I often feel that ultrasonic and subsonic filtering is like slicing a grape with an axe in a darkened room.
Would it be possible to create a plugin which can detect when these issues occur and automatically filter them?
I would like to be able to see what is going on in these areas too, but most spectrum analysers don’t extend far enough.

I agree with the grape and axe.

In case of IMD specifically, it's not easy to objectively measure their relevance. It's dangerously close to snake oil, and we definitely advise to bypass often and remain critical. The ideal is minimalism, not "filter everything". Ultrasonic filtering only really makes sense in case of highly oversampled chains. Only in this case it has the potential for a clear positive effect (that isn't snake oil).

Ultrasonic will come with a specialized analyser covering the HF to system Nyquist and beyond (for the filter curve itself). It makes it very easy to detect the presence of Ultrasonic content, pre and post filtering. But It won't tell you how much IMD your chain will produce out of this ultrasonic content. We added a "bypass all" allowing to bypass all Ultrasonic filters in the project at once.
The new Ultrasonic also has a dynamic mode that gates the HF whenever Ultrasonics appear beyond the noise floor, plus flexible slope and mix options. Even near rectangular slopes if one likes them.

I have a rule. If I can’t hear ultrasonic improve things in any way (and there are many ways it can) then I bypass it.

I don’t need a measurement device to tell me what I can or can’t hear. That’s kiddie stuff in my world. If I can’t hear it, I turn it off. This applies to any process. Sorry to be so blunt, but I didn’t get this far in the game using visuals.

That falls in the same camp as listening to which amount of oversampling best suits a plugin.
Referring to a graph for that task would not only be silly but d@mn that eats up a lot of wasted time.
The Sonnox plugs almost always sound best at 8xOS on a 44k sesh.
Lots of newer plugs require far less oversampling.
If you went by a graph you would pick the highest OS every time!

I would like to see where inaudible frequencies are being added regardless of whether I can hear any effect in MY auditory range.
These can cause issues downstream when further processing is happening and it can be useful to see where the problem is being generated.

Speaking of HF Limiters, Fabien, what's going on here? I found some hidden features in Limiter 6:

When Lim6 is run inside Metaplugin, I am able to tweak the HF Limiter Thresh, Width Thresh, Range and Width Range, all to audible effect. (And it sounds good)

So, I do basically everything in 48k and use plug-in OS when seen fit. I get the gist of Ultrasonic being for projects with higher sample rates and those wanting deeper/manual control over the sort of filtering OS would have done. If I recall, the original version doesn't work on 48k, will this be the same for the update?

My desire for using Ultrasonic is perhaps more for creative purposes than practical. It would be nice to have such a clean + steep utility LPF even if it's just to preview audible bandwidth before resampling a master to 44.1k for example. Though I can think of many other uses as well.

I see that subsonic filtering got mentioned above as well. Currently I use MAAT theEQorange as a go to for non destructive super low HPF (there's also a great free alternative, lkjb QRange). I prefer it over DC filters I've tried that certain plug-ins offer. Filtering subsonic content I find myself doing often because of samples from various sources and their various noises (turntable rumble, DC hums, etc), so if there were ever a reason Tokyo Dawn explored a Subsonic plug-in as well that would be very intriguing.

Hear Hear.

It's that old Tchad Blake quote that paraphrased goes something like 'if after three times switching back and forth I can't hear a difference, I lose interest in that difference'. ;-)

Fair point of seeing what happens where.

I suppose this could be useful for prepping masters for Vinyl LP.

Filtering out any Ultrasonics… and giving the lacquer cutter some relief.

JT

Definitely, but with a little bit of care and a little bit of tweaking, it's an excellent tool that improves program material more often than not.

Any news on the Ultrasonic filter ?

I could use it a lot right now, as I'm processing a lot of different sample rate files for a sound design session, working in 192kHz, using some compression and saturation, and exporting to 44kHz (most files are 192kHz, and the destination format needs to be 44kHz).

I'm amazed at the quantity of ultrasonic information in these files, in the source like after processing.
Even if I try to filter it with some standard EQ high cut there's still a lot left at the output.
I truly don't know what Live is doing about that when downsampling...

We're already beta testing a new TDR Ultrasonic, it's probably a matter of 1-3 weeks until release.

Great !
And what about the other plugins you were speaking about in a previous post ?

Some of these as well, at least 4 in this first push.

By the way, in the case of processing some foley sounds at 88kHz (with non linear processing producing extra ultrasonic content) and exporting them in a "ready to be used" form, without downsampling but planning these 88kHz files may be used in a different sample rate session (48kHz for exemple), would you recommend using a plugin like your ultrasonic filter (legacy one or "to be released" one) at the end of the processing, or letting the DAW do its work when downsampling (in this case, Ableton Live) ?

And is it useful to use such plugins when not using non linear fx, for exemple basic EQing ?

I guess it's not a big deal when it's to be used at the end of all processing, because your Ultrasonic is to be used between non linear processing (for exemple between a compressor and a saturator, and then again after the saturator for exemple), but who knows, I may learn something new...

