SIGNi\L PURITY CONS IDERI\TIONS FOR FREQUEN<:Y Sn1 TilES I ZED llEl\DEND EQU I Pr1ENT

David L. Kel:na

General Instrument, Jerrold Division

ABSTRACT BASIC OVERVIEW OF THE PLL

As cabl~ television system band­ Before we evoluate the system for wirlths increase and frequency plans pro­ its noise performance, let us first re­ liferate, more manufacturers are turning view the b~sic operation of a chase­ to synth~sized frequency agile head~nd locked loop. A simple phase-locked loop channel converters. Uith this new ap­ consists of a voltage controlled oscil­ proach using phased lo~ked loops and dual lator, or VCO, a digital divider, a chase conversion come spurious signals and detector, a reference frequency sour~e, nois~ sources not encountered before in and an integrator or loop filter. Figure crystal controlled channel converters. 1 represents such a system.

Important characteristics of these These system components function ~s headend converters including phas~ noise, a s~rvo loop sue~ that when the·vco is spurious sign3ls generated by the compar­ phase-locked to the reference, the output ison frequ~ncy, and residual frequency frequency and phase of the digital divi­ a~d phase modulation, are evaluated for der is equal to the frequency and phase their subjective impact on the output of the referenc~. This makes the average signal to the cable. D3ta is presented output of the phase detector zero and, which shows the correlation between sub­ therefore, the output of the loop filter jective picture degradation and measured remains unchanged. Should a disturbance headend synthesizer noise contribution. cause the VCO oscillating at "N" times the reference to shift frequ~ncy or ~hase, the digital divider output would not be coherent with t~e reference. 7his makes the output of the phase detector INTRODUCTION nonzero causing the loop filter to change its averag~ DC output voltage, which In the past, designers of cable sys­ forces the VCO back to the proper fre­ tem headend equipment hardly concerned quency and phase (REF. 1). themselves with phase noise. This is be­ cause most systems relied on crystal NOISE SOURCES IN A PHASE-LOCKED LOOP oscillators wh0re the phase noise is not a major concern due to the inherent low Now let us take a look at the mech­ noise an~ stability of such circuitry. anisms that can oroduce noise within the The current trend is toward greater use phase-locked loop. Signals that are of frequency synthesis and phase-locked integer multiples of th~ reference fre­ loop controlled conversion processes. quency will inevitably be present at the This is due to the attractiveness to both output of the phase detector. These ref­ the manufacturer and use of frequency erence frequency components can cause programmability in the multichannel en­ spurious outputs by modulating the VCO. vironment ~nd is further driven by the Optimum phase detector characteristics decreasing cost of associated components. and the loop filter design can reduce these signals to an acceptable level. Noise is influenced by each section (REF. 2, 6, 8.) of the phase-lock8d loop system an~ ade­ quate performance is obtain~d only by Loop Filter Noise careful ctesign of all circuits. Roth product and system designer must deter­ Since the loon filter drives t~e VCO mine the phase noise performance level tuning line, any noise at the output of that is required for subjectively accept­ the loa~ filter produces phase noise on able signal quality in the intended ap­ the oscillator output. Furthermore, any plication. noise on the phase detector output is

39 C4APC OUTPUT PLL SYSTEM (SIMPLIFIED)

CHANNEL 768 MHZ FREQUENCY IF SIGNAL OUTPUT (MODULATED) (MODULATED)

OSCILLATOR POWER AMPLIFIER

------~

vco PRE SCALER r-...1-"" 800-1400 MHZ +32

LOOP FIL TEA AND AMPLIFIER

PROGRAMMABLEr-~-~ DIVIDER th REFERENCE FREQUENCY DIVIDER f~EFERENCE PROGRAMMING GENERATOR INPUT '------"""i U68 ....f---- INPUT 48 MHZ

