10.
'9-E. Calculation of the Behavior of the
‘Second-Order Loop
‘We now show the performance of a phase-locked loop
‘whose loop filter is the passive integrator of (5-13). This
SUAS LINEAR.
‘REPLI
‘UREAR Sera,
RP RTSTON
Ae Bitten
Slter function inserted into the spectral equation (9-11)
produces the density
Suid) =
N= 2
WPT F TF a Ga) + ART
(928)
‘The density (9-26) is quite different from what one ob-
tains from a simple linear model of the loop. Still, we
cean use the same types of ideas to express the output
phase variance. For one thing, Sev(fe) has a Sducial
‘bandwidth
-) f%
ue = Begey [Sede (027
‘Note that this is computed in the same way as 1, ercept
that we must use Seo(je) as given in (9-26) rather than
| Ue), as one would do if the linear model were in
effect. In these terms
oo Nea (928)
Again, it is evident that there are two effects in the
second-order loop that deviate a from the value it would
take in ¢ linear analysis: wei.) must be used rather than
ty, and A is reduced to Ay.
‘The value of ws) can be found by integrating (9-27):
uNeaR
Wehxnuarion
Noe/* « Noh/24
Fig. 9-3. Comparison of linear, quati-linear, and lineer-
spectral approximate methods with the actual
behavior of the first-order lo
‘The approximation follows the natural assumption
my? te
‘A “damping factor” can be defined almost in the same
way as itis for the linear loops, a way that reduces to
the linear case for high input signal-to-noise ratios. We
hhave already derived Sua; it is of the form
Ina
Sel) = Soper *
‘The roots of the denominator govern the “damping be-
havior” of 440). If these roots are fi, tip —fy, Fo then JPL TECHNICAL REPORT NO. 32-819.
wwe can define, for the left-half-plane factor 3+ (r,+12)
+ rr, an equivalent damping coeficient
(ntn)
fe any”
(931)
‘We relate rir; and (r, + 1.) to the parameters of Sex(s) by
noting
tony = (450
32)
(6s — Boy = [LE BAR os ~ 1) + (AK).
‘Hence, the equivalent output damping factor of the loop is
a OL rt + MaKe =
~ ois
1. Linear Analysis
It is easy to see when y = y= 1 (the linear analysis
parameters) that (9-28), (9-29), and (9-33) reduce to (5-8),
(6-18), and (5-20), respectively.
2 Quastlineur Analysis
‘The quasilinear analysis replaces A everywhere by
Aes” (ie, y= 9 = €*°7), Thus the behavior predicted
in this ease is given as
eee
esp
‘From this it is important to note that while bandwidths
and damping are less than the value predicted by the
linear model, the overall value uf a° is larger than its
linearly predicted value.
34
3. Linear Spectral Analysis
In a still more general analysis, we set » = €*” and
7 =e? (1 — e+), The loop behavior is then approxi-
mated very closely by
Nave _ (= ewe
a Wiieer
te Lt rear
we ple ds eT
a+ +See
(935)
Again, the same type of behavior results: ‘ii 804 fo,
are less than their linearly predicted value, while o° i
sreate. Although the quantities in (0-5) are generally
transcendental in a, nevertheless Newton's rule is easly
applied by a digital computer for solutions. Figure 9-4
relates various parameters above to corresponding values
of et, based ona constant value ofr (viz, r = 2). Figure
85 then shows low o? compares with the value pre-
nivale
Thea nea tN
. nontiean TERT
.
¥ ol
i
: |
3 +
% eof |
3 t
ij i
—
i |.
| T rl
SIGNAL STRENGTH, 2 ABOVE 43 + Mam
Fig. 9-4. Comparison of
{for second-ords
the value
wear and nonlinear theor dicted by a linear analysis as a function of signal power
for a fixed Ng; since r varies with input signal level, we
have arbitrarily"* chosen a value r, = 2 at Ai
Reasons for this choice are given in Chapter 10.
:
{—} |
| _-=umean-seecraat
eerste
JPL TECHNICAL REPORT NO. 32-819
Figure 96 illustrates the way Wess Wir Lean and £
‘vary; parameters here are the same as those in Fig. 95.
