Date   : Mon, 12 Feb 1996 17:44:50 +0000 (GMT)
From   : Roger Walker <Roger.Walker@...>
Subject: Re: Decoding BBC tapes...

On Mon, 12 Feb 1996, Robert Schmidt wrote:
>
[<Discussion about sampling BBC tapes cut>]
> 
> I didn't get this.  If you sample at 2400 Hz, you should be lucky to detect
> anything at all.  The "0" frequency is exactly 2400 Hz, so your sample can
> (in theory) be all zeroes.  If you're lucky, "0" will show as "++" (or
> "--"), while "1" will show as either "+-" or "-+".
> 
> But actually it's not too bad an idea, but I would rather sample at a
> greater multiple of 2400 Hz, maybe 4800 will do, though 9600 sounds quite
> safe.  This would reduce the amount of data to process substantially - I
> guess 44100 is not really neccessary.  I wish I knew more of the theory
> behind all this.
> 
> robert
The BBC tape format used two frequencies, either 1 cycle of 1200Hz for a 0
or 2 cycles of 2400Hz for a 1 (both of which take the same time).
Shannons sampling theorm specifies that you should sample at greater than
twice the maximum frequency you want to sample (i.e. 4800Hz).  But because 
we don't need to look too closely at the signal just detect 2x2400Hz or
1x1200Hz you could actually drop to a lower frequency (say 3KHz).

Sampling at higher frequencies will probably make the software easier to
write, to reduce the amount of data to be stored/processed the sample
accuracy could be dropped to 8 bits (lower if the soundcard allows).

Scan the sample and look for -ve slopes (or +ve) and look at the time between
them.

Good luck, Roger Walker.