Download Digital Communication Systems: Handwritten Exercises with Solution and more Exercises Digital Communication Systems in PDF only on Docsity! hype 7
Gl () C= W hog. (i+ sue)
C= 3000 Leg, (1001) 2 30 000 bik fy.
(2) 2 = j+sNR
SVR= 27-1 = Qtr? = 2.03
= 3,08 dB
~| = $3,45= 17,248
192%C/g,00
| (O swR= 2
1,92 ce W log. Cit SMR) = Be00 fog (")
= 3000 * 3.46 = !0, 380 bE fa,
“s zt 4 (H)
uf
toa 4+ ¢ og, e+ the 16
44 pe Oe B®
7, H = Z fe toy- C4: )
= 15-(0.06) bog, (Ki) + 26S (0.00035 lo, ar
= 3653+ LIGS =
= £ dg tee
zy
I
= 4,196
Ree = 4.148 x 1000 = 4798 bebe /a.
45 (a)
He = BR hen OH)
6 (qe gat) = FM fader
Ha, x Beminde [tte
4
Rae
= bela /etmondt * 32x2«/0° amine fre.
= 2.56 xo* Ate fate
() Haw = GFxi6 (ime Lag, (6*64)) * 10 bE [aly
Rar = Ha * 32x 2810alemedily = 6.44n/0" lols /a
9-2
With 4 Mis dale Pll, moe Gf
fir ce Pte foeturtr- bo eednc/
Ant it fe getsumted ha? thes 6
Quaclatle pn he bancluntl buley
“a
tae f- PK
bn ly, Ait rele By 7
ecrtaston fanolartlX (47 Cf,
Cypaes GY ineuseed Cefn) for a
0 LEE.
9-6
7.9 fn EIRP ys
t & EIRP Gp
> a> 2 2+ =f“
N Riel. ? ® ” RELL
Eh )ap = ZIRP +6. (R th +L,+L,) all en AR
INTELSAT We, tak, = 176 dB By = Ace
Dscs 1:
pscs Tr:
Eum, = 20.648 5 Ky = 0.002
ELg, = 2148; Ry = 1 44
GAPSAT: ELA, = 3.648, Ry = 0,00!
Raw
¥
6 tyrewsay Te
4 SEVERELY BW LiMiTED
psc IC
: @ 8 UnueD
SEE Fe, OD
*
oar
one
Oneee
8. D3iar
bore
O0O7E
0.0834
6, 900g ® Dses
aed tL © garsar Pewen mire
SEVERELY
B.000NS Pow LiMiTED
9-7
=0
mR
a,
8
~
as
%
—
‘
&
e
=
~
mo (97) Coen) Goan)
- 23S x IO”
9-10
3°” ahle for
712 Barc ti -. Q
fe Z GA9 He (bang
Wh (5 ry GAY 0b bit, ane
9-11
Pitnemsllid po 9b00 bee
4 bets fogmtel
Wilf. the (122 G2) BCH cole the
2400 will be bupandid
ayoo «22 ~ 3313 (wheed
fuse he 3400 Jeg usta tnd )
The enete provides thy follerng tren
efit anes ;
fin -(@) (tara) (1-4 az x10"*) re!
= 103% °
R= 1-G-& ye? = BI x10
Wheih rl the ape. Y fe < tot
Tnbobensing in Aregucree i hanoll.
9-12
For the (127, 64) BCH code, we note that the rate of the code is
k/n=64/127~1/2. Hence, repeating the above computation yields
approximately the same channel bit-error probability ( p, =1.18x10-?)
as before. However, in this case of entering p, on the abscissa
of Figure 6.21, we are using the transfer function of the (127, 64)
code, which yields a decoded bit-error probability that meets the
required P, 10-7.
We now consider the final candidate, the (127, 36) BCH code, and
Tepeat the same computations shown above. (Remember to use the
given transfer-function intercepts to guide you in making graphical
estimates.) The computations now yield p,~3.8x10%. When
entering this point on the abscissa of Figure 6.21, and examining the
transfer function of the (127, 36) BCH code, we find that P, >10-’.
