Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Aug 22, 2023 19:11:45 GMT -6
It's an interesting video but I think it comes at oversampling from the wrong direction. We use oversampling because it eases the requirements for antialiasing filters, not the other way around. You'd upsample the original waveform (which will likely already be band-limited), which is not hard, but it is time and resource intensive as they mention in the video. However, the oversampling moves the Nyquist frequency up considerably, which allows the use of lower-order filtering. This reduces the need for filters with fast cutoffs but also have high over/undershoot. Yes, that's one use of oversampling, but the video is talking about creating frequency headroom (increasing the working bandwidth) for subsequent non-linear processing. With the added headroom, the aliasing doesn't extend back into the audio band as quickly/easily, and is removed on the downsample back to the base rate. The video goes overboard on the FUD (ear, uncertainty, doubt) factor. Seemingly imagining that we typically use 6-8 non-linear plugins on a track, and the plugin designers may have made decisions that won't stack well with each other, etc. (But we'll happily live with vintage analog gear chains up to our level of noise tolerance, and celebrate the sound.) Sometimes all of your processing is very non-linear and the recording gear is noisy but if you are good with gates and expanders then there will not be any perceived noise  |
|
|
|
Post by gouge on Aug 23, 2023 4:32:09 GMT -6
,One of the points discussed was not needing to overssmple if using higher sample rates in the first place. It was suggested just use higher sample rates Not so simple with 192 khz. See fabientdr's thread on gearslutz where he showed that sometimes lower sample rates are cleaner from not intermodulating high frequency garbage. gearspace.com/board/mastering-forum/968641-some-thoughts-quot-high-resolution-quot-audio-processing.htmlI switched to 88.2 and 96 kHz to make MDWEQ zero latency and to clean up mixes that have everything automated all the time Typically I prefer higher sample rates when I do blind tests. Can’t say why, it’s just where I landed. Some of the plugs I use don’t do 192 so I use 96, and kinda moved beyond thinking about it any more.
|
|
|
|
Post by mcirish on Aug 23, 2023 9:56:23 GMT -6
I hope this doesn't turn into a sample rate war. :-) I figure it's best to work at the highest sample rate you can and still get work done. For me, that's 48K/24. I do so much stuff with super high track counts that I need the processing power of 48K vs 96K, even though I have a monster computer with a 13th gen CPU. If I was dealing with only a couple dozen tracks, I could get away with a higher sample rate, but that's never the case.
As far as aliasing, I've just become more aware of what plugins start making the sound worse. It's usually things that are supposed to distort, but come off very harsh and non-musical. I just grab something else to use. I use oversampling on plugins that can. Even better is when they can be setup to use a higher oversampling rate on mixdown only. Best of both worlds, as long as the results are pretty similar. Then again, I tend toward much cleaner mixes anyway.
|
|
|
|
Post by earlevel on Aug 24, 2023 0:51:51 GMT -6
I remember George Massenburg once suggested that it was a bit foolish to assume we know everything there is to know about hearing and how high frequencies contribute to our perception over larger periods of time. And he shares my skepticism of A-B methodology. I hear a difference using 192k. Better? More real? Full of artifacts? Can't say. Different sound? Yes. Of course, I could be fooled in a momentary double-blind test, but as I'm working, I can feel something, whether that's a non-linearity or something else. I don't think it makes sense for a lot of, maybe most, records but if it's a naturalistic recording with not too many tracks, I track it at 192k. I'm in severe danger of being misinterpreted on this...I think people should do what they are comfortable with. But I want to present a counter argument for the benefit of the thought process for the board. People will argue all day long about whether higher sample rates are better. (I say "better", because "different" shouldn't be a sufficient reason to require more processing power and storage space.) There are—at least in theory—reasons why 96 kHz might be better than 48 kHz. One would be the possibility we somehow sense higher harmonics, and wiping them out above 20 kHz may be causing us to lose something. But it's probably safe to say this argument runs out of steam quickly if we try to extend that argue much above 40 kHz. Real instruments, including electrified and electronic as well as acoustic, and real music, just doesn't have anything higher, and most are much lower. The second main reason would be filter characteristics. At 44.1/48 kHz sample rate, sampling filters must be steep to allow 20 kHz bandwidth and still have deep attenuation by half the sample rate. If phase around the cutoff area and steep-filter characteristics like ringing could be a problem we could sense, then 96 kHz gives ample room to use choices for which it would be hard to argue that they are deleterious. In fact, some have said (I can provide references) that at around 66 kHz sample, we have all the room we could possibly need. Maybe you see where I'm heading: While we can make an argument for double-rate (88.2/96k)—whether it stands up to audibility tests or not—past that point it's difficult to justify yet higher sample rates. Sample rates above double rate have no apparent justification, even theoretically. Of course, I'm talking about delivery formats. For recording/mixing/mastering there are reasons to run higher rates, for non-linear processes in particular, especially inside plugins where they can be done as needed. But for that, 192 kHz is not enough anyway (it could be sufficient for milder non-linearities, but not for amp sims, distortion plugins). Better to let plugin designers implement oversampling as needed, inside the plugin, and run at 44.1/48, or the double rates at most. There are other potential downsides to the higher rates (above 96k), besides the data increase, but there doesn't seem to be any conceivable upside. I'm not saying you don't hear a difference, or that it doesn't sound better on your converters. But if so, it's hard to understand why those converters can't be tweaked to sound just as good at the lower rates, since there is no theoretical basis for it sounding better. |
|
|
|
Post by gouge on Aug 24, 2023 5:12:28 GMT -6
i gotta say i can't believe people still argue over sample rates.