4. Conclusions About the Second-Order Loop
‘The behavior depicted in Figs. 9-4, 95, and 9-6 indi-
cates that a linear theory can be expected to yield satis-
factory accuracy whenever A?/Nyw, > 10. Beyond this
int, the linear spectral approximation probably agrees
‘with actual results, if we may extrapolate the results
‘obtained for the frstorder loop, for of < 15 rad, This
figure lies beyond the useful range of most receivers.
PHASE NOISE VARIANCE of, rod?
8
|
SIGNAL LEVEL, @ ABOVE 43 « Hom
and the signal leval It varied. The valve of r
Is taken as r, = 2 of a reference signal lavel of
= INJ/2r, = Nywig. Note that the ultimate
Is Sdb/decade, rather than 10 db/¢
Fig. 9-4. Nete alto thot even the linear approxi-
mation produces some curvature of o near AZ.
i ‘
t
j
. i
SONAL LEVEL, 0 ABEVE AB * Ae
Fig. 9-6, Variation in bandwidth and damping parem-
ters as a function of signal strength. The value of rot
© reference signal level AZ = 3N./2r: = Newig
was token ast.
35 JPL TECHNICAL REPORT NO. 32-819.
56
|. Tourworthe, R. C.,"A New Method for Calculating Phase-Locked Loop Behavior,
}- Boren, J. F, and Lomperd, D. G.,
|. Price, Robert, “A Useful Theorem for Nonlineor Devices Having Goussian Input,
REFERENCES FOR CHAPTER 9
Space Programs Summery No. 37-31, Vol.
Laboratory, Posadeno, Colt, February 1965.
IV, pp. 292-300, Jet Propulsion
Burigang, J. J, "Crot-Cortelation Functions of Amplide-Distorted Gaussian
‘Signals,” MIT Research Laboratory of Electronics Technical Report, No. 216, The
MALT. Press, Combridge, Mass., March 1952.
"An Expansion for Some Second-Order Probo-
bility Distributions and its Applications to Noite Problem," IRE Transactions of the
Professional Group on Information Theory, Vol. T-1, March 1955, pp. 10-15.
[RE Trontactions of the Professional Grovp on Information Theory, Vol. [T-4, June
1958, 9p. 69-72. JPL VECHNICAL REPORT NO. 32-819
CHAPTER 10
DESIGNING A DOUBLE-HETERODYNE TRACKING LOOP
So far in this work, we have concemed ourselves, for
the most part, in predicting the performance of a receiver
when its parameters were specified. Exceptions to this
dealt only with ways of choosing some of the receiver
parameters, given the others, We may now pick a set
of parameters, and, through the formulas that have been
presented, predict quite accurately how the loop will
‘operate, Once the receiver is built, it performs just so.
What we requice for design, however, is an effective
smethod for picking a nominal set of values that produces
a desirable tracking function over a reasonable variation
‘of the input signal power.
‘As long asthe operation of the loop isto remain within
the Linear-analysis region, the task is much simplifed,
because there have been analyses put forth for extracting
an optimum result from the linear loop. Nonlinear op-
timization in most cases is difficult, if not analytically
impossible.
Tis customary to optimize, insofar as possible, receiver
performance atthe “worst-case” parameter values in this,
‘way one issue that, while perhaps not operating optimally
at any other set of values, the loop will do the best that
it possibly can in those cases that require it mest. This
procedure is purely an arbitrary one, in that it is apt to
change in accordance with the philosophy of the design
‘engineer, and with the particular mission for which the
receiver is intended.
‘What design rules are given here will be somewhat
abstract, but perfectly general, so that the designcr may
10-A. Definition of Receiver Threshold
‘The words “receiver threshold” conjure up a different
image to each engineer: to one, it is that shaky, dll-
defined signal level at which the receiver transits from
operability to ronoperability; to another, it is some point
at which the receiver operates with the least acceptable
reliability; to still another, it may mean the signal level
at which the receiver exits {rom its linear range. There
are no absolutes when it comes to defining such 2 point
{in each of the cases above, the engineer meant that the
system hit “threshold” when its performance was in some
way no longer tolerable, when the receiver no longer
produced meaningful data. The trouble with such defini
tions tied to performance is that they tend to be sub-
jective — what is tolerable in one case is not in another.
‘The word “threshold,” as we shall use it here, is a
precisely defined quantity, which can be subjectively
interpreted as desired. Some give it the namo “design-
point threshold”and, in some cases, it has been called the
“absolute receiver threshold." The concatenation of modi-
fiers does not seem to be necessary at all, and we shall
‘use the word “threshold” in oaly one sense.
