PAM-5: Tx waveshaping at 5 Gbaud
Hello 10G'ers,
In their presentation:
"10 Gb/s PMD using PAM-5 Trellis
Coded Modulation"
by: Oscar Agazzi, Nambi Seshadri
and Gottfried Ungerboeck
March 2000
The authors (in what follows, "Oscar et al")
add Tomlinson-Harashima Precoding in the
transmitter in order to lessen the burden
on the receiver analog front end and the
receiver equalizer.
The basic idea of Tomlinson-Harashima is
simple: provide elaborate shaping of the
launched waveform such that the transmitted
waveform can travel through the optical fiber
almost undisturbed (low introduced distortion).
Or, in Oscar et al's words: "the precoder is
essentially a filter with a response equal
to the inverse of the channel".
In this way the receiver could be a less
challenging circuit that could process and
recover the original raw bits of data.
Towards this end, the original PAM-5 symbols
are transformed, before transmitting, into
15 possible levels ("expanded alphabet" in
Oscar et al's nomenclature) using a 5 GHz DAC
and a 5 Gbaud laser driver so that an adequate
shaping of the waveform can be achieved.
A similar idea was presented almost a year
ago by Transcendata. It was coined "T-waves".
However, perhaps the most difficult task
in this kind of approaches is the synthesis
of this waveform in the analog domain, since
it involves not only a 5 GHz many-level DAC
but also the subsequent analog 5 Gbaud laser
driver.
It is important to note here in this respect, what
Mike Wincn, a former member of Transcendata,
wrote on the Reflector a few days ago:
> ... finally, there are probably many ways
> one can build a fast ADC but the more
> difficult challenge - for 10 Gbd x-PAM -
> is in building a fast linear DAC."
(email to Reflector, March 1st, "PAM-5 what
are your BERs", Mike Wincn, March 1st, 2000)
It seems that it does not matter where you
try to put the processing, either in the receiver
or partly in the transmitter and partly in
the receiver. The results will be similar:
a PAM-5 serial at 5 Gbaud claiming to support
500 meters over installed MMF (or even 300 meter)
is not reallistic in present technologies.
Jaime
Jaime E. Kardontchik
Micro Linear
San Jose, CA 95131
email: kardontchik.jaime@xxxxxxxxxxx