44.1 kHz vs 96 kHz 7.0a render differences -ACID Product Family

Subject:44.1 kHz vs 96 kHz 7.0a render differences Posted by: Angels Date:3/5/2009 1:24:29 PM I had an odd glitching problem with 7.0a which I traced to a plugin that didn't work at 96 kHz, and as part of the troubleshooting I was comparing setting the project at 44.1 kHz vs 96 kHz and I think I found a bug: When setting the project sampling rate at 96 kHz, a render is consistently over 1 dB hotter than when rendering with the project setting at 44.1 kHz. Shouldn't the renders be exactly the same given that nothing else changes?
Subject:RE: 44.1 kHz vs 96 kHz 7.0a render differences Reply by: Geoff_Wood Date:3/5/2009 7:49:54 PM If there is any content above 22KHz in the source media, it will be lost in the 44k1 render. Possibly mixing some signals may result in additonal artefacts over 22KHz, that will not show in a 44k1 sr render, especially if some plugin is generating additional spectral content.. If there isn't content audio over 22KHz then why even bother with 96KHz ;-) ?!!!! geoff
Subject:RE: 44.1 kHz vs 96 kHz 7.0a render differences Reply by: Angels Date:3/6/2009 9:59:37 AM I did the renders at the project's frequency, so one's 44.1 and the other 96. I actually got it backwards. It's the 44.1 kHz render that's softer (RMS) than at 96 kHz: 44.1=============================================== ................................................................Left Channel.........Right Channel Minimum sample value (dB).............-21.068..................-22.394 Maximum sample value (dB)............-20.660..................-23.132 RMS level (dB).....................................-48.657...................-49.493 Average value (dB)............................-110.515..................-112.899 96================================================ ..............................................................Left Channel...........Right Channel Minimum sample value (dB)............-20.895...................-21.949 Maximum sample value (dB)............-22.015...................-21.345 RMS level (dB).....................................-47.295...................-47.860 Average value (dB)............................-109.847................-112.899 (BTW: Looks like there's a bug in SF: stats for 96 kHz Max/Min values are interchanged...) Though the source material sampling rate doesn't exceed 44.1 kHz, the principle is that a higher sampling rate will process effects at that rate and provide more resolution for any processing performed. More troubling than levels is that there's more detail in the 96 kHz render than in the 44.1 kHz render. And this difference survives after resampling the 96 kHz render to 44.1. Try it yourself with material that has a lot of dynamic high frequency noise components (fast LPF filtered noise sweeps for example). It's undeniable. Knowing the ear will not hear the same at different volumes, I compensated for the level difference and still heard the same thing. So either this is me discovering the essential truth of 96 kHz processing, or there's a problem somewhere else. I have to admit that while undeniable to a trained ear, the difference is subtle. It could be that the plugins I used don't have the same transfer function at both sampling rates. Like you, I've been a 96kHz skeptic for a long time which is why I expected the renders to be indistinguishable. I just did a test: and took a -12 dB RMS 44.1 kHz 24-bit noise sample and rendered it in both 44.1 and 96 kHz 24-bit with no processing and no envelopes, and there's still a level difference. It seems to me there shouldn't be any. So I'm just looking for confirmation that the math is good because the evidence says otherwise. If Acid's math is all good, then it's something else. ITB, it's all math in the end. Message last edited on3/6/2009 10:04:49 AM byAngels.
Subject:RE: 44.1 kHz vs 96 kHz 7.0a render differences Reply by: Geoff_Wood Date:3/6/2009 3:30:46 PM "Though the source material sampling rate doesn't exceed 44.1 kHz, the principle is that a higher sampling rate will process effects at that rate and provide more resolution for any processing performed." If this is in fact the case, then presumably audio content over 22KHz is in fact being generated. I don't KNOW, but I'm curious. Of course if there is new content (that would explain the small level change) this could be'normal and expected' or it could be 'error' information from the plugin. Any scientists here ? geoff

Subject:44.1 kHz vs 96 kHz 7.0a render differences

Posted by: Angels

Date:3/5/2009 1:24:29 PM

I had an odd glitching problem with 7.0a which I traced to a plugin that didn't work at 96 kHz, and as part of the troubleshooting I was comparing setting the project at 44.1 kHz vs 96 kHz and I think I found a bug:

When setting the project sampling rate at 96 kHz, a render is consistently over 1 dB hotter than when rendering with the project setting at 44.1 kHz. Shouldn't the renders be exactly the same given that nothing else changes?