------FIGURE 1 modified by the filter transfer function, VCO Noise added to the loop filter output, and modulated onto the VCO. The amount of One of the more significant cont~i­ oscillator phase noise is a function of butors to output noise is the VCO itself, the tuning sensitivity, Kv, and the although all of the phase-locke~ loop amount of noise reaching the VCO's tuning components add noise to the output s~ec­ port. trum. The noise analysis of oscillators is difficult because the active device is For example, let's look at a VCO operating in its nonlinear region. How­ with a sensitivity of +35 MHZ per volt ever, if we examine the oscillator as an and assume that the VCO output frequency amplifier that has as much gain as the is 1000 MHZ with 0 volts on the control feedback network has loss, we will be line. At this sensitivity, a positive 1 able to get a rEasonable aoproximation volt de average value on the control line of the phase noise performance. Using would give an output frequency of: the basic r=lationship for thermal nois0, we note that: kTB = -174 dBm/Hz (deci­ F out F nominal + (Vdc times Kv) bels relative to 1 milliwatt per hertz) is the noise floor of any amplifier in­ F out lCJo;, MHz + (1 volt times put. For a 1 Hz bandwdth, adding the 3 5 ~m z 1 vo 1 t l amplifier noise figure, and ajding tho gain gives the corresponding amplifier F out = 1035 MHZ output noise floor. This will be the oscillator noise floor far removed from If the 1 volt signal had been l volt the carrier. The phase noise performance RMS random noise then the output would close into the carrier will depend upon have been 1000 MHz plus and minus 35 MHz whether it is a bipolar or field effect RMS of residual phas~ and frequency noise transistor. Rather than give a rigorous modulation. (REF. 8) mathematical description here, let us

40 take a look at some measureo ohase noise Divider Noise characteristics. Figure 2 sh;ws that there are 3 basic areas of phase noise: Increasing the output frequency of the VCO inherently leads to a larger fre­ 1) The close in portion, where the quency divider which, of course, ~eans oscillator phase noise is propor­ higher divider noise. Generally sp~ak­ tional to the transistor's low ing, the output phase noise of a digital frequency nois~ characteristic. divider is equal to the input phase noise divided by the circuit divisor plus the 2) A region that is a little farther inherent divider noise. This manifests removed from the carrier where the itself as a ~inimu~ attainable noise noise is related to the Q of the floor. The practical noise limits are frequency determining circuit. -l7G dBc/KHz (decibels relative to the carrier per kilohertz) for TTL types and -155 dBc/KHz for ECL dividers. CMOS OPEN LOOP VCO PHASE NOISE dividers, working up to a frequency of about 10 MHZ are similar to TTL devices, except that they have slightly higher noise between the carrier frequency and about 10 Hz away from the carrier. (REF. 3) Unfortun3tely, digital circuits have another side effect, crosstalk. This causes the input signal to appear at the output as both feedthru and stray pickup. Most synthesizer systems are limited by 0 other factors and this effect adds less 0 dB/OCTAVE than l dB to the noise level. (REF. 3)

Phase Detector Noise Response -+------~------~------~~F Fo In most synthesizer applications, FOR REGION: the phase detector is chosen for reasons other than noise performance, such as CD ( ;~ r( F:, ) acquisition and hold-in range. This is because the synthesizer phase detector 0 ( :~ r (~~:\~T2 )_ does not normally operate near its noise threshold, For this reason, phase ~e­ Fosc kT tectors are evaluated for their noise re­ 0 2 Ps sponse and not as a noise source them­ selves. (REF. 9, ll, 12, 13)

