable of directly tuning Other factors that we explored in de- complex in the situation here, so we will up to 50 GHz using a coaxial input. Even veloping the tuning equation were the use a graphical approach to see what is higher frequencies can be measured choice of LO and intermediate frequen- happening. The low band is the simpler using external mixing techniques. This cies. We decided that the IF should not case, so we’ll start with that. In all of our chapter describes the techniques used to be within the band of interest because graphs, we will plot the LO frequency enable tuning the spectrum analyzer to it created a hole in our tuning range along the horizontal axis and signal fre- such high frequencies. in which we could not make measure- quency along the vertical axis, as shown ments. So we chose 5.1 GHz, moving the in Figure 7-2. We know we get a mixing Internal harmonic mixing IF above the highest tuning frequency product equal to the IF (and therefore of interest (3.6 GHz). Our new tuning a response on the display) whenever In Chapter 2, we described a single- range will be above 3.6 GHz, so it seems the input signal differs from the LO by range spectrum analyzer that tunes to logical to move the new IF to a frequency the IF. Therefore, we can determine 3.6 GHz. Now we want to tune higher below 3.6 GHz. A typical irst IF for these the frequency to which the analyzer in frequency. The most practical way higher frequency ranges in Keysight is tuned simply by adding the IF to, or to achieve an extended range is to use spectrum analyzers is 322.5 MHz. We subtracting it from, the LO frequency. To harmonic mixing. will use this frequency in our examples. determine our tuning range, we start by

Low band path Analog or 3.6 GHz 5.1225 GHz 322.5 MHz 22.5 MHz digital IF

Input signal

High band path 3.8 to 8.7 GHz To external mixer 4.8 GHz 300 MHz

322.5 MHz

Preselector

Sweep generator Display

Figure 7-1. Switching arrangement for low band and high bands

62 plotting the LO frequency against the signal frequency axis, as shown by the dashed line in Figure 7-2. Subtracting 13 the IF from the dashed line gives us a tuning range of 0 to 3.6 GHz, the range 9 we developed in Chapter 2. Note that +IF LO frequency, this line in Figure 7-2 is labeled “1−” GHz 5 to indicate fundamental mixing and 1- –IF the use of the minus sign in the tuning 1+ Signal frequency (GHz) equation. We can use the graph to de- 1 termine what LO frequency is required to receive a particular signal or to -3 what signal the analyzer is tuned for a 3 4 5 6 7 8 9 given LO frequency. To display a 1-GHz LO frequency (GHz) signal, the LO must be tuned to 6.1 GHz. For an LO frequency of 8 GHz, the Figure 7-2. Tuning curves for fundamental mixing in the low band, high IF case spectrum analyzer is tuned to receive a signal frequency of 2.9 GHz. In our text, we round off the irst IF to one decimal 25 place; the true IF, 5.1225 GHz, is shown on the block diagram. 20 Now let’s add the other fundamental- mixing band by adding the IF to the LO line in Figure 7-2. This gives us the solid 15 1- upper line, labeled 1+, that indicates 1+ a tuning range from 8.9 to 13.8 GHz. 10 2- Note that for a given LO frequency, the two frequencies to which the analyzer 2+ is tuned are separated by twice the IF. LO

Signal frequency (GHz) 5 Assuming we have a low-pass ilter at 2xLO the input while measuring signals in the low band, we will not be bothered by 0 signals in the 1+ frequency range. 3 4 5 6 7 8 9

-5 Next let’s see to what extent harmonic LO frequency (GHz) mixing complicates the situation. Har- monic mixing comes about because the Figure 7-3. Signals in the “1 minus” frequency range produce single, unambiguous responses in the LO provides a high-level drive signal to low-band, high-IF case the mixer for eficient mixing, and because the mixer is a non-linear device, it generates harmonics of the LO signal. Let’s add second-harmonic mixing to not really complicate the measurement Incoming signals can mix against LO our graph in Figure 7-3 and see to what process. In other words, signals in the harmonics just as well as the funda- extent this complicates our measure- 1− tuning range produce unique, un- mental, and any mixing product that ment procedure. As before, we irst plot ambiguous responses on our analyzer equals the IF produces a response on the LO frequency against the signal display. The same low-pass ilter used the display. In other words, our tuning frequency axis. Multiplying the LO fre- in the fundamental mixing case works (mixing) equation now becomes: quency by two yields the upper dashed equally well for eliminating responses line of Figure 7-3. As we did for funda- created in the harmonic mixing case. fsig = nfLO ± fIF mental mixing, we simply subtract the IF (5.1 GHz) from and add it to the LO The situation is considerably differ- where n = LO harmonic second-harmonic curve to produce the ent for the high-band, low-IF case. As (Other parameters remain the 2− and 2+ tuning ranges. Since neither before, we start by plotting the LO fun- same as previously discussed.) of these overlap the desired 1− tuning damental against the signal-frequency range, we can again argue that they do axis and then add and subtract the IF,

63 producing the results shown in Figure 7-4. Note that the 1− and 1+ tuning 10 ranges are much closer together, and in fact overlap, because the IF is a Image Frequencies much lower frequency, 322.5 MHz in this case. Does the close spacing of the tuning ranges complicate the mea- 5.3 1- surement process? Yes and no. First 5 4.7 of all, our system can be calibrated for 1+ only one tuning range at a time. In this LO Frequency Signal frequency (GHz) case, we would choose the 1− tuning to give us a low-end frequency of about 3.5 GHz, so we have some overlap with the 3.6-GHz upper end of our 0 4.4 5.6 3 4 5 6 7 8 9 low-band tuning range. So what are LO frequency (GHz) we likely to see on the display? If we enter the graph at an LO frequency of 5 GHz, we ind there are two possible Figure 7-4. Tuning curves for fundamental mixing in the high-band, low-IF case signal frequencies that would give us responses at the same point on the ated on the 1− tuning curve for which a displayed signal that appears to be at display: 4.7 and 5.3 GHz (rounding the our analyzer is calibrated and those pro- 13.0 GHz. numbers again). On the other hand, if duced on the 1+ tuning curve. However, we enter the signal frequency axis at before we look at signal identiication The displayed signals created by the 5.3 GHz, we ind that in addition to the solutions, let’s add harmonic-mixing responses to the 3+ and 3− tuning 1+ response at an LO frequency of 5 curves to 26.5 GHz and see if there are curves are known as in-band multiple GHz, we could also get a 1− response. any additional factors we must consider responses. Because they occur when This would occur if we allowed the LO in the signal identiication process. Fig- the LO is tuned to 4.63 GHz and 4.43 to sweep as high as 5.6 GHz, twice the ure 7-5 shows tuning curves up to the GHz, they will produce false responses IF above 5 GHz. Also, if we entered fourth harmonic of the LO. on the display that appear to be genu- the signal frequency graph at 4.7 GHz, ine signals at 8.96 GHz and 8.56 GHz. we would ind a 1+ response at an LO In examining Figure 7-5, we ind some frequency of about 4.4 GHz (twice the additional complications. The spec- Other situations can create out-of- IF below 5 GHz) in addition to the 1− trum analyzer is set up to operate in band multiple responses. For example, response at an LO frequency of 5 GHz. several tuning bands. Depending on suppose we are looking at a 5-GHz Thus, for every desired response on the the frequency to which the analyzer is signal in band 1 that has a signiicant 1− tuning line, there will be a second tuned, the analyzer display is frequency third harmonic at 15 GHz (band 3). In response located twice the IF below it. calibrated for a speciic LO harmonic. addition to the expected multiple pair These pairs of responses are known as For example, in the 8.3- to 13.6-GHz caused by the 5-GHz signal on the image responses. input frequency range, the spectrum 1+ and 1− tuning curves, we also get analyzer is calibrated for the 2− tuning responses generated by the 15-GHz With this type of mixing arrangement, curve. Suppose we have an 13.6-GHz signal on the 4+, 4−, 3+, and 3− tuning it is possible for signals at different signal present at the input. As the LO curves. Since these responses occur frequencies to produce responses at sweeps, the signal will produce IF re- when the LO is tuned to 3.7, 3.8, 4.9, the same point on the display, that is, sponses with the 3+, 3-, 2+ and 2− tun- and 5.1 GHz respectively, the display at the same LO frequency. As we can ing curves. The desired response of the will show signals that appear to be see from Figure 7-4, input signals at 4.7 2− tuning curve occurs when the LO located at 3.4, 3.5, 4.6, and 4.8 GHz. and 5.3 GHz both produce a response frequency satisies the tuning equation: This is shown in Figure 7-6. at the IF when the LO frequency is set to 5 GHz. These signals are known as Multiple responses generally always 13.6 GHz = 2 fLO – 0.3 image frequencies, and they are also come in pairs1, with a “plus” mixing fLO = 6.95GHz separated by twice the IF frequency. Similarly, we can calculate that the product and a “minus” mixing product. response from the 2– tuning curve oc- When we use the correct harmonic Clearly, we need some mechanism to mixing number for a given tuning band, curs when fLO = 6.65 GHz, resulting in differentiate between responses gener- the responses will be separated by 2

1. Often referred to as an “image pair.” This is inaccurate terminology, since images are actually two or more real signals present at the spectrum analyzer input that produce an IF response at the same LO frequency. The numbers for your analyzer may differ.

