Re: [tlug] Questions about specifications for some cutting edge video recording via laptop

[Ian Wells <ijw@example.com>]

On 9/10/05, Dave Gutteridge <dave@example.com> wrote:

>The most important thing would be how long are the feeds going to run?

Each of the three feeds would probably need to be a maximum of 5
minutes, and most usually about one and a half minutes. If I could pull
this system off with each camera recording up to 3 minutes, that would
be acceptable.
Wouldn't the three feeds be writing directly to the HD? Might that not
require 3 hard drives?

15 minutes at 10MBit/s (DVD quality) would be 900MB - a big chunk of that could be buffered into memory if speed were the issue.  30MBit/s shouldn't exceed even laptop HD bandwidth, anyway - I'd imagine that 10MByte/s (80Mbit/s) is easily sustainable, but the test is easy and left as an exercise to the reader.  (A quick experiment on my modern Dell laptop with Windows 2k and Cygwin says you're looking at 19MByte/s but that's an entirely useless comparison ;-) .  You might require software that streams 3 feeds to a single file in order to avoid hard disk hunting when you're writing - dunno how good modern FSes are at avoiding hunting, although I know NTFS is crap from past experience. Such software is easy to write: e.g., wait for a full block to be received from one of the streams, write block to file, write stream identifier to control file.

The whole proposition depends entirely on what your cameras would send.  Firewire tops out at 800Mbit/s (80MB/s-ish) but I wouldn't expect a camera to use all of it.  There's bound to be some sort of compression but it's unlikely to be as good as a DVD, so that's the biggest variable, I reckon.

Firewire devices can, theoretically, be daisychained.  Bet cameras don't have the extra socket though.  USB ones could work over a hub but then you're probably limited by the availability of good cameras.


--
Ian.