>On Sun, 27 Sep 1998, Thomas Sailer wrote:
> A friend of mine here at work was intrigued by this idea and has done
>some investigation, using the cheapest tape deck he had. Unfortunately,
>most of these multi-MB storage calculations for audio tape are unfounded.
> His cheapo tape recorder loses phase for mid-to-high frequencies. This
>kills QAM/TCM coding really fast. The automatic volume control (AVC)
>distorts amplitude in a totally nonlinear fashion. The stereo coding in
Did he try to deal with AVC by using the most significant bit as a clock
bit, i.e. toggling it for every other sample? This should give a strong
signal all the time, lessening teh effects of AVC considerably.
>modern tape decks steals bandwidth. About the only thing reliably
>preserved is the frequency distribution. At present, with a bit of work,
>he estimates he can get at most a couple of megabytes on a 90 minute
>tape.
> A better tape deck might have better qualities, but it'd probably take a
>*lot* of manual tuning for each individual tape deck to handle the wide
>variety of hardware quality out there.
Tuning will be necessary in order to get maximum bandwith from generic
recorders. But surely it can be done automatically:
Record a nice long "test pattern" that have all sorts of problem bit
sequences. Then read it back 10-20 times. Let a program analyze the
stuff and see how much this particular recorder destroy data and in what
ways. The program can then set up a config file with how many bits per
sample and what sampling rate is safe for that recorder. And any other
parameters relevant to the coding used.
Reading data back with a higher sampling rate may allow better recovery.
> A pity, but possibly sound-card networking might work a little better,
>given decent cables. One limitation is that SoundBlaster hardware (the
>most common type of sound card from what I can see) either isn't capable
>of full-duplex, or can only do 16-bit one way and 8-bit the other. This
>means you could have a two-way 8-bit channel, or a 16-bit downstream and
>an 8-bit upstream.
This would have to be user selectable then. Not all of us have
soundblasters. :-)
Helge Hafting
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/