4+ 3+ 25.00 Apparent location of an input signal 4- 3- resulting from the response to the Band 4 2- tuning curve 20.00 2+ In-band multiple

responses 2- 15.00 Band 3 13.6 13.0 Band 2 10.00 1+ 8.96 8.56 Signal frequency (GHz) 1- Band 1 5.00

Apparent locations of in-band Band 0 multiples of a 13.6 GHz input signal (lowband) 0.00 6.65 6.95 3 4 4.43 4.63 5 6 7 8 9 LO frequency (GHz)

Figure 7-5. Tuning curves up to 4th harmonic of LO showing in-band multiple responses to a 13.6-GHz input signal

25.00 4+ 3+

4- 3- Band 4 20.00 Out-of-band 2+ multiple responses

2- 15.00 Band 3

Band 2 10.00 1+

1- Signal frequency (GHz) Band 1 5.00

Band 0 (lowband) 0.00 8.3 7.3 8.3 4.9 5.1 3.5 3 4 5 6 7 8 9 LO frequency (GHz)

Figure 7-6. Out-of-band multiple responses in band 1 as a result of a signal in-band 3

times fIF. Because the slope of each pair 2fIF (Nc/NA) of tuning curves increases linearly with the harmonic number N, the multiple where Nc = the correct harmonic num- pairs caused by any other harmonic ber for the desired tuning band mixing number appear to be separated NA = the actual harmonic num- by: ber generating the multiple pair

65 25.00 2+ LO Doubled

Band 4 20.00 2- LO Doubled 2+

2- 15.00 1+ LO Doubled Band 3

1- LO Doubled Band 2 10.00 1+ Signal frequency (GHz) 1- Band 1 5.00

Band 0 (lowband) 0.00 3 4 5 6 7 8 9 LO frequency (GHz)

Figure 7-7. X-Series analyzer harmonic bands using LO doubling

In X-Series analyzers, the LO is doubled Fortunately, there is a way to essen- the dashed lines are rejected. Let’s to produce a new, higher-frequency LO tially eliminate image and multiple re- continue with our previous example of for harmonic mixing. As a result, the LO sponses through a process of preilter- 4.7- and 5.3- GHz signals present at harmonics are twice as far apart as they ing the signal. This technique is called the analyzer input. If we set a center would otherwise be and likelihood of mul- preselection. frequency of 5 GHz and a span of 2 tiple responses is signiicantly reduced. GHz, let’s see what happens as the Compare Figure 7-6 and Figure 7-7. Preselection analyzer tunes across this range. As the LO sweeps past 4.4 GHz (the frequency Can we conclude from this discussion What form must our preselection take? at which it could mix with the 4.7-GHz that a harmonic mixing spectrum ana- Referring back to Figure 7-4, assume input signal on its 1+ mixing mode), lyzer is not practical? Not necessarily. we have two signals at 4.7 and 5.3 GHz the preselector is tuned to 4.1 GHz In cases where the signal frequency present at the input of our analyzer. If and therefore rejects the 4.7-GHz is known, we can tune to the signal we were particularly interested in one, signal. The input signal does not reach directly, knowing that the analyzer will we could use a band-pass ilter to allow the mixer, so no mixing occurs, and select the appropriate mixing mode that signal into the analyzer and reject no response appears on the display. for which it is calibrated. In controlled the other. However, the ixed ilter does As the LO sweeps past 5 GHz, the environments with only one or two sig- not eliminate multiple responses; so if preselector allows the 4.7-GHz signal nals, it is usually easy to distinguish the the spectrum is crowded, there is still to reach the mixer, and we see the real signal from the image and multiple potential for confusion. More important, appropriate response on the display. responses. However, there are many perhaps, is the restriction that a ixed il- The 5.3-GHz image signal is rejected, cases in which we have no idea how ter puts on the lexibility of the analyzer. so it creates no mixing product to many signals are involved or what their If we are doing broadband testing, we interact with the mixing product from frequencies might be. For example, certainly do not want to be continually the 4.7-GHz signal and cause a false we could be searching for unknown forced to change bandpass ilters. display. Finally, as the LO sweeps past spurious signals, conducting site sur- 5.6 GHz, the preselector allows the veillance tests as part of a frequency- The solution is a tunable ilter 5.3-GHz signal to reach the mixer, and monitoring program or performing conigured such that it automatically we see it properly displayed. Note in EMI tests to measure unwanted device tracks the frequency of the appropriate Figure 7-8 that nowhere do the various emissions. In all these cases, we could mixing mode. Figure 7-8 shows the mixing modes intersect. So as long as be looking for totally unknown signals effect of such a preselector. Here we the preselector bandwidth is narrow in a potentially crowded spectral envi- take advantage of the fact that our enough (it typically varies from about ronment. Having to perform some form superheterodyne spectrum analyzer 35 MHz at low frequencies to 80 MHz of identiication routine on each and is not a real-time analyzer; that is, at high frequencies) it will greatly every response would make measure- it tunes to only one frequency at attenuate all image and multiple ment time intolerably long. a time. The dashed lines in Figure responses. 7-8 represent the bandwidth of the tracking preselector. Signals beyond The word “eliminate” may be a little

66 strong. Preselectors do not have ininite rejection. Rejection in the 70- to 80-dB range is more likely. So if we are 1+ looking for very low-level signals in the presence of very high-level signals, we 6 might see low-level images or multiples of the high-level signals. What about 5.3 the low band? Most tracking preselec- 1– tors use YIG technology, and YIG ilters 4.7 do not operate well at low frequencies. Fortunately, there is a simple solution. Figure 7-3 shows that no other mixing 4 mode overlaps the 1− mixing mode in the low-frequency, high-IF case. So a

simple low-pass ilter attenuates both Signal frequency (GHz) 3 image and multiple responses. Figure Preselector 7-9 shows the input architecture of a bandwidth typical microwave spectrum analyzer. 2 3 4 4.4 5 5.6 6 LO frequency (GHz)

Figure 7-8. Preselection; dashed gray lines represent bandwidth of tracking preselector

Low band path Analog or 3.6 GHz 5.1225 GHz 322.5 MHz 22.5 MHz digital IF

Input signal

High band path 3.8 to 8.7 GHz To external mixer 4.8 GHz 300 MHz

322.5 MHz

Preselector

Sweep generator Display

Figure 7-9. Front-end architecture of a typical preselected spectrum analyzer

67 Amplitude calibration So far, we have looked at how a harmonic mixing spectrum analyzer responds to various input frequencies. What about amplitude?

The conversion loss of a mixer is a function of harmonic number, and the loss goes up as the harmonic number goes up. This means that signals of equal amplitude would appear at different levels on the display if they involved different mixing modes. To preserve amplitude calibration, something must be done. In Keysight spectrum analyzers, the IF gain is changed. The increased conversion loss at higher LO harmonics causes a loss of sensitivity Figure 7-10. Rise in noise loor indicates changes in sensitivity with changes in LO harmonic used just as if we had increased the input attenuator. And since the IF gain change occurs after the conversion loss, the gain change is relected by a corresponding change in the displayed noise level. So we can determine analyzer sensitivity on the harmonic-mixing ranges by noting the average displayed noise level just as we did on fundamental mixing.

In older spectrum analyzers, the increase in displayed average noise level with each harmonic band was very noticeable. More recent models of Keysight spectrum analyzers use a double-balanced, image-enhanced harmonic mixer to minimize the increased conversion loss when using higher harmonics. Thus, the “stair step” Figure 7-11. Phase noise levels for fundamental and 4th harmonic mixing effect on DANL has been replaced by a gentle sloping increase with higher mental mixing, phase noise (in decibels) tion is very small, as in the situation frequencies, as shown in Figure 7-10. increases by: here, the amplitude of the modulation side bands is directly proportional to Phase noise 20 log(N), the deviation of the carrier (LO). If the In Chapter 2, we noted that instability of where N = harmonic of the LO deviation doubles, the level of the side an analyzer LO appears as phase noise bands must also double in voltage; around signals that are displayed far For example, suppose that the LO fun- that is, increase by 6 dB or 20 log(2). enough above the noise loor. We also damental has a peak-to-peak deviation As a result, the ability of our analyzer noted that this phase noise can impose of 10 Hz. The second harmonic then to measure closely spaced signals that a limit on our ability to measure closely has a 20-Hz peak-to-peak deviation; are unequal in amplitude decreases as spaced signals that differ in amplitude. the third harmonic, 30 Hz; and so on. higher harmonics of the LO are used for The level of the phase noise indicates Since the phase noise indicates the mixing. Figure 7-11 shows the difference the angular, or frequency, deviation of signal (noise in this case) producing in phase noise between fundamental the LO. What happens to phase noise the modulation, the level of the phase mixing of a 5-GHz signal and fourth- when a harmonic of the LO is used in noise must be higher to produce greater harmonic mixing of a 20-GHz signal. the mixing process? Relative to funda- deviation. When the degree of modula-