Subject:RE: 44.1 kHz vs 96 kHz 7.0a render differences

Reply by: Geoff_Wood

Date:3/5/2009 7:49:54 PM

If there is any content above 22KHz in the source media, it will be lost in the 44k1 render.

Possibly mixing some signals may result in additonal artefacts over 22KHz, that will not show in a 44k1 sr render, especially if some plugin is generating additional spectral content..

If there isn't content audio over 22KHz then why even bother with 96KHz ;-) ?!!!!

geoff

Subject:RE: 44.1 kHz vs 96 kHz 7.0a render differences

Reply by: Angels

Date:3/6/2009 9:59:37 AM

I did the renders at the project's frequency, so one's 44.1 and the other 96. I actually got it backwards. It's the 44.1 kHz render that's softer (RMS) than at 96 kHz:

44.1===============================================
................................................................Left Channel.........Right Channel
Minimum sample value (dB).............-21.068..................-22.394
Maximum sample value (dB)............-20.660..................-23.132
RMS level (dB).....................................-48.657...................-49.493
Average value (dB)............................-110.515..................-112.899

96================================================
..............................................................Left Channel...........Right Channel
Minimum sample value (dB)............-20.895...................-21.949
Maximum sample value (dB)............-22.015...................-21.345
RMS level (dB).....................................-47.295...................-47.860
Average value (dB)............................-109.847................-112.899

(BTW: Looks like there's a bug in SF: stats for 96 kHz Max/Min values are interchanged...)

Though the source material sampling rate doesn't exceed 44.1 kHz, the principle is that a higher sampling rate will process effects at that rate and provide more resolution for any processing performed.

More troubling than levels is that there's more detail in the 96 kHz render than in the 44.1 kHz render. And this difference survives after resampling the 96 kHz render to 44.1. Try it yourself with material that has a lot of dynamic high frequency noise components (fast LPF filtered noise sweeps for example). It's undeniable. Knowing the ear will not hear the same at different volumes, I compensated for the level difference and still heard the same thing.

So either this is me discovering the essential truth of 96 kHz processing, or there's a problem somewhere else. I have to admit that while undeniable to a trained ear, the difference is subtle. It could be that the plugins I used don't have the same transfer function at both sampling rates. Like you, I've been a 96kHz skeptic for a long time which is why I expected the renders to be indistinguishable.

I just did a test: and took a -12 dB RMS 44.1 kHz 24-bit noise sample and rendered it in both 44.1 and 96 kHz 24-bit with no processing and no envelopes, and there's still a level difference. It seems to me there shouldn't be any.

So I'm just looking for confirmation that the math is good because the evidence says otherwise. If Acid's math is all good, then it's something else. ITB, it's all math in the end.

Message last edited on3/6/2009 10:04:49 AM byAngels.

Subject:RE: 44.1 kHz vs 96 kHz 7.0a render differences

Reply by: Geoff_Wood

Date:3/6/2009 3:30:46 PM

"Though the source material sampling rate doesn't exceed 44.1 kHz, the principle is that a higher sampling rate will process effects at that rate and provide more resolution for any processing performed."

If this is in fact the case, then presumably audio content over 22KHz is in fact being generated. I don't KNOW, but I'm curious. Of course if there is new content (that would explain the small level change) this could be'normal and expected' or it could be 'error' information from the plugin. Any scientists here ?

geoff