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

RE: [802.3ae] RE: [802.3ae_Serial] 802.3ae PRBSs are upside down


As Tim points out, we don't specify where the pattern starts. We only
specify that the initial value for the PRBS generator must be non-zero. We
don't control exactly when the PRBS generator or the PRBS checker is turned
on either. So, it doesn't make any operational difference whether we specify
using the input to the generator or the output of the last stage. They will
have the same pattern with different phase and we exercise absolutely no
control over the phase of either of them.

My understanding is that the test equipment acquires lock to the pattern by
looking for a point in the pattern and starting the test when it gets there,
so it doesn't require us to control the phase either.

The only question is whether to use the current pattern or to invert it. My
reading of O.150 is that it implies the inverted pattern is used, but it
doesn't state it and some parts of the spec appear to say that either
inverted or normal might be used for a given pattern. Inverted and
non-inverted should perform an equally effective test. The concern with
inverting the signal is that people are trying to get these chips done and
implementations under way will end up with the wrong polarity. It would be
helpful if someone could verify whether BERT testers support both polarities
or just one.


-----Original Message-----
From: Tim Warland [mailto:twarland@xxxxxxxxxxxxxxxxxx]
Sent: Tuesday, March 05, 2002 8:44 AM
To: stds-802-3-hssg@xxxxxxxxxxxxxxxxxx
Subject: [802.3ae] RE: [802.3ae_Serial] 802.3ae PRBSs are upside down

"Nepple, Bruce" wrote:

> hehe --- we definitely have a bit of a mess here.

But it's easy to fix.

> You interpret the "inverted signal" statements in O.150 to mean
> that the output must be inverted.  Makes sense to me.
> In addition the PRBS31 specified in 802.3ae 4.01 is also not O.150
> compatible since it's output sequence does not start at the proper place
> in the sequence.  The output should be the inverted MSB, not the input.

However, the starting point is trivial relative to having the correct data.
Starting the PRBS at the wrong place impacts the time required to
to the pattern not the quality of the results.

> Section 4 states that the digital signal  is take from the output of the
> shift register.
> If you read Section 5.1 you will see that this sequence starts with the
> 1 of a sequence of 9 ones.  That is most easily done by resetting to all
> 1's and taking the output from the MSB.

FWIW the correct reference is section 5.8 which contains an error. It should
recommend a thirty-one stage shift register not a twenty-nine stage shift
register. Furthermore, section 5.8 does not suggest a starting value. The
remains that regardless of where you take the output, the pattern generated
by a 2^31 PRBS generator should match the expected PRBS pattern which
the majority of test equipment recognize. That pattern is 2^31 using the
described in clause 49 and 50, with the output inverted.

> <snip>
> In summary:
> 1.  The PRBS31 scrambler specified in 4.01 is drawn at odds with O.150 in
that the
> output is taken from the input, rather than from the MSB, and it is not

But by adding a simple inverter, the two patterns become compatible
(although as
you say, perhaps not identical). Compatibility with existing test equipment

Tim Warland  P. Eng.
Applications Engineer
Quake Technologies   (613)270-8113 ext 2311

Tough Times don't last, tough people do