Receiver Dynamic Range: Part 2

The Communications Edgeª Tech-note Author: Robert E. Watson

Part 1 of this article reviews receiver mea- the receiver can process acceptably. In sim- NF is the receiver noise figure in dB surements which, taken as a group, describe plest terms, it is the difference in dB This dynamic range definition has the receiver dynamic range. Part 2 introduces between the inband 1-dB compression point advantage of being relatively easy to measure comprehensive measurements that attempt and the minimum-receivable signal level. without ambiguity but, unfortunately, it to characterize a receiver’s dynamic range as The compression point is obvious enough; assumes that the receiver has only a single a single number. however, the minimum-receivable signal signal at its input and that the signal is must be identified. desired. For deep-space receivers, this may be COMPREHENSIVE MEASURE- a reasonable assumption, but the terrestrial MENTS There are a number of candidates for minimum-receivable signal level, including: sphere is not usually so benign. For specifi- The following receiver measurements and “minimum-discernable signal” (MDS), tan- cation of general-purpose receivers, some specifications attempt to define overall gential sensitivity, 10-dB SNR, and receiver interfering signals must be assumed, and this receiver dynamic range as a single number noise floor. Both MDS and tangential sensi- is what the other definitions of receiver which can be used both to predict overall tivity are based on subjective judgments of dynamic range do. receiver performance and as a figure of merit signal strength, which differ significantly to compare competing receivers. They DESENSITIZATION DYNAMIC from author to author. They are mentioned include: 1-dB compression dynamic range, RANGE here because of their historical significance, desensitization dynamic range, spur-free but the uncertainty limits their value as a Desensitization dynamic range (DDR) mea- dynamic range, and NPR (noise-power part of receiver dynamic-range specifications. sures the receiver degradation effects due to a ratio) figure of merit (NPRFOM). In gener- A more repeatable measurement is 10-dB single, dominant, out-of-band interferer. In al, they are based on the primary measure- SNR; but this, too, has disadvantages many “real world” signal environments, a ments of receiver performance, but the because of the variations of SNR due to type single, strong signal may be the major source NPRFOM test attempts to simulate the and percentage of modulation. The least of interference due to the effects of receiver actual signal environment in a way that phase noise and out-of-band signal compres- combines all of the receiver dynamic range ambiguous indicator of minimum receivable sion. In this test, a signal that produces an characteristics (see Table 1). This test is pro- signal is probably receiver noise floor. This output SNR of 10 dB is injected at the posed as a practical and realistic measure- can be defined in two ways: noise floor in a receiver input. An interfering sinusoid is ment of receiver dynamic range. 1-Hz bandwidth and total equivalent input noise power in the narrowest receiver band- added to the input at a particular frequency 1-DB COMPRESSION DYNAMIC width. The first is simply -174 dBm plus the offset from the tuned frequency and its mag- RANGE receiver noise figure in dB; while the second nitude is increased until the output SNR has the additional factor of 10 times the log degrades 1 dB. The DDR is then the power The receiver 1-dB compression dynamic of the receiver bandwidth. For most purpos- ratio (in dB) of the undesired signal power range defines the range of signal levels that es, the inclusion of the receiver bandwidth (in dBm) to the receiver noise floor in dBm yields a better estimator of usable dynamic per Hertz. The DDR can be calculated using 100 kHz 10 MHz range. Using this definition, receiver dynam- the equation: 200 kHz 20 MHz ic range can be expressed as: DDR = Pi - NF + 174 500 kHz 50 MHz CDR = Pic + 174 dBm -10 log BW - NF where: 1 MHz* 100 MHz* where: DDR is the desensitization dynamic 2 MHz 200 MHz range in dB CDR is the compression dynamic range Pi is the interfering signal power in 5 MHz* 500 MHz* in dB dBm 1 GHz P is the 1-dB input compression ic NF is the receiver noise figure *Recommended minimum set. power in dBm BW is the narrowest receiver bandwidth DDR is a true measure of dynamic range Table 1. Recommended standard filter frequencies for NPRFOM measurements. in Hz because it includes both noise figure and

