Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
DPSK SIGNALS
K.T. WOO
20.3-1
CH 1734-3/82-0001/$00.75© 1 982 IEEE
Authorized licensed use limited to: Defence Electronics Application Lab. Downloaded on August 16,2010 at 07:33:20 UTC from IEEE Xplore. Restrictions apply.
inphase and quadrature low-pass signals can be - k k2 (C) + a2( )
written as (5)
y1(t) = LJd(t)sin0 + where
nc(t)J
yQ(t) = d d (t) cosf + n (t)J ak(s) = fy1(t)P(t-kTs + s)dt
where S is the received signal power; d(t) is the Sk(s) = fYQ(t)P(t-kT5 + s)dt
bi-phase data modulation; n , n are quadrature
components of the narrow-bafid Gaussian noise addi- are the inphase tgd quadrature integrate and dump
tive to the received signal; and * is the carrier outputs of the k symbol. The MAP estimator e
phase error which can be assumed to be a random will then satisfy, for the NDA case, the following
variable uniformly distributed between 0 and 2sr equation
for this non-coherent operation under consideration.
To a good approximation the noise processes nc aA(°)I0= t
and ns can be assumed to be independent, white
Gaussian processes. The sampled values of y,
ag
-
) + k ake Yk(
(7)
Bks 0k (c)
where a, Bk are partial derivatives of a and B
Yn wi ll then have a joint probability densit; of (6) Kespect
fInction of the form to c, respectively.
p(YW, YQ/e, , A)-exp p(cosf Y TD + sinf YQTD)
In the with-data-aiding (DA) case the bit stream
(2) ak is assumed known _+1 numbers. The gradient equa-
where p = 2/z/N0depends tion which the DA estimator e must satisfy can
only on SNR and then be found from (4a) to be
where YI, YQ, and D are L-dimensional vectors
which are sampled values ofYi,YQ, and d(t) re-
spectively, T denotes transpose. The aboveisthen
the conditional probability density of YI, YQ
kRID(g)"lk k(e) + RQD(t)AkAk (e) = 0 (8)
where RID, RQD are thecorrelation of YI, YQ with
given the symbol sync error £, carrier phase error the known data pattern:
*, and the sequence of data polarities A = (al, a2,
... ) corresponding the data modulation RID(s) f y1(t) Z a.P(t-jT + s) dt
d(t) = z akP(t-KT5 + c) (3)
K RQD(s) = f yQ(t) z a.P(t-jTs + s) dt (9)
where P(t) is the rectangular pulse function with jJ
symbol time duration T RID, RIQ actually themselves the estimators
The MAP estimation We is given by the value
are
9 that maximizes p(YI, Y?/c), which can be ob-
of coso and sino respectively, which serve to
eliminate the * dependence on ak and Bk in (8).
tained by averaging eq. 2) over the random var-
iables * and A. Averaging (2) over * we obtain Closed loop symbol synchronizers, motivated by
the MAP estimator equations (7) and (9), can be
P(YV, YQ I£ A)- I (P2'Y ITD)2 + (YQTD)21) realized as in Figures (2a) and (2b), for the NDA
and DA cases respectively. A positive and a neg-
(4) ative pulse at the edges of each symbol, as shown
in Figure (2c), can be used to approximate the
whereIois the modified Bessel function. Further- partial derivative of P(t). The data modulation
more, as the sample interval approaches zero, we polarity estimates ak can be obtained from the
obtain formally the continous-time version of (4). DPSK data detector outputs ak (+1 valued) by the
operation
Io (P Y(t)d(t)dt]2 -+ Iry (t)d(t)dtl 2 ak = ak ak-1 (10)
(4a) FREQUENCY HOP ACQUISITION AND DPSK FRAME SYNC
Two forms of symbol synchronizers may result Coarse acquisition of frequency hopped signals
depending on whether the data is assumed to be is usually accomplished by either a stepped ser-
known (or, decision-directed with data detector ial search or matched filter detection. In the
outputs), or to be random. These are conmnonly re- stepped serial search scheme the received signal
ferred to as with and without data aiding [2). is dehopped according to the local pre-set fre-
For non-data-aided (NDA) solutions we need to av- quency hop pattern to an IF signal which is band-
erage (4a) over the data randomness to obtain a pass filtered and envelope detected. Accumulated
likelihood function to be minimized to obtain e. sum of the envelope values over an examination
With ak's assumed to be independent random vari- interval is then compared to a threshold to de-
bles taking value ±1 with probability ½ it can be termine if frequency hop sync has been acquired.
shown that the likelihood functional is given by The entire FH code phase uncertainty will be
searched until sync is found. The matched filter
A(e/YI, YQ) = ln p (y1., yQ/e) scheme, on the other hand, is a passive correlator
which searches the code phase in real time by
20.3-2
Authorized licensed use limited to: Defence Electronics Application Lab. Downloaded on August 16,2010 at 07:33:20 UTC from IEEE Xplore. Restrictions apply.
