I don't know the exact answer but I suspect it's mainly because of what you mentioned, you'd need a critical mass of plugins that support it, otherwise one popular plugin would break the flow. This could probably be implemented as a DAW feature between each fx instance.
But up/downsampling is relatively cheap. And you only need to do it on nonlinear effects. So oversampling your guitar amp simulator, then downsampling back to 44.1k before your reverb probably uses less CPU than running both your guitar amp sim and reverb at 88.2k in order to avoid repeated conversion. > On Jul 4, 2023, at 17:16, brianw <[email protected]> wrote: > > Not to stray from the original question too far, but why perform all of that > upsampling and downsampling? According to your description, it should be > sufficient to low-pass nonlinear effect inputs, and then low-pass their > outputs before passing further along the signal chain. It should be just fine > to run the entire processing chain at 2x or 4x with judicious application of > filtering ... without any SRC. > > Of course, this would require coordination of all effects designers so that > one sample rate could be selected for the entire recording + mixing + > mastering chain (especially since everything can be done in one pass these > days). > > Brian Willoughby > >> On Jul 4, 2023, at 2:04 PM, N G (NG) <[email protected]> wrote: >> Sorry for double-posting but just to be clear my prior message only counts >> for applying non-linear processes to the signal. Like mixing with >> compression, saturation, analog modeled plugins. For just playback of audio >> it doesn't matter except you'll have frequencies in your signal that go >> above the human hearing range. And probably above the frequency response of >> whatever speakers/headphones you're listening through. >> >>> On Jul 4, 2023, at 16:58, N G (NG) <[email protected]> wrote: >>> There's this video by Dan Worrall / Fabfilter about higher sample rates. >>> 44.1k/48k is too small of a difference to really be noticeable, it's more >>> about things like 44.1k vs 88.2k. >>> https://urldefense.proofpoint.com/v2/url?u=https-3A__youtu.be_-2DjCwIsT0X8M&d=DwIFAg&c=009klHSCxuh5AI1vNQzSO0KGjl4nbi2Q0M1QLJX9BeE&r=TRvFbpof3kTa2q5hdjI2hccynPix7hNL2n0I6DmlDy0&m=zDM0oTd8nJ5kOSCIug10aOXbraJoVHQ1Ri3yMbiAmmA-xepTlnXrbK4xCTTG7XmC&s=TlEIUmDN7rtQCZIcsGrB4UC2STKWJvTAvPPfN-j0zDc&e= >>> <maxresdefault.jpg> >>> Samplerates: the higher the better, right? >>> youtu.be >>> >>> >>> Basically a higher sample rate will push most aliasing to happen outside >>> the audible range, which makes it easier to filter those artifacts. But >>> doing your entire mix at a higher sample rate can lead to a wider signal >>> bandwidth for intermodulation distortion to occur in, which will cause >>> artifacts in the audible range. So the most effective way to mix, if you're >>> specifically worried about avoid aliasing and IMD, is per-plugin over >>> sampling. Upsample by 2x or 4x, do your nonlinear processing, filter >>> everything above 22k, downsample back to the session's native sample rate. >>> >>> Or basically, you want your audio bandwidth to be as much as you need and >>> not more. >>> >>> -Neil >>> >>> On Jul 4, 2023, at 12:50 PM, Kevin Dixon <[email protected]> wrote: >>>> I don't know what you're referring to in the subject, but 44.1khz is just >>>> a left over standard, picked for convenience with existing mediums. 48k is >>>> more convenient for film use, since it aligns evenly with 24fps. But >>>> that's also an irrelevant medium at this point! >>>> >>>> The Red Book was the initial standard for compact disc digital audio >>>> >>>> https://urldefense.proofpoint.com/v2/url?u=https-3A__en.wikipedia.org_wiki_44-2C100-5FHz&d=DwIFAg&c=009klHSCxuh5AI1vNQzSO0KGjl4nbi2Q0M1QLJX9BeE&r=TRvFbpof3kTa2q5hdjI2hccynPix7hNL2n0I6DmlDy0&m=zDM0oTd8nJ5kOSCIug10aOXbraJoVHQ1Ri3yMbiAmmA-xepTlnXrbK4xCTTG7XmC&s=KCD_gKgHgM4BTwXnyuR_68Dc-t0iUyJSvq0AYNN8Sbg&e= >>>> >>>> https://urldefense.proofpoint.com/v2/url?u=https-3A__en.wikipedia.org_wiki_Rainbow-5FBooks-23Red-5FBook-5F-281980-29&d=DwIFAg&c=009klHSCxuh5AI1vNQzSO0KGjl4nbi2Q0M1QLJX9BeE&r=TRvFbpof3kTa2q5hdjI2hccynPix7hNL2n0I6DmlDy0&m=zDM0oTd8nJ5kOSCIug10aOXbraJoVHQ1Ri3yMbiAmmA-xepTlnXrbK4xCTTG7XmC&s=as5SwIPavlYwnJpkZuS3ak7PYSDw59GXPGn0qsWbct4&e= >>>>