Fo= AVERAGE OSCILLATOR FREQUENCY Q =LOADED QUALITY FACTOR OF OSCILLATOR'S RESONATOR Reference Signal Noise a= A CONSTANT WHICH IS PROPORTIONAL TO THE FLICKER NOISE AMPLITUDE Finally, we should consider the Fm = F- Fo phaselock reference signal and realize Fosc"' EFFECTIVE NOISE FIGURE OF THE AMPLIFIER USED IN THE that the output signal c~n be no more OSCILLATOR CIRCUIT stable than the r2ference frequency k = BOL TZMAN'S CONSTANT T = TEMPERATURE IN DEGREES KELVIN stability times the digital divider Ps"' RF OUTPUT POWER OF THE OSCILLATOR ratio. Generally, we can dismiss this as a problem, by recognizing that the FIGURE 2 reference frequency in a synthesizer system is fixed. This allows us to use a high stability, low phase nois2 circuit 3) The far removed noise floor of the such as a crystal controlled source. oscillator. SYSTEM H1PACT The second item is significant be­ cause a VCO that has a wide tuning range In order to discuss how we expect necessarily has a lower "Q" and there­ the perceived signal to be degraded in fore, has more phase noise. the presence of phase noise, we must consider how the receiver circuits re­ Also, ell three r~gions are relateo spond to this type of noise. First, we to frequency. This means that the higher will determine what common effects re­ the frequency of o~eration in an oscilla­ ceivers will experience. Second, we will tor, the higher the phase noise, all take a look at the video and sound demod­ othP.r parameters being the same. (RFF. ulators. It is n?.cessary to ev3luate th2 4, 5, 7, 10) video and sound signals separately, be-

41 ca~se the difference in the modulation We can derive an equation for the type for the two carriers will cause signal to phase noise ratio of a carrier their respective demodulators to respond by defining Fl and F2 as the lowest and differently to phase noise. Finally, we highest frequency offsets of interest will look at the effect of a signal with with respect to the carrier. From this phase noise on broadband system distor­ we will make the assumption that between tion. frequencies Fl and F2 the slope of the noise power versus frequency is a Nyquist Filtering straight line. Next, we will let Fl = 0 and F2 = Fe, the phase-locked loop cutoff All television receivers have a Ny­ frequency. In looking at Figure 3, we quist IF filter in front of their video notice that the synthesizer has a white detector circuits for proper demodulation noise characteristic below the loop of the vestigial side band video signal. cutoff frequency of 500 Hz, i.e., the The slope of this filter will translate slope is 0. This is typical of a closed the residual FM noise into an amplitude loop phase-locked system. (REF. 14) Our modulation. The vioeo demodulator cir­ derivation leads us to the following cuit will then detect this AM noise just equation from which we calculate the as any other portion of the video signal. phase noise floor. Assuming a carrier to noise ratio objec­ tive of 60 dB and using the ideal Nyquist filter slope, we can perform a simple 3 (Frms) (Frms) calculation of the allowable residual FM NP 10 LOG ------at the input of the Nyquist filter. The 2 (Fe) (Fe) (Fe) ideal Nyquist filter slope starts at 45.00 MHz with 0 dB attenuation and has 3 (750) (75CJ) infinite attenuation at 46.5 MHz. This represents 100% amplitude modulation as 2 (530) (50G) (500) the result of an FM input signal with 750 KHz deviation. A noise level of 60 dBc - 21.4 dBc/Hz on the detector output would then corres­ pond to AM modulation of 0.1%. To find the residual FM input necessary to pro­ Although this noise floor seems to duce this amount of AM, use the following be very high, we must remember that this equation: is phase noise ~nd will not be directly demodulated by the video detector. RFM = (Residual FM) = (750,300 Hz) (% AM on filter output)

100 VECTOR DIAGRAM OF CARRIER, VIDEO 75G Hz AND PHASE NOISE Thus, a synthesizer design objective might be to obtain 750 Hz RMS of FM noise or better. PHASE NOISE VECTORS

CLOSED LOOP OUTPUT PHASE NOISE C= 100 Vz 87.5 y T CARRIER VECTOR VIDEO VECTOR 21.4 dBc/Hz FIGURE 4A Envelope Demodulation

We would expect an envelope demoju­ lator to be the most sensitive type of detector to phase noise because it will detect the video amplitude n~ise caused by the phase noise spectrum plus the Ny­ quist slope converted AM noise. Figure Fo Fo+Fc 4a shows the complete VPctor diagram of Fo= VCO OUTPUT FREQUENCY the carrier, video, and phase noise. We Fe= LOOP FILTER CUTTOFF FREQUENCY will let the carrier amplitude be 100 Frms• ROOT MEAN SQUARE RESIDUAL Fm units and the video equal to 87.5 units. FIGURE 3