Hence, of the three candidate codes, only one, the (127, 64) BCH code
meets the requirements.
(b) From Figure 6.21 it should be clear that the (127, 36) BCH code
is the most capable of the group. Even its label states that t= 15,
meaning that within a block of 127 bits, this code can comnect any
combination of 15 or fewer errors. A natural initial guess might be to
choose the (127, 36) BCH code-—but that would have been incorrect.
The reason that the more capable (127, 36) code does not meet our
requirements is related to the fact that in a real-time communications
system, there are two mechanisms at work: 1) more redundancy
makes for more powerful codes and, 2) more redundancy results in
Jess energy per channel bit. As the rate of a code decreases (from 1
to 0), there will be an improvement in Ps due to mechanism 1. But
eventually, mechanism 2 results in more errors out of the demodulator
and outweighs the benefits of mechanism 1, (See Section 9.7.7.2.)
(c) Using Equations (9.25) and (9.27) we compute the uncoded
received E,/No corresponding to p,=10-7, as follows:
9-15
2h, a (H
20) Ny wn( 5)
log, M log, Af
1.5x10°7 ala} 3827) 5 |: Q(x)
where Q(x) = seo 2/2)
E, Ey [E) (1
By trial and error: x = 5. 13, Thus, = N, =89.85 and W Fe #| (3
0 a
E.
and thus, E: W we, =29.2=14.8 dB. Therefore, the coding gain from
Equation (9.32) is: G(dB) = 14.8 - 10 = 4.8 dB.
9.14
4ad
A
§ T, =T,+(LF-1)290 = 290+(20~-1)290 = 5800K (or37.6dB-K).
Since (Z,/N,), = Mx(£,/No)
for (E,/No ) with the parameters given and the basic link margin
___EIRP G/T
(ETN naga R kLs Ly
= 5.2 dB (or 3.31). Since the channel is bandlimited, we choose
MPSK modulation. To meet the bandwidth requirement of 3000 Hz, i
and at the same time conserve power, we choose the smallest M/-ary i
value for MPSK, which is 16-PSK. This modulation (with perfect
filtering) will require a symbol rate (and transmission bandwidth of i
Rilog, M = 9600/4 = 2400 Hz, We also need to select a BCH code |
Space loss is Z, = a i =3.94x10" (or 156 dB). System temperature
rgd? aNd margin is 0 dB (or 1), we solve
equation M = which yields the value of (£,/N,)
9-16 |
such that the 2400 Hz modulation bandwidth is not expanded beyond
3000 Hz. Hence, this places a restriction on the code redundancy n/k,
and for a (127, k) code, the smallest value allowable for kis dictated
by the fact that 127/A must not exceed 3000/2400. Therefore, & must
be larger than 102, and our choice from Table 9.2 to provide the most
redundancy (and still meet our bandwidth constraint) is the (127, 106)
triple error correcting code.
The chosen 16-PSK modulation dictates that the available (&, / Ny),
will yield an £,/Nq=(log, M)(k/n)(E,/Nq)=4x(106/127)»3,31=11.05.
We use this value for finding symbol-error probability, P;, in the
following relationship:
P,(0)=20| ae ea
where Q( ) is the complementary error function. With the computed
value of £,/N,=11.05, we find that Py = 0.359. Since Gray coding is
called for, the probability of channel bit error is p, =a 0.09.
2
Now, we can calculate the decoded bit-error probability, Pp, using
the approximation of Equation (6,46). Note, that if p, is small (or
E,/N, reasonably large) then the first two terms in the summation of
Equation (6.46) are usually adequate because of rapid convergence.
However, in this problem, convergence occurs after about 25 terms,
yielding a decoded Pz = 0.09 (There is no coding gain!) If only the
first two terms in the sunmmation of Equation (6.46) had been used,
the result would give the erroneous appearance of acceptable error
performance.
