keith
Max Output Level: -36.5 dBFS
- Total Posts : 3882
- Joined: 2003/12/10 09:49:35
- Status: offline
RE: Using S8 with sample rates higher than the hardware supports? Oversampling the project
2009/02/21 11:44:31
(permalink)
ORIGINAL: dcastle
I guess I am going to have to reread the literature to get my head wrapped around this, but the best explanation I can come up with is that I have digitized a perfect triangle waveform (that has harmonics extending out to infinity) inside my computer without any alias filtering — when I think about it this way it is obvious that there is going to be aliasing. Right, so the next logical step is to add a good filter... But the practical theory is that if you want to filter right up against 20kHz or whatever you need a steep, high-order filter. Doubling/tripling/quadrupling the bandwidth allows for a much simpler filter with a gentle slope (i.e., less expensive, more stable). That covers the waveform generation side. On the effects plugin side you're going to have similar concerns regarding alias filtering and whatnot for nonlinear type manglings. In my mind one important factor that is kinda glossed over is the "oversampling" employed by those plugins that implement it. An "oversampling" plugin must "upsample" the source content (oversampling is missapplied in this case, I think). Upsampling is sample rate conversion. So what are the best software SRCs available? There's a few (voxengo being one of them) -- but clearly: not all software SRCs are created equally! A centralized "oversampling"/"upsampling" mechanism would allow the end user to select an SRC implementation to be used across the audio pipeline, relative to cost/quality tradeoffs.
|
dcastle
Max Output Level: -49 dBFS
- Total Posts : 2623
- Joined: 2004/11/15 12:40:02
- Location: Inland Empire
- Status: offline
RE: Using S8 with sample rates higher than the hardware supports? Oversampling the project
2009/02/21 12:06:55
(permalink)
Right, so the next logical step is to add a good filter... But the practical theory is that if you want to filter right up against 20kHz or whatever you need a steep, high-order filter. Doubling/tripling/quadrupling the bandwidth allows for a much simpler filter with a gentle slope (i.e., less expensive, more stable). I'm not sure a filter will work because I don't know how to filter before I digitize — if you get what I mean — because this ideal waveform only exists in a 64-bit floating point calculation that is sampled at 44100 samples per second — so it already has aliases. I think I will have to sample at a higher rate and filter that before downconverting to 44100, but that's just a guess because I'm still in the middle of reading the literature on the subject. I feel like I'm walking down the path trodden by giants ahead of me. I'm curious to find what's at the end of the path. Regards, David
ASUS M3A78 AMD 9950 Quad 2.6G 8GB Shure • Rhode • Audio-Technica • Allen&Heath GL2200-24 MOTU 24i • Presonus Firepod • E-MU 1212m • Zoom H2 SONAR 2XL-8PE • Sound Forge 1-9 • Audacity 0.1-1.3
|
cheater
Max Output Level: -90 dBFS
- Total Posts : 28
- Joined: 2007/04/09 15:14:46
- Status: offline
RE: Using S8 with sample rates higher than the hardware supports? Oversampling the project
2009/02/22 15:40:20
(permalink)
ORIGINAL: dcastle
Right, so the next logical step is to add a good filter... But the practical theory is that if you want to filter right up against 20kHz or whatever you need a steep, high-order filter. Doubling/tripling/quadrupling the bandwidth allows for a much simpler filter with a gentle slope (i.e., less expensive, more stable).
I'm not sure a filter will work because I don't know how to filter before I digitize — if you get what I mean — because this ideal waveform only exists in a 64-bit floating point calculation that is sampled at 44100 samples per second — so it already has aliases. I think I will have to sample at a higher rate and filter that before downconverting to 44100, but that's just a guess because I'm still in the middle of reading the literature on the subject. I feel like I'm walking down the path trodden by giants ahead of me. I'm curious to find what's at the end of the path.