42 ~ext, we will take the phase noise level because it will track out the residual of -21.4 dBc/Hz and convert it to a FM. In looking at the vector diagram of linear form, remembering to multiply by Figure Sb, we see that the synchronous the signal level. This is done in the detector only responds to the modulation following equation. vector which is in phase with the car­ rier: however, the detector follows the ( ( Np/ 20 angle produced by the phase noise compo­

Rpn (1!3 ( 8 7 0 5 ) 7.45 nent as well. The phase noise directly units contributes nothing to the amplitude com­ ponent; therefore, none of the phase Figure 4c shows, by application of noise is directly detected. the Pythagorian theorem, that the de­ tected level of the noise is: Unlike the envelope detector, the synchronous Jetector has a threshold Rl - R2 which determines the level of noise No 20 LQG - 48.7 dBc induced phase - frequency deviation it R2 can track. This threshold is determined by parameters within its own phase or where Rl square root ( (Rpn) (Rpn) + (V) frequency locked loop. At noise levels (V)) and R2 = V. below this threshold, we would expect the subjective effect of FM noise to be similar to that of the envelope detector, PHASE NOISE RESULTANT but to a lesser degree. This is because the detector doesn't actually hold zero phase to the carrier as the FM noise QUADRATURE COMPONENTS ADD approaches the hold-in threshold, but instead has a small offset. This offset allows the detection of a small amount of the phase noise, which increases toward the threshold. Exactly at the hold-in threshold, the synchronous detector will IN PHASE RESULT ANT EQUAL ZERO jump in and out of lock following the peak noise induced F~. The result will FIGURE 4B be cycle slipping, an effect which should be familiar to anyone who has operated a This shows that the main contributor satellite receiver under poor signal to to picture degradation in an envelope noise conditions. This appears as ran­ detector is the directly detected compo­ dom tearing of the picture horizontally, nent of the phase noise, and not the Ny­ associated with random loss of vertical quist FM to ~M noise at - 50 dBc. ~ore­ sync. over, noting that the phase noise is predominently low frequency due to the Serious degradation should occur at phased locked loop in the headend synthe­ a lower input ?hase noise level than that sizer, the detected noise will be of low of the envelope detector because of the frequency. This suggests that the sub­ dependence of the synchronous demodula­ jective video degradation for the enve­ tion loop on the ability to follow the FM lope detector will be in the form of noise as a result of the synchronous horizontally streaked noise. demodulator threshold.

PHASE NOISE INDUCED SYNCRONOUS DEMODULATION

AMPLITUDE ERROR PHASE NOISE VECTORS

R 1 =87.82

~ Rpn=7.45

~1 R2 ... CARRIER VECTOR FIGURE 4C

Synchronous Detection PHASE In theory, we would exoect the syn­ NOISE chronous detector to be better in noise VECTORS performance than the envelope detector FIGURE SA