68 Improved dynamic range

A preselector improves dynamic range –45 if the signals in question have suficient –50 frequency separation. The discussion of dynamic range in Chapter 6 assumed that both the large and small signals –60 were always present at the mixer and their amplitudes did not change dur- –70 ing the course of the measurement. But as we have seen, if signals are far –80 enough apart, a preselector allows one to reach the mixer while rejecting the –90 others. For example, if we were to test

a microwave oscillator for harmonics, Internal distortion (dBc) –100 a preselector would reject the fundamental when we tuned the analyzer to one of the harmonics. –110 –115 Let’s look at the dynamic range of a –120 second-harmonic test of a 3-GHz oscil- –90 –80 –70 –60 –50 –40 –30 –20 –10 0 lator. Using the example from Chapter Mixed level (dBm) 6, suppose that a –40-dBm signal at the mixer produces a second harmonic Figure 7-12. Second-order distortion graph product of –75 dBc. We also know, from our discussion, that for every 1 dB the level of the fundamental changes at the mixer, measurement range also changes by 1 dB. The second-harmonic (harmonic) end only by the noise loor conservative igure, we might use 18 distortion curve is shown in Figure (sensitivity) of the analyzer. dB per octave of bandwidth roll-off of 7-12. For this example, we assume a typical YIG preselector ilter beyond plenty of power from the oscillator and What about the upper, high-level the 3 dB point. So to determine the set the input attenuator so that when end? When measuring the oscillator improvement in dynamic range, we we measure the oscillator fundamental, fundamental, we must limit power at must determine to what extent each of the level at the mixer is –10 dBm, below the mixer to get an accurate reading of the fundamental tones is attenuated the 1-dB compression point. the level. We can use either internal or and how that affects internally gener- external attenuation to limit the level of ated distortion. From the expressions in From the graph, we see that a –10-dBm the fundamental at the mixer to some- Chapter 6 for third-order intermodula- signal at the mixer produces a second- thing less than the 1-dB compression tion, we have: harmonic distortion component of point. However, the preselector highly 2 –45 dBc. Now we tune the analyzer attenuates the fundamental when we (k4/8)VLOV1 V2 cos[ωLO– (2ω1 – ω2)]t to the 6-GHz second harmonic. If the are tuned to the second harmonic, so and preselector has 70-dB rejection, the we can remove some attenuation if (k /8)V V V 2cos[ω – (2ω – ω )]t fundamental at the mixer has dropped we need better sensitivity to measure 4 LO 1 2 LO 2 1 to –80 dBm. Figure 7-12 indicates that the harmonic. A fundamental level of for a signal of –80 dBm at the mixer, +20 dBm at the preselector should Looking at these expressions, we see the internally generated distortion is not affect our ability to measure the that the amplitude of the lower distor- ω – ω ) varies as –115 dBc, meaning 115 dB below the harmonic. tion component (2 1 2 new fundamental level of –80 dBm. the square of V1 and linearly with V2. This puts the absolute level of the Any improvement in dynamic range for On the other side, the amplitude of ω harmonic at –195 dBm. So the differ- third-order intermodulation measure- the upper distortion component (2 2 ω ence between the fundamental we ments depends upon separation of the – 1) varies linearly with V1 and as the square of V . However, depending on tuned to and the internally generated test tones versus preselector band- 2 second harmonic we tuned to is 185 width. As we noted, typical preselector the signal frequencies and separation, the preselector may not attenuate the dB! Clearly, for harmonic distortion, dy- bandwidth is about 35 MHz at the low namic range is limited on the low-level end and 80 MHz at the high end. As a two fundamental tones equally.

69 Consider the situation shown in Figure 7-13 in which we are tuned to the lower 3 dB distortion component, and the two 21 dB fundamental tones are separated by half the preselector bandwidth. In this case, the lower-frequency test tone lies at the edge of the preselector pass band and is attenuated 3 dB. The upper test tone lies above the lower distortion component by an amount equal to the full preselector bandwidth. It is attenu- 27 dB ated approximately 21 dB. Since we are tuned to the lower distortion component, internally generated distortion at this frequency drops by a factor of two relative to the attenuation of V1 (2 times 3 dB = 6 dB) and equally as fast Figure 7-13. Improved third-order intermodulation distortion; test tone separation is signiicant relative to preselector bandwidth as the attenuation of V2 (21 dB). The improvement in dynamic range is the preselector with that of the input mixer conigurations, the tuning ramp for the sum of 6 dB + 21 dB, or 27 dB. As in the can cause a degradation of frequency preselector and local oscillator come case of second harmonic distortion, the response. You must use proper calibra- from the same source, but there is no noise loor of the analyzer must be con- tion techniques to compensate for this feedback mechanism to ensure the sidered, too. For very closely spaced ripple. Another approach to minimize preselector exactly tracks the tuning of test tones, the preselector provides this interaction would be to insert a the analyzer. Another source of post- no improvement, and we determine matching pad (ixed attenuator) or tuning drift is the self-heating caused dynamic range as if the preselector was isolator between the preselector and by current lowing in the preselector not there. mixer. In this case, sensitivity would be circuitry. The center of the preselector degraded by the full value of the pad or pass band will depend on its tempera- The discussion of dynamic range in isolator. ture and temperature gradients, which Chapter 6 applies to the low-pass- depend on the history of the preselec- iltered low band. The only exceptions Some spectrum analyzer architectures tor tuning. As a result, you obtain the occur when a particular harmonic eliminate the need for the match- best latness by centering the prese- of a low-band signal falls within the ing pad or isolator. As the electrical lector at each signal. The centering preselected range. For example, if length between the preselector and function is typically built into the spec- we measure the second harmonic of mixer increases, the rate of change of trum analyzer irmware and is selected a 2.5- GHz fundamental, we get the phase of the relected and re-relected either by a front-panel key in manual beneit of the preselector when we signals becomes more rapid for a given measurement applications or program- tune to the 5-GHz harmonic. change in input frequency. The result matically in automated test systems. is a more exaggerated ripple effect on When activated, the centering function Pluses and minuses latness. Architectures such as those adjusts the preselector tuning DAC to of preselection used in PSA Series analyzers include center the preselector pass band on the mixer diodes as an integral part of the signal. The frequency response We have seen the pluses of preselec- the preselector/mixer assembly. In such speciication for most microwave tion: simpler analyzer operation, un- an assembly, there is minimal electri- analyzers applies only after centering cluttered displays, improved dynamic cal length between the preselector and the preselector, and it is generally a range and wide spans. But there are mixer. This architecture thus removes best practice to perform this function also some disadvantages relative to an the ripple effect on frequency response (to mitigate the effects of post-tuning unpreselected analyzer. and improves sensitivity by eliminating drift) before making amplitude mea- the matching pad or isolator. surements of microwave signals. First of all, the preselector has insertion loss, typically 6 to 8 dB. This loss Even aside from its interaction with In our discussion of sweep time, we comes prior to the irst stage of gain, the mixer, a preselector causes some found that analyzers such as the PXA so system sensitivity is degraded by degradation of frequency response. use FFTs when the narrower resolu- the full loss. In addition, when a prese- The preselector ilter pass band is tion bandwidths are selected. Because lector is connected directly to a mixer, never perfectly lat, but rather exhibits the LO is stepped and ixed for each the interaction of the mismatch of the a certain amount of ripple. In most FFT segment, the preseletor must be

70 stepped and ixed as well. Since the the fourth harmonic (N=2–), with the LO to the external mixer, which uses the preselector takes several milliseconds to doubled, to tune to 26.5 GHz. However, higher harmonics to mix with the high- tune and stabilize, sweep time may be what if you want to test outside the frequency signals. The external mixer’s negatively impacted relative to similar upper frequency range of the signal IF output connects to the analyzer’s IF settings in the low band. The X-Series analyzer? Some analyzers provide the “in” port. The latest analyzers have only signal analyzers allow you to select the ability to use an external mixer to make one front-panel port, and this is possible width of each step to minimize the num- high-frequency measurements, where because the LO frequency supplied from ber of steps. (For details, see the operat- the external mixer becomes the front the analyzer is between 3 and 14 GHz, ing manual for your particular analyzer.) end of the analyzer, bypassing the input while the IF output frequency from the If your analyzer has Option MPB, you attenuator, the preselector and the irst external mixer to the analyzer is 322.5 may bypass the preselector to eliminate mixers. The external mixer uses higher MHz. Because of the wide frequency its impact on sweep time. However, be harmonics of the analyzer’s irst LO, difference between the LO and IF sig- sure your signal is such that no images or and in some cases, the irst LO fre- nals, both signals can exist on the same multiples can cause confusion. quency is doubled before being sent to coaxial interconnect cable that attaches the external mixer. Higher fundamental the analyzer and the mixer. As long as External harmonic mixing LO frequencies allow for lower mixer the external mixer uses the same IF as conversion loss. Typically, a spectrum the spectrum analyzer, the signal can be We have discussed tuning to higher analyzer that supports external mixing processed and displayed internally, just frequencies within the signal analyzer. has one or two additional connectors like any signal that came from the inter- For internal harmonic mixing, the X- on the front panel. Early analyzers had nal irst mixer. Figure 7-14 illustrates the Series signal analyzers use the second two connectors. An LO “out” port routes block diagram of an external mixer used – harmonic (N=2 ) to tune to 17.1GHz and the analyzer’s internal irst LO signal in conjunction with a spectrum analyzer.