WJ Communications, Inc. ¥ 401 River Oaks Parkway ¥ San Jose, CA 95134-1918 ¥ Phone: 1-800-WJ1-4401 ¥ Fax: 408-577-6620 ¥ e-mail: [email protected] ¥ Web site: www.wj.com The Communications Edgeª Tech-note Author: Robert E. Watson

measurement of overload/ interfering signal effect seen in both the XX-R7000 and XXX- receiver is tuned to center the test signal in power. The use of receiver input attenuation 500 receivers. the IF passband and to produce an audio will improve large signal-handling capability, “beat note” of 1 kHz. The significance of DDR is somewhat but noise figure will be degraded commen- dependent on signal environment. If the If the receiver does not have a “predetection” surately. The DDR, however, is not affected interfering signals have significant phase demodulation mode like “cw” or ssb which by input attenuators. When it is desirable to noise of their own, it is only necessary for uses a bfo frequency conversion to audio, the determine the absolute signal power in dBm the receiver’s phase noise to be better than narrowest available IF output may be used required to cause desensitization for a partic- the interferer’s. Most radio transmitters have with a spectrum analyzer. In this case, the ular receiver configuration, the following significant amounts of phase-noise sideband signal is monitored for a 1-dB amplitude equation can be used: energy at modest offsets from the carrier fre- decrease due to compression, and the noise Pi = DDR + NF - 174 quency. This is especially true for variable floor is monitored for a 1-dB increase due to frequency oscillator (vfo) and most frequen- phase-noise reciprocal mixing. Note that the noise figure must include the cy-synthesized frequency sources. A notable effects of input attenuation as the receiver is The interfering signal generator must have exception, which may have very low levels of intended to be used. phase noise much better than that of the small offset phase noise, is crystal oscillator receiver under test. The tunable bandpass fil- DDR is strongly affected by the frequency signal sources. At large frequency offsets, ter will help eliminate any residual generator offset of the interfering signal. At small fre- many transmitters will have low phase noise phase noise at large frequency offsets. The quency offsets, the DDR is dominated by because of the filtering properties of tuned audio lowpass filter is not required, but the effects of receiver phase noise reciprocal power output stages and narrow antenna serves to minimize the effects of variations in mixing. In this region, the DDR is approxi- bandwidths. For this reason, more attention audio response from receiver-to-receiver. mately 6 dB less than the magnitude of the should be given to obtaining a good DDR at single-sided phase noise spectrum in dB per large frequency offsets. SPUR-FREE DYNAMIC RANGE Hertz below the “carrier” (dBc). For example, if a receiver’s phase noise at 100 kHz A test setup for DDR measurement is shown Spur-free dynamic range (SFDR), as general- from the tuned (carrier) frequency is -130 in Figure 2. The receiver is tuned to the test ly used, attempts to define receiver dynamic dBc, the DDR at 100 kHz offset will be frequency and set for maximum gain in the range in terms of two undesired interferers about 124 dB. In general, receiver phase narrowest available bandwidth with a bfo and the receiver noise floor. As with the 1- noise improves with frequency offset so that detection mode. In some receivers, it will be dB compression dynamic range, it is based in some receivers, interfering signals well necessary to use the ssb mode to activate the on a mathematical manipulation of the pri- removed from the tuned frequency, will bfo and to achieve narrow bandwidth. The mary measurements of receiver range. In this begin to cause signal compression before the effects of phase noise reciprocal mixing are 190 observed. In this case, the DDR will be 180 worse than 6 dB less than the magnitude of WJ-8615 W/PRE 170 the phase-noise suppression at these frequencies. Because of these frequency effects, it is 160 necessary to specify the DDR at several dif- 150 ferent offset frequencies. The best presenta- 140 GERMAN XXX-500 tion of this data would be in the form of a 130 JAPANESE XX-R7000 graph, as shown in Figure 1. 120

This figure compares the DDR of three pop- 110 TEST FREQUENCY 100 MHz ular vhf/uhf receivers. At small frequency 100 offsets, the DDR is typically dominated by RANGE (dB) DYNAMIC DESENSITIZATION 90 receiver phase noise. At larger frequency off- 10 20 50 100 200 500 1 2 5 10 20 50 100 kHz MHz MHz sets, in receivers with modest signal input fil- INTERFERING FREQUENCY OFFSET tering (rf preselection) 1 dB compression due to signal overload may occur. This is the Figure 1. Desensitization dynamic range (DDR) as a function of frequency offset of the interfering signal.

possible candidate is the NPR figure-of- INTERFERING SIGNAL ± FREQ = FTEST FOFFSET merit (NPRFOM).

