Sample rate consequences for third party plugins

[2pauluzz2]

Hi there,

For years now I've been under the impression that Live's sample rate does not make a major difference to overall sound quality of a piece of music made in Live.

Numerous audiophile debates online would make you want to believe otherwise, but many notable audio engineers and artists who I admire say it doesn't matter much, and since I never noticed a big difference I just stuck with working on 48000 for all my projects.

One of the few third party plugins that I decided to keep alongside Live's stock plugins is the OPX-PRO II. It's a stunning Oberheim-inspired synth. Today I watched a video of one of the audio banks designed by a user, and buried in the comments section he wrote "don't forget to run this plugin at 96000 for the best experience".

Since he clearly has some experience with the plugin I thought I'd try it out. Boy was I surprised.. check this out:

https://www.dropbox.com/s/huxh02xw8nz3e ... 2.wav?dl=0

You're hearing internal recordings (resampling the audio coming from the plugin to an audio track inside Ableton Live). First take is Ableton's sample rate setting at 48000hz, then 96000hz, then 192000hz.

The difference is huge.

For the tech–savvy among us: is this common for third party plugins? I'm under the impression it doesn't matter much for Live's own instruments. Could this just be a design decision for this particular plugin?

With my current computer I can't really be working at 192000 constantly and switching sample rates everytime I want to bounce an instance of OPX down seems rather silly

[Angstrom]

Most plugins use oversampling in the filter sections at least,
[edited]
Abletons effects are all mostly 2x oversampled. So that means if the computer is running at 41.1kHz then the oversampled filter is calculated 2 times faster. So 48k becomes 96k internally, oversampled. Anything which can create harmonics really benefits from oversampling, filters do too for complex reasons.
[/edited]
Lots of the Live devices like EQ8 and the distortion devices are switchably oversampled, thats the "hi quality" switch. the Cytomic filters must be oversampled as drive, saturation and filtering combined are an area where there's never enough frequency spectum to calculate in.

most VSTs oversample their filters, some oversample other areas too. Personally I think ringmod always needs to be oversampled, and any audio rate modulation.

I suspect the synth you are mentioning does not have any oversampling, which is unusual.

[Tarekith]

That was my thought as well. There's some sort of aliasing going on with the synth that is getting filtered out at higher sample rates, which used to be the case more frequently quite a few years ago. These days as mentioned, most effects and synths have some form of oversampling or other DSP tricks to make differences in sample rate less audible, or not audible at all.

Nothing wrong with a manufacturer saying that they prefer a sepcific sample rate for the sound it gives their plug in either though. Heck, evan the Roland Aira stuff was initially limited to only working in 96k.

[Stromkraft]

is this common for third party plugins? I'm under the impression it doesn't matter much for Live's own instruments. Could this just be a design decision for this particular plugin?

I run all software synths I have, that can be, in 88.1 kHz sample rate**. Projects are in 44.1 kHz**. Twice the sample rate sounds notably better.

I assume that if the synth in question doesn't have any settings for sample rate, then there is no guarantee that they will switch with the project either, as your synth seems to be doing. Naturally that would often be worse. Of course, some may just be set to a higher rate no matter what. I'm not sure if there is any way to know for sure, as this is internal. One can always ask the developer I suppose.


**Actually it seems I typically use 96kHz oversampling as projects are in 48kHz for some issue this resolved. I forgot which issue now.
I export to 44.1 kHz / 16bit (final) or 48kHz/24bit (for mastering) nevertheless.

[2pauluzz2]

Thanks guys, that's very helpful. It's good to be aware of this now.

The OPX seems to be a bit older which could explain the lack of internal oversampling. I've found a freeware alternative in OBXd that sounds surprisingly good on first listen. I'll do some testing with it tonight or later this week. If I like the sound and don't detect differences with different sample rate settings, I'll leave it at that and will not reach out to the developer of OPX. With plenty of choice in great sounding software I'll just use what works best for my setup (unfortunately for some hard working devs, it's a hard world ).

[andy_cytomic]

Hi there,

For years now I've been under the impression that Live's sample rate does not make a major difference to overall sound quality of a piece of music made in Live.

Numerous audiophile debates online would make you want to believe otherwise, but many notable audio engineers and artists who I admire say it doesn't matter much, and since I never noticed a big difference I just stuck with working on 48000 for all my projects.