External mixer

Waveguide IF out input IF in

Low band path Analog or 3.6 GHz 5.1225 GHz 322.5 MHz22.5 MHz digital IF

Analyzer input

LO 3.8 to 8.7 GHz High band path To external mixer 4.8 GHz 300 MHz

322.5 MHz

Preselector

Sweep generator Display

Figure 7-14. Spectrum analyzer and external mixer block diagram

71 Table 7-1. Harmonic mixing modes used by X-Series analyzers with external mixers

Band Keysight 11970 Series Keysight M1970 Series Other Manufacturer’s Other Manufacturer’s Mixers Mixers mixers mixers (LO range 3–7 GHz) (LO range 6–14 GHz) (LO range 3–7 GHz) (LO range 6–14 GHz) A (26.5 to 40.0 GHz) 6− and 8− Q (33.0 to 50.0 GHz) 8− and 10− U (40.0 to 60.0 GHz) 10− V (50.0 to 75.0 GHz) 12− and 14− 6− E (60.0 to 90.0 GHz) N.A. 6− and 8− W (75.0 to 110.0 GHz) 18− 8− F (90.0 to 140.0 GHz) 16− 10− D (110.0 to 170.0 GHz) 20− 14− G (140.0 to 220.0 GHz) 26− 18− Y (170.0 to 260.0 GHz ) 30− 20− J (220.0 to 325.0 GHz) 38− 24− (325.0 to 500.0 GHz) 58− 36− (500.0 to 750.0 GHz) 86− 54- (750.0 to 1,100.0 GHz) 80−

Table 7-1 shows the harmonic mixing external mixers from other manufactur- produces mixing products that can be modes used by the X-Series analyzers ers require a bias current to set the mix- processed through the IF just like any at various millimeter-wave bands for er diodes to the proper operating point. other valid signals. both the Keysight M1970 Series and the The X-Series analyzers can provide earlier 11970 Series external mixers. For up to ± 10 mA of DC current through A tunable ilter that performs preselec- ease of use and low conversion loss, the front-panel external mixer port to tion of the signals reaching the irst the M1970 Series mixers provide a USB provide this bias and keep the measure- mixer in the internal signal path is com- connection that is used to automatically ment setup as simple as possible. mon in most signal analyzers. External identify the mixer model number and mixers that are unpreselected will serial number, perform an LO adjust- Whether you perform harmonic mixing produce unwanted responses on screen ment to optimize performance, and with an internal or an external mixer, that are not true signals. A way to deal download the mixer conversion loss the issues are similar. The LO and its with these unwanted signals has been table into the analyzer memory. You harmonics mix not only with the desired designed into the signal analyzer. This also can use external mixers from other RF input signal, but also with any other function is called “signal identiication.” manufactures if you know the mixer’s signal, including out-of-band signals, conversion loss with frequency. Some that may be present at the input. This

72 Signal identiication It is quite possible that the particular response we have tuned onto the display has been generated on an LO harmonic or mixing mode other than the one for which the display is calibrated. So our analyzer must have some way to tell us whether or not the display is calibrated for the signal response in question. For this example, assume we are using a Keysight M1970V 50- to 75-GHz unpreselected mixer, which uses the 6− mixing mode. The full V-Band measurement can be seen in Figure 7-15.

Keysight X-Series signal analyzers offer two different identiication methods: image shift and image suppress. Let’s irst explore the image shift method. Looking at Figure 7-15, let’s assume we have tuned the analyzer to a frequency Figure 7-15. Which ones are the real signals? of 62.50 GHz. The 6th harmonic of the LO produces a pair of responses, frequency of 61.85 GHz, which is 2 where the 6− mixing product appears times fIF below the real response. The on screen at the correct frequency of X-Series analyzer has an IF frequency 62.50 GHz, while the 6+ mixing product of 322.5 MHz, so the pair of responses produces a response with an indicated is separated by 645 MHz.

Harmonic mixing tuning lines 120 8+

8- 100

Signal frequency, GHz 80 6-

62.500 61.855 60 Apparent location of image 40

20 6 8 10.36 10.47 12 14 LO frequency, GHz

Figure 7-16 Harmonic tuning lines for the X-Series analyzers using the M1970 Series mixers

73 Let’s assume we have some idea of the characteristics of our signal, but we do not know its exact frequency. How do we determine which is the real signal? The image-shift process retunes the LO fundamental frequency by an amount equal to 2fIF/N. This causes the Nth harmonic to shift by 2fIF.

If we are tuned to a real signal, its corresponding pair will now appear at the same position on screen that the real signal occupied in the irst sweep. If we are tuned to another multiple pair created by some other incorrect harmonic, the signal will appear to shift in frequency on the display. The X-Series signal analyzer shifts the LO on alternate sweeps, creating the two displays show in Figures 7-17a and 7-17b. In Figure 7-17a, the real signal (the 6− mixing product) is tuned to the center Figure 7-17a: 6− centered (yellow trace) of the screen. Figure 7-17b shows how the image shift function moves the corresponding pair (the 6+ mixing product) to the center of the screen.

Figures 7-17a and 7-17b display alternate sweeps taken with the image shift function.

Figure 7-17b: 6+ centered (blue trace)

74 Let’s examine the second method of signal identiication, image suppres- sion. In this mode, two sweeps are taken using the minimum hold function, which saves the smaller value of each display point, or bucket, from the two sweeps. The irst sweep is done using normal LO tuning values. The second sweep offsets the LO fundamental frequency by 2fIF/N. As we saw in the irst signal ID method, the image product generated by the correct harmonic will land at the same point on the display as the real signal did on the irst sweep. Therefore, the trace retains a high amplitude value. Any false response that shifts in frequency will have its trace data replaced by a lower value. Thus, all image and incorrect multiple responses will appear as noise, as shown in Figure 7-18.

Figure 7-18. The image suppress function displays only real signals Note that both signal identiication methods are used for identifying correct frequencies only. You should not attempt to make amplitude measurements while the signal identiication function is turned on. Once we have identiied the real signal of interest, we turn off the signal ID function and zoom in on it by reducing the span. We can then measure the signal’s amplitude and frequency. See Figure 7-19.

To make an accurate amplitude measurement, it is important that you irst enter the calibration data for your external mixer. This data is normally supplied by the mixer manufacturer, and it is typically presented as a table of mixer conversion loss, in dB, at a number of frequency points across the band. This data is entered into a correction table on the signal analyzer, and the analyzer uses this data to compensate for the mixer conversion Figure 7-19. Measurement of a positively identiied signal loss. If you are using the M1970 Series harmonic mixers, the mixer conversion loss is automatically transferred from the mixer memory to the X-Series signal analyzer memory, which eliminates manual entry into a correction ile. The spectrum analyzer reference level is now calibrated for signals at the input to the external mixer.

75 Chapter 8. Modern Signal Analyzers

In the previous chapters of this ap- Application-speciic power of the signal exceeds the aver- plication note, we have looked at the measurements age power by a certain number of dB. fundamental architecture of spectrum This information is important in power analyzers and basic considerations for In addition to measuring general signal ampliier design, for example, where it making frequency-domain measure- characteristics like frequency and is important to handle instantaneous ments. On a practical level, modern amplitude, you often need to make signal peaks with minimum distor- spectrum or signal analyzers must also speciic measurements of certain signal tion while minimizing cost, weight and handle many other tasks to help you parameters. Examples include chan- power consumption of the device. meet your measurement requirements. nel power measurements and adjacent These tasks include: channel power (ACP) measurements, Other examples of built-in measure- which we described in Chapter 6. Many ment functions include occupied band- – Providing application-speciic signal analyzers now have these built-in width, TOI, harmonic distortion, ACP measurements, such as adjacent functions available. You simply specify and spurious emissions measurements. channel power (ACP), noise igure, the channel bandwidth and spacing, The instrument settings – such as and phase noise then press a button to activate the center frequency, span and resolution –Providing digital modulation automatic measurement. bandwidth – for these measurements analysis measurements deined by depend on the speciic radio standard industry or regulatory standards, The complementary cumulative distri- to which the device is being tested. such as LTE, GSM, cdma2000®, bution function (CCDF), which shows Most modern signal analyzers have 802.11, or Bluetooth® power statistics, is another measure- these instrument settings stored in – Performing vector signal analysis ment capability increasingly found in memory so you can select the desired – Saving, printing or transferring modern signal analyzers, as you can radio standard (LTE, MSR, GSM/EDGE, data see in Figure 8-1. CCDF measurements cdma2000, W-CDMA, 802.11a/b/g/n/ – Offering remote control and opera- provide statistical information showing ac and so on) to make the measure- tion over GPIB, LAN or the Internet the percent of time the instantaneous ments properly. –Allowing you to update instrument irmware to add new features and capabilities, as well as to repair defects – Making provisions for self-calibration, troubleshooting, diagnostics and repair –Recognizing and operating with optional hardware or irmware to add new capabilities –Allowing you to make measurements in the ield with a rugged, battery-powered handheld spectrum analyzer that correlate with data taken with high-performance bench-top equipment

Figure 8-1. CCDF measurement

76 RF designers are often concerned with the noise igure of their devices, as noise igure directly affects the sensitivity of receivers and other systems. Some signal analyzers, such as the X-Series, have optional noise igure measurement capabilities available. This option provides control for the noise source needed to drive the input of the device under test (DUT) as well as irmware to automate the measurement process and display the results. Figure 8-2 shows a typical measurement result, with DUT noise igure (upper trace) and gain (lower trace) displayed as a function of frequency.

The need for phase information Phase noise is a common measure of oscillator performance. In digitally Figure 8-2. Noise igure measurement modulated communication systems, phase noise can negatively impact bit error rates. Phase noise can also degrade the ability of Doppler radar systems to capture the return pulses from targets. The X-Series signal analyzers offer optional phase noise measurement capabilities. These options provide irmware to control the measurement and display the phase noise as a function of frequency offset from the carrier, as shown in Figure 8-3.

Figure 8-3. Phase noise measurement

77 As communications have become more complex, phase has been introduced to speed the transmission of data. An example of a simple model is QPSK (quadra-phase-shift keying) in which two data values are transmitted at once. See Figure 8-4.