50 Ω NPR FIGURE OF MERIT TUNABLE 2 kHz DISTORTION FILTER RF LOWPASS ANALYZER -3 dB ADUDIO The NPR figure-of-merit dynamic range IN OUT HP8640 RECEIVER F = 1 kHz measurement attempts to overcome some of RF GENERATOR FREQ = FTEST HP334A the shortcomings of earlier measurements by GAIN = MAX TEST SIGNAL BFO ON better simulating the signal environment 50 Ω FREQ = FTEST with high-power white noise. NPR is an abbreviation for Noise Power HP8640 Ratio, a term familiar to those involved in RF GENERATOR FDM telephone work. NPR testing simu- lates the signal environment by a broad band Figure 2. Desensitization dynamic-range test configuration. of “white noise with the total noise power case, the spur-free dynamic range is the dif- SFDR specification overlooks several impor- adjusted to equal the total signal power that ference in dB between the receiver noise tant factors which influence dynamic range. can be expected at the receiver input. The floor and the level of each of two equal- First, it attempts to model interference by noise is removed at the receiver’s tuned fre- amplitude, out-of-band interfering tones using just two interfering signals. This over- quency by a notch filter. Due to receiver dis- that produce an in-band spurious product comes some of the objections to single-tone tortion, this notch tends to fill with inter- equal in power to the noise floor. Generally, testing, but the real signal environment is modulation products. The apparent notch the receiver third-order intercept point is usually populated by a multitude of signals. depth, as seen by the receiver, is the noise used to predict the spurious product, but Second, it does not reveal the effects of reci- power ratio; that is, the ratio of out-of-notch often the second-order distortion dominates. procal mixing or compression like the desen- noise to in-notch noise measured in dB (see In any case, the SFDR can easily be sitization dynamic-range test. Third, it does Figure 3). This test has been used for many expressed as: not effectively test the effects of receiver years in FDM telephone measurements and is specified by the CCIR and CCITT. SFDR3 = 2/3 (IIP3 + 174 - NF - 10 log BW) input filtering (preselection). Finally, SFDR, or as it is ordinarily specified, considers only NPRFOM is defined as the sum of the SFDR2 = 1/2 (IIP2 + 174 - NF - 10 log BW) the third-order distortion. In fact, for many input noise spectral density which produces receivers, especially those with modest input a receiver NPR of 40 dB plus 174 dB minus where: filters, the second-order products may domi- the receiver noise figure. For this measure- SFDR3 is the third-order spur-free nate. For example, for a receiver with a ment, noise figure is measured at the manual dynamic range in dB bandwidth of 100 Hz, a noise figure of 10 gain setting, which produces nominal receiv- IIP3 is the receiver third-order input dB, a third-order intercept of +20 dBm, and er output when tuned to the notch frequen- intercept point in dBm a second-order intercept of +50 dBm, the cy. NPRFOM can be expressed as: NF is the receiver noise figure in dB second- and third-order SFDRs will be 97 NPRFOM = Pnpr + 174 - NF BW is the narrowest receiver band- dB and 109.3 dB, respectively. Surely, the width in Hz lesser of the two values is more valid; howev- where: SFDR2 is the second-order spur-free er, it is often not specified. NPRFOM is the Noise Power Ratio Figure dynamic range in dB Because spur-free dynamic range is derived of Merit in dB IIP2 is the receiver second-order from the primary measurements, it would P is the input power spectral densi- input intercept point in dBm npr seem to provide no new information. ty in dBm/Hz for the white noise Spur-free dynamic range has become a very Instead, it merely adds to the confusion of that produces a 40 dB NPR popular specification because it seems to give receiver dynamic range specifications. NF is the receiver noise figure in dB a single number which can be used to com- However, there is a continuing need - or at measured at the receiver-gain set- pare the overall dynamic-range performance least a desire - for a truly comprehensive ting which produces the nomi- of competing receivers. Unfortunately, the measurement of receiver dynamic range. A nal receiver output in the nar-

rowest available predetection output The 40-dB value for NPR was chosen because it is somewhat inconvenient to obtain noise sources with notch depths significantly greater than -50 dB.

