Best sound quality - run 96khz samplerate, do not oversample.

[@midnight]

So I've been thinking, as everything has a tradeoff the only way to really get good analog response from plugins is to run your project natively at 88.2 or 96khz. Because running at 44.1 or 48khz, without oversampling if the plugin generates harmonic distortion it will also generate aliasing, unless it uses a steep low pass filter which causes phase distortion.

Or you can run @ 44.1 or 48khz and use plugins that use oversampling, but then you will still get phase distortion cause by either the minimum phase anti-aliasing filters, or you will get linear-phase artifacts from using linear phase oversampling filters.

So really, the only way to minimize aliasing AND phase distortion AND linear-phase artifacts is to simply run your DAW at 24/96 (or 88.2) and use plugins that do not oversample.

You still might get a little aliasing but it will likely be inaudible and not effect your mix.

The only drawback here is you will use more CPU. There are no sonic drawbacks, as there are with oversampling or running 44.1khz w/o oversampling.

[valhallasound]

So I've been thinking, as everything has a tradeoff the only way to really get good analog response from plugins is to run your project natively at 88.2 or 96khz. Because running at 44.1 or 48khz, without oversampling if the plugin generates harmonic distortion it will also generate aliasing, unless it uses a steep low pass filter which causes phase distortion.

Or you can run @ 44.1 or 48khz and use plugins that use oversampling, but then you will still get phase distortion cause by either the minimum phase anti-aliasing filters, or you will get linear-phase artifacts from using linear phase oversampling filters.

So really, the only way to minimize aliasing AND phase distortion AND linear-phase artifacts is to simply run your DAW at 24/96 (or 88.2) and use plugins that do not oversample.

You still might get a little aliasing but it will likely be inaudible and not effect your mix.

The only drawback here is you will use more CPU. There are no sonic drawbacks, as there are with oversampling or running 44.1khz w/o oversampling.

To me, this is one of the main reasons that people would run at 88.2/96 kHz. You summarize the signal processing issues nicely (phase distortion, linear phase artifacts, aliasing).

Another good reason is that audio interfaces that can run at 96 kHz won't have the steep antialiasing filters that the older 44.1/48 KHz interfaces had to use. In addition, interfaces that run at the higher sampling rate probably have better characteristics overall. It reminds me of why you use 24-bit convertors instead of 16 bits: for most mid-range convertors, the last few bits are crap, so a 24-bit convertor pushes those bits below the range of audibility.

A 3rd reason: many digital filters that emulate analog filters (VCFs, EQs) have weird rolloffs near Nyquist. Put Nyquist above where you can hear things, and the problem is solved.

The same issue pops up with delay interpolation (chorusing/flanging/any sort of modulated delays). Most interpolation types rolloff around Nyquist, but have good behavior up to 1/4 the sampling rate. Put Nyquist up to 88.2/96 KHz, and even the cheapest interpolation types will have an almost flat amplitude response through the audible range.

Back around the turn of the century, I worked on physical models for video games. Some of these had weird oscillating behavior around Nyquist. At the time, we had to use 22 KHz sampling rates for efficiency, and the oscillations sounded horrible. Run things at 44.1 KHz, and the problems largely went away.

So there are a lot of good reasons to run at 88.2/96 KHz, that have nothing to do with the idea that people can hear frequencies at some level up beyond 20 KHz. I see this bandied about in forums on occasion. Presumably by dolphins with computers.

Meanwhile, I like to run things at LOWER sampling rates inside my plugins, just to make things sound like crappy old digital effects processors. Is there phase distortion? You bet there is! Weird behavior around the lower Nyquist? Absolutely!

Sean Costello

[A.M. Gold]

Does any of this matter, though, when we all deliver any audio we produce as mp3's (or worse) or, at best, "CD quality" audio format?

[Shabdahbriah]

Does any of this matter, though, when we all deliver any audio we produce as mp3's (or worse) or, at best, "CD quality" audio format?

Been recording @ 32/88.2 since I could do so. What you are working with before the 'degradation' to a lesser 'rate', I find to be a significant plus.

Having - that is "retaining" a master copy of your original in a higher resolution, is simply "smart" from my perspective, given that the formats you mention are what you are "delivering" it in - TODAY.

YMMV

[Aleksey Vaneev]

I think with 6-core CPUs currently being released we should be talking about working at 192kHz as a standard - which is like an "ultimate solution" for high-quality "in the box" sound processing. Going higher than 192 kHz is usually only marginally better. Working at 384kHz like Pyramix does covers all the possible bases, but it's too much I think.

[khanyz]

Even at that level I would be considering my destination rate and use 176.4. Which, to me at least, means music rather than movie use.

I currently use 88.2 as my standard, but drop to 44.1 for demo tracks.

[A.M. Gold]

Does any of this matter, though, when we all deliver any audio we produce as mp3's (or worse) or, at best, "CD quality" audio format?

