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

Re: Framing packets over the link




Hello Rich and Hon,

There is a Multilevel Signalling approach based on the 1000BASE-T
PCS and 4-WDM, using a baud rate of 1.25 Gbaud/sec in the optical
fiber (the "10G-BASE-T" proposal). This approach optimizes the
reuse of the installed, bandwidth limited, optical fiber used in 1 GbE.

The corresponding entries in the table would be:

        1.1981 Gbaud/sec   and 1.25 Gbaud/sec

for signalling rates of OC-192 and 10 Gb/s, respectively.

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


Rich Taborek wrote:

> Hon Wah,
>
> The purpose of this note is to complete the entries for multilevel and
> multichannel signaling which don't appear to be complete in your note:
>
>                       Fixing the Signalling           Fixing the Throughput
>                       rate at OC-192                 Rate at 10Gb/s
>
> Multilevel           4.79232 GBaud*               5.0 GBaud**
>   signalling
>
> Multichannel      2.39616 GBaud/                 2.5 GBaud/
>   signalling         2.9952 GBaud***               3.125 GBaud****
>
> All entries assume an OC-192 PAYLOAD rate of 9.58464 Gbps and not any true
> OC-192 line rate, since the current HSSG speed objective only considers a
> MAC/PLS rate of ONLY 9.58464 Gbps OR 10 Gbps. There is no HSSG objective to
> consider a line or signaling rate.
>
> * Given an OC-192 MAC/PLS rate of 9.58464 Gbps, the fully encoded MAS line rate
> proposed is half the MAC/PLS rate, or 4.79232 GBaud.
>
> ** Given a MAC/PLS rate of 10 Gbps, the fully encoded MAS line rate proposed is
> half the MAC/PLS rate, or 5 GBaud.
>
> *** Given an OC-192 MAC/PLS rate of 9.58464 Gbps, the unencoded multichannel (4
> x WWDM or parallel optics) line rate proposed is one fourth the MAC/PLS rate, or
> 2.39616 GBaud per channel. The 8B/10B encoded multichannel (4 x WWDM or parallel
> optics) line rate proposed is five fourths the unencoded  rate, or 2.9952 GBaud
> per channel.
>
> **** Given a MAC/PLS rate of 10 Gbps, the unencoded multichannel (4 x WWDM or
> parallel optics) line rate proposed is one fourth the MAC/PLS rate, or 2.5 GBaud
> per channel. The 8B/10B encoded multichannel (4 x WWDM or parallel optics) line
> rate proposed is five fourths the unencoded  rate, or 3.125 GBaud per channel.
>
> Best Regards,
> Rich
>
> --
>
> Hon Wah Chin wrote:
>
> > We still have to consider how to format and frame the packets
> > once we get the speed of the MAC/PLS settled.  Here is a start for a
> > comparison
> > table (read with a fixed width font) for some schemes.
> >
> >                 Fixing the Signalling           Fixing the Throughput Rate
> >                 rate at OC-192                  at 10Gb/s
> >                 (get to use regenerators)       (Might have to do new PHY/PMD)
> >
> > FRAMING
> >
> > Manchester      approx 5Gb/s throughput         Must signal at 20Gb/s
> >                         (sucks)                         (really hard)
> >
> >           -- tradeoff about fast locking PLLs (hard)
> >                 vs some scheme like running quarter speed transitions
> >                 to indicate "no carrier"
> >           -- This would not be a serious contender
> >
> > 8B/10B          approx 8 Gb/s throughput        must signal at 12.5 Gb/s
> >
> > PPP/POS         variable throughput             variable througput
> >                         10Gb/s - some delta             10Gb/s +- some delta
> >                                                 (depending on signalling rate)
> >
> >           -- some kind of scrambling required to reduce chance
> >                 of long runs
> >
> > Header/Length   10 Gb/s - epsilon throughput    Can be 10 Gb/s throughput
> > (SDL et al)
> >           -- would require changing MAC/PLS definition to provide
> >                 packet length early
> >           -- framing acquisition logic required - parallel CRC
> >                 checkers to search for byte/packet alignment
> >           -- some kind of scrambling required to reduce chance
> >                 of long runs
> >
> > Multilevel                     NA                       ?
> >   signalling
> >
> > Multichannel                   NA                       ?
> >   signalling
>
> -------------------------------------------------------------
> Richard Taborek Sr.    Tel: 650 210 8800 x101 or 408 370 9233
> Principal Architect         Fax: 650 940 1898 or 408 374 3645
> Transcendata, Inc.           Email: rtaborek@xxxxxxxxxxxxxxxx
> 1029 Corporation Way              http://www.transcendata.com
> Palo Alto, CA 94303-4305    Alt email: rtaborek@xxxxxxxxxxxxx