Please don't quote Gilder as a technical reference, his prose are
intentionally too hand sweeping. I like the Shaw reference and I'll
comment on that.
Here's some facts.
1) Motorola and Com21 have been running on all coaxial plants just
fine. All coax is not reserved by Terayon. I pointed this out in
one of my earlier notes. Please stop quoting that the
others you mention cannot. It's a false statement.
2) The DOCSIS requirement for upstream noise floor is 25dB S/N or
better.
3) The Com21 specification for upstream noise floor for it's
proprietary system is 16dB S/N or better. That means spec-wise,
Com21 can work in a plant that is 8 times more noisy than DOCSIS.
Com21 has a *very* well engineered upstream demodulator in the
ComCONTROLLER. It is superior to DOCSIS, but that's another story.
See com21.com
4) An independent company measured Com21's performance down
to 12.9 dB before seeing packet loss in the upstream. This
information was presented to analysts as part of either the IPO
or secondary road show (I can't recall at this time.) Compared
to your reference, Com21 can continue at 100% with 0% loss where
Terayon begins to suffer at 15dB. Yes, S-CDMA can work in more
severe environments, but I'll talk about that later.
5) At the time Shaw replaced their "TDMA" modems, they were likely
using either Zenith or LANCity. Both of these suffer from the
lack of FEC and using a dumb head-end transverter. These have
the characteristic of taking any upstream noise and modulation
disturbance and echo'ing it right back down the downstream channel.
They can only run in very clean portions of spectrum. Hence the
35dB requirement for these types of "TDMA" modems.
6) Shaw and Rogers are both major participants with CableLabs
in the development of the DOCSIS specifications. I know that George
Hart, from Rogers, was personally very involved. They want DOCSIS.
7) Com21 uses a Reed Solomon fixed T=2 FEC encoding for each ATM
cell upstream packet burst. In contrast, DOCSIS is very flexible
and requires that cable modems be able to support Reed-Solomon
coding from T=0 (off) to T=10. T=10 is a lot of FEC. I'm assuming
that with this flexibility and under certain conditions, DOCSIS
upstream can do better than 12.9dB. I need to go find the analytical
curves for noise performance of QPSK and 16QAM with different FEC
encodings. Can't put my finger on it right now.
8) Yes, S-CDMA, like CDMA, has the ability to reduce data carrying
capacity in exchange for better error performance in the presence of
certain types of narrow band interference. With the S/N facts
presented above, the plants and/or portions of spectrum that
need this performance from the cable modem system are in the
diminishing nitch class. Diminishing, because as plants are tuned
and or upgraded, much more spectrum with S/N above 25dB (and 16dB)
are created, paving the way for the symmetrical system to be replaced
with a more economical system when the business plan of the cable
operator calls for it.
9) The statement I gather that S-CDMA in upgraded plant can
reliably operate while passing many times the homes per node
than competitors can. is not true. S-CDMA and DOCSIS are
more or less equivalent here due to how the cable operator sizes
their nodes and then combines upstream returns. There is a limit
on the number of upstream returns that can be combined that
is not a function of the cable modem equipment. I get at some
of the issues in my November 1997 CED article which can be
found at: cedmagazine.com
10) Back to your false statement. I suggest going looking at
Com21's Reverse Path Multiplexor (RPM) product. It was designed
specifically for HFC return paths and all coax return trunks. In
fact with it, a Com21 ComCONTROLLER can pass significantly more
homes on any plant than any other manufacturer, including TERN.
Note that there are other issues regarding the physical plant
topology that go beyond just cable modems and modulations.
com21.com
Mark |