Noise figure is included in the NPRFOM to dB/Hz NPR give a relative indication of receiver dynamic FILLING OF NOTCH range. It is measured with receiver gain set DUE TO RECEIVER for a nominal output when tuned to the NON-LINEARITY noise notch because this eliminates the effects of receiver input attenuation. Adding f attenuation to the input of a receiver lineari- Figure 3. Noise power ratio (NPR). ty increases its power-handling capability, but adding attenuation also linearly increases nominal output, or until the gain is maxi- 5. Decrease the attenuation until the output the receiver noise figure. Consequently, there mum. (Nominal output is the output is at the nominal level. is no net change in receiver dynamic range produced by a strong signal in agc mode, due to input attenuation. The factor of 174 6. Tune the receiver to a frequency outside or the output specified by the manufac- dB normalizes the measurement to the theo- of the notch and increase the attenuation turer.) retical noise floor of -174 dBm/Hz. The by 40 dB. Adjust the receiver manual gain NPRFOM measurement produces a single 3. Reconnect the noise source and decrease for nominal output and return to Step 4. measurement of dynamic range which can the attenuation until the nominal output Steps 7 and 8 measure the noise level and be used to directly compare the effective per- level is indicated on the meter. receiver noise figure which are then used to formance of competing receivers. This com- Now that the receiver and test setup have calculate NPRFOM. parison is not affected by the nonideality of been preset, Steps 4, 5 and 6 are repeated third-order intercept extrapolations, or by until the noise level at the receiver input 7. Without changing the receiver settings, the vagaries of preselector specifications, but produces a 40-dB NPR. disconnect the signal from the receiver instead highlights the differences between input and measure the out-of-notch noise receivers that appear in operation, but not 4. Tune to the notch frequency and decrease spectral density at the attenuator output the attenuation by 40 dB. If the output is on a manufacturer’s data sheet. with a spectrum analyzer. Enter this value less than the nominal level, go to Step 5. into the NPRFOM equation. A basic NPRFOM test setup is shown in If the output is equal to the nominal Figure 4. The test is performed using the fol- value, go to Step 7. If the output is 8. Without changing the receiver settings, lowing step-by-step procedure (Steps 1,2 and greater than the nominal value, the measure the receiver noise figure. Enter 3 preset the receiver and the noise level): receiver has insufficient dynamic range to this value into the NPRFOM equation 1. Set the receiver for the minimum avail- achieve an NPR of 40 dB. and calculate NPRFOM. able bandwidth and a predetection mode (CW or SSB). Monitor the audio output level with the meter. If these predetection PREDETECTED OUTPUT modes are not available, monitor the nar- rowband IF output. Set the audio gain for mid-range and tune the receiver to a fre- NOISE RECEIVER quency outside of the notch. SOURCE 2. Set the attenuator for maximum and tem- NOTCH VARIABLE METER porarily disconnect the noise source. FILTER ATTENUATOR Increase the receiver RF/IF gain until the meter indicates -3 dB with respect to the Figure 4. NPRFOM test configuration.

