Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: PAM-5, what are your BERs ?




Jamie:

Thanks very much for your very elaborated explanation.


Regards,

Edward S. Chang
NetWorth Technologies, Inc.
EChang@xxxxxxxxxxxxxxxx
Tel: (610)292-2870
Fax: (610)292-2872


-----Original Message-----
From: owner-stds-802-3-hssg@xxxxxxxx
[mailto:owner-stds-802-3-hssg@xxxxxxxx]On Behalf Of Jaime Kardontchik
Sent: Thursday, March 02, 2000 12:16 PM
To: stds-802-3-hssg@xxxxxxxx
Subject: Re: PAM-5, what are your BERs ?



Edward Chang wrote:

> Jaime:
>
> I agree that your analyses are solid.  No question in my mind.
>
> My real question is what caused the PAM-5 link to yield BER of 10^-3 or
> 10^-4?
>
> I believe the only answer is that the complex data recovery requirements
of
> the multilevel coding caused the BER to go up.  Therefore, .....
>
>
> Edward S. Chang

Edward,

Your question is incorrect and this leads to the incorrect answer.

If the electrical SNR at the input of the receiver is about 17 dB you
will get a BER of about 10^(-12). Period. If you do not design your
receiver properly and the effective SNR at the input of the receiver
is only 11 dB instead of 17 dB, then you will get a BER of about
10^(-4). (see my presentation in Kauai, Nov 99, Part II).

I do not think that the answer would  be essentially  different if
the modulation scheme were on-off instead of PAM-5.

There are many things that could go wrong in a 10 Gbps
receiver that could cause the BER to drop to unnaceptable
levels. This is true for all proposals using both on-off or
PAM-5 modulation. It could be a low dynamic range design of
the analog front end of the receiver, it could be the clock
and data recovery, it could be a nominal 6-bit ADC that
has an effective number of bits of only 4 at the frequencies
of interest, and so on. And to make it even more clear:
the problem of a low BER does not have to be necessarily
in the analog front end or the Viterbi algorithm. At these
high frequencies a simple logic state machine could cause
from time to time errors and lower the BER to unnaceptable
values if the synthesized gates do not have enough timing
margin or you have a race condition.

For more details about what could cause the BER to
drop to 10^(-4), I think that the reachest source of
information are the poor serial guys that are trying
desperately to design a 10 Gbaud transceiver and are
fighting over whether they could design a 10.3 Gbaud
receiver or not (12.5 Gbaud seems to have dropped
already) ... or whether there is a way to modify the
eye safety requirements so that they can increase
the signal power and get a better SNR ...

Jaime

Jaime E. Kardontchik
Micro Linear
San Jose, CA 95131
email: kardontchik.jaime@xxxxxxxxxxx