43 this coherency is modified by phase VIDEO VECTOR RESULTANT DEMODULATED PHASE NOISE noise, it was expected that the subjec­ WITH PHASE NOISE ______, tive intermodulation improvement would be degraded. As with the other phase noise effects, this should appear as horizon­ tally streaked low frequency noise. IN PHASE VIDEO Summary of Phase Noise Effects FIGURE SB When signals with phase noise are introduced to a cable system, the re­ sound Degradation ceivers exhibit perceptible video and sound degradation depending on the types Considering that phase noise pro­ of demodulation circuits employed in any duces a noise related FM deviation and particular receiver. The vid2o and sound that the sound carrier is FM modulated, signals were presented separately because we expect that the perceptibility of the of the difference in modulation proces­ noise will depend on its spectrum. The ses. Both envelope and synchronous video noise spectrum of the sound carrier will detectors, along with split and intercar­ be the same as that of the video carrier. rier sound detection schemes were consi­ This makes the level of perceptibility dered with their perceptual appearances. mostly dependent upon whether a split or Also, the problems of Nyquist residual F~ an intercarrier sound detection scheme is to AM conversion and system triple beat used. reduction were covered. Now let's see the results of our testing. Split sound TEST RESULTS In a split sound system, the sound carrier is detected independently from The video and audio tests were per­ the video, therefore demodulating all the formed using a phase-locked modulator to noise on the sound carrier. The noise produce controlled phase noise condi­ spectrum at the output of the detector, tions. A synthesized signal generator will have a parabolic spectral shape. provided the phaselock ref2rence signal. After deemphasis, th~ noise spectrum is The generator was frequency modulated by relatively constant at the low frequen­ a continuously variable white noise sig­ cies, falling off at the mid audio fre­ nal. The noise level control was quencies at approximately 20 dB per calibrated for root mean square FM noise decade. This noise response is similar deviation by using a modulation analyzer. to pink noise which sounds like the The subjective test results are the aver­ rushing noise made by a running shower. age of 10 expert and non-expert viewers randomly selected from laboratory person­ Intercarrier Sound n,21.

In the intercarrier sound demodu­ System testing for intermodulation lator, the phase noise spectrum is distortion and triple beat performance cancelled by mixing the sound carrier was done using a "typical" cascade of 17 with the video carrier, both of which trunk amplifiers, followed by one line have the same phase noise modulation. extender amplifier. The cascade was Indeed, the singular advantage of the loaded with 52 HRC phase-locked channels. intercarrier process is the cancella­ Phase noise was added to the carriers of tion of frequency and phase modulations the channel selected for viewing. This common to picture and sound carriers. was done by injecting the calibrated The degree of cancellation will be re­ noise source directly into the phase­ duced by processes that independently locked loop of the associated IF to chan­ modify the phase or frequency modulations nel converter. The program material on of each carrier. the channels not being viewed included both live video and a standard color bar Reduction of Intermodulation Benefits pattern. of Coherent Carriers Video Test Results It was anticipated that the presence of phase noise would reduce the benefi­ As oredicted, the two ~nvelope de­ cial effects of HRC operation on the tectors tested were the least sensitive visibility of system intermodulation. to phase noise. A residual FM of 1565 Hz The improvement factor results from RMS was the average level of perceptibi­ "hiding" the carrier intermodulation pro­ lity for both envelope detectors tested. ducts behind a given picture carrier by This corresponds to a demodulated signal causing the distortion to be coherent to noise ratio of 36.7 dB, as calculated with the carrier. To the extent that by the formulas presented.

44 A precision demodulator was used to quality. The resulting FM noise devia­ perform thP- synchronous detector testing. tion_;.;as measured at 876 Hz RMS. The It was operated in thrP.e different Qhase­ cascade signal l2vel was then increased locked loop sampling modes and with two until intermodulation ~istortion was just loop ba~dwidths. The results presented perceptible. The effect of the phase below are the average for these different noise was seen to increase as u result of operating connitions. The synchronous increasing the cascade signal level; how­ detector displayed the same subjective ever, no new types of degradation appear­ noise characteristic as the envelope ed. In order to reduce the picture de­ detector, but perceptibility occured at a gradation du~ to the phase noise back to lower residual FM level. Also, as sus­ the previous perceptibility, it was nec­ pected, the synchronous detector lost essary to reduce the residual FM to 349 lock very quickly after the appearance of Hz RMS. This brings us to the conclusion noise in the picture due to the failure that additional degradation does occur of its phase-locked loop to remain when phase noise interferes with the co­ stable. The results of the video tests herent carrier intermodulation process. are tabulated below. Furthermore, this degradation is on the order of 60% of the tolerable phase noise Residual FM for Definitely Perceptible with no intermodulation distortion. Noise env. det. l env. det. 2 synchronous SUMMARY AND CONCLUSIONS

134 7 Hz RMS 1783 Hz RMS 306 Hz RMS The phenomenom of phase noise in synthesized headend equipment has been As we can see from the table, the discussed and shown to be a problem if envelope detectors can withstand the not properly attended to early in the greatest amount of phase noise before the design stage of such equipment. ' brief picture is perceptually degraded. overview of the phase-locked loop and the major contributors to phase noise within Sound Test Results the loop have been presented.

