(b) Table 2, presented below, describes the mapping from the four bits I) ».172 p-1+ 03,04,» t0 the
PI tnetl2n-t» 0312s,
output coordinates of the V.32 modem.
Table 2
Present output | Present input Output
dibit dibit coordinates
Tin lon Q3n 24. o 2
0 1 0 0 1 -l
0 1 0 I 1 a
0 = 1 0 3 -1
0 BE 7 I a 3
0 0 0 0 “ll “1
0 0 0 1 3 Bt
0 0 1 0 1 3
0 0 1 1 3 3
1 0 0 0 <1 1
1 0 0 1 <1 3
1 cv) 1 0 3 1
1 0 1 1 3 3
1 1 0 0 1 1
1 1 0 1 3 1
1 1 1 0 1 3
1 1 1 1 3 3
(b) We are given the current input quadbit:
21, nQ2,nQ3,nQs,n = 0001
and the previous output dibit
Tin-tlan
From Table 1, we find that the resulting present output dibit is
Tinton =
4th quadrant
3rd quadrant
}
/ 2nd quadrant
Ist quadrant
Hence, using this result, together with the given input dibit Q3 ,04, = 01 in Table 2, we find
that the coordinates of the modem output are as follows:
357o, = 3, and = 1
We may check this result by consulting Table 6.10 and Fig. 6.49 of the textbook With
Qj, ,Q2y = 00 we find from Table 6.10 that the modem experiences a phase change of 90°.
With Fry Foyer = Ol, we find from Fig. 6.49 that the modem was previously residing in the
fourth quadrant. Hence, with a rotation of 90° in the counterclockwise direction, the modem
moves into the first quadrant. With Qs ,Qs,, = O1, we readily find from Fig. 6.49 that
6, =3, and = 1
which is exactly the same as the result deduced from Tables 1 and 2 of the solutions manual.
For another example, suppose we are given
Qj, nr, nQ3,nOa,n = LOU
and
Ty nal
‘Then, from Table 1, we find that
Fnlajn = 00
Next, from Table 2, we find that the output coordinates are $y =-3 and 2 = -3. Confirmation that
these results are in perfect accord with the calculations based on Table 6.10 and Figure 6.49 is left
as an exercise for the reader.
Problem 6.40
(a) The average signal-to-noise ratio is defined by
Pay
(SNR) gy = @
°
where Pyy is the average transmitted power, and 62 is the channel noise variance. The transmitted
signal is defined by
sylt) = a,cos(2nft)—Bysin(Q2n fF), OStSTwhere (ap, bs) is the kth symbol of the QAM signal, and T is the symbol duration. The power
spectrum of s;(t) has the following graphical form:
Power
spectrum
of 540)
Main lobe
Fig. |
On the basis of this diagram, we may use the null-to-null bandwidth of the power spectrum in Fig.
1 as the channel bandwidth:
Py = 2Ey = —™ @
where E,, is the average signal energy per symbol.
To calculate the noise variance 6, refi» the following figure:
Pox
spectrum
of noise | No/2
0 ie 5
Lo .s_
Fig. 2
359‘The noise variance is therefore
o = NB GB)
Hence, substituting Eqs. (2) and (3) into (1)
BE,,/2
(SNR)qy = 5
NB
1/E aw
* alm)
Expressing the SNR in decibels, we may thus write
Ew
10log (SNR), = —3 + 10log (se).08
10\No.