Been recording @ 32/88.2 since I could do so. What you are working with before the 'degradation' to a lesser 'rate', I find to be a significant plus.

Having - that is "retaining" a master copy of your original in a higher resolution, is simply "smart" from my perspective, given that the formats you mention are what you are "delivering" it in - TODAY.

YMMV

Well, I generally agree, and I don't fall in line with some of the people who have (supposedly) scientifically argued that nothing above 44.1 will ever make any audible difference.

But you pay a huge price for working at 24/96, and obviously a bigger one for going all the way up to 32/192. I have a quad core now and am on the verge of going to Win 7 64 (still on XP x86), but even then I'm not sure I want to sacrifice the power to go to 24/96 as a working standard.

I think a reasonable alternative, though, is to just make sure you have all your projects and VST's archived so you can render them at higher depth/rate, later. If you track live instruments, you have the option to make a master recording at a high resolution and then render down to work at a lower resolution, while retaining the hi res files for later use.

[bezusheist]

...So there are a lot of good reasons to run at 88.2/96 KHz, that have nothing to do with the idea that people can hear frequencies at some level up beyond 20 KHz. I see this bandied about in forums on occasion. Presumably by dolphins with computers...

Eric, maybe...?

[tony tony chopper]

So really, the only way to minimize aliasing AND phase distortion AND linear-phase artifacts is to simply run your DAW at 24/96 (or 88.2) and use plugins that do not oversample.

& then what do you do with your 88k signal for your listeners? You downsample it to 44k, because that's all they will handle. But wait, that's also when you'll get your linear-phase artefacts (which don't exist for real, because the place where it could possibly cause ringing, is inaudible).

Also, oversampling isn't the only way to prevent aliasing. Lots of synths don't produce aliasing by other means. Lots of effects don't produce aliasing at all.

Been recording @ 32/88.2 since I could do so. What you are working with before the 'degradation' to a lesser 'rate', I find to be a significant plus.

you should post an example showing how audible that "degradation" is. It's not any audible, but prove me wrong, it's easy.

So there are a lot of good reasons to run at 88.2/96 KHz

these are good reasons for a plugin to process at a higher rate, yes. Not a good reason for a host to process everything at a higher rate.

[bmanic]

If you use a lot of samples and sample libraries then you will still get problems running at higher sample rates because most samples are stuck at 44.1kHz or 48kHz.

It's really tricky to justify high sampling rates, even with todays processing power. I've done a lot of A/B tests in the past and came to the conclusion that it's just not worth it, especially when todays music is mostly sold on itunes or some other compromised media. Not to mention the way most music is mastered extremely hot, causing all kinds of nasty artifacts.

For high fidelity specialty records.. sure, go right ahead but for the typical pop/dance/rock/heavy/whatever music of today I'd say it's worth putting more effort on the actual music than the technical aspects of it.

Cheers!
bManic

[khanyz]

There's the accumulation of artifacts, rather than just the single stage when mastering.

You can't always disregard them as insignificant (inaudible) and then just assume that they will stay so as they accumulate.

[A.M. Gold]

I had a feeling this would happen. We need Ethan Winer now. Now we will have the back and forth game of "Nyquist proved 44.1 is always adequate" vs. "I don't care, I can hear a difference". I've seen this play itself out to very caustic, argumentative results on other forums (but I'm assuming it's happened here before as well).

I've seen knock down drag out fights between professional audio engineers vs. people with a background in digital processing who have viciously disagreed about whether a basic Sound Blaster card from 2003 is adequate for D/A conversion (i.e. there is "no discernible difference") vs. high end converters.

The bottom line is this: no one will be able to resolve this without extensive ABX testing, and even then it may completely depend on who is tested.

Otherwise we are doomed to just go around and around again, and believe me this has happened many times before with no appreciable constructive result.

[ermi]

I'd be more willing to use higher sample rates if all plugins worked properly at higher-than-44.1k rates.
Sadly, they don't. It's something that I keep an eye on now, when trying out new plugins.

Does any of this matter, though, when we all deliver any audio we produce as mp3's (or worse) or, at best, "CD quality" audio format?

Have you actually tried ABXing good MP3 or 16/44.1 audio with 24/96 files? While I have heard MP3 artifacts with a few problematic samples, I have yet to hear the benefits of higher than 44.1/16 resolutions for the final product. But I remain open minded.

[A.M. Gold]

Does any of this matter, though, when we all deliver any audio we produce as mp3's (or worse) or, at best, "CD quality" audio format?

Have you actually tried ABXing good MP3 or 16/44.1 audio with 24/96 files? While I have heard MP3 artifacts with a few problematic samples, I have yet to hear the benefits of higher than 44.1/16 resolutions for the final product. But I remain open minded.

See my post above yours.

[khanyz]

Yes, unfortunately hearing is incredibly subjective (especially on the myriad of systems) and the maths is very stupid.

Either way, it's not something you can show and get everyone to understand, or agree.