Reaper now has built-in oversampling

[LittleStudios]

I haven't run tests, but if it functions correctly, could be game changing:

https://youtu.be/H9YwW6bNuek

I know Cakewalk by Bandlab has a similar feature, but it's quite buggy, which I have posted about and uploaded videos to YouTube as well.  This appears to be very straight forward.

[Craig Anderton]

At one point, oversampling was important. I've been an advocate of it, but selectively.  It only makes a difference with sounds or effects generated in the box. In other words, if you're just recording a vocal, oversampling won't make a difference. It can make a difference with some virtual instruments (and to a lesser extent, amp sims and limiters), but no difference with others. 

Bear in mind that more and more, effects and virtual instruments have internal oversampling. Also, you can "oversample" by changing to a higher sample rate, rendering, then returning to the original sample rate. This saves a lot of CPU compared to running everything at a higher sample rate. Finally, real-time oversampling will not necessarily be as good as the algorithms used for offline oversampling. 

In a way, I think calling it a "game-changer" is being a little late to the party. These days, when rendering virtual instruments, I'll likely just render as audio at 192 kHz. If I can hear a difference, I'll keep the audio. Of course, this likely won't work if the plug-in is already oversampling internally, so expecting to do, for example, 4X internal oversampling at 192 kHz is not going to end well. In that case, I'll render at a lower sample rate. A lot of plug-ins simply cannot oversample above 96 kHz.

But here's another consideration. I've done workshops as far back as 2015 where I compared the purer sound of an oversampled synth vs. a sound that had aliasing. It was not always a given that people would prefer the oversampled version. Although I never conducted a formal test over multiple workshops, I would say that at least half the audience preferred the sound of the version with aliasing.

[LittleStudios]

I don't want to get into a discussion on whether or not high sample rates make an audible difference.  For those who prefer to work at higher sample rates or prefer to have certain plugins run at higher sample rates, this is a definite workflow game changer.  Especially for those who have weaker computers.  This can eliminate the extra steps of bouncing out tracks at a higher sample rates, then downsampling to match the project sample rate, taking up more drive space, etc. to have the result of a higher sample rate.

This also helps address an issue that some may run into while working at higher sample rates, and that's the fact that not all plugins are created equally, having different sample rate limitations.  Having the option to control the sample rate of plugins on an individual basis is another workflow plus.

So, yeah, if you're in the camp that questions whether or not high sample rates are worth it, this feature is definitely not something to write home about.

If you're in the camp that finds that working at higher sample rates produces a desired outcome, then this could be considered game changing, especially from a workflow perspective.

 

 

[Resonant Serpent]

Oversampling seems to vary from DAW to DAW. It recently came out that Ableton doesn't recognize the internal sampling of plugs. Other DAWs seem to be hit or miss on implementation and recognizing internal sampling.

Given how buggy I found Reaper, especially when it came to MIDI, I have a hard time trusting that it would work perfectly.

[Esteban Villanova]

REAPER’s oversampling is per FX or per FX chain, ?

Edited April 29, 2022 by Esteban Villanova

[Will.]

3 minutes ago, LittleStudios said:

 

This also helps address an issue that some may run into while working at higher sample rates, and that's the fact that not all plugins are created equally, having different sample rate limitations.  Having the option to control the sample rate of plugins on an individual basis is another workflow plus.

 

(Not trying to ruin your excitement or whatever?  This is already in Cakewalk and it's exactly as crsig said: not a game changer in workflow. 

[LittleStudios]

6 minutes ago, Will. said:

(Not trying to ruin your excitement or whatever?  This is already in Cakewalk and it's exactly as crsig said: not a game changer in workflow. 

I'm aware that it's in Cakewalk, but I've shown that it's buggy as hell. To quote myself from my original post:

17 hours ago, LittleStudios said:

I know Cakewalk by Bandlab has a similar feature, but it's quite buggy, which I have posted about and uploaded videos to YouTube as well.  This appears to be very straight forward.

Here are links to videos I've made showing Cakewalk's mediocre implementation of per plugin oversampling:

https://youtu.be/CfMU93SCigU

https://youtu.be/LgupFqtLDHc

I have yet to run tests on Reaper's implementation.  So far activating and controlling the sample rate is much more intuitive than Cakewalk's approach.