In this case, the data are plotted on an IQ diagram in which the in-phase data are plotted on the I (horizontal) axis and the quadrature data are plotted on the Q (vertical) axis. The combined values represent the phase and magnitude of the transmitted signal. As seen here, equal values for both I and Q should have produced only dots at 45, 135, 225 and 315 degrees. However, our analysis shows that neither the amplitude nor the phase is consistent Figure 8-4. Modulation analysis of a QPSK signal measured with Keysight’s 89600 VSA software as the signal changes quadrants. On the other hand, it is quite unlikely that the receiver would be confused as to which quadrant a particular data point belongs, so the transmission would be received without error.

A newer and much more compli- cated system is 802.11ac, which uses 256QAM (quadrature-amplitude modulation). See Figure 8-5. The maximum power is limited, so the data points are much closer in both phase and magnitude than for QPSK. The analyzer you use to evaluate the transmitted signal must be suficiently accurate that it does not lead you to a false conclusion about the quality of the transmission. Pure amplitude measurements are also required to determine signal attributes such as latness, adjacent-channel Figure 8-5. Modulation analysis of WLAN 802.11ac signal using Keysight 89600 VSA software power levels and distortion.

78 Digital modulation analysis The common wireless communication systems used throughout the world today all have prescribed measurement techniques deined by standards-development organizations and governmental regulatory bodies. Optional measurement personalities are available on the X-Series signal analyzers to perform the key tests deined for a particular wireless communication format. For example, if we need to test a transmitter to the Bluetooth wireless communication standard, we must measure parameters such as:

–Average/peak output power – Modulation characteristics –Initial carrier frequency tolerance –Carrier frequency drift – Monitor band/channel – Modulation overview Figure 8-6. EVM measurement of LTE FDD downlink signal –Output spectrum – 20-dB bandwidth Not all digital communication systems – Adjacent channel power are based on well-deined indus- More information try standards. If you are working on Additional information is These measurements are available on nonstandard proprietary systems or available on the following: the Keysight X-Series signal analyzers the early stages of proposed industry- with appropriate options. standard formats, you need more Noise igure measurements, see lexibility to analyze vector-modulated Keysight Noise Figure Measurements of Other optional measurement capabili- signals under varying conditions. You Frequency Converting Devices Using the ties for a wide variety of wireless com- can achieve that lexibility two ways. Keysight NFA Series Noise Figure Ana- munications standards that are avail- First, modulation analysis personalities lyzer − Application Note, literature able on the X-Series signal analyzers: are available on the X-Series signal an- number 5989-0400EN. alyzers. Alternatively, you can perform – LTE/LTE-Advanced more extensive analysis with software Measurements involving phase, – WLAN running on an external computer. For see Vector Signal Analysis Basics – – Multi-standard radio (MSR) example, you can use Keysight 89600 Application Note, literature num- – GSM/EDGE VSA software with X-Series signal ber 5989-1121EN. – W-CDMA analyzers to provide lexible vector – HSDPA signal analysis. In this case, the signal Bluetooth measurements, see – cdma2000 analyzer acts as an RF downconverter Performing Bluetooth RF Measurements – 1xEV-DO and digitizer. The software can run Today – Application Note, literature – 1xEV-DV internally on the signal analyzer or number 5968-7746E. – cdmaOne communicate with the analyzer over –NADC and PDC a GPIB or LAN connection. IQ data is – TD-SCDMA transferred to the computer, where it performs the vector signal analy- Figure 8-6 illustrates an error vector sis. Measurement settings, such as magnitude (EVM) measurement per- modulation type, symbol rate, iltering, formed on a LTE FDD downlink signal. triggering and record length, can be This test helps you diagnose modula- varied as necessary for the particular tion or ampliication distortions that signal you are analyzing. lead to bit errors in the receiver.

79 Real-time spectrum analysis For the capable RF engineer, continuous- wave (CW) and predictably-repeating signals are no great challenge – but today’s complex and agile signals and multi-signal environments are proving to be another matter. To keep up with evolving analysis needs, new types of signal analyzers and application software have emerged in recent years. The Keysight PXA and MXA signal analyzers now offer a combination of swept spectrum, real-time and vector signal analysis capability – all in one instrument.

Design and troubleshooting tasks are much more dificult when dealing with agile signals, and the challenges are often made more dificult when these Figure 8-7 . Even when you use fast sweeps and max hold over a period of many seconds, the swept signals are in an environment of other spectrum analyzer view of the radar signal is not very informative agile signals. Even the analysis of a single signal can be a challenge when that signal is very agile or complex. You can use the Keysight PXA and MXA real-time spectrum analysis capability to capture the behavior of dynamic and elusive signals with true gap-free spectrum analysis.

An example of a single complex signal is the agile S-band acquisition radar signal. The signal at the receiver varies widely in amplitude over a period of several seconds, and this long-duration characteristic, combined with the short-duration characteristics of its pulse length and repetition interval (and therefore short duty cycle) make it agile and dificult to measure well. A basic spectrum analysis of this signal Figure 8-8. Real-time capture of S-band acquisition radar signal with a swept spectrum analyzer shows the measurement dificulty it poses, as illustrated in Figure 8-7. Even after and frequent events, with an indication More information many sweeps and the use of of relative frequency of occurrence. a max hold function, the signal is not For additional information on clearly represented. The PXA’s real-time analyzer mode measurements involving real-time and density display provide a fast and spectrum analysis, see Measur- The Keysight PXA real-time spectrum insight-producing representation of this ing Agile Signals and Dynamic Signal analyzer screen shown in Figure 8-8, wideband, dynamic and agile signal. Environments – Application Note, in contrast with the swept spectrum The blue color of all but the noise loor literature number 5991-2119EN. screen, readily shows the main charac- indicates that the pulses, while promi- teristics of the signal using the density nent, have a very low frequency-of- or histogram display. The density or occurrence. This is the principal charac- histogram display collects a large teristic that makes it dificult to measure amount of real-time spectrum data into (or even to rapidly and reliably ind) this a single display that shows both rare signal with a swept spectrum analyzer.

80 Chapter 9. Control and Data Transfer

Saving and printing data Data transfer and remote A variety of commercial software instrument control products are available to control After making a measurement, we spectrum analyzers remotely over an normally want to keep a record of the In 1977, Keysight Technologies (part I/O bus. Also, you can write your own test data. We might simply want to of Hewlett-Packard at that time) software to control spectrum analyz- make a quick printout of the instrument introduced the world’s irst GPIB- ers in a number of different ways. One display. Depending on the particular controllable spectrum analyzer, the method is to directly send program- analyzer and printer model, we might 8568A. The GPIB interface (also known ming commands to the instrument. use the USB or LAN ports to connect as HP-IB or IEEE-488) made it pos- Older spectrum analyzers typically the two units. sible to control all major functions of used proprietary command sets, but the analyzer from an external computer newer instruments, such as Keysight’s Very often, we may want to save mea- and transfer trace data to an external X-Series signal analyzers, use industry- surement data as a ile, either in the computer. This innovation paved the standard SCPI (standard commands spectrum analyzer’s internal memory or way for a wide variety of automated for programmable instrumentation) on a USB mass-storage device. There spectrum analyzer measurements that commands. A more common method are several different kinds of data we were faster and more repeatable than is to use standard software drivers, can save this way: manual measurements. By transferring such as VXI plug&play drivers, which the raw data to a computer, it could be enable higher-level functional com- – An image of the display – Prefer- saved on disk, analyzed, corrected and mands to the instrument without the ably in a popular ile format, such operated on in a variety of ways. need for detailed knowledge of the as bitmap, GIF, PNG or Windows SCPI commands. Most recently, a new metaile. Today, automated test and measure- generation of language-independent – Trace data – Saved as X-Y data ment equipment has become the norm, instrument drivers, known as “inter- pairs representing frequency and and nearly all modern spectrum ana- changeable virtual instrument,” or amplitude points on the screen. lyzers come with a variety of standard IVI-COM drivers, has become avail- The number of data pairs can vary. interfaces, including Ethernet LAN, able for the X-Series signal analyzers. Modern spectrum analyzers such as USB 2.0 and GPIB. Ethernet LAN con- The IVI-COM drivers are based on the the X-Series allow you to select the nectivity is the most commonly used Microsoft Component Object Model desired display resolution by setting interface, as it can provide high data- standard and work in a variety of PC a minimum of 1 up to a maximum transfer rates over long distances and application development environments, of 40,001 frequency sweep points integrates easily into networked envi- such as the Keysight T&M Program- (buckets) on the screen. This data ronments such as a factory loor. Other mers Toolkit and Microsoft’s Visual format is well suited for transfer to a standard interfaces used widely in the Studio .NET. spreadsheet program on a com- computer industry are likely to become puter. available on spectrum analyzers in the Some applications require you to – Instrument state – To keep a record future to simplify connectivity between control the spectrum analyzer and of the spectrum analyzer settings, instruments and computers. collect measurement data from a very such as center frequency, span, long distance. For example, you may reference level and so on, used in Keysight’s X-Series signal analyzers lit- want to monitor satellite signals from the measurement. This informa- erally have computer irmware running a central control room, collecting data tion is useful for documenting test USB ports and a Windows operating from remote tracking stations located setups used for making measure- system. These features greatly simplify hundreds or even thousands of kilome- ments. Consistent test setups are control and data transfer. In addition, ters from the central site. The X-Series essential for maintaining repeatable the X-Series analyzers can be operated signal analyzers have software options measurements over time. remotely, and the analyzer’s display ap- available to control these units, capture pears on the remote computer. Details screen images and transfer trace data are beyond the scope of this applica- over the Internet using a standard Web tion note; see the operating manual for browser. your particular analyzer.

