RE: [EFM-OAM] RE: [EFM] Re: Loopback and BERT
I am not sure where the "error-correcting endpoint" came from. Most CPE
demarc equipment does not have any error-correcting functionality at the
lower physical layer.
The conformance test is to determine if bit error detection works correctly
at the lower physical layer before upper layer error-correction takes
over. I think that you might be getting back into confusing the the OAM
for EFM is and is not responsible for.
As for being able to distinguish between the upstream and downstream
problem. Normally the OAM overhead will indicate whether the problem is in
the upstream or the down stream. This is the same function that in the Far
End Error Reporting in the overhead of SONET and T1/T3 framing. We
included that functionality in 802.3ae(r).
At 05:31 PM 2/24/2002 -0800, Horne, David M wrote:
>I don't know Roy, if I were a network operator I'd have a tendency to want
>to leave a mysterious "error-correcting endpoint" in place if it made it
>through that deliberate error-injection test ;o). Seems like a functional
>test of the silicon more than anything, i.e. not a system-level test.
>Bob: In your whitepaper regarding bit error rate test, have you calculated
>how long such a test would take for BER <= 10^-12? If you wanted 95%
>confidence the measured error rate was within 10% of the true error rate,
>back of the napkin (and only on a P2P link where entire BW is dedicated to
>BERT, and ignoring header overhead)would be over 100 hours. If you're
>talking about a crude test just to see that it's better than say 10^-7,
>that's a different story. I think you were talking about characterizing the
>link at install time though. Just wondering what shortcuts are taken when
>they run such a test, and what other hooks we might need to provide since
>100+ hours isn't going to cut it. There are ways to accelerate it by adding
>calibrated stress (e.g. precision attenuator), but that definitely can't be
>a required addition for EFM products.
>Also was wondering how you'd distinguish between a downstream problem versus
>one on the upstream if loopback is used? Do you insert a downstream bit
>error count in a header before you loop a corrected test pattern back
>One function that was always part of BERT testing when I was working in
>field operations was the deliberate introduction of errors in the encoded
>data stream to be sure that the customer premise equipment was properly
>able to detect errors. This is part of the originating conformance
>testing, and part of the conformance testing that was done after a
>repair. I do not think that your "loop back plug" would qualify for that
>level of conformance BERT testing.
>At 08:55 PM 2/24/2002 +0000, Bob Barrett wrote:
> >Dear all
> >Rather than put up a presentation on this at the March meeting I have
> >written a white paper expressing what I think are the key issues (and the
> >key reasons and benefits) of including loopback and BERT functions in the
> >EFM subscriber access standard. It can be found at:
> >or from a link at www.rjbarrett.net
> >I have written this as an individual, and not included references to my
> >company or products, and I have put it on my personal web site. Inevitably
> >do mirror my views into the technolgy that my company produces, as do most
> >of we equipment vendors.
> >By putting this paper up for discussion now I hope to reach the broad
> >audience of all EFM streams and stimulate debate in advance of the March
> >meeting, to help us to reach consensus.
> >Bob Barrett