Higher Sample Rates. Good, Bad, both ?

[KevWind]

Here is an interesting and seems a pretty good video at explaining (in understandable layman's terms) about higher sample rates and oversampling ( granted it is produced by a plugin company and is indirectly advertising ) But I think the video is a still pretty good general informational piece, on what has and is still an often a misunderstood, and sometimes highly debated subject .

I particularly like how it address, and demonstrates graphically ,,, how (in digital recording) what is happening above the audible range can and does effect what is in the audible. And pretty much dispels the overly simplistic argument that up to 20k is the "only thing" and "all that matters".... And at the same time dispels the equally simplistic notion that "higher is always better".

[keith.rogers]

Haven't waded through the whole thing, but early on it gets to the thing that has always bothered me about these ultra high sample rates, and that's the fact that it's creating data points for audio which 1) we can't possibly hear, 2) could be garbage if nothing was really being input, 3) is more likely to be noise than anything real, and if it was real, there's no distinguishing between supersonic content from your Steinway, or some minute leak in your HVAC system, and 4) as the video shows, the impact of that content in processing is generally unknown, with the risk being audible changes in the processed audio > 20kHz may be due not to actual audio, but artifacts of processing data not comprehended in the plugin/DAW design.

Back to the video. Thanks.

P.S. Good ad for FabFilter!

[Dirk_Z]

[KevWind]

Originally Posted by Dirk_Z

Well lots opinion based conclusions. He mentions in passing but more or less dismisses the very real (and noticeable) effects of aliasing , being flatter with less depth and presence to the soundstage (actually pretty evident in the OP video even over Youtube and ear buds on my laptop).

Also he completely ignores 48 k now considered by many, if not most, digital sound engineers, digital converter and digital audio systems designers, as being optimum in terms of results to file size and what's available today.

And once again, the entire argument about what highest single frequency humans can hear, is not entirely relevant when talking about complex digital multi frequency musical audio

[jim1960]

When first I got my hands on a interface capable of 96, I started recording everything in 96. I heard no difference but I figured higher was better and didn't see a down side. Once I realized that I was making my computer work twice as hard for differences that are, for the most part, inaudible to me and inaudible to the general public which doesn't care a bit about the 48 vs 96 thing, I went back to 48 and don't even think about it anymore.

[KevWind]

Originally Posted by jim1960

When first I got my hands on a interface capable of 96, I started recording everything in 96. I heard no difference but I figured higher was better and didn't see a down side. Once I realized that I was making my computer work twice as hard for differences that are, for the most part, inaudible to me and inaudible to the general public which doesn't care a bit about the 48 vs 96 thing, I went back to 48 and don't even think about it anymore.

Ya personally I do 48k 24 bit and see no reason to change. But I think the video is interesting non the less

[rick-slo]

Moral of the story: use good gear and plugins that have high quality frequency filters and plugins that up sample and down sample cleanly.

I usually record at 88200 versus 44100 as it gives some more room above 20000 hertz in case those less than perfect "brick wall" frequency filters
allow aliasing into the audible range when recording at 44100. Also 88200 converted to 44100 (CD quality) is more pure (at least mathematically)
than converting from 48000.

Interestingly in the video showing the FabFilter Q he had it set to Zero Latency mode processing which is not nearly as clean sounding as the
other modes. I use Natural Phase mode.

[jim1960]

Originally Posted by KevWind

Ya personally I do 48k 24 bit and see no reason to change. But I think the video is interesting non the less

If I had Doug's new Mac Pro, maybe I'd be recording everything in 192, but my current iMac wouldn't like that very much.

[Brent Hahn]

For the kind of sound design where I'm slowing stuff down a lot and, say, turning pig noises into space monsters, 192 is the only way to go. Otherwise I'm fine with 44.1.

[gwlee7]

Originally Posted by Brent Hahn

For the kind of sound design where I'm slowing stuff down a lot and, say, turning pig noises into space monsters, 192 is the only way to go. Otherwise I'm fine with 44.1.

LOL.

“How was your day Brent?”

“Fine. Spent most of the morning turning pig noises into space monsters. After that pretty quiet.”

[Doug Young]

Originally Posted by jim1960

If I had Doug's new Mac Pro, maybe I'd be recording everything in 192, but my current iMac wouldn't like that very much.

I've not tried 192, seems like overkill. I did just mix a project at 96K, with 10 tracks, certainly no problem for the Mac, tho my Apollo ran out of memory for not all that many UAD plugins. So there's always some bottleneck. At 192, I'd probably only be able to use one plugin :-)

[rick-slo]

https://people.xiph.org/~xiphmont/de...g.html#toc_1ch

also read "Digital Audio Explained for the Audio Engineer" by Nick Aldrich.

[paulp1960]

I record at 44.1 24 bit. I've tested my ears with the frequency generator utility in Logic Pro X and my upper limit these days seems to be around 11 Khz so I don't worry about the higher frequency stuff anymore.

[keith.rogers]

Originally Posted by rick-slo

...
I usually record at 88200 versus 44100 as it gives some more room above 20000 hertz in case those less than perfect "brick wall" frequency filters
allow aliasing into the audible range when recording at 44100. Also 88200 converted to 44100 (CD quality) is more pure (at least mathematically)
than converting from 48000....

The video does suggest a good reason why 48k might be a better choice from low-pass filter standpoint.

I do not know how down-sampling is done. If it was done via a D/A, followed by A/D, it should not really matter, if you trust in the math of reconstruction for audio sampled at or above the Nyquist rate, right? Now, it may be more likely that it's done with just math, but I'm in the 48k is just fine club, and anything going to 44.1/16 is likely headed to MP3 anyway, even if it ever went to a CD first...

Originally Posted by rick-slo

https://people.xiph.org/~xiphmont/de...g.html#toc_1ch

Great link.

[keith.rogers]

Originally Posted by paulp1960

I record at 44.1 24 bit. I've tested my ears with the frequency generator utility in Logic Pro X and my upper limit these days seems to be around 11 Khz so I don't worry about the higher frequency stuff anymore.

Yeah, "get hearing checked" item is on my "honey-do" list. Between closing bars as a kid in a R&R band in front of a 6x10 cabinet, summer job running a radial-arm saw with no hearing protection, and 1000mi days on a motorcycle, it's a wonder I hear at all.