81 Firmware updates Calibration, troubleshooting, Summary Modern spectrum analyzers have much diagnostics and repair This application note has provided a broad survey of basic spectrum more software inside them than do Spectrum analyzers must be periodi- analyzer concepts. However, you may instruments from just a few years ago. As cally calibrated to insure the instru- wish to learn more about many other new features are added to the software ment performance meets all published topics related to spectrum analysis. An and defects repaired, it becomes highly speciications. Typically, this is done excellent place to start is to visit the desirable to update the spectrum ana- once a year. However, between these Keysight Technologies Web site at lyzer’s irmware to take advantage of the annual calibrations, the spectrum www.keysight.com and search for improved performance. analyzer must be aligned periodically to signal or spectrum analyzer. compensate for thermal drift and aging The latest revisions of spectrum and effects. Modern spectrum analyzers signal analyzer irmware can be found such as the X-Series have built-in align- on the Keysight Technologies website. ment routines that operate when the You can download this irmware to a instrument is irst turned on and during ile on your local computer. A common retrace (between sweeps) at predeter- method to transfer new irmware into mined intervals. The alignment routines the spectrum analyzer is to copy the also operate if the internal tempera- irmware onto a USB drive and then in- ture of the instrument changes. These sert it into one of the spectrum analyz- alignment routines continuously adjust er’s USB ports. Some models, such as the instrument to maintain speciied the X-Series, allow you to transfer the performance. new irmware directly into the spectrum analyzer using the instrument’s Ethernet Modern spectrum analyzers usually LAN port. have a service menu available. In this area, you can perform useful diagnostic It is a good practice to periodically functions, such as a test of the front- check your spectrum analyzer model’s panel keys. You also can display more Web page to see if updated irmware is details of the alignment process, as well available. as a list of all optional hardware and measurement personalities installed in the instrument. When you upgrade a spectrum analyzer with a new measurement personality, Keysight provides a unique license key tied to the serial number of the instrument. You install this license key through the USB port or enter it on the front-panel keypad to activate the measurement capabilities of the personality.

82 Glossary of Terms

Absolute amplitude accuracy: The Average detection: A method of detec- Delta marker: A mode in which a ixed, uncertainty of an amplitude measure- tion that sums power across a frequency reference marker has been established ment in absolute terms, either volts or interval. It is often used for measuring and a second, active marker is avail- power. Includes relative uncertainties complex, digitally modulated signals and able that we can place anywhere on the (see Relative amplitude accuracy) plus other types of signals with noise-like displayed trace. A read-out indicates the calibrator uncertainty. For improved characteristics. Modern Keysight spec- relative frequency separation and ampli- accuracy, some spectrum analyzers have trum analyzers typically offer three types tude difference between the reference frequency response speciied relative of average detection: power (rms) aver- marker and the active marker. to the calibrator as well as relative to aging, which measures the true average the midpoint between peak-to-peak power over a bucket interval; voltage Digital display: A display technology extremes. averaging, which measures the aver- where digitized trace information, stored age voltage data over a bucket interval; in memory, is displayed on an instru- ACPR: Adjacent channel power ratio is and log-power (video) averaging, which ment’s screen. The displayed trace is a measure of how much signal energy measures the logarithmic amplitude in a series of points designed to present from one communication channel spills dB of the envelope of the signal during a continuous-looking trace. While the over or leaks into an adjacent channel. the bucket interval. default number of display points varies This is an important metric in digital between different models, most modern communication components and sys- Average noise level: See Displayed spectrum analyzers allow the user to tems, as too much leakage will cause average noise level. choose the desired resolution by con- interference on adjacent channels. It is trolling the number of points displayed. sometimes also described as ACLR, or Bandwidth selectivity: A measure of The display is refreshed (rewritten from adjacent channel leakage ratio. an analyzer’s ability to resolve signals data in memory) at a licker-free rate; the unequal in amplitude. Also called shape data in memory is updated at the sweep Amplitude accuracy: The uncertainty factor, bandwidth selectivity is the ratio rate. Nearly all modern spectrum analyz- of an amplitude measurement. It can be of the 60-dB bandwidth to the 3-dB ers have digital lat-panel LCD displays, expressed either as an absolute term or bandwidth for a given resolution (IF) rather than CRT-based analog displays relative to another reference point. ilter. For some analyzers, the 6-dB that were used in earlier analyzers. bandwidth is used in lieu of the 3-dB Amplitude reference signal: A signal of bandwidth. In either case, bandwidth Display detector mode: The man- precise frequency and amplitude that selectivity tells us how steep the ilter ner in which the signal information is the analyzer uses for self-calibration. skirts are. processed prior to being displayed on screen. See Neg peak, Pos peak, Normal Analog display: A display technology Blocking capacitor: A ilter that keeps and Sample. where analog signal information (from unwanted low-frequency signals (in- the envelope detector) is written directly cluding DC) from damaging circuitry. Digital IF: An architecture found in to an instrument’s display, typically A blocking capacitor limits the lowest modern spectrum analyzers in which the implemented on a cathode ray tube frequency that can be measured ac- signal is digitized soon after it has been (CRT). Analog displays were once the curately. downconverted from an RF frequency standard method of displaying informa- to an intermediate frequency (IF). At tion on spectrum analyzers. However, CDMA: Code division multiple access that point, all further signal processing modern spectrum analyzers no longer is a method of digital communication in is done using digital signal processing use this technology; instead, they now which multiple communication streams (DSP) techniques. use digital displays. are orthogonally coded, enabling them to share a common frequency channel. It Display dynamic range: The maximum is a popular technique used in a number dynamic range for which both the of widely used mobile communication larger and smaller signal may be viewed systems. simultaneously on the spectrum analyzer display. For analyzers with a maximum Constellation diagram: A display type logarithmic display of 10 dB/div, the ac- commonly used when analyzing digitally tual dynamic range (see Dynamic range) modulated signals in which the detected may be greater than the display dynamic symbol points are plotted on an IQ range. graph.

83 Display scale idelity: The degree of un- Error vector magnitude (EVM): A quality Frequency resolution: The ability of a certainty in measuring relative differenc- metric in digital communication systems. spectrum analyzer to separate closely es in amplitude on a spectrum analyzer. EVM is the magnitude of the vector spaced spectral components and display The logarithmic and linear IF ampliiers difference at a given instant in time be- them individually. Resolution of equal found in analyzers with analog IF sec- tween the ideal reference signal and the amplitude components is determined tions never have perfect logarithmic or measured signal. by resolution bandwidth. The ability to linear responses, and therefore they resolve unequal amplitude signals is a introduce uncertainty. Modern analyzers External mixer: An independent mixer, function of both resolution bandwidth with digital IF sections have signiicantly usually with a waveguide input port, and bandwidth selectivity. better display scale idelity. used to extend the frequency range of spectrum analyzers that use external Frequency response: Variation in the Display range: The calibrated range mixers. The analyzer provides the LO displayed amplitude of a signal as a of the display for the particular display signal and, if needed, mixer bias. Mixing function of frequency (latness). Typically mode and scale factor. See Linear and products are returned to the analyzer’s speciied in terms of ± dB relative to the Log display and Scale factor. IF input. value midway between the extremes. Also may be speciied relative to the Displayed average noise level: The FFT (fast Fourier transform): A math- calibrator signal. noise level as seen on the analyzer’s ematical operation performed on a time- display after setting the video bandwidth domain signal to yield the individual Frequency span: The frequency range narrow enough to reduce the peak-to- spectral components that constitute the represented by the horizontal axis of peak noise luctuations such that the signal. the display. Generally, frequency span displayed noise is essentially seen as a See Spectrum. is given as the total span across the full straight line. Usually refers to the ana- display. Some earlier analyzers indicate lyzer’s own internally generated noise as Fast sweep: A digital signal processing frequency span (scan width) on a per- a measure of sensitivity and is typically technique that implements complex- division basis. speciied in dBm under conditions of valued resolution bandwidth iltering for minimum resolution bandwidth and a sweeping spectrum analyzer, allowing Frequency stability: A general phrase minimum input attenuation. faster sweep rates than a traditional that covers both short- and long-term analog or digital resolution bandwidth LO instability. The sweep ramp that Drift: The very slow (relative to sweep ilter would allow. tunes the LO also determines where time) change of signal position on the a signal should appear on the display. display as a result of a change in LO Flatness: See Frequency response. Any long term variation in LO frequency frequency versus sweep voltage. The (drift) with respect to the sweep ramp primary sources of drift are the tem- Frequency accuracy: The degree of causes a signal to slowly shift its hori- perature stability and aging rate of the uncertainty with which the frequency zontal position on the display. Shorter- frequency reference in the spectrum of a signal or spectral component is term LO instability can appear as ran- analyzer. indicated, either in an absolute sense or dom FM or phase noise on an otherwise relative to some other signal or spec- stable signal. Dynamic range: The ratio, in dB, be- tral component. Absolute and relative tween the largest and smallest signals frequency accuracies are speciied Full span: For most modern spectrum simultaneously present at the spectrum independently. analyzers, full span means a frequency analyzer input that can be measured to span that covers the entire tuning range a given degree of accuracy. Dynamic Frequency range: The minimum to of the analyzer. These analyzers include range generally refers to measurement of maximum frequencies over which a single -band RF analyzers and micro- distortion or intermodulation products. spectrum analyzer can tune. While the wave analyzers such as the ESA, PSA maximum frequency is generally thought and X- Series that use a solid-state Envelope detector: A circuit element of in terms of an analyzer’s coaxial input, switch to switch between the low and whose output follows the envelope, but the range of many microwave analyzers preselected ranges. not the instantaneous variation, of its can be extended through use of external input signal. In a superheterodyne spec- waveguide mixers. NOTE: On some earlier spectrum analyz- trum analyzer, the input to the envelope ers, full span referred to a sub-range. detector comes from the inal IF, and the For example, with the Keysight 8566B, a output is a video signal. When we put microwave spectrum analyzer that used our analyzer in zero span, the envelope a mechanical switch to switch between detector demodulates the input signal, the low and preselected ranges, full span and we can observe the modulating sig- referred to either the low, non-preselect- nal as a function of time on the display. ed range or the high, preselected range.