Regards,
David
The revelation that the only way to make your current algorithm work better is to run it at 1.5 MHz, then downsample. Try it, you'll see the things from my point of view :) Granted, there are lots of other algorithms that don't do that, BLT for one that would fix your problems easily... but the statistical developer doesn't rewrite his algorithm. So 'you' should be as 'lazy' as that developer and - forbid - never ever try and actually do some work to optimize. Therefore 'I' am stuck with this crappy plugin that I have to work with because some pr0duc4h has me work on his project to make it sound better, and he *is insisting* on using pro-53. To my woe, the only way to not lose sanity is to render at awesome sampling rates. If you do want to find out though, google for band limited transients: BLT/BLIT. There's also a technique called Blep. If you look at a sonogram, you'll notice that there are vertical lines at the peaks of your wave, where the singularities happen. This is aliasing. The simplest way to alleviate the problem is to not generate that peak. Play back a sample of the peak instead. HTH
|
RTGraham
Max Output Level: -57 dBFS
- Total Posts : 1824
- Joined: 2004/03/29 20:17:13
- Location: New York
- Status: offline
RE: Using S8 with sample rates higher than the hardware supports? Oversampling the project
2009/02/23 23:56:32
(permalink)
ORIGINAL: cheater
Therefore 'I' am stuck with this crappy plugin that I have to work with because some pr0duc4h has me work on his project to make it sound better, and he *is insisting* on using pro-53. To my woe, the only way to not lose sanity is to render at awesome sampling rates.
And finally, we find out just *why* this is so important to you.  I kept wondering why someone who seemed to have so much insight into the processing quality of the plugins they used wouldn't just choose to use only plugins whose processing quality they were happy with.
~~~~~~~~~~ Russell T. Graham Keys, Vocals, Songwriting, Production russell DOT graham AT rtgproductions DOT com www DOT myspace DOT com SLASH russelltgraham
|
cheater
Max Output Level: -90 dBFS
- Total Posts : 28
- Joined: 2007/04/09 15:14:46
- Status: offline
RE: Using S8 with sample rates higher than the hardware supports? Oversampling the project
2009/02/25 03:00:51
(permalink)
ORIGINAL: RTGraham
ORIGINAL: cheater
Therefore 'I' am stuck with this crappy plugin that I have to work with because some pr0duc4h has me work on his project to make it sound better, and he *is insisting* on using pro-53. To my woe, the only way to not lose sanity is to render at awesome sampling rates.
And finally, we find out just *why* this is so important to you.
I kept wondering why someone who seemed to have so much insight into the processing quality of the plugins they used wouldn't just choose to use only plugins whose processing quality they were happy with.
You can never know. Even the seemingly best plugin can have aliasing, and with stuff that happens at the audio rate (compressors, audio-rate filter modulation etc) it might happen for 10 miliseconds at a time. You will never find out, and it will just sound 'OK' while it might sound much better at a higher rate. The point is that almost all plugins sound better at higher sampling rates. And if Sonar starts supporting high sample rate transports, then those plugins can stop using upsampling and downsampling, and those processes also degrade quality. They could be eliminated.
|
altima_boy_2001
Max Output Level: -55 dBFS
- Total Posts : 2033
- Joined: 2005/11/04 17:48:01
- Location: Central Iowa
- Status: offline
RE: Using S8 with sample rates higher than the hardware supports? Oversampling the project
2009/02/25 04:49:22
(permalink)
The only potential problem I see is that you have mentioned some very oversampling numbers. Right now Sonar supports up to 384 kHz rates (equivalent of 8x oversampling at 48 kHz) so obviously they'd have to change that to go higher. However, the Sonar developers seem to code things useful for 99% of users rather 1% of users so I don't really see them pushing the envelope on something like this. I could see them allowing 2/4/8x oversampling for 48 kHz hardware, 2/4x oversampling for 96 kHz hardware, and 2x oversampling when hardware is running 192 kHz thus maintaining the 384 kHz max sample rate. Or would they subscribe to the idea that any amount of oversampling should reduce aliasing issues enough for most people/situations and then just implement 2x oversampling for whatever the hardware rate is. Also, some plugins/instruments also list supported max sample rates of 96/192/384. What does Sonar do when a plugin is incapable of running at a chosen rate, or rather, do plugins/instruments report max supported sample rates so Sonar could disable them if needed?
post edited by altima_boy_2001 - 2009/02/25 04:59:27
You can use me as your eSoundz referral (altima_boy_2001).
|
cheater
Max Output Level: -90 dBFS
- Total Posts : 28
- Joined: 2007/04/09 15:14:46
- Status: offline
RE: Using S8 with sample rates higher than the hardware supports? Oversampling the project
2009/02/25 20:58:17
(permalink)
ORIGINAL: altima_boy_2001
The only potential problem I see is that you have mentioned some very oversampling numbers. Right now Sonar supports up to 384 kHz rates (equivalent of 8x oversampling at 48 kHz) so obviously they'd have to change that to go higher. However, the Sonar developers seem to code things useful for 99% of users rather 1% of users so I don't really see them pushing the envelope on something like this. 99% of all users of some popular algorithms use oversampling at 3 MHz or more. 64x oversampling for clipping distortions is the standard. 48kHz * 64 = 3.072 MHz I could see them allowing 2/4/8x oversampling for 48 kHz hardware, 2/4x oversampling for 96 kHz hardware, and 2x oversampling when hardware is running 192 kHz thus maintaining the 384 kHz max sample rate. Or would they subscribe to the idea that any amount of oversampling should reduce aliasing issues enough for most people/situations and then just implement 2x oversampling for whatever the hardware rate is.