having a bank of correlators which consists of where a, s are the quadrature integrate and
dehop/BPF/envelope-detector circuits whose LO dump ou putg. (see, e.q., [4], [5]). Figure 4
frequencies correspond to a sequence of N con- is a functional block diagram of a DPSK detector.
secutive frequencies in the frequency hop code. For the Gaussian noise channel under perfect time
The envelope values are added in real time to be and frequency synchronization the probability of
compared with a threshold. bit error of the DPSK detector is well known:
Performance of either acquisition algorithms
depends on the probabilities PD and PFA of signal Pe = h exp (-R) (12)
detection and false alarm in the threshold com-
parison tests described above. For example, with where R = STA/N0 is the symbol SNR.
q cells to be searched,each with an examination
time of Te,the mean acquisition time can be shown Assuming independent, equally likely +1 data
to be (31 the probability of DPSK detection error with a
normalized symbol sync error x (Ixl < h) can be
T- = q (2-PD) (1 + KPFA) Te (11) shown to be given by the following:
2PD
where KTe is the additional delay required by the
Pe (X) = 1 [eR +eR(1-21xj)2]
receiver to return to the search mode after the
occurrence of each false alarm. +¼ + ¼ Q [J lI l, 2R I( 1- I X) -
;J) j (15)
of the symbol duration. Thus one of these de- Io
tectors will be in symbol sync with the received
data stream to within ¼a of a symbol, and can thus Numerical results of this evaluation are given in
detect the DPSK data with a degradation whish is Figure 6.
not too significant (< 4 dB at a BER of 10-3, see
Figure 5), provided that the hopping frequencies REFERENCES
are in alignment. The detected data can then be
used to wipe off the data modulation, forming, (1) Simon, M.K. and Lindsey, W.C., "Tracking
essentially, envelope detectors whose bandwidths Performance of Symbol Synchronizers for
are matched only to the hop rate. The maximum of Manchester Coded Data", IEEE Trans. on Comm.,
the two envelope outputs will be accumulated to VOL. COM-25, No. 4, April 1977.
test against a fixed threshold for frequency hop
sync detection. (2) Frank, L.E., "Carrier and Bit Synchronization
If a known data preamble is available during in Data Communication -A Tutorial Review",
initial acquisition it can also be used, as in- IEEE Trans. on Comm., VOL. COM-28, No. 8,
dicated in Figure 3. The same circiut can be used August, 1980.
for the detection of DPSK frame sync, i.e., for
the determination of the first DPSK bit within a (3) Holmes, J.K., "Coherent Spread Spectrum Sys-
hop, by testing all possible orders of the known tems", Wiley Interscience, 1982.
pattern assuming a repeated DPSK pattern is sent
during the initial acquisition phase. (4) Park, J.H., Jr. "On Binary DPSK Detection",
IEEE Trans. on Comm., VOL. COM-26, No. 4,
EFFECTS OF SYMBOL SYNC ERROR ON DPSK DETECTION Apr., 1978.
Detection of DPSK signals is usually achieved
by performing the operation (5) Simon, M.K. , "Comments on "On Binary DPSK
Detection"," IEEE Trans. on Comm., VOL. COM-
k 9 [k ak-1 +k Sk-1] 26, No-10, Oct. 1978
(11)
(6) Whalen, A.D. "Detection of Signals in Noise",
20.3-3
Authorized licensed use limited to: Defence Electronics Application Lab. Downloaded on August 16,2010 at 07:33:20 UTC from IEEE Xplore. Restrictions apply.
Academic Press, 1971.
Figure 28 SKibol Smac Phase Dtector-ODta Aided Case
DPSK
(7) Lindsey, W.C. and Simon, M.K., "Teleconmnuni-
cations Systems Engineering", Prentice Hall,
1973.
ikP(t-kT S+)
Loop Filter
y0a(t).
akP(t-kTS+ )
Figure 2c Approximation of a
P(t-kT5 4)
P(t-kTs ) K- T.
kTs - (k+I)TS c
du
1.P(t-kT +5)
-
To Loop Filter
Hard Decision
-I Output
20.3-4
Authorized licensed use limited to: Defence Electronics Application Lab. Downloaded on August 16,2010 at 07:33:20 UTC from IEEE Xplore. Restrictions apply.
Figure 3 Frequency Hop Sync Detection With OPSK Deta Aiding
1.0
Figure.5 DPSK Probd i i f Bit Err b-1O Figure 6 DPSK Probabilityof Bit Error
I X
1-
0
1
L.
LAJ
L l- 01
.102io!-i
; >\ l ;\ \ .l N naie RI .s.
0
Symb Sync 0 O.,15 j2\25 35 0
Error
D 15s9 3 6
F/N (dB)
12 1 18 21
l 6
IN
21
1 u
20.3-5
Authorized licensed use limited to: Defence Electronics Application Lab. Downloaded on August 16,2010 at 07:33:20 UTC from IEEE Xplore. Restrictions apply.