84 Gain compression: That signal level at Image response: A displayed signal that Linear display: The display mode in the input mixer of a spectrum analyzer is actually twice the IF away from the which vertical delection on the display at which the displayed amplitude of the frequency indicated by the spectrum is directly proportional to the voltage of signal is a speciied number of dB too analyzer. For each harmonic of the LO, the input signal. The bottom line of the low due just to mixer saturation. The sig- there is an image pair, one below and graticule represents 0 V, and the top nal level is generally speciied for 1-dB one above the LO frequency by the IF. line, the reference level, some nonzero compression and is usually between +3 Images usually appear only on non- value that depends upon the particular and –10 dBm, depending on the model preselected spectrum analyzers. spectrum analyzer. On most modern of spectrum analyzer. analyzers, we select the reference level, Incidental FM: Unwanted frequency and the scale factor becomes the refer- GSM: The global system for mobile modulation on the output of a device ence level value divided by the number communication is a widely used digital (signal source, ampliier) caused by (in- of graticule divisions. Although the standard for mobile communication. It is cidental to) some other form of modula- display is linear, modern analyzers allow a TDMA-based system in which multiple tion, e.g., amplitude modulation. reference level and marker values to be communication streams are interleaved indicated in dBm, dBmV, dBuV, and in in time, enabling them to share a com- Input attenuator: A step attenuator some cases, watts as well as volts. mon frequency channel. between the input connector and irst mixer of a spectrum analyzer. Also called LO emission or feedout: The emer- Harmonic distortion: Unwanted fre- the RF attenuator. The input attenuator gence of the LO signal from the input of quency components added to a signal is used to adjust level of the signal inci- a spectrum analyzer. The level can be as the result of the nonlinear behavior of dent upon the irst mixer. The attenuator greater than 0 dBm on non-preselected the device (e.g., mixer, ampliier) through is used to prevent gain compression due spectrum analyzers but is usually less which the signal passes. These unwant- to high-level or broadband signals and than –70 dBm on preselected analyzers. ed components are harmonically related to set dynamic range by controlling the to the original signal. degree of internally generated distortion. LO feedthrough: The response on the In some analyzers, the vertical position display when a spectrum analyzer is Harmonic mixing: Using the LO harmon- of displayed signals is changed when the tuned to 0 Hz, i.e., when the LO is tuned ics generated in a mixer to extend the input attenuator setting is changed, so to the IF. The LO feedthrough can be tuning range of a spectrum analyzer the reference level is also changed ac- used as a 0-Hz marker, and there is no beyond the range achievable using just cordingly. In modern Keysight analyzers, frequency error. the LO fundamental. the IF gain is changed to compensate for input attenuator changes, so signals Log display: The display mode in which IF gain/IF attenuation: Adjusts the remain stationary on the display, and the vertical delection on the display is a vertical position of signals on the display reference level is not changed. logarithmic function of the voltage of without affecting the signal level at the the input signal. We set the display input mixer. When changed, the value of Input impedance: The terminating calibration by selecting the value of the reference level is changed accord- impedance that the analyzer presents to the top line of the graticule, the refer- ingly. the signal source. The nominal imped- ence level and scale factor in dB/div. On ance for RF and microwave analyzers is Keysight analyzers, the bottom line of IF feedthrough: A raising of the base- usually 50 ohms. For some systems, e.g., the graticule represents zero volts for line trace on the display due to an input cable TV, 75 ohms is standard. The de- scale factors of 10 dB/div or more, so the signal at the intermediate frequency gree of mismatch between the nominal bottom division is not calibrated in these passing through the input mixer. Gener- and actual input impedance is given in cases. Modern analyzers allow reference ally, this is a potential problem only on terms of VSWR (voltage standing wave level and marker values to be indicated non-preselected spectrum analyzers. ratio). in dBm, dBmV, dBuV, volts, and in some The entire trace is raised because the cases, watts. Earlier analyzers usually signal is always at the IF; mixing with the Intermodulation distortion: Unwanted offered only one choice of units, and LO is not required. frequency components resulting from dBm was the usual choice. the interaction of two or more spectral Image frequencies: Two or more real components passing through a device Marker: A visible indicator we can place signals present at the spectrum analyzer with nonlinear behavior (e.g., mixer, anywhere along the displayed signal input that produce an IF response at the ampliier). The unwanted components trace. A read out indicates the abso- same LO frequency. Because the mixing are related to the fundamental compo- lute value of both the frequency and products all occur at the same LO and IF nents by sums and differences of the amplitude of the trace at the marked frequencies, it is impossible to distin- fundamentals and various harmonics, point. The amplitude value is given in the guish between them. e.g. f1 ± f2, 2f1 ± f2, 2f2 ± f1, 3f1 ± 2f2, and currently selected units. Also see Delta so forth. marker and Noise marker.