[Will.]

23 minutes ago, LittleStudios said:

 I've shown that it's buggy as hell. To quote myself from my original post:

Strange! It's been working fine myside. Maybe that because i understand how to use all my FX ceiling limitations in conjunction with my project sample rate before i start to put down the first note/chord. 

Not saying you dont - of course. 

25 minutes ago, LittleStudios said:

Here are links to videos I've made showing Cakewalk's mediocre implementation of per plugin oversampling:

? Oh-boy! You're gona get heat on this. 

I'll watch your video's when I get a chance though. 

[Craig Anderton]

I've never had a problem with upsampling when used as intended according to the documentation, and wondered if maybe a recent update broke it or something. So I did a test.

1. Duplicated a drum track so there were two parallel tracks of the same events.

2. Inserted the TS-64 Transient Shaper on one track, and verified that the two tracks nulled when one was thrown out of phase (I chose the TS-64 because it's possible to have settings that don't alter the sound, which allowed for an accurate null test).

3. I enabled Upsample on Playback for the TS-64, and the two tracks still nulled. Then I did Upsample on Render, rendered, and the two tracks still nulled.

So it worked as expected, although I didn't try more exotic use cases like sidechaining, using plug-ins with different internal oversampling rates, using plug-ins whose internal oversampling doesn't allow upsampling, plug-ins with widely varying latencies, etc. 

[Esteban Villanova]

If a plugin is oversampled already REAPER will just ignore it if it’s at a higher sample rate than specified. 

[David Baay]

31 minutes ago, Craig Anderton said:

I chose the TS-64 because it's possible to have settings that don't alter the sound, which allowed for an accurate null test

Transient Shaper itself induces PDC so not the best choice. When this was first reported by the OP last year, I verified it using Channel Tools and the issue still reproduces for me in 22.02. I had to nudge the non-effected track 140 samples later to get them to null. I reported it to the Bakers but did not get any confirmation back.

[Craig Anderton]

45 minutes ago, David Baay said:

Transient Shaper itself induces PDC so not the best choice. When this was first reported by the OP last year, I verified it using Channel Tools and the issue still reproduces for me in 22.02. I had to nudge the non-effected track 140 samples later to get them to null. I reported it to the Bakers but did not get any confirmation back.

Well, it nulled, so what can I say? Maybe PDC is being applied properly to the dry track, which I believe is the way it should work. It's not hard to reproduce the test if you want to see if I'm doing something wrong in the test itself.

[David Baay]

3 minutes ago, Craig Anderton said:

Maybe PDC is being applied properly to the dry track, which I believe is the way it should work.

Yes, I think that's why it's not a good test of the Upsampling issue. With Upsampling alone, some delay is introduced but PDC is not invoked. Your test suggests that having some plugin in the project that's reporting delay to CbB causes PDC to be invoked, masking the problem with Upsampling,  especially in the case that the plugin introduces more than 140 samples of delay which I suspect Transient Shaper does since the added latency is noticeable.

I'll give it a try when I have some time.

 

 

[Craig Anderton]

I repeated the same test with Waves Sibilance Live, which they claim has no latency or lookahead, and is suitable for live use.  I obtained the same results, regardless of whether PDC was enabled or not.

So at least in these two basic, intended use cases (playback and rendering, which is how I use upsampling), I still don't know how to introduce buggy behavior so I can experience what the problem is. Upsampling seems to do what it's supposed to do. 

[Esteban Villanova]

Do you guys have a strategy for using up/oversampling? Use your ears and choose on a per-track bases? How to easily detect cramping in an eq, for example?

[LittleStudios]

Using Reaper, I set up the same test I conducted in Cakewalk.  Everything functioned as expected.  Here's a link to the Reaper oversample test:

https://youtu.be/cmDFNrf5vQk

And once again here is a link to the Cakewalk test:

https://youtu.be/CfMU93SCigU

[Craig Anderton]

 

 

50 minutes ago, LittleStudios said:

And once again here is a link to the Cakewalk test:

https://youtu.be/CfMU93SCigU

I did look at your videos, but didn't quite get the point. So I looked at the one you referenced above again, and it seems the main issue is the sidechain. So, I tried a new test where I fed a drum track into the Waves Metafilter's sidechain, which was processing a guitar track. I figured this would be a good test because the Metafilter has internal oversampling you can turn on and off. I wanted to see whether upsampling an oversampled plug-in broke the process.

