🚀 go-pugleaf

In article <830610958.20155.0@ttelmah.demon.co.uk>
           roger@ttelmah.demon.co.uk "Roger Hamlett" writes:

> Actually, when digitising, you have to consider that you are converting from
> an analogue 'bandwidth', to a digital sample. Hence the reason, why audio
> CD's, use 44kHz sample rate, for 'only' a 20kHz used bandwidth. The comments
> about moving images, are _very_ right. However, I would have to say, that a
> sample at 220 pixels/line, definately 'throws away' some of the useable
> information from VHS. Sampleing at @320pixels/line, seems to give a real
> resolution that is as good as you can expect from VHS.

This starts to get really complicated. Yes, to totally reconstruct the 
original signal without hitting the Nyquist limit requires a sample rate of 
double the maximum frequency to be recorded (and even then the response rolls 
off at the higher frequencies). However, since we are recording the 
encoded information rather than the whole signal, some technical 
jiggery-pokery allows you to store the data at the actual data rate.

The 220 pixels/line implies ignoring any phase difference. The phase 
differences (i.e. the elements in the image may not lie exactly on the pixel 
centres - another manifestation of Nyquist) allows a higher perceived 
resolution (to 440) to be obtained.

> > What would also be useful is some help to compute what sort of bandwidth
> > is required to capture full PAL. I reckon it's over 5MB sec to disk which
> > means a going for something like a 4GB barracuda which does not thermally
> > recalibrate, hog CPU resources, and can sustain sequential throughput at
> > over 7MB/sec.

Well over 5MB. Standard PAL-I is (I've looked it up) 5.5MHz bandwidth. Full 
signal digitisation at one byte/sample should be at least 11MB/s. Ouch.
I can easily understand how higher bandwidth, higher bit-depth uses 20MB/s.
This is actually nowhere near the information rate. Even without using fancy 
MPEG compression schemes, a simple hardware implemented differencing 
algorithm or similar should be able to cut this to 5MB/s with absolutely no
loss of information. I would recommend one of the AV drives that several 
manufacturers do, they really are optimised for long data streaming tasks, 
and some sort of RAID system is almost inevitable.

Mike