85 Measurement range: The ratio, ex- Noise marker: A marker whose value Preampliier: An external, low-noise- pressed in dB, of the maximum signal indicates the noise level in a 1-Hz noise igure ampliier that improves system level that can be measured (usually the power bandwidth. When the noise (preampliier/spectrum analyzer) sensi- maximum safe input level) to the dis- marker is selected, the sample display tivity over that of the analyzer itself. played average noise level (DANL). This detection mode is activated, the values ratio is almost always much greater than of a number of consecutive trace points Preselector: A tunable bandpass ilter can be realized in a single measurement. (the number depends upon the analyzer) that precedes the input mixer of a spec- See Dynamic range. about the marker are averaged, and trum analyzer and tracks the appropriate this average value is normalized to an mixing mode. Preselectors are typically Mixing mode: A description of the equivalent value in a 1-Hz noise power used only above 2 GHz. They essentially particular circumstance that creates a bandwidth. The normalization process eliminate multiple and image responses given response on a spectrum analyzer. accounts for detection and bandwidth and, for certain signal conditions, im- The mixing mode, e.g., 1+, indicates the plus the effect of the log ampliier when prove dynamic range. harmonic of the LO used in the mixing we select the log display mode. process and whether the input signal is Quasi-peak detector (QPD): A type of above (+) or below (–) that harmonic. Noise power bandwidth: A ictitious detector whose output is a function of ilter that would pass the same noise both signal amplitude as well as pulse Multiple responses: Two or more re- power as the analyzer’s actual ilter, repetition rate. The QPD gives higher sponses on a spectrum analyzer display making comparisons of noise mea- weighting to signals with higher pulse from a single input signal. Multiple re- surements among different analyzers repetition rates. In the limit, a QPD will sponses occur only when mixing modes possible. exhibit the same amplitude as a peak overlap and the LO is swept over a wide detector when measuring a signal with a enough range to allow the input signal Noise sidebands: Modulation sidebands constant amplitude (CW) signal. to mix on more than one mixing mode. that indicate the short-term instability Normally not encountered in analyzers of the LO (primarily the irst LO) system Raster display: A TV-like display in with preselectors. of a spectrum analyzer. The modulat- which the image is formed by scanning ing signal is noise, in the LO circuit itself the electron beam rapidly across and Negative peak: The display detection or in the LO stabilizing circuit, and the slowly down the display face and gating mode in which each displayed point sidebands comprise a noise spectrum. the beam on as appropriate. The scan- indicates the minimum value of the video The mixing process transfers any LO ning rates are fast enough to produce signal for that part of the frequency span instability to the mixing products, so the a licker-free display. Also see Vector or time interval represented by the point. noise sidebands appear on any spectral display and Sweep time. component displayed on the analyzer far Noise loor extension: Developed by enough above the broadband noise loor. Real-time spectrum analyzer: A method Keysight Technologies, Inc., a modeling Because the sidebands are noise, their of signal analysis in which all signal algorithm of the noise power in a signal level relative to a spectral component samples are processed for some sort of analyzer which can be subtracted from is a function of resolution bandwidth. measurement result or triggering opera- the measurement results to reduce the Noise sidebands are typically speciied tion. There are no gaps between time effective noise level. in terms of dBc/Hz (amplitude in a 1-Hz acquisitions while nonreal-time opera- bandwidth relative to the carrier) at a tions leave gaps. Noise igure: The ratio, usually ex- given offset from the carrier, the carrier pressed in dB, of the signal-to-noise being a spectral component viewed on Reference level: The calibrated vertical ratio at the input of a device (mixer, the display. position on the display used as a refer- ampliier) to the signal-to-noise ratio at ence for amplitude measurements. The the output of the device. Phase noise: See Noise sidebands. reference level position is normally the top line of the graticule. Positive peak: The display detection mode in which each displayed point in- Relative amplitude accuracy: The un- dicates the maximum value of the video certainty of an amplitude measurement signal for that part of the frequency span in which the amplitude of one signal is or time interval represented by the point. compared to the amplitude of another regardless of the absolute amplitude of either. Distortion measurements are relative measurements. Contributors to uncertainty include frequency response and display idelity and changes of input attenuation, IF gain, scale factor and resolution bandwidth. 86 Residual FM: The inherent short-term Sensitivity: The level of the smallest Spurious responses: The improper frequency instability of an oscillator in sinusoid that can be observed on a responses that appear on a spectrum the absence of any other modulation. In spectrum analyzer, usually under opti- analyzer display as a result of the input the case of a spectrum analyzer, we usu- mized conditions of minimum resolution signal. Internally generated distortion ally expand the deinition to include the bandwidth, 0-dB RF input attenuation products are spurious responses, as are case in which the LO is swept. Residual and minimum video bandwidth. Keysight image and multiple responses. FM is usually speciied in peak-to-peak deines sensitivity as the displayed aver- values because they are most easily age noise level. A sinusoid at that level Sweep time: The time to tune the LO measured on the display, if visible at all. will appear to be about 2 dB above the across the selected span. Sweep time noise. does not include the dead time between Residual responses: Discrete responses the completion of one sweep and the seen on a spectrum analyzer display Shape factor: See Bandwidth selectivity. start of the next. In zero span, the spec- with no input signal present. trum analyzer’s LO is ixed, so the hori- Signal analyzer: A spectrum analyzer zontal axis of the display is calibrated in Resolution: See Frequency resolution. that also uses digital signal processing time only. In nonzero spans, the horizon- to perform other more complex mea- tal axis is calibrated in both frequency Resolution bandwidth: The width of surements such as vector signal analysis. and time, and sweep time is usually a the resolution bandwidth (IF) ilter of a function of frequency span, resolution spectrum analyzer at some level below Signal identiication: A routine, either bandwidth and video bandwidth. the minimum insertion loss point (maxi- manual or automatic, that indicates mum delection point on the display). For whether or not a particular response on Time gating: A method of controlling the Keysight analyzers, the 3-dB bandwidth the spectrum analyzer’s display is from frequency sweep of the spectrum ana- is speciied; for some others, it is the the mixing mode for which the display is lyzer based on the characteristics of the 6-dB bandwidth. calibrated. If automatic, the routine may signal being measured. It is often useful change the analyzer’s tuning to show the when analyzing pulsed or burst modu- Rosenfell: The display detection mode signal on the correct mixing mode, or it lated signals’ time-multiplexed signals in which the value displayed at each may tell us the signal’s frequency and and intermittent signals. point is based upon whether or not the give us the option of ignoring the signal video signal both rose and fell during the or having the analyzer tune itself prop- TDMA: Time division multiple access is a frequency or time interval represented erly for the signal. Generally not needed digital communication method in which by the point. If the video signal only on preselected analyzers. multiple communication streams are in- rose or only fell, the maximum value is terleaved in time, enabling them to share displayed. If the video signal did both Span accuracy: The uncertainty of the a common frequency channel. rise and fall, then the maximum value indicated frequency separation of any during the interval is displayed by odd- two signals on the display. Units: Dimensions of the measured numbered points, the minimum value, quantities. Units usually refer to am- by even-numbered points. To prevent Spectral purity: See Noise sidebands. plitude quantities because they can be the loss of a signal that occurs only in an changed. In modern spectrum analyzers, even-numbered interval, the maximum Spectral component: One of the sine available units are dBm (dB relative to 1 value during this interval is preserved, waves comprising a spectrum. milliwatt dissipated in the nominal input and in the next (odd-numbered) interval, impedance of the analyzer), dBmV (dB the displayed value is the greater of Spectrum: An array of sine waves of relative to 1 millivolt), dBuV (dB rela- either the value carried over or the maxi- differing frequencies and amplitudes and tive to 1 microvolt), volts, and in some mum that occurs in the current interval. properly related with respect to phase analyzers, watts. In Keysight analyzers, that, taken as a whole, constitute a we can specify any units in both log and Sample: The display detection mode in particular time-domain signal. linear displays. which the value displayed at each point is the instantaneous value of the video Spectrum analyzer: A device that ef- Vector diagram: A display type com- signal at the end of the frequency span fectively performs a Fourier transform monly used when analyzing digi- or time interval represented by the point. and displays the individual spectral tally modulated signals. It is similar to components (sine waves) that constitute a constellation display, except that in Scale factor: The per-division calibration a time-domain signal. Phase may or may addition to the detected symbol points, of the vertical axis of the display. not be preserved, depending upon the the instantaneous power levels during analyzer type and design. state transitions are also plotted on an IQ graph.

87 Vector display: A display type used in Video average: A digital averaging of Video ilter: A post-detection, low-pass earlier spectrum analyzer designs, in a spectrum analyzer’s trace informa- ilter that determines the bandwidth of which the electron beam was directed so tion. The averaging is done at each the video ampliier. Used to average or that the image (trace, graticule, annota- point of the display independently and is smooth a trace. See Video bandwidth. tion) was written directly on the CRT completed over the number of sweeps face, not created from a series of dots selected by the user. The averaging al- Zero span: That case in which a as in the raster displays commonly used gorithm applies a weighting factor (1/n, spectrum analyzer’s LO remains ixed today. where n is the number of the current at a given frequency so the analyzer sweep) to the amplitude value of a given becomes a ixed-tuned receiver. The Video: In a spectrum analyzer, a term point on the current sweep, applies an- bandwidth of the receiver is that of describing the output of the envelope other weighting factor [(n – 1)/n] to the the resolution (IF) bandwidth. Signal detector. The frequency range extends previously stored average, and combines amplitude variations are displayed as from 0 Hz to a frequency typically well the two for a current average. After a function of time. To avoid any loss beyond the widest resolution bandwidth the designated number of sweeps are of signal information, the resolution available in the analyzer. However, the completed, the weighting factors remain bandwidth must be as wide as the signal ultimate bandwidth of the video chain is constant, and the display becomes a bandwidth. To avoid any smoothing, the determined by the setting of the video running average. video bandwidth must be set wider than ilter. the resolution bandwidth. Video bandwidth: The cutoff frequency Video ampliier: A post-detection, DC- (3-dB point) of an adjustable low-pass coupled ampliier that drives the vertical ilter in the video circuit. When the video delection plates of the CRT. See Video bandwidth is equal to or less than the bandwidth and Video ilter. resolution bandwidth, the video circuit cannot fully respond to the more rapid luctuations of the output of the envelope detector. The result is a smoothing of the trace, i.e., a reduction in the peak-to-peak excursion of broadband signals such as noise and pulsed RF when viewed in the broadband mode. The degree of averaging or smoothing is a function of the ratio of the video bandwidth to the resolution bandwidth.

88 89 | Keysight | Spectrum Analysis Basics - Application Note

myKeysight For more information on Keysight Technologies’ products, applications or www.keysight.com/find/mykeysight services, please contact your local Keysight A personalized view into the information most relevant to you. office. The complete list is available at: www.lxistandard.org www.keysight.com/find/contactus LAN eXtensions for Instruments puts the power of Ethernet and the Web inside your test systems. Keysight is a founding member of the LXI Americas consortium. Canada (877) 894 4414 Brazil 55 11 3351 7010 Three-Year Warranty Mexico 001 800 254 2440 www.keysight.com/find/ThreeYearWarranty United States (800) 829 4444 Keysight’s commitment to superior product quality and lower total cost of ownership. The only test and measurement company with three-year Asia Paciic warranty standard on all instruments, worldwide. Australia 1 800 629 485 China 800 810 0189 www.keysight.com/go/quality Hong Kong 800 938 693 Keysight Technologies, Inc. India 1 800 112 929 DEKRA Certified ISO 9001:2008 Japan 0120 (421) 345 Quality Management System Korea 080 769 0800 Malaysia 1 800 888 848 Keysight Channel Partners Singapore 1 800 375 8100 www.keysight.com/find/channelpartners Taiwan 0800 047 866 Get the best of both worlds: Keysight’s measurement expertise and product Other AP Countries (65) 6375 8100 breadth, combined with channel partner convenience. Europe & Middle East Austria 0800 001122 Belgium 0800 58580 Finland 0800 523252 France 0805 980333 Bluetooth and the Bluetooth logos are trademarks owned by Bluetooth SIG, Inc., U.S.A. Germany 0800 6270999 and licensed to Keysight Technologies, Inc. Ireland 1800 832700 cdma2000 is a US registered certification mark of the Telecommunications Industry Association. Israel 1 809 343051 Italy 800 599100 www.keysight.com/find/SA Luxembourg +32 800 58580 Netherlands 0800 0233200 Russia 8800 5009286 Spain 800 000154 Sweden 0200 882255 Switzerland 0800 805353 Opt. 1 (DE) Opt. 2 (FR) Opt. 3 (IT) United Kingdom 0800 0260637

For other unlisted countries: www.keysight.com/find/contactus (BP-09-23-14)

This information is subject to change without notice. © Keysight Technologies, 1971 - 2014 Published in USA, August 4, 2014 5952-0292 www.keysight.com