9-17
2 NO)
a ty, im Cn")
aM~ = Avtrage porte for MK
EO ee
(e) for « fod & ama As Enns, oe
SA ee Oe tg
fronelin of MT Hyena the cane Y GAA]
avenge frmeay wees te A ntar fumelion
R _ 28.8 kbits/s :
W 3439 Ha 8-4 bits/s/Hz
S) ER C E, R|
(b) C=Whog (1+) =Wt0g see © Liog [r |
) 2 N ON 0 | Ltt
W NW
E, C) w(20" 1) g
Ha: CL ne abe \ =)
LetR=C: Then, 2 iff | and a xt 10
342920738294
ae =l0 By trial and error, C ~ 20,300 bits/s,
weer)
soo —1}=40.1 (or = 16 4B)
a 3
N, Cc 28,
tynatittelerrn zh 4 ton -234]
Arey Argmel powen for the Halt apoare
eo. part & her anplilede OS ngs”
ft b har ancplide fasts se
peer 2.57
pre © a>
Prewege Pree» Boss Lae) $ (ue)
-F)
SD The _ conetllalicn Nepures
ar, ancl Tun orbhorpnel
. tlh 2 Am pltictin Oud
[sar Vb (PRee, BL Mienerbita
bmp llirtes wth lawn tg gn 0
paacec.
the,
9-21
9.20
For this problem, the circuit, trellis, and Ungerboeck partitioning
diagrams are seen in Figures 9.29, 9.30, and 9.32, respectively. The
error-event path with the minimum distance is seen in Figure 9.24 as
the darkened path Jabeled with waveform numbers 2, 1, 2. The
minimum distance-squared ay =36as is shown in Equation (9.61).
Note that in this example, the parallel paths do not characterize the
etror-event with the minimum distance. Using Equation (9.63), the
average power for the signal waveforms is S,,=21. In this example,
the reference waveform set was chosen such that @?.. =4, and the
1? +167
5 =128.5.
average power for this reference set is Si, =
Coding gain is computed using Equation (9.62) as follows:
d2/S, ( 36/21 \
G(AB) = 10 log, oe | = 10 logy oes jr i74ap
ref ¢ a \ .
Of course this answer seems to violate the Shannon prediction of
coding gain. Would anyone use such a paradoxical reference set as
the one given here? Absolutely not. However, sometimes the choice
of a reference set involves judgement. Generally, any reasonable
choice for a reference set yields similar coding gains. But, this
problem purposely starts with a very unreasonable reference choice
to emphasize that the resulting 17.4 dB coding gain teflects two
different nechanisms: 1) the improvement due to coding, and
2) the improvement due to the better signal-waveform set (compared.
to the sub-optimum reference set).
Because trellis-coded modulation involves coding in conjunction
with modulation, the “so called” coding gain can be made to appear
arbitrarily large by simply starting with a reference set that is
sufficiently poor.
9-25
9.21
We draw the trellis diagram, and we assign waveforms to trellis
branches according to the Ungerboeck assignment rules. We then
label each state according to the waveforms assigned to the branches
that emanate (top to bottom) from that state. For example, the first
state and its branches are labeled as follows:
STATE D
0426
4
2
6
After drawing and labeling three sections of the cight-state machine, a
methodical search for the error-event path with the minimum distance
yields the darkened path shown below.
We find d ; using Figure 9.22 as
dP =d? +d? +d? =2+0.585+2=4.585 i
For the 4-PSK reference set, we find d?,,=2 from Figure 9.23.
Te
9-26
1e5
1537
4062
5173
2604
3715
F351 s a
The asymptotic coding gain relative to a 4-PSK reference set is found.
using Equation (9.62) as follows:
ft
21S,
G(4B)=10 log,, (Z a |
a? Sa,
ref8¥
The average power is unity for both the signal waveform set and the
reference waveform set. Therefore, G(dB)=10 lob | =3.6 dB.
9-27