The idea isn't to have an 'oversampling amount', it's to have a high native sampling rate for the project (perhaps minus the files - I can see how that could easily get cumbersome) A fixed oversampling rate would be very detrimental. A limited oversampling rate would not address the issues at hand.
Also, some plugins/instruments also list supported max sample rates of 96/192/384. What does Sonar do when a plugin is incapable of running at a chosen rate, or rather, do plugins/instruments report max supported sample rates so Sonar could disable them if needed?
Exactly what it does right now in such a situation. Cheers
Of course, a smarter way would be to allow the user to run each plugin in the chain at a separate sampling rate, and upsample and downsample accordingly between the plugins. Out of all possibilities, this would create the least distortion in a mixed sample-rate setup. It could be a bit more challenging to keep track of the pre-delays and other stuff like that, but it's not really a hurdle. Also, if Cakewalk expose an API for the sample rate converter, I'm sure people will write some good sample rate converters, which will improve the sound even further, since we're not stuck with just one algorithm. And an API like this wouldn't be too difficult or too complicated... on afterthought you could probably just use VST - perhaps slightly modified, since processreplacing obviously works in a way that assumes the input sample rate is the same as the output sample rate
|
Mr. Ease
Max Output Level: -71 dBFS
- Total Posts : 960
- Joined: 2003/11/24 18:44:01
- Location: West Sussex, UK
- Status: offline
RE: Using S8 with sample rates higher than the hardware supports? Oversampling the project
2009/02/26 13:33:10
(permalink)
ORIGINAL: dcastle
Right, so the next logical step is to add a good filter... But the practical theory is that if you want to filter right up against 20kHz or whatever you need a steep, high-order filter. Doubling/tripling/quadrupling the bandwidth allows for a much simpler filter with a gentle slope (i.e., less expensive, more stable).
I'm not sure a filter will work because I don't know how to filter before I digitize — if you get what I mean — because this ideal waveform only exists in a 64-bit floating point calculation that is sampled at 44100 samples per second — so it already has aliases. I think I will have to sample at a higher rate and filter that before downconverting to 44100, but that's just a guess because I'm still in the middle of reading the literature on the subject. I feel like I'm walking down the path trodden by giants ahead of me. I'm curious to find what's at the end of the path.
Regards,
David
Well I'm back from my short holiday! I see the discussion about DDS is now redundant - I had quickly seen your last response re. this and had some responses ready in my head. I was also about to say why I did not see that it was relevant to the point in question but there is no point now as you have already found. With regard to the "filtering requirement" there is a simple way to obviate the problem. Reduce whatever waveform you want to it's polynomial and then only generate the terms that are within the nyquist frequency. No need to filter then as none of the "super nyquist" frequencies are ever generated to alias.... Of course the number of terms required will depend on the frequency needed. In essence this is the same requirement as the anti aliasing filter of a real analogue signal into an A-D. The advantage though is that we can generate the entire signal, as accurately as required by the sample frequency, without ever generating anything high enough to alias. All the "standard" synth. waveforms can be easily reduced to their polynomials so they should be quite easy to generate. You could even have some fun by adding a phase control on the harmonic terms to generate even more basic waveforms! Of course this does not get round Cheater's problem with the crappy synth. and his suggestion for resampling to high rates would be the only real way round this (although I'm not convinced that this technique would work for all VSTI's).
|
cheater
Max Output Level: -90 dBFS
- Total Posts : 28
- Joined: 2007/04/09 15:14:46
- Status: offline
RE: Using S8 with sample rates higher than the hardware supports? Oversampling the project
2009/02/28 16:25:07
(permalink)
ORIGINAL: Mr. Ease
Of course this does not get round Cheater's problem with the crappy synth. and his suggestion for resampling to high rates would be the only real way round this (although I'm not convinced that this technique would work for all VSTI's).
Definitely not all.. but surely most.
|