Given the value 1010819, (Eay/No ) = 20 dB or E,y/Np = 100, we thus have
10log jg(SNR),y = 17 4B
(b) With M= 16, the average probability of symbol error is,
P,= 1 we farim|
1 100
E ot serte( Hm)
= 1.16% 10°
Problem 6.41
We are given the following set of passband basis functions:
i
{O(t)eos(2nf,1), $(4)sin(27f,1)},
360Property 1
fi (o(Heos(2nf,M)PU(e)sin(2mf,N))dt = 0 for all n a
To prove this property, we use the following relation from Fourier transform theory:
feos ans, O(sin(2xf,)de = 0 for all n @
where g,(t)G,(f) for i = 1,2, and the asterisk denotes complex conjugation. For the problem at
hand, we have
att) = [Psine((A)cos(2nF,0)
a(t) = Fsine((2)sineens,0)
‘The Fourier transform of the sine function is
F [sine{2)] = Trect( fT)
where
1 1
hspsh
rect(fr) = | | for -aRSfS5p
0 otherwise
rect((f—f,)T) + rect((f + f,)T)]
361rect((f-f,)T) — rect((f + f,)T)]
Let 7, denote the integral on the left-hand side of Eq. (1). We may then use Eq. (2) to write
AG) tech -F,7) ~ reel + far ®
where the integrand is depicted as follows:
4 : f
Fig. 1 fy ° :
— VT le— VT —s
From Fig. | we immediately see that the areas under the two rectangular functions are exactly
equal. Hence, Eq, (3) is zero, thereby proving Property 1 for any n.
Property 2
fi Jaf! 1 PRE 1 for k= n
ole on dt = 4)
Late (Z oe) { 2 for ken ‘
Let J denote the integral in Eq, (4). When k = n, we have
~ Loop
= af ear
= slo (9)
3621 of
= Zf_sine (Za
Using Rayleigh’s energy theorem, namely,
fear =
we may write
6nPar
3A sine?[Z)at = sine" an, hewT
fi reo (nar
which proves Property 2 for k =
To prove Property 2 for k#n, let
par,
gilt) = one”
_P Ward
: FFsne(S)e
Dap
a2) = OMe
fesine($) ttle
Then applying the following relation from Fourier transform theory,
Feanewo a =f ene af
we may rewrite the integral J, of Eq. (4) asar F[sine (pe) F[sine (2) eo ap
Since f,, = n/T by definition, we may depict the two Fourier transforms constituting the integrand
of Jy as shown in Fig. 2 for the worst possible case of k= n+1:
Flsince ye nly
Fig.2
Hf gt ft
2 2s
From Fig. 2 we immediately see that the Fourier transforms of (1) and 9(t)e will
never overlap for k #. Hence, the integral J; is zero, proving the rest of Property 2.
Property 3
Forney * oayen(nye yah =0 for ken
where the star denotes convolution. From the convolution theorem, we have
FLO@)*A()] = O(AHCS)
where ®(f) = F[(t)] and ®(f) = F[h(1)]. For k #7, the picture portrayed in Fig. 2 remains
equally valid except for the fact that the basic rectangular spectrum is now replaced by that
rectangular spectrum multiplied by the frequency response H(f). This multiplication does not
affect’ the nonoverlapping nature of -—sthe_~— two spectra._—_representing
Coens ana (QUp*ACNE”
*) for k #n, hence, proving Property 3.
Problem 6.42
Step 1 - Set k = 0 and the initial noise-to-signal ratio NSR(K) = 0. Sort the subchannels used in
ascending order (i.., from the smallest to largest ones),
Step 2 - Update the number of subchannels used by setting k= k+1.
364Step 3 - Compute NSR(k+1) = NSR(O) +
8
Step 4 - Set 2(k) = PP ESRD)
%
Step 5-If Py = A(k)-T] | <0
Bk.
(st)
then compute P; = A(k-1) A 5
Pi)
and
for 1 = 1,2,-,k-1
Otherwise, go to step 2.
For notations, refer to Section 6.12.
(The algorithm presented here is adapted from T. Starr, .M. Cioffi, and PJ. Silverman (1999);
see the bibliography.)
Pr 43
(a) P| +P, +P3 a
o 2
| -K = TS =To* @
81
P,-K = )
365@)
‘Adding Eqs. (2), (3) and (4), and then using Eq, (I):
2 lot
3K = P40 ee
40° (1+ i +7)
Solving for K, we thus have
10,1
yt 3(1 +3-3)
a
3
366