[mpeg2] Driver project: a new 4-channel MPEG decoder card

[Warren Young <warren@etr-usa.com>]

I just found some very interesting things out regarding multi-channel
MPEG decoder cards for Linux.  (These are commonly used in digital video
servers.)

A company called Visual Circuits Corporation (VCC) has an existing
4-channel MPEG decoder that works under Linux.  We're using their boards
in existing products, and they work reasonably well.

Our main problem with VCC has been support.  We've gotten very mixed
signals from them in the past regarding OEM support, and unfortunately
they're not interested in opening up the source code for their driver to
allow the community to support it.  By "mixed signals" I mean that we've
spoken with various people at VCC and some will repeatedly profess that
they will help us in any way possible, and others will tell you that VCC
has a hard time supporting OEMs because they're busy working on their
own digital video server line.

I'm not faulting VCC for this, since building their own digital video
servers is a better business model for them.  Meanwhile, however, this
uncertainty means we must keep our eyes open for alternative decoder
cards.

Recently, VCC has been working on a new 4-channel decoder which will be
cheaper than their present one.  The new board is based on Sigma
Designs' EM8400 chip; Sigma already has a Linux-compatible 1-channel
board based on that chip that costs US$220 street, so even after
factoring in the lower likely sales volume, a 4-channel board based on
the same chip should cost around US$1000.  The current VCC board is a
bit over US$2000.

What I found out today is that the new VCC board is actually designed by
Sigma and simply resold by VCC.  Once the exclusive marketing agreement
between Sigma and VCC for this board expires -- very soon, I'm told --
you will be able to buy the board from Sigma as well as VCC. 
Unfortunately, Sigma didn't write the driver for the board, VCC did. 
Given VCC's current treatment of the OEM channel, it's possible that VCC
won't provide the new driver to OEMs, in either closed or open source
form.  (I don't know this for a fact, it's just a possibility I'm aware
of.)

I said that Sigma already has a 1-channel decoder with a Linux driver. 
I suspect that this driver could be reworked to control the new Sigma
4-channel board, which will be available from both companies once the
exclusive marketing agreement expires.  I don't have the skills to do
this conversion, though I suspect if pressed I could acquire them.  I
once helped write a simple parallel port bit-banging driver for a
different *ix and then later ported it to Linux.  I'm sure that writing
an MPEG decoder card driver, even given the source to a simpler working
one, is a far more difficult task.

Ny company still has options for decoder cards: VCC is still giving us
some support, and recently a company called Enseo put out a Linux driver
for their cards.  Only if those dried up would I have a pressing case to
spend my company's time learning how to write device drivers.  At the
moment, saving $1000 per card isn't a good enough reason for me to spend
development time writing a driver for the card; our volume just isn't
high enough and I have many other things that need doing anyway.

That said, my company would be willing to help in a group effort to
write this driver.  At the very least my company would be willing to buy
one or more of these boards and to help test drivers out.  It's possible
my company would let me help out on the driver some, too, but I'm sure I
couldn't go it alone.  

Anyone interested in this project?
-- 
= Warren -- Video articles: http://www.cyberport.com/~tangent/video/
= 
= ICBM Address: 36.8274040 N, 108.0204086 W, alt. 1714m