Specifically, the loop threshold is defined as that input
signal power AY given by
= Nas +1)
a (14 2
en
= Nave, - (a4)
Here, ws, isthe value of livear-loop fiducial bandwidth
at threshold, ie, computed by (6-18) asiny: the threshold
‘value ofr, r. One must not be mistakenly le. to believe
that the loop is operating Hnearly; w., is merely the
bandwidth a Toop would have if A, were to lie in a inear-
analysis region. In actuality, A, isa signal level at which
the linear theory does not apply, for, in fact, @ > 1.
Jn the notation that is to follow, all quantities sub-
scripted by “O° refer to the value of that particular
parameter at threshold.
10-B. Tracking Loop Performance of the
Double-Helerodyne Receiver
‘Once the values of the receiver's physica! parameters
K.kreoMF, n, ry and Wn have been determined, per-
formance proceeds as specified by the formulss”” in
‘Chapter 8. But with a fixed receiver, the loop bandwidth
changes as a function of signal o noise level Rather than
although Wa a measured physical quantity, it my not be
‘equal to we ifthe bop is not operating at the frequency pro-
acing” H (ju) ae Cate must often be taken fo seare that H(3)
{is properly tuned for maximum response at the lop operating
frequen,
37 JPL TECHNICAL REPORT NO. 32-819.
calculate each receiver characteristic separately, it is
‘easier to evaluate the performance in terms ofa reference
signal level and to relate the remainder of the loop be-
hhavior to this point. One convenient reference point is the
threshold signal level. Specifically, we define the ratio
(0-2)
as the loop performance margin. In terms of this m,
mo aic (AJC) =nms- con
As a result, all of the loop characteristics are expressible
in terms of m, w:,, tox, and z2/, (and this latter quantity
is not needed if r,» +.). Thus a design consists really of
specifying these three (or four) quantities.
However, itis often necessary to compute the perform-
ance of a receiver when wy, KikrcoMF, 1», and ry are
sgiven— quantities that can be measured. Most of the
formulas we have given are in terms of rs, ws, etc. How
does one proceed with the set of parameters above? We
rmerely need to find r, and the rest then follow. When
10w4,< on, a3 is very well approximated by x01, /4u4;
rom this, a slight manipulation of (8-14) produces the
desired result:
HK KrcoMt PF)?
TS we
(25P*) (0+ [1+ gg]
~$ [ae (e4)']. 10
‘We can also express U in terms of Fx
v
10-C. Nonlinear Behavior of the
Doubie-Heterodyne Receiver
‘The formulas in preceding chapte-s can be made to
apply to loops with predetection bandpass liters by
making 2 few minor alterations, and we have scen how
these changes come about in Chapter 8: fist, AK is re-
placed by aK,KycoMF, as in (83); and second, the ex-
pression (9-28) must be multiplied by the limiter
performance factor = wapu/tcips, shown in Fig. 83.
‘These steps lead to a set of equations which charactenze
the double-heterodyne receiver:
w=a(S)
« [ O7854p, + 047685, if
Tip, ¥ CATES,
pa 1+ 0345p,
‘Tae F 06Hp,
™Pny (10-6a)
eKiKrosM FO)?
ac
A + Tern)
Liable eth
~ Fen c (0.88) (io!)
(106m)
(2060)
“The ute of yas given in (1068) requires x,» rs otherwise,
‘1 of (8-18) should be substituted in its place. Notice that
the only quantities needed to specify everything in (10-6)
are the margin m, the predetection SNK pg = tng/tim fo
and n/n.
‘The performance observed here is very much the same
as that exhibited in loops without limiters, exeept thatthe
value of KiKrcoMF required is drastically less than the
value of K required when there is no limiter in the loop.
‘This, of cours, is due tothe fact that while Ay is very small
(the threshold rms signal level) a, is many orders of magni-
‘tude greater, approximately equal tothe predetection SNR.
10-D. The Signol Level Producing o* = 1
Since it is doubtful that one would ever expect a loop
to operate usefully with a value of o* greater than unity,
the specific value ofa is more a point of ucademic interest
than anything else. Furthermore, if we may judge from
the first-order loop result, the threshold value of phase
noise predicted by the linear spectral theory would
probably not be a very accurate one, for the theory
begins to fail somewhere near this point.
