Analog summing emulation idea
-
- KVRAF
- 2029 posts since 21 Jul, 2004
Allright, I just did a bit of reading. But aren't there frequencies produces that are not UPPER harmonics, but LOWER harmonics that get lost when converting the highest human-heard frequency to a square-wave through the Nyquist theorum? (I'll probably read this tomarrow and delete it, LOL!).
Do not lick the fablanky
-
- KVRAF
- 6478 posts since 16 Dec, 2002
I have no idea what you mean by that question. The answer is no, probably after the weekend, and definitely not on sundays. 
-
- KVRist
- 148 posts since 2 Sep, 2004
I've got a question which I think Aleksey or Kingston might be able to help me with - is the smallest phase difference that can be represented accurately, quantized by the bit depth? What I'm thinking is that the smallest phase shift would have to cause at least one sample to change by one bit. If this is true then low amplitude and low frequency sines would be more phase quantized than high amplitude, high frequency ones, wouldn't they? My rough calculations say that a 1khz full amplitude sine sampled at 16 bits would have a minimum phase resolution of ~0.036ms. Is this right, or have I missed something?
-
- KVRAF
- 4030 posts since 7 Sep, 2002
Samb, I do not think bit depth is closely related to phase precision. If phase difference is treated as a statistical entity, its resolution is pretty infinite. So, even if you quantize two sinewave that have a infinitely small phase difference to 1 bit resolution, at some point of time they'll change relative 1 bit pattern telling you that their relative phases were indeed different.
From what I understand, Fourier series analysis tells that phase measurement precision depends on how you measure it and how long your time series are.
However, on this very subject I may be wrong. It is just how I understand it.
From what I understand, Fourier series analysis tells that phase measurement precision depends on how you measure it and how long your time series are.
However, on this very subject I may be wrong. It is just how I understand it.
-
- KVRist
- 148 posts since 2 Sep, 2004
Thanks for your reply Aleksey, I'm sure you understand these things a lot better than me.
Now you point it out it's obvious I was over-simplifying, but I think my example would still apply (theoretically anyway)for sines that are perfect divisors of the sample rate?
As you say, at some point in time the bit pattern will change, so I suppose in practice this depends on how well the D/A converter comes to being perfect.
Now you point it out it's obvious I was over-simplifying, but I think my example would still apply (theoretically anyway)for sines that are perfect divisors of the sample rate?
As you say, at some point in time the bit pattern will change, so I suppose in practice this depends on how well the D/A converter comes to being perfect.
-
- KVRAF
- 4030 posts since 7 Sep, 2002
Samb, this has nothing really to do with the D/A. If the original analog waveform had this tiny phase difference, D/A will represent it in any case - maybe you just won't notice it immediately, but this phase difference won't go away just because of D/A bit depth resolution (look at 1 bit example). So, phase information is never lost at any bit depth and it is always exact.
However, if you are trying to represent phase information as a static delta value, it will be, of course, giving a progressive error or may turn to 0 if it is too small. When you sample, this is not the case: due to dithering delta value may become 0 and then may become 1.
However, if you are trying to represent phase information as a static delta value, it will be, of course, giving a progressive error or may turn to 0 if it is too small. When you sample, this is not the case: due to dithering delta value may become 0 and then may become 1.