Re: Wide Area Networking for the Rest of US - the debate on BER and other issues
- To: Rohit Sharma <RSharma@xxxxxxxxxxxxxxxxxxx>
- Subject: Re: Wide Area Networking for the Rest of US - the debate on BER and other issues
- From: Yongbum Kim <yongbum@xxxxxxxxxxxxx>
- Date: Fri, 04 Jun 1999 11:59:59 -0700
- CC: "'stds-802-3-hssg@xxxxxxxx'" <stds-802-3-hssg@xxxxxxxx>
- Organization: Beacon Systems
- References: <B123A833D5BDD1119776006008AE250B650911@xxxxxxxxxxxxxxxxxxxxxxxx>
- Reply-To: YongbumKim@xxxxxxxxxxxx
- Sender: owner-stds-802-3-hssg@xxxxxxxxxxxxxxxxxx
Subject of BER and FCS.
I agree. If you assume simple binary serial link, the difference is
less than a dB of dynamic range.
I do have bigger issue. We are not analyzing the system performance
at all. Let's go all the way back to the early days of 802.3 and
what analysis was done back then. This same BER and FCS subject
have been discussed at every task force -- only the players are
One of the goal for a physical layer is ensure that there is no
undetected data corruption. The requirement for 802.3 was that
the undetected data corruption rate is so low that system performance
approaches this goal. With the low BER, say 10**-9 (physical link
BER), 32 bit FCS provides reasonably high protection from undetected
error. As you all know that the 32 bit FCS used in IEEE 802 is fairely
robust in detecting single bit error, although its protection is frame
Now this consideration changed when we started doing 4B/5B, etc, coding
where single bit error in a symbol could cause multiple bit changes
in the decoded bits. FCS used in IEEE 802 is not as robust as before
because the robustness of the FCS algorithm is greately decreased as you
increase the # of error bits in a frame. So this was addressed and
deemed good enough by the commitees who wrote standards using the symbol
coding schemes so far.
My recomendation is to deal w/ this issue systematically in the commitee
w/ proper attention and come up w/ the total solution that deals w/
this requirement that Layer 1/Layer 2 provides a reliable trasmission
medium -- this, together w/ the DC balance issue.
Rohit Sharma wrote:
> Its hard to see whether such a "knob" would do anything for optical gear
> (Tx, Rx, amps) under operation. I agree with Drew that the most one can
> gain is looser specs on the design end...
> Also, remember that the difference between a 10^-12 and 10^-10 BER is merely
> a few dB at best (in absence of floors) if you look at the 'waterfall' BER
> curves. Once the Tx, Rx, amps are designed (and installed) for a certain
> BER objective, operating them for a worse BER (e.g. reducing the launch
> power at the transmitter) doesn't result in cost savings.... so the "knob"
> is fairly static in any case.
> Optical Networks
> > -----Original Message-----
> > From: Drew Perkins [mailto:drew.perkins@xxxxxxxxxxxx]
> > Sent: Friday, June 04, 1999 10:24 AM
> > To: 'bill.st.arnaud@xxxxxxxxxx'; 'Larry Miller';
> > 'stds-802-3-hssg@xxxxxxxx'
> > Subject: RE: Wide Area Networking for the Rest of US - the
> > debate on BER
> > a nd other issues
> > Bill,
> > I concur completely with your analysis of BER vs. RED.
> > But I still
> > don't believe that a BER "knob" would be very useful to anyone. BER is
> > really a side-effect of controlling something else, so you
> > really can't
> > control BER directly. The other things that could be
> > controlled include:
> > output launch power, loss (distance) between transmitter and receiver,
> > receiver sensitivity, bit rate, and probably a few other
> > things. Some of
> > these things, such as output launch power or receiver sensitivity, are
> > controllable at the time equipment is designed. Thus I don't
> > see these as a
> > "knob" so much as simply a looser spec. Other things, such as loss
> > (distance) between are in the direct control of the end-user, but are
> > fairly static. I don't see these as "knobs" either.
> > How are you defining "knob"? Is it a setting that can be controlled by
> > network management? Is it a design parameter that can be
> > controlled by the
> > equipment designer? The network designer?
> > Drew
> > ---------------------------------------------------------
> > Ciena Corporation Email: ddp@xxxxxxxxxxxx
> > Core Switching Division Tel: 408-865-6202
> > 10201 Bubb Road Fax: 408-865-6291
> > Cupertino, CA 95014 Cell/Pager: 408-829-8298
adr:;;107 Bonaventura Drive;San Jose;CA;95134;USA