Holding the rest of the loop parameters constant, we
‘can solve forthe value of margin (callit m,) that produces
a'=a} = 1. Thic is a more meaningful quantity, and
certainly much more accurately calculable by the methods
in Chapter 9. 1 follows from {10-6} that when a? = 1, the
value of A\/A. = mi satisfies the equation
JPL TECHNICAL. REPORT NO. 32-819
‘The subscript “I” on Ty, 9, ete, refers to evaluation at
@° = 1. Under the usual assumption ysror > rs, the approxi-
mate value for 7, is
me[sten) +E + Ace]:
(10.8)
‘The value of o* is within 05% of unity when o* = 1;
hence to the accuracy we need, the values of m, andr,
necessary to replace y by yi, (see 9-18). The values given
in the Figure result by assuming that 7, in (10-8) takes
a (2) *
na(t
= san (1 —
somal ) ‘ (10-9)
MARGIN m ABOVE THRESHOLD,
MARGIN mb
rots
Fig. 10-1, Variation in margin producing o? = Vas
function of threshold design parameter,
4 = aK deco
UAy + Uren)
(o7
59 SPUCTECHNIGAL REEORT/NO: aa Gg eee eee eee eee eee
10-E. Choice of Receiver Parameters
(Choosing values of Ne, trey KiKyooMF, and wy speci-
fes a typical receiver design. Values for r,r2, Pg do, and
TT follow directly from (10-1) and the first four parts of
(10-6). The value of 12/r, is a measure of how well the
imperfect integrating loop filter performs in the loop; with
to/r, = 1, for exzmple, the loop Biter is merely F(s) = 1,
and all the results in (10-6) reduce to the first-order loop
‘equations. On the other hand, the results in (10-6) are
approximately the same for all values of r2/ny less than
about ro/10. There is some flexibility in choosing loop gain
parameters because Ka, Kreo, M, and F appear as a single
factor in the theory.
However, the only things an engineer needs to know
in order to know hovr well any receiver will perform are
‘Fo, Pgs ane perhaps 1»/,. Ths latter set of quantities
depends on several things: 1, sets the loop damping factor,
to be chosen in accordance with the discussion in Section
the doppler tracking capability, the state ofthe art in VCO
design, the spectral purity of the received carrier, the
expected incoming A?/N,, and the amount of phase jitter
that can be tolerated at that operating level. In addition,
‘7 should be chosen (Section 6-C3) so that the phase error
due to the expected doppler rate is increasing slowly
enough that only infrequent retunings of the VCO are
required.
‘Once re and pug (and perhaps r2/n) are given, the re-
ceiver performance in terms of margin can be found from
(10-6):a typical plot of evs mis given in Fig. 10-2, wherein
Tinear approximations are compared with actual behavior,
Fig. 103 shows how bandwidth and damping factor
change as a function of margin. With assumed values of
'N, and tw, threshold is specified (10-1), so the eurves are
easily referenced to the actual incoming. signal power.
CA; the value of pus depends on wz. and wy; wr, is T
dtnined by contring the Serve sequen ae asneror|
Be
7 ' eer
t :
Dommean ream | — t
S oe ete
. SS ‘LINEAR THEORY, i —_
3 ef EE Srey LS
fo re
5 | west necer, vo
y= See en :
t nant
i ree
t te | aa
Fig. 10-2. Comparison of linear and nonlinear approxi-
motions to loop rms phase errer, es @ function
of loop mars
10-3. v,
fectors, 0
7 of loop bandwidths and damping
tunction of loop margin, PART It
SUMMARY OF PHASE-LOCKED LOOP DESIGN CONCEPTS
In this second part, the important definitions, concepts, and formulas are col-
lected chapter by chapter. The notation used in these formulas is listed in the
‘Appendix with names, units, and text references. Many of the formulas here are
only approximations of more accurate ones in Part I.
Since not all of the subdivisions in Part I are referred to in Part If, there are
some discontinuities both in the headings and in the figure and equation numbers
of Part Il.
et JPL TECHNICAL REPORT NO. 32-819.
CHAPTER 1
INTRODUCTION AND HISTORY OF THE PHASE-LOCKED LOOP
“The first serious application of the phase-ock principle was asa synchronizing
You might also like