Working at either 88/96 khz is the best bang for buck in terms of quality. If most of your previous recordings are done at 44.1khz then use 88.2khz, as playback of unpitched signals will hide the aliasing even for average resampling fitlers. If you have a blank slate and are targetting video and/or encoded audio files like mp3 then use 96 khz. This way all plugins get the benefit of a some oversampling to push nyquist warmping and a bit of aliasing out of the audio range, but without introducing latency and extra cpu load because each plugin has to oversample. Remember to double your audio card buffer sizes when doubling your sample rate.

One of the few third party plugins that I decided to keep alongside Live's stock plugins is the OPX-PRO II. It's a stunning Oberheim-inspired synth. Today I watched a video of one of the audio banks designed by a user, and buried in the comments section he wrote "don't forget to run this plugin at 96000 for the best experience".

Since he clearly has some experience with the plugin I thought I'd try it out. Boy was I surprised.. check this out:

https://www.dropbox.com/s/huxh02xw8nz3e ... 2.wav?dl=0

You're hearing internal recordings (resampling the audio coming from the plugin to an audio track inside Ableton Live). First take is Ableton's sample rate setting at 48000hz, then 96000hz, then 192000hz.

The difference is huge.

Yep. Synths should have oversampling settings so that you can set a realtime preview and offline render amount for the synth so you don’t have to change your project sample rate to get decent quality. I recomend 88/96 for realtime, and 352 (44*8) / 384 (48*8) khz for render. This will cover some pretty hard clipping and/or FM. If you are really pushing FM and non-linearities to the point of turning smooth shapes into square waves then you may need up to 2816 (44*64) / 3072 (48*64) khz to contain the harmonics / sidebands created. All Cytomic software supports these types of oversampling settings.

For the tech–savvy among us: is this common for third party plugins? I'm under the impression it doesn't matter much for Live's own instruments. Could this just be a design decision for this particular plugin?

With my current computer I can't really be working at 192000 constantly and switching sample rates everytime I want to bounce an instance of OPX down seems rather silly

It matters for all devices, those built into Live or third party ones - dsp is developer agnostic. 192khz is overkill for almost all audio production, but if your particular plugin doesn’t have its own oversampling settings then you either need to get a new computer than can hadle that rate, or a new plugin that has oversampling, or have the horrible workflow of changing the sample rate of the project to 192 khz, then bouncing the track down to 96 khz (Live will first bounce at the project rate and then resample that file to the target rate - which is what you want when bouncing to a lower than project sampler rate), then drag those files back into your project. Probably best just to grab a new plugin that supports oversampling. If a developer takes the time to implement decent oversampling then it’s a good sign they may have spent the time also working on other parts of the plugin as well.

[andy_cytomic]

Most plugins use oversampling in the filter sections at least, Abletons stuff is all mostly 4x oversampled. So that means if the computer is running at 41.1kHz then the oversampled synth filter is calculated at 176.4kHz. 4 times faster. So 48k becomes 192k internally, oversampled.
Anything which can create harmonics really benefits from oversampling, filters do too for complex reasons.

I am not aware of any of “Abletons stuff” that is 4x oversampled. Can you please be specific about some Live devices that do this?

Lots of the Live devices like EQ8 and the distortion devices are switchably oversampled, thats the "hi quality" switch. the Cytomic filters must be oversampled as drive, saturation and filtering combined are an area where there's never enough frequency spectum to calculate in.

Yep, this enages x2 oversampling from the current project rate. This is equivalent to running those devices at 96 if your project is 48, so when switching your project to a higher rate most of these “Oversampling” switches can be disabled and you get the same results. Alternatively you could view this as adding the oversampling option to ALL plugins and devices at once with no added CPU to do the oversampling (up sampling then downsampling high order FIR/IIR filters) or added latency. The CPU will double because you are processing twice as many samples per device, but with per device oversampling it would more than double and each plugin will introduce around 2-50 samples of latency (depending on the type of oversampling used).

most VSTs oversample their filters, some oversample other areas too. Personally I think ringmod always needs to be oversampled, and any audio rate modulation.

I suspect the synth you are mentioning does not have any oversampling, which is unusual.

