10G-BASE-T SNRs

[Jaime Kardontchik <kardontchik.jaime@xxxxxxxxxxx>]

Hello 10G'ers,

I would like to thank from my heart the HP researcher
that sent me yesterday evening a private email pointing
out a mistake I made in my SNR calculations.

I replied to him and he was very kind to send me back a
very detailed explanation of where my mistake was and
how to calculate correctly the SNR.

Briefly speaking, I mixed up "optical SNR" with
"electrical SNR"

Although I still have to do a lot of learning, I prefer
not to delay his main conclusions. Here they are, if
I understood correctly what he explained to me:

1) Step 1: calculate the optical SNRs

The "optical SNRs" of the architectures are:

1) 1000BASE-X                                - 1 dB
2) 8b/10b + 4-WDM                            -9 dB
3) PAM-5 unencoded + 4-WDM        -7 dB

("unencoded" means before any coding gain)

2) Step 2: convert to electrical SNRs:

1) 1000BASE-X                                    -2 dB
2) 8b/10b + 4-WDM                            -18 dB
3) PAM-5 unencoded  + 4-WDM        -14 dB

3) Step 3: add electrical coding gain

1) 1000BASE-X                                    - 2 dB
2) 8b/10b + 4-WDM                            -18 dB
3) 10G-BASE-T - EVEN coding    -14 + 3 = -11 dB
4) 10G-BASE-T - Trellis coding    -14 + 6 = - 8 dB

4) Step 4: convert back to optical units to get
    the correct SNR table that replaces the one
    I showed in my presentation:

1) 1000BASE-X                                    -1 dB
2) 8b/10b + 4-WDM                            - 9 dB
3) 10G-BASE-T - EVEN coding        - 5.5 dB
4) 10G-BASE-T - Trellis coding        - 4 dB

                *****************

I enjoy being a member of IEEE Task Forces. It is
for me the best learning experience to have the
opportunity to interact with technical people,
present  not completely debugged new ideas and
listen to their inputs and corrections.

Jaime

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