Recent Finds & Thoughts

From: Philip.Belben_at_pgen.com <(Philip.Belben_at_pgen.com)>
Date: Thu Nov 26 05:59:56 1998

> I have many old computer cassettes, and have been thinking of recording
them
> onto a computer in order to preserve their contents. The signal from
computer
> tapes oscillates between two levels, right? This being the case, it
should be
> possible to record them using 1-bit sampling. Perhaps record with 8- or
16-bit
> sampling and then convert down to single bit.
This is one of those things that depends on the hardware. The PET and its
descendants did indeed use one bit in that manner. There are quite full
descriptions of the waveforms it generates in several manuals.

One-bit sampling is not necessarily a good idea. What you then have is:

PET writes to tape nice square signal of well defined timing.

Signal comes off tape with rounded corners, noise etc.

You sample it at 1 bit and write to a file.

Later you read the file on a modern machine and reconstruct the audio.
This is a nice square signal, with timing dependent on (a) the distortion
of the original recording/playback process, (b) the noise added by tape
ageing etc., (c) your sampling at 22 kHz.

It goes through some audio system with not necessarily a good bandwidth and
arrives at the PET cassette port. The PET then has to interpret it...

Worse still if you're using an audio CD. The CD player will be filtering
according to what's best for the human ear, not what's best for the
computer.

My advice: If you're going to do 1-bit sampling, process _immediately_ to
try and reconstruct the original signal. That is what the machines do
(often in hardware), after all.

Then remember that some machines _don't_ use 1-bit encoding. The BBC
micro, for example, uses a standard modem chip of the time, and the outputs
look (I think) fairly sinusoidal...

> Are there any programs to do this conversion? I imagine the equivalent of
a
> Schmitt trigger (in software) would work. What about playing back a 1-bit
> audio signal? Are there any standard audio file formats that can be used
to
> store 1-bit data?
>
> The final 1-bit audio files should be highly compressible, so they could
be
> archived with zip etc. to reduce size.
Agreed. 1-bit sampled audio - or multiple bit sampled audio processed to
reconstruct the 1-bit waveform - should be many times more compressible
than ordinary sampled sound. In theory, the compressed file need be only a
few times the size of the original program, but I doubt this will be
achievable.

> Actually recording tapes to audio CDs is quite wasteful since you can
only get
> 70 minutes or so on a CD (an issue if you have hundreds of cassettes). My
> approach would be to archive tapes as described above; of course burning
an
> audio CD is useful for transferring the data back to the source computer.

I agree with whoever pointed out that you can use the two tracks of the
stereo audio for different programs. Cross talk is negligible. That's 140
min. After all, those of us with reel-to-reel tape have been doing this
for years...

Audio CD is necessary in my opinion. The point of this exercise is to plug
your computer into your sound system and load programs _without_ the need
for a modern machine to retrieve archives and things.

If it wasn't for copy protection, turbo loaders and the like I'd favour
sampling the output of the computer's cassette port...

Philip.

PS PET and many later C= machines have one more problem: they didn't use
standard audio cassette machine, but one with a special Commie board in it
and a custom interface.
Received on Thu Nov 26 1998 - 05:59:56 GMT