I would be wary of a plugin that doesn’t support its own oversampling. There are the rare plugins that automatically handle the oversampling for you, but if so then both the cpu and audio quality should be no different between 44/48 and 88/96 since internally it will be always running at 88/96. Some plugins may even only run at a fixed sample rate and then reample their outputs to the project rate. I would guess that’s what anything by Roland would do since they are mainly hardware guys and would probably only get things working at one sample rate as this is all that is needed for hardware synths.

[Angstrom]

Most plugins use oversampling in the filter sections at least, Abletons stuff is all mostly 4x oversampled. So that means if the computer is running at 41.1kHz then the oversampled synth filter is calculated at 176.4kHz. 4 times faster. So 48k becomes 192k internally, oversampled.
Anything which can create harmonics really benefits from oversampling, filters do too for complex reasons.

I am not aware of any of “Abletons stuff” that is 4x oversampled. Can you please be specific about some Live devices that do this?

Uh, crap, I misremembered!
I think the Saturator, Overdrive, Chorus, Flanger, Valve, and obviously the Glue and EQ8 use 2x oversampling. Which I misremembered as 4x. Doh! Probably some wishes leaking into my memory module.
I think the manual only talks about oversampling in nebulous terms like "hi-quality mode"so there's little clue. My assumptions about what "Hi-Quality" means is mostly derived from what other developers have said on this forum. Hence my crappy memory of it.

[2pauluzz2]

Andy and Angstrom, thanks a lot for adding further details and continuing the discussion. I've learned a great deal and am happy to have some specific pointers for sample rate settings to work with

[jlgrimes]

Hi there,

For years now I've been under the impression that Live's sample rate does not make a major difference to overall sound quality of a piece of music made in Live.

Numerous audiophile debates online would make you want to believe otherwise, but many notable audio engineers and artists who I admire say it doesn't matter much, and since I never noticed a big difference I just stuck with working on 48000 for all my projects.

One of the few third party plugins that I decided to keep alongside Live's stock plugins is the OPX-PRO II. It's a stunning Oberheim-inspired synth. Today I watched a video of one of the audio banks designed by a user, and buried in the comments section he wrote "don't forget to run this plugin at 96000 for the best experience".

Since he clearly has some experience with the plugin I thought I'd try it out. Boy was I surprised.. check this out:

https://www.dropbox.com/s/huxh02xw8nz3e ... 2.wav?dl=0

You're hearing internal recordings (resampling the audio coming from the plugin to an audio track inside Ableton Live). First take is Ableton's sample rate setting at 48000hz, then 96000hz, then 192000hz.

The difference is huge.

For the tech–savvy among us: is this common for third party plugins? I'm under the impression it doesn't matter much for Live's own instruments. Could this just be a design decision for this particular plugin?

With my current computer I can't really be working at 192000 constantly and switching sample rates everytime I want to bounce an instance of OPX down seems rather silly

This depends on the synth design.

Generally, running at double sample rate should be similar to "2X Oversampling" if your plug-in supports it. On many plug-ins running at double sample rate can sound "cleaner" but that comes with a caveat.

1. Some plug-ins are designed to sound best at 44100 or 48000.

2. There could be bugs or features that don't work at double sample rate on some plug-ins.

3. More CPU intensive.


Generally though it is probably better to just use plug-ins with oversampling switches built-in if you can which gives you the ability to "pick and choose" what processes you want to oversampling as some might sound better than others or not even produce any noticeable effect (Some plugins run at 2x oversampling by default and not all of this is probably documented). Also some plugins can go a lot beyond double (2x) sampling. That can save some CPU cycles (or at least give you a little more control).


But that said it is an experimental thing. I was doing projects at 88.2 for a while. Using Sonar and Reason most stuff did sound "cleaner" especially when using most standard stock plugins in Sonar. But I also remember in Reason one time using a distortion plug-in where it didn't sound "right" at 88.2. It definitely lost some of its oomph. I think most of the reason people use higher sample rates is to avoid aliasing (I could be wrong here) but there are various methods of getting rid of aliasing and sometimes aliasing might be a part of the "character" of a plug-in so that's probably why experimentation is important.


I'm back at 44.1 kHz now and just tend to pick and choose plug-ins now when mixing.