When i initially started using digital i preferred higher sample rates. i was quickly told on forums that it was in my head because sample theory meant that 44.1 was all i needed. on the purple site people were ridiculed for suggesting higher sample rates sounded better.
over time people started to identify they preferred higher rates too and over time the industry slowly moved from 44.1 to 96
With time, a few leaders in the industry started to address the idea that plugins performed better at higher sample rates. So we had a fundamental shift. It was becoming accepted that higher sample rates had benefits. higher sample rates was now a thing.
while this was going on we still had the "plugins sounded plastic" argument going on. Personally, i thought they sounded plastic. Again bodies of people on forums let me know how i was misinformed, had amateur ears, didn't understand the science etc. etc......
again, over time we were informed that actually, plugins can sound plastic because aliasing adds up. so that plastic sound i was hearing was in fact aliasing. now, oversampling is a thing.
throughout the journey, we also had the idea pushed that because sample theory can recreate a wave file perfectly you can simply resample everything that needs it. of course, doing loop back tests kinda suggested something different but that was put down to poor designs and analogue input/ouput circuits.
now we have people starting to identify that sample theory is great if you have unlimited computing power but as we don't have unlimited computing power it is possibly not as perfect as suggested in the real world so now there are other ways of resampling being approached.
you know It really stands out to me that for years people could hear the flaws in digital while being told there were none or their ears couldn't hear that high.. don't get me wrong, I've got plugs i rate really highly and i think that modern digital has come a long way and i love using it, but i have not forgotten the journey and for everyone whos tried to tell me 44.1 is all i need i can only think surely we've moved beyond this because history has not been kind to that idea.
|
|
|
|
Post by earlevel on Aug 24, 2023 12:19:29 GMT -6
i gotta say i can't believe people still argue over sample rates. When i initially started using digital i preferred higher sample rates. i was quickly told on forums that it was in my head because sample theory meant that 44.1 was all i needed. on the purple site people were ridiculed for suggesting higher sample rates sounded better. over time people started to identify they preferred higher rates too and over time the industry slowly moved from 44.1 to 96 With time, a few leaders in the industry started to address the idea that plugins performed better at higher sample rates. So we had a fundamental shift. It was becoming accepted that higher sample rates had benefits. higher sample rates was now a thing. while this was going on we still had the "plugins sounded plastic" argument going on. Personally, i thought they sounded plastic. Again bodies of people on forums let me know how i was misinformed, had amateur ears, didn't understand the science etc. etc...... again, over time we were informed that actually, plugins can sound plastic because aliasing adds up. so that plastic sound i was hearing was in fact aliasing. now, oversampling is a thing. throughout the journey, we also had the idea pushed that because sample theory can recreate a wave file perfectly you can simply resample everything that needs it. of course, doing loop back tests kinda suggested something different but that was put down to poor designs and analogue input/ouput circuits. now we have people starting to identify that sample theory is great if you have unlimited computing power but as we don't have unlimited computing power it is possibly not as perfect as suggested in the real world so now there are other ways of resampling being approached. you know It really stands out to me that for years people could hear the flaws in digital while being told there were none or their ears couldn't hear that high.. don't get me wrong, I've got plugs i rate really highly and i think that modern digital has come a long way and i love using it, but i have not forgotten the journey and for everyone whos tried to tell me 44.1 is all i need i can only think surely we've moved beyond this because history has not been kind to that idea. Respectfully, there a lot of "I thought this...then someone said it was true" ("we were informed"—by whom?). There is no validity in saying that sampled audio doesn't really live up to expectations because we need unlimited computing power. The sampling itself works exactly, perfectly encoding bandlimited audio, to within the limits of electronics (the limits of analog). Differences in digital are primarily the choices and implementation of the lowpass filters. And maybe "plastic sound" is just sound unaided by the distortions of analog media. If you're trying to cure aliasing in plugins, 96k is not nearly enough, and 192k would only be good for mild aliasing from non-linear processing. I'm not saying it's not possible that a particular plugin would sound better at a higher rate, but that would be because they didn't do their job. And if you like to work at 192k or 384k, I'd still say it makes little sense to distribute the final product at more than 96k (not convinced it's needed, but acknowledging it has potential benefits theoretically). People should work the way they want to. I'm just responding because the post has a "and we now know..." quality, when there is nothing to back up some of these assertions. Especially claims that digital sampling doesn't quite measure up to theory. I hope this doesn't come off as argumentative, I'm just presenting counterpoint. |
|
|
|
Post by gouge on Aug 24, 2023 18:12:50 GMT -6
all good. i respect your counterpoint and understand this is a conversation with many different ideas and experiences
|
|