The measured values for NPRFOM are width characteristic tends to produce The LC filter of Figure 6 is most useful for somewhat dependent on the test configura- NPRFOM values which do not vary greatly frequencies below 50 MHz. It is relatively tion. While the relative ranking of several with test frequency. Also, this type of filter easy to construct and requires components receivers is generally unaffected by the test can be realized with limited values of com- of only modest Q. The capacitor values are configuration, direct comparisons of ponent Q. Figure 5 gives recommended specified to produce constant percentage NPRFOM values are possible only when the bandwidth values for both LC filters and bandwidth of approximately 6%. The induc- test conditions are standardized. In particu- crystal filters. Table 1 lists recommended tor should have a Q greater than 50 and the lar, the characteristics of the notch filter and center frequencies. Where feasible, the nar- capacitors should be matched within 5%. noise source must be specified. rower crystal filters are preferred for the rea- The delay-line filter of Figure 7 is most use- The noise source must be spectrally “flat” or sons stated above. These filters are relatively ful for frequencies from 50 to 1000 MHz. “white.” A desirable limit of noise amplitude simple, but a full set is somewhat expensive. Its major advantage is repeatability, but a spectrum flatness is 1 dB. The band-of-noise Lower-cost notch filters are shown in Figures major disadvantage is the production of frequencies, in the most desirable case, 6 and 7. multiple notches (Figure 8). However, the would cover the range from less than one- third the receiver’s lowest tunable frequency to greater than twice the highest tunable frequency. For wide-tuning receivers, this range 7% 0.35% -1 -1 dB -1 -1 dB may be difficult to obtain. A more reason- 5% 0.25% -3 -3 -3 -3 able noise frequency range would be from one-third the test frequency to two times the 2% 0.1% -25 -25 -25 -25 test frequency. For some receiver frequencies, 0.5% 0.025% -50 -50 dB noise generator limitations make an even -50 -50 dB more restricted noise bandwidth necessary. In this case, a noise band covering the range LC FILTER CRYSTAL FILTER of ±25% of the tuned frequency is recommended. Figure 5. Recommended notch-filter bandwidths for NPRFOM measurements. The notch filter is the most critical component of the test configuration. The ideal L 16.6 QL 1 notch filter would have a -3 dB bandwidth C = π only slightly wider than the receiver final IF QL -16.6 200 f 1 bandwidth, and a -50 dB bandwidth equal L = to the final IF bandwidth. Such a narrow 2C 2C (2 πf)2 C R NULL notch would allow the noise to simulate a 3 ADJUST R = uniformly dense signal environment. This 2 QL would also maximize the test’s ability to dis- criminate between the performances of dif- Figure 6. Notch filter with 6%, -3 dB bandwidth for 50-ohm source and load. fering receivers. This is because a narrow notch maximizes the stress on the receiver’s input filter and first IF stages. Unfortunately, NULL ADJUST the component Q’s for such a filter would POWER Ω DIVIDER 50 make it difficult or impossible to realize. It is 27 Ω Ω more practical to use a set of standardized 62 filters similar to those specified by the CCI- TIT for FDM NPR measurements. These DELAY LINE filters typically have -3 dB bandwidths, LENGTH = 8.5 λ which are a constant percentage of their center frequencies. The constant percent band- Figure 7. Delay-line notch filter.

power loss due to the extra notches is only figure. The first stage of preselection is low pact vhf/uhf receiver, without the preselector about 2.5 dB, and this small amount can be loss and rather broad. Also the tube-technol- option, is overloaded by the broadband subtracted from the resulting NPRFOM ogy input amplifier does not have the signal- input noise because of both second- and measurement without causing a major error. handling capability of more modern high third-order intermodulation effects. Receiver The ninth notch in the series (counting from zero frequency), was selected for this dynamic-range designs. Receiver B is a com- C has a modest tracking preselector, but its test because it is a reasonable compromise considering notch -3 dB bandwidth, -50 dB bandwidth, and cable loss. Higher notch numbers have narrower -3 dB bandwidths, but the -50 dB bandwidth becomes very

narrow and the increased cable length causes H(f) excessive frequency loss “tilt.” For the ninth notch, the -3 dB bandwidth is approximately 6.06% of the center frequency, and the - f 50 dB bandwidth is approximately 0.024% 1 2 3 4 5 6 7 8 9 10 11 NULL NUMBER

(see Figure 9). The out-of-notch frequency fo/3 fo/2 fo for measuring noise spectral density and for tuning during the NPRFOM test is at plus Figure 8. Frequency response of delay-line filter. or minus 6% from the notch frequency. These frequencies correspond to the nearest filter transfer maxima. The effectiveness of NPRFOM in differenti- 6.06% ating the dynamic range performance can be -3 dB seen in Table 2. Several communications

H(f) quality vhf/uhf receivers were tested for -50 dB 0.024% NPRFOM using filters similar to those in

Figure 6. The differences in performance can f be readily explained by the differences in fo - 6%fo fo + 6% receiver design. Receiver A is a proven, older design which was optimized for low noise Figure 9. Detail of delay-line filter frequency response for ninth null.

