Author |
Message |
asb2m10
Joined: Jan 08, 2006 Posts: 7 Location: Montréal
|
Posted: Fri Aug 15, 2014 11:48 pm Post subject:
Using a quantizer as noise generator |
|
|
I am currently writing a DX7 VST emulator and one of my reference implementation besides the DX7 is the g2ools converted patches. The FX section contains a "Vintager" modules that I am curious on what it does:
The 'noise' module is a note quantizer. This one I am not really sure what it does besides noise. The quantizer acts like a noise generator that gets excited (amplitude) from the output signal ? And the purple signal gets rendered at 22Khz ? So the quantizer is just a "economical" noise generator ? Why not just use a noise generator ?
There is also the "Fat" button mixer; the output is connected to the input. Does this means that the signal gets added one sample later ? What is the voodoo about this one ?
Thanks!!! |
|
Back to top
|
|
|
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
Posted: Sat Aug 16, 2014 2:00 am Post subject:
|
|
|
The note quantizer at this setting truncates the numeric resolution of the incoming signal to 7 bits, thus adding quantisation noise. And since it is one of those modules that cannot run at 96kHz audio rate (red cables), it also reduces the sample rate to 24kHz (control rate, blue cables) without bandlimited downsampling, thus adding aliasing. All in all, it's an uber-cheap low-fi generator, and a neat example of using G2 modules in unusual ways to save DSP.
The feedbacked mixer channel ("Fat") is the most primitive and economical LF shelving EQ in the digital universe. |
|
Back to top
|
|
|
asb2m10
Joined: Jan 08, 2006 Posts: 7 Location: Montréal
|
Posted: Sat Aug 16, 2014 11:52 am Post subject:
|
|
|
Thanks a lot for clearing this up. |
|
Back to top
|
|
|
varice
Joined: Dec 29, 2004 Posts: 961 Location: Northeastern shore of Toledo Bend
Audio files: 29
G2 patch files: 54
|
Posted: Sat Aug 16, 2014 5:09 pm Post subject:
|
|
|
Tim Kleinert wrote: | ...The note quantizer at this setting truncates the numeric resolution of the incoming signal to 7 bits... |
Hmmm… Are you sure about that (with the Notes parameter set to Off) _________________ varice |
|
Back to top
|
|
|
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
Posted: Sun Aug 17, 2014 12:48 pm Post subject:
|
|
|
varice wrote: | Tim Kleinert wrote: | ...The note quantizer at this setting truncates the numeric resolution of the incoming signal to 7 bits... |
Hmmm… Are you sure about that (with the Notes parameter set to Off) |
Oops sorry, lol , you're absolutely right. In the "Off" setting the numeric resolution is unaffected.
So in this case the module is only a cheap aliasing generator. Too bad they missed out on the formidable opportunity to use it, in true lo-fi spirit, as a cheap quantisation noise generator as well. Given the fact that the DX7 clocked at a spiffy 50kHz system frequency, but had a numerical resolution of only 9 bits, modeling quantisation noise would have been my personal priority. |
|
Back to top
|
|
|
varice
Joined: Dec 29, 2004 Posts: 961 Location: Northeastern shore of Toledo Bend
Audio files: 29
G2 patch files: 54
|
Posted: Sun Aug 17, 2014 1:44 pm Post subject:
|
|
|
Tim Kleinert wrote: | ...So in this case the module is only a cheap aliasing generator. Too bad they missed out on the formidable opportunity to use it, in true lo-fi spirit, as a cheap quantisation noise generator as well. Given the fact that the DX7 clocked at a spiffy 50kHz system frequency, but had a numerical resolution of only 9 bits, modeling quantisation noise would have been my personal priority... |
Yes, in this case, using the G2 Note Quantizer module like the example posted is probably not the best way to emulate the sound of the DX7, since it appears to pass the full 24 bit signal and it drops the audio bit rate too low (from 96kHz to 24kHz). It would probably be better to use the Digitizer module which can be set to a bit rate of 50.2kHz and also set to lower bit resolutions to generate more aliasing noise. _________________ varice |
|
Back to top
|
|
|
asb2m10
Joined: Jan 08, 2006 Posts: 7 Location: Montréal
|
Posted: Sun Aug 17, 2014 8:43 pm Post subject:
|
|
|
Tim Kleinert wrote: | So in this case the module is only a cheap aliasing generator. Too bad they missed out on the formidable opportunity to use it, in true lo-fi spirit, as a cheap quantisation noise generator as well. Given the fact that the DX7 clocked at a spiffy 50kHz system frequency, but had a numerical resolution of only 9 bits, modeling quantisation noise would have been my personal priority. |
I'm curious of the source about the 9 bits numerical resolution of the DX7 since the DX7-9 service manual states that the YM2128 chip was 16-bit. What is a mystery although is the sin LUT lookup bit resolution. Some say it is 10-11 bits, others say 8 bits with interpolation; but to my opinion, interpolation in 1983 was a little bit too expensive.
Is the 9 bits about the sin LUT lookup bit resolution ? |
|
Back to top
|
|
|
ian-s
Joined: Apr 01, 2004 Posts: 2669 Location: Auckland, New Zealand
Audio files: 42
G2 patch files: 626
|
Posted: Sun Aug 17, 2014 10:08 pm Post subject:
|
|
|
asb2m10 wrote: | Tim Kleinert wrote: | So in this case the module is only a cheap aliasing generator. Too bad they missed out on the formidable opportunity to use it, in true lo-fi spirit, as a cheap quantisation noise generator as well. Given the fact that the DX7 clocked at a spiffy 50kHz system frequency, but had a numerical resolution of only 9 bits, modeling quantisation noise would have been my personal priority. |
I'm curious of the source about the 9 bits numerical resolution of the DX7 since the DX7-9 service manual states that the YM2128 chip was 16-bit. What is a mystery although is the sin LUT lookup bit resolution. Some say it is 10-11 bits, others say 8 bits with interpolation; but to my opinion, interpolation in 1983 was a little bit too expensive.
Is the 9 bits about the sin LUT lookup bit resolution ? |
I'm not sure about the DX7 specifically but some of the early Yamaha OPL chips have been reverse engineered. The 1/4 sine table is 256 words long with a maximum value of 2137 which requires 12 bits. The values stored are actually the log of the sine. They used the old trick of adding log values to save having to do a multiply. I guess with 6 operators times 16 voices they would probably need quite a lot of resolution internally to avoid clipping. |
|
Back to top
|
|
|
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
Posted: Mon Aug 18, 2014 1:22 am Post subject:
|
|
|
I read about the 50k sample rate and 9 bits headroom. I have no first hand experience. So please correct me if this isn't true.
Maybe the final DAC is 9 bit and internally it's higher. (Like in modern 24/32 bit systems that have a 56bit accumulators.) |
|
Back to top
|
|
|
|