An audio output signal to noise Perceptibility tests were performed ratio criterion of 50 dB was used as the for a virteo color bar pattern and tests basis for the phase noise analysis. The show that the envelope detector was most phase ~oise level required to produce insensitive to phase noise for video. this signal to noise ratio was found to These tests illustrate that a residual be 126 Hz RMS and 1530 Hz RMS for split noise relateJ FM of 750 Hz RMS should be and intercarrier sound detection, respec­ subjectively acceptable when receiver tively. This was done by modulating a l envelope detertion is used. KHz tone onto a channel with phase noise added as previously done for the video Sound testing was ~erformed for a 50 test. The aural carrier deviation was dB signal to noise ratio and showed, as set to 25 KHz peak, demodulated using a expected, that the split sound detector precision detector, and a reference set was inferior to the intercarrier detec­ on an audio voltmeter. The l KHz tone tor. Furthermore, the type of noise was then removed and the signal to noise heard was list~ned to 0nd des~ribed. ratio measured. The noise modulation was increased until a 50 dB ratio was ach­ System cascade testing to determine ieved. The residual FM noise level was the impact on intermodulation performance then recorded. in an HRC situation was also ~erformed. Although not specifically proven, our ex­ Listening tests confirmed that the pectations of a reduction in coherent subjective quality is one of random noise carrier system advantage were partially predominated by low frequencies. The fulfilled by the a~parent increase in the sound has a "rain falling on a drum" level of phase noise observed; the ab­ characteristic. sence of new distortion products was not expected. System Testing for Intermodulation Per­ formance Reduction Since the data presented here repre­ sents only a limitPd number of tests, an~ After setting the amplifier cascade was obtained from a limited nu~ber of signal level at a point where the inter­ viewers, the results must be taken as modulation distortion was not a factor, preliminary. However, these results can phase noise was added to give a defini­ serve as a relative basis for the evalua­ tely perceptible d~gradation in picture tion of synthesized headend equipment.

45 ACKNOWLEDGMENTS

The author is appreciative of G. J. Palladino, P.E. for his time, guidance and technical assistance in the prepara­ tion of this paper.

The author would also like to recog­ nize William T. Homiller who has been generous with his advice in the testing and editing phases of this paper.

REFERENCES

l Blanchard, Alain, Phased Lock~d Loops Application to Coherent Receiver De­ sign (New York: John Wiley & Sons, l976),pp3-7.

2 Egan, William F., Frequency Synthesis by Phaselock (New York: John Wiley & Sons, 1981), pp 115 - 123.

3 Rhode, Ulric~ L., Digital PLL Fre­ quency Synthesizers Theory and Design (Englewood Cliffs, N.J.: Prentice - Hall, Inc. 1983), pp 85 - 88.

4 Egan, pp 81 - 92.

5 Mannassewitsch, Vadim, Frequency Syn­ thesizers Theory and Design (New York: John Wiley & Sons, 1980), pp 104 - 119.

6 Egan, pp 47 - 49.

7 Gardner, Floyd M., Phaselock Techni­ ques, 2nd Edition (New York: John Wiley & Sons, 1979), pp 100 - 106.

8 Ibid, pp 198 - 201.

9 Blanchard, pp 139 - 145.

13 Vandelin, George o., Design of Ampli­ fiers and Oscillators by the S - Parameter Method (New York: John Wiley & Sons, 1982), pp 145 - 162. ll Blanchard, pp 145 - l53.

12 Egan, pp 103 - 115.

13 Gardner, pp 116- 123.

14 Martin, Harry, "Noise-Property Anal­ ysis Enhances PLL Designs", EON Sept. 16, 1981, pp. 91 - 98.

46