close-circle
Close

Still alive + infinite sample resolution ponderances

A project log for Esot's Trenches

True tales (Trigger Warning!)

esot.eric 10/18/2017 at 01:182 Comments

nothing too much to update... really just not feeling too talkative at the moment... finally installed the new heater core... so cozy and sleepy. Also figured out why the vents smelled of engine odors, the air intake had a *huge* hole in the seal... so was sucking almost as much from outside as from under the hood!

Spent a day on the floppy-audio project...  been on the mind for years. playback is as simple as tying a few pins together (e.g. index to step) and hooking a speaker to the read-data pin. Recording is a bit more difficult. But I managed to make a line-in to pwm converter from a couple comparators... was way easier than I'd've thought, being so used to microcontrollers. So have been contemplating comparator-logic to handle some of the other timing-intricacies. E.g. index can't be tied directly to step when the write-gate is open. That... could be a problem requiring replacement of the original circuitry, which could be interesting in its own right (e.g. constant microstepping for spiral recording rather than concentric circles, spindle speed-control, etc....) but so far the idea is a plug-in that uses the original circuitry so any drive could do it... (otherwise... skip pwm altogether and record in analog! But... that was the original plan... long ago... so obviously plans change.)

The use of pwm... hadn't really thought about it before, but it *can* be a discrete-time, infinite-sample-resolution (forget the official name) system... CDs being discrete time And discrete sample-resolution (44.1ksps, 16bits/sample) pwm could make for 44.1ksps with infinite sample resolution. There was a whole class before dft on the subject of the math behind that... which seemed weird 'cause I couldn't think of such a system. (Was that the difference between laplace and fourier?) Well, a decade later, here it is. Another example might be those bucket-brigade analog delay chips, also digital camera ccds... sample-and-hold circuits, and maybe vcrs? But... most of those are fed into an adc so the sample-resolution is discretized before stored. But storing pwm on a magnetic media wouldn't discretize the sample resolution. Huh.

E.g. a floppy drive can reliably record at 500kbps (I think that's *transitions* rather than data bits, so with mfm would be 250k data-bits per second). Storing transitions any closer than that can cause the magnetized particles to repel each other, as I understand)

A cd, in mono (one channel) is 44.1ksps * 16bits = 705kbps... too fast for a floppy, right? But the two transitions per pwm cycle is no more than 88.2ktps (t transitions)... so way less than the 500ktps limit. Then the only limitation being that both sides of the pulse have to be longer than 1/500000 sec... so bump a zero value to say 30% duty cycle and a 100% value to say 70% duty cycle... hmmmm...

Taken that far... the sample rate/pwm-frequency could be much higher, as well... hmmm. A flip-flop might make for stereo... hmmm...

Then, I suppose there's a limit to the positional resolution based on the size of magnetic particles, etc... but surely much much smaller than the 500ktps limit... similar, I suppose, to the "resolution" limit from film cameras, and grain size... (have we yet come up with enough megapixels  and megabytes to rival that?)

(Heh, guess it's pretty obvious I turned down the heater a while back...)

...several flip-flops (a counter?) Could allow for interleaved samples such that the spindle could rotate several times on the same track... hmmm... so, say cycling through each track five times... that'd give one second per track, and 160 seconds per disk. Way more complicated playback system, and nearing the 500ktps limit... ...

Discussions

Dr. Cockroach wrote 10/18/2017 at 01:39 point

That project sounds promising and glad your able to get heat now :-)

  Are you sure? yes | no

esot.eric wrote 10/18/2017 at 01:58 point

thankya sir!

  Are you sure? yes | no