finally finished my PT80

[Brian Marshall]

ok, so if we have 44k, and each byte represents one sample.  if we set our delay time at 1 sec we would have a 44khz sample rate (or 35.2 if the bytes referred to are 8 bit) so arent we somewhere near CD quality?

anyways, i am pretty sure this delay is rolling highs off well below 15k, probably more like 10k, so even a 35.2 sample rate should be acceptable.

I seem to remember a lot of digital processors from teh 80's having sample rates arround 32 khz......   What am i missing?

[Mark Hammer]

ok, so if we have 44k, and each byte represents one sample.  if we set our delay time at 1 sec we would have a 44khz sample rate (or 35.2 if the bytes referred to are 8 bit) so arent we somewhere near CD quality?

anyways, i am pretty sure this delay is rolling highs off well below 15k, probably more like 10k, so even a 35.2 sample rate should be acceptable.

I seem to remember a lot of digital processors from teh 80's having sample rates arround 32 khz......   What am i missing?

I'm certainly not the digital guy here, but it is important to consider that, ultimately, *usable* bandwidth is a function not of the sample rate but of what you have to do to eradicate aliasing stemming from the sampling process.  For instance, if we had a delay sampling at 50khz with a 4-bit resolution, the stairstep-like nature of the resulting sampling, even though higher in sampling *frequency/rate* than CD standard, would be so gritty as to need serious lowpass filtering to get it to sound reasonable.  I have no idea what the formal equations are, but I am certain they are packed away somewhere in a back issue of JAES (Journal of the Audio Engineering Society, a fabulous read if you're ever in a university library).  What we need to understand at this point is that higher sampling rates (and the RAM capacity that allows them to turn into longer delay/storage times) assist a given A/D resolution in yielding better sound, but do not, of themselves result in high quality sound.

I will repeat my earlier comment that I have no idea what exactly the PT2399 is packing 44k *of*.  If it's 44k of 10-bit memory locations to complement the 10-bit A/D resolution, then we're talking.  If it is *nominally* 44k of 8-bit RAM, and the 10-bit sampling process turns that into something of a functionally lower capacity, then that is a separate matter.  Itis also worth noting, with reference to thepreceding paragraph, that 10-bits is not God's gift to high fidelity, and faster sampling rates (which reduce what that same memory can deliver in terms of delay time) maybe the only way to make it yield reasonable sound quality.

As for the clock-frequency-equals-double-the-max-audio-frequency, there are realistic limits to that. which are never articulated terribly well  While the limits of human hearing make it such that we probably cannot easily tell the difference between an 18khz tone represented by a simple on-off/high-low pulse and the original source, you can bet your sweet ass that once you move down into the sub-10khz range it starts to become VERY noticeable, and reasonable reproduction of those frequencies (and lower) requires something muchmore than merely a sampling rate double the maximum audio frequency in the passband.  Indeed, attempting to represent/reconstruct a 1000hz tone by anything less than *maybe* 10khz sampling rate is simple foolishness.  

This gets even more complicated when you realize that there are no guarantees that any portion of the signal is aligned with the sampling process in time such that all the snapshots the sampling process is *able* to take are taken at those times that will faithfully represent the waveform.  Consider the foolish example of sampling a 2hz signal at a 10hz rate.  If the first sample isn't taken at the "right moment" you've essentially missed it.  That 2hz waveform could come at ANY time and you'd miss it.

The only way to assure faithful reconstruction without being to guarantee the wave starts only when you're "ready" for it is to up the sampling rate again so that you're just sampling like crazy in the assumption that you'll always cover it.

This is all the long way of saying that 44k of RAM may sound like a lot, but when it's a 10-bit word, and as complicated a signal to reconstruct as an audio signal, 44k doesn't stretch very far unless you rein in bandwidth a LOT..

[Brian Marshall]

sometimes a 16 bit audio sample, if at lower volume can be only working with 10 bits of bandwidth.....  I think that is why the compander is useful.... it make use of as much bandwidth as possible.

your right though about digial timing an wave form peaks, but the artifacts created by that are sub harmonics, and cant be filtered out with a multipole low pass filter.  they really cant be filtered out by anything, because they are so unpredictable.  as your sample rate goes up subharmonics reduce significantly, but mathematically they are always there.

I actullally have a pretty decent digital recording studio, and if you record somthing at 8 bit, and 16 bit, and 24 bit at 41KHZ you can tell a difference between all of them.....

If you compress before you record it is much less noticable.

If you cut all the highs about 5 K or so, you cant even hear the difference..... but if you reduce the sample reat to say.... 22khz. you cant filter out the nastyness.

that is my experience anyways.

brian

[rocahopo]

Brian,

I am about to start on a PT80 build, but I cant loacte any source for the PT2399. I am in Europe but I jave searched far and wide.

No sources whats so ever.

Do you know of one that will ship to France.

Cheers,
Robert