Noise PWA dim/Hz Noise Figure at NPRFOM (dB) Receiver Configuration For 40 dB NPR Test Gain @39 MHz @111 MHz @39 MHz @111 MHz @39 MHz @111 MHz A Normal -118 -114 5 5.5 51 54.5 B No Preselector -99 -104 10 10 65 60 C Normal -95 -98.5 13 13 66 62.5 D Normal -88 -82 8.5 9 77.5 77 E Normal -83 -86 10 10 81 78 F External Suboctave Preselector -86 ÐÐ 7 ÐÐ 81 ÐÐ G Tracking Preselector -82 -82 10 10 82 81 H Internal Tracking Preselector -77 -79 10 10 87 85

Table 2. NPRFOM measurements for several communications receivers.

poor third-order performance allows only a al parameters. The primary receiver measure- November/December 1981. slight improvement in NPRFOM. Receiver ments of noise figure, second-order inter- 2. Dexter, Charles E., “Digitally Controlled D performs much better due to its internal cept, third-order intercept, 1-dB compres- VHF/UHF Receiver Design,” Tech- sion, phase noise and internal spurious sig- switched suboctave preselector and very Notes, Vol. 7 No. 3, June 1980. good third-order performance. The Receiver nals can provide information for determin- 3. McDowell, Rodney K., “High Dynamic E performs slightly better than D receiver ing receiver dynamic range. Secondary because it has a tracking preselector which is receiver measurements, such as sensitivity, Range Receiver Parameters,” Tech-Notes, slightly narrower than a switched suboctave cross modulation, intermodulation distor- Vol. 7 No. 2, March/April 1980. preselector. Receiver F, with an external sub- tion, and reciprocal mix, can provide supple- 4. Erst, Stephen J., Receiving Systems octave filter, has a NPRFOM similar to E mental information, but they should not be Design, Artech House, 1984. substituted for the primary measurements. because it has a significantly better noise fig- 5. Hsieh, Chi, “GaAs FET IMD Demands Comprehensive measurements of receiver ure. Receiver G has significantly improved Better Standard,” Microwaves, Vol. 21 dynamic range include noise-power ratio fig- dynamic range because of the relatively nar- No. 6, June 1982. row bandwidth of the tracking preselector. ure-of-merit and desensitization dynamic 6. Hirsh, Ronald B., “Knowing the Finally, receiver H, with the tracking prese- range. Together, these two measurements Meaning of Signal-To-Noise Ratio” lector option, performs best of all. This is and the noise figure give an excellent indica- Microwaves and RF, Vol. 23 No. 2, because the preselector eliminates most of tion of receiver dynamic range. February 1984 the broadband noise of the test signal. In REFERENCES addition, because the preselector is integral 7. Fisk, James R., ‘Receiver Noise Figure to the receiver, the entire signal path has The following list of books and articles rep- Sensitivity and Dynamic Range - What been optimized for maximum dynamic range. resents a sampling of the more readable liter- the Numbers Mean,” Ham Radio, ature relating to the subject of this paper. October 1975. CONCLUSION 1. Grebenkemper, C. John, “Local Oscillator 8. Schwartz, Mischa, Information Satisfactory assessment of receiver dynamic Phase Noise and Its Effect on Receiver Transmission, Modulation and Noise, range requires careful measurement of sever- Performance,” Tech-Notes, Vol. 8 No. 6, McGraw-Hill, 1980.

WJ Communications, Inc. ¥ 401 River Oaks Parkway ¥ San Jose, CA 95134-1918 ¥ Phone: 1-800-WJ1-4401 ¥ Fax: 408-577-6620 ¥ e-mail: [email protected] ¥ Web site: www.wj.com