TKK | Tietoverkkolaboratorio | Opetus
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Some comments on the problems and answers obtained:
1. The small questions were answered well. Many guessed wrong
what the 'A' in
ADSL stands for, it is not Asynchronous or Advanced, but Asymmetric:
the
main idea is that the downstream (towards the user) data rate is much
higher
than that for upstream.
2. The answers to questions 2a) and 2b) were more diverse. It
was apparently
a problem that the questions were based on guest lectures, the echo
canceler
on the network and xDSL lectures and the multipath RAKE on the CDMA
lecture.
Only the basic ideas with pictures were required for full points.
3. The capacity calculations in Problem 3 were interesting. Most
solutions
for a) gave optimum power spectrum Sx,opt(f) = Px/W correctly and also
obtained for b) the capacity C = W log2(1+PxK2/Pn). In c) the notch
means
that the power must be distributed over the remaining band, giving
Sx,opt(f) = Px/(W-delta f) and the capacity C = (W-delta f) log2(
1+PxK2W/(Pn(W-delta f)) ), some solutions failed to notice this.
4. Also the adaptive equalizer problem seemed to be easier than
I expected.
The most frequent omission was that no solution was given for the step
parameter beta, and I decided to give the extra 2 points only if beta
was
solved correctly as beta= 2/lambda,max = 2/R0.
5. The DFE equalizer was perhaps the most tricky problem. It
was modified
from the problem sessions. The linear ZF equaliser in a) is easy
to get from
the Nyquist criterion
HT(z)C(z)HR(z)=C(z)HR(z)=1 (HT(z)=1) or HR(z)=1/C(z)= 1+kz-1. However,
in b)
one must first decide the criterion of optimality (which was not specified).
If ZF criterion is taken, then solution for a) is one possibility which
meets the criterion for the linearized model C(z)HR(z)-FR(z)=1 when
the
feedback is set to zero (FR(z)=1). However, if the MMSE criterion is
used,
then the noise power or integral of HR(f)2Sn(f) must be minimized.
If the
noise is white, then the optimum choice for HR(z)=1 and the feedback
filter
is solved from the Nyquist criterion as FR(z)=C(z)-1=-kz-1/(1+kz-1).
As stated in the exam paper, problem 4 could give 2 extra points. In
addition, some extra point were given for other exceptionally brilliant
solutions.
Thank you for great performance in the exam! I hope you will be able
to
complete the MATLAB project as well.
The next exam has been tentatively scheduled for December 13, but I
feel it
would be better to have it in September instead. Please contact me
by e-mail
if you have any wishes for the date of the exam.
June 6, in Otaniemi
Timo Laakso
Networking laboratory is now part of Department of Communications and Networking. Content on this page may be obsolete.
Check MyCourses for upto date information about cources.
This page is maintained by
and
Webmaster.
Last update on the page 16.06.2000 14:43. URL: http://www.netlab.tkk.fi/opetus/s38411/exam.shtml [ TKK > Electrical and Communications Engineering > Networking Laboratory > Studies ] |
Ask |