Regardless of whether Metafilter had oversampling turned on and off, or the PDC status, upsample on playback worked fine. It also apparently upsampled correctly when rendered, however I couldn't just bounce to clip, because the bounce wouldn't take the sidechain signal into account. I had to export the rendered file, then bring it back into the project.

Mostly I'm looking for a) a repeatable problem in Cakewalk so that if I run into it, I'll know it's not pilot error, and b) of course I'm always interested in something if it truly is game-changing. However, I suppose there are too many variables (for example, I don't have the plug-ins you used, there's VST2 vs. VST3, the extent to which an oversampled plugin is oversampling, etc.), and I guess I just keep hitting on combinations that work. If I run into a situation where the upsampling in Cakewalk doesn't work, I'll look into it further.

[Craig Anderton]

5 hours ago, Esteban Villanova said:

Do you guys have a strategy for using up/oversampling? Use your ears and choose on a per-track bases? How to easily detect cramping in an eq, for example?

Great question. My attitude is to use oversampling if I can hear a difference. I prefer not to run high-sample-rate sessions (unless requested by clients), because of the CPU overhead.  I certainly don't want to run sessions at 192 kHz! But I have noticed with some instruments that changing the project sample rate temporarily to 192 kHz, rendering as audio, then bringing back into a lower-sample-rate project does sound better. 

Bear in mind that the audible difference can be non-existent (most of the time), subtle (sometimes), or dramatic (occasionally). Here's a good example of how much sample rate can change a virtual instrument's sound. I think you'll be shocked. If you can't hear the difference, stop reading this post immediately, and make an appointment with an audiologist However, note that this is the most extreme example I could find.

https://www.instagram.com/p/B2MXG6rgalg/

 

[LittleStudios]

3 hours ago, Craig Anderton said:

I tried a new test where I fed a drum track into the Waves Metafilter's sidechain

I decided to run a very similar test where I ran a kick sample as a sidechain to Waves Metafilter, and ran pink noise through the filter.  Here's a link to the video which provides a better explanation of the steps I took to create the result.  Keep in mind, the volume does jump.  No fear though, I provide a warning before the volume spikes.  @Craig Anderton, I'm not sure what you're doing, but so far I'm able to reliably able recreate this "bug" by simply enabling "upsampling" on playback and activating it with the "2X" button.

Here's the link:  https://youtu.be/uu2-_rmkJLs

Reasons for wanting oversampling can vary, which is not my point.  But the point of these tests is to show that Cakewalk's upsampling/oversampling on a per plugin basis needs some work.

I provide these videos to help remove as much confusion about the tests I'm running as possible.  I understand these tests aren't perfect and that there are many variables at play. 

That being said, Reaper's implementation of a per plugin oversample feature, so far, appears to be better than Cakewalk's.  I'd love to see Cakewalk improve this feature as Cakewalk is my DAW of choice.

[Starship Krupa]

4 hours ago, Craig Anderton said:

I don't have the plug-ins you used

That's easy to remedy; MCompressor is part of the Meldaproduction FreeFXBundle. A (slightly cut down) Unfiltered Audio G8 is part of the Computer Music plug-in collection.

I bought the upgrade to G8 because it's the best all-around gate I've yet to encounter, and cut my compression teeth on MCompressor. I've also since upgraded my FreeFXBundle to the "pro" version.

MCompressor in "pro" form has internal oversampling (BTW, @LittleStudios/Chris, if you want SERIOUS oversampling, Meldaproduction plug-ins, including the registered versions of the FreeFXBundle let you internally oversample to 1024X ?).

I don't think REAPER improving its oversampling is going to change any games for me. I hit a wall with perceiving any difference at 2X. Also, I don't use REAPER.

Edited April 30, 2022 by Starship Krupa