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

Re: PAM-5, what are your BERs ?




Hello Sean Leighton (Lucent),Ed Grivna (Cypress) and 10G'ers,

PAM-5 using 4-WDM at 1.25 Gbaud uses FEC with
zero overhead: there is no need to increase the signalling
rate by any amount. FEC decreases the BER from a raw
BER of about  10^(-3)  to a BER of 10^(-12).

For details, see my presentation in Kauai, Nov 99.
Part II, "BERs and RINs". A couple of graphs there
show explicitly the BERs with and without FEC as
a function of the SNR.

The simulations shown in Part II use an ideal model.
Part IV of the same presentation, titled "Coding Gain",
shows the actual coding gains that can be obtained
using a receiver analog front end with an 18-level ADC
at 1.25 Gbaud. The expected actual coding gain using
this ~4 bit ADC sampled at 1.25 GHz is around 5 dB,
compared with the ideal coding gain of 6 dB. Not bad.

The FEC capability is included in the PCS of the
1000BASE-T standard. The PAM-5 at 1.25 Gbaud
architecture I proposed reuses this PCS.

Jaime

Jaime E. Kardontchik
Micro Linear
San Jose, CA 95131

Ed Grivna wrote:

> Hi Sean,
>
> unfortuantely, FEC has additional drawbacks beyond just latency.  the
> primary concern here would be that you need to ratchet-up the signaling
> rate even faster to include the overhead of sending the FEC bits.
> Depending on the amount of redundancy in the FEC information,
> you may need to increase the signalling rate as much as 5-to-10%.
>
> Regards,
>
> Ed Grivna
> Cypress Semiconductor
>
> >
> > Hello,
> >
> > With regards to the BER problem, are we open to the possible use of forward
> error
> > correction to correct for bit errors in the transmission stream. Currently in
> > T1X1.5, there is a submission for inband FEC based on a BCH-3 code that turns
> a
> > 1x10**-7 BER into a corrected BER of 1x10**-17. Prehaps a BCH-1 code could be
> used?
> > Using FEC would introduce the need for some additional bits within the frame
> format
> > and it would probably introduce some latency, but it would introduce some
> coding
> > gain.
> >
> > Cheers,
> > Sean
> >
> >