From: Billy Biggs
Subject: Re: Studio-grade hardware support?
Date: Sat, 13 Apr 2002 19:14:49 -0500
I really appreciate the reply, BTW.
Hi Matt,
To start, check out http://www.linuxmediaarts.com/ for a company
selling an SDI card that seems supported under linux, along with some
other 'high end' goodies. But the cost is prohibitively expensive to
end-users, since we're not the market.
Yeah, I had run across them, but it's a bit outside of my budget.
Even if manufacturers provided this information, what linux developer
could a) afford the cards, or b) have equipment to use them? I know I'd
I'd drop a grand on a used piece, like a digisuite card, if I had enough
information to write a driver for it. I personally know the designer of all
the Media100 hardware (I almost went to work there), so if I could score one
of their cards, I should be able to get enough info to write a driver.
Well the V4L2 API (http://www.thedirks.org/v4l2/) is supposedly good
enough for what it wants to do: provide an API to consumer-level video
equipment. It has support for timecode, but I don't think anything
implements that (but you can read VBI data and correlate it with video
frames, and I wrote a timecode decoder for this API). The API has no
support for controlling genlocks, like if you wanted to genlock two
Yeah, it worries me to see stuff that no one has tried to push into the
studio. I know that a staff of over 100 hacked away on Avid's Media
Composer, for about a decade, and there are just some issues you can't
foresee until you stumble into them. But some of their effort is due to the
code being un-modular and poor reuse. Then, consider what platform they
developed it for (and what they ported it to). Also, you have the issue of
"professional" software quality, to contend with, too. So, maybe it's not
that hard. (I didn't work at Avid, but know several people who have.)
Audio sync is a bit funny, what hardware did you want to support? The
V4L2 API doesn't speak at all about audio, but I'm not sure what
hardware you'd be talking about here anyway. For field-correct output
sync is also a bit weird, but again nothing supports the TV output API,
so I think this is mostly untested.
Well, I'm no expert, here, but it seems like you should at least know the
latency of the audio and video capture cards. Then, you also want to be
able to deal with drift, in case the two clocks aren't in synch.
A co-worker who used to work in Media100's support department (I didn't work
for them, either, BTW) once told me a story about the fact that they
captured audio at the correct rate. The players of the time weren't written
to account for the fact that video was 59.96 fps (or some such issue), and
the video & audio would get badly out of synch, after a few minutes. They
got continual tech support calls, until they finally decided to ship their
own player that got it right.
I guess, if V4L knows nothing about audio, that this must be solved by the
existing apps, because it sounds like most users use a separate sound card
that's not synched to the video, in any way. Do you know what they do to
address this? Perhaps cross-fading, at every field boundary, to resynch
with the latest samples from the sound card? If the two aren't genlocked,
then you probably can't do anything much better, but I wonder if that could
ever result in an audible 60 hz buzz.
recording. Most PVR efforts etc focus on recording every second field
at low resolution, like 253x240@xxxxxxxxx
That's a real shame. A high-quality PVR, with footage I can dump to DVD, is
one of my long-term goals.
That doesn't mean we don't care about quality, but I don't see a big
push to get some of the more higher-end code done, even on our consumer
grade hardware.
You might consider raising your standards, for hardware. The stuff is
getting quite cheap (see my next post).
Since most developers aren't working on V4L professionally, I don't
see why we'd want to target the post/broadcast environments. Sure
The more work it is, the more unlikely it'll happen. I hear things about
big 3D houses using linux, and I wonder about video. But then, high-end 3D
has been entrenched in UNIX for much of its history, whereas video was on
proprietary systems, and first appeared on Macs, when it hit the desktop
(Avid originally developed the Media Composer on an Apollo workstation. The
story goes that Apple saw it at a convention, and their quicktime team
showed up on Avid's doorstep, a week later).
> Note: I would be willing to contribute code & documentation to driver
> or API projects in line with the above goals. I have no (current)
> commercial interests in this, BTW.
So you're a home user and you want to do what?
Like you, I want a high-quality PVR, and I'm interested in video editing.
There's some video restoration work I want to do (some of my friends are
video collectors, and have irreplaceable content on VHS). I don't really
need genlock or timecode, but I insist on quality.
I also like hacking around with audio and video. I come from more of a
compositing/effects/3D background (I suppose I can probably work on
compositing & effects, since my non-compete agreement, from my previous job,
has expired (and they're out of business, anyway)). I think I may like to
do it full-time, again. I don't know how Linux might fit into that, but the
more professional uses and applicability it has, the better.
Linux has the potential to be rock-solid, and is very automation-friendly.
I see no reason why it can't be the platform of choice, for large media
outlets. Perhaps things are a bit closer in the audio world, where you
actually have things like cutting-edge, surround-sound streaming
demonstrations, at professional conferences, built on linux.
Anyway, as I'm sure everyone knows, it already has lots of use for embedded
video applications. Things should eventually reach a point where
semiconductor vendors provide linux drivers with all their chips. Then,
board vendors will have virtually no work, for their products to be
supported under Linux. (hmm... I guess the MPAA will have to put a stop to
that.)
Well, thanks for taking these ideas seriously.
Is there support for any uncompressed capture cards? Computers are getting
fast enough (I hear intel already has a faster-than-realtime MPEG4 encoder)
that it'd be nice to just capture uncompressed and do software encoding.
IMO, hardware encoding is quickly becoming an anachronism.
I don't have a whole lot of free time, but I am interested in contributing
to making linux the video platform of choice. Are there any lists of ways
to get involved in V4L development? I should note that I recently deployed
XML DocBook, at my job (including some scripts to automatically generate
parts of our reference manuals from comments extracted from source code), so
I could help with coding as well as documentation infrastructure.
If there aren't any uncompressed cards supported, I'd like to start there,
since it's closest to my immediate needs. (I refuse to do anything that
requires me to boot windows. Anything worth doing, including video capture,
editing, processing, and output, shouldn't require it. :-)
Matt
_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp.