Ive found that most tweeters sound/perform better without excess ultrasonic information, so my 88/96k client exports are usually filtered. I know this somewhat defeats the purpose of native high resolution files (with visible spectrum above 22khz), but it is my job to ensure the masters sound as intended at all rates so I don't care.

If the files are downsampled to 1fs later on, they will be filtered in the SRC process again, but that potential double filter situation doesn't bother me as I'm now confident the 96k and 44k files will playback near identically depending on the quality of the SRC algorithm.

If I am mistaken here, I would love to be corrected to know better.

No, not usually because they aren't generating harmonics, but I recommend you do some quick listening tests for yourself to be sure.
The real question is whether or not to use ultrasonic filtering before or after or both before and after your nonlinear plugins.

There is no clear answer to this because plugins seem to handle incoming and outgoing harmonics differently. Some plugins will exacerbate incoming harmonics into blatant IMD or aliasing, while others seem to filter before processing, but not filter afterwards, while others filter both before and after. So unfortunately it's a case by case, ear based decision.

If your plugin is using internal oversampling, and you are ok with its sound, then ultrasonic filtering will likely have little to no effect. If you use an oversampling wrapper like Metaplugin or Reaper that bookend the plugins with up and down sampling SRC, the filtering during both up and downsampling (beginning and end) might be enough to remove excess harmonics that would lead to further IMD and aliasing. I say 'might' because in some cases I notice a sonic improvement when applying ultrasonic filters inside these oversampling wrappers, so again, its a case by case basis.

The main thing is to learn to trust your ears as you audition various ultrasonic filter locations. If you cant hear a change, it may not be worth it. Some DAWs let you bypass groups of plugins (like how pro tools does on horizontal insert lines in the mixer), and I find using that to bypass ALL ultrasonic filters at once can really let my ear lock into what is happening. And then I may start bypassing one at a time to confirm their value. Some instances will have an obvious impact, while others not so much, though I may still leave them on.

@ Ghost

Thanks a lot for the detailed answer.

Is your new HF limiter going to have a mode or a preset that compliments the action of the Acceleration limiter on Simulathe?

The new version allows to "Bypass All" (instances) with a click.

Until release, we'll find an elegant way to communicate the concept. Ghost's post is spot on.

It's important to understand where, why and how such a filter is worth a try. Filtering for fun generally is a bad idea. But "oversampled projects" processing a chain of nonlinear plugins, are a valid exception, though. I think it's definitely a good idea to place an ultrasonic filter in front of the main limiter, just to have access to an alternative sound. It's also worth checking for ultrasonics, even if you decide not to touch them.

EBU R128 assumes this stuff is audible. Thus, it's maybe wise making sure your material does not contain too much of it (it's probably irrelevant in real life, though ).

Having a global bypass on any filter plugin (a bit like the global option in the UAD Studer A800) would be nice to accommodate DAWs that don't have that feature.

I agree.

I think that is what Fabien said is coming?

What kind of filter do you use for such a task ?

(By the way, I noticed with great surprise that the 2014 old Ultrasonic Filter was 100% functionnal on my M1 pro running Monterey (using the AudioUnit version in Ableton Live, so Rosetta is doing the translation job).
When I see that Plugin Alliance are not even able to update correctly their plugins for Ventura, I'm truly amazed by the TDR coding.)

Awesome - I missed that - sorry!

TDR Ultrasonic!

Fabien, are the filters in a typical SRC / oversampling algo at all similar to your ultrasonic filter? They seem to do a similar job sonically.

And are oversampling filters applied only on the way down?

Ive never read anything about SRC/OS filters, and they are not easy to measure so I'm curious.

btw I wasn't being facetious by my question back to you.

My brain didn't want to accept that this awesome 'all instance bypass' is coming so I first read it as something else entirely.

Generally, yes. They are linear phase, of the SINC type (= flat spectrum). They differ in implementation, though:

- Resampling filters are somewhat steeper than what is necessary for effective IMD suppression. An ultrasonic filter doesn't have to be that steep.

- In case of 2x upsampling for example, it's possible to skip half of the work, because only half of the bins have to be calculated (as half of the input will be zero).

- In case of 2x downsampling, it's also possible to skip half of the work. This time because every second output sample will be thrown away, so we don't have to calculate it.

- In case of resampling for processing, we even like to use so called halfband filters, which half the CPU load again. These have special properties useful in this context.

Thus, interestingly, Nyquist filtering for resampling is much easier (on the cpu) than filtering at samplerate!

Oh good question. In depends what you are doing, or more precisely, it depends whether there is a nonlinearity in between two conversions or not.

- In pure SRC (i.e. here's no nonlinearity in between), only one filter step is needed.

- When using resampling to fight Nyquist warping in IIR filters, only one filter is needed. Here, the design method is nonlinear, not the filtering itself. This means that only either the upsampling or downsampling processes really have to filter.

- When resampling for general processing, or for analogue insert (incl. nonlinear processing), one must filter both the upsampling and downsampling

Thanks for the great info. Any updates about when the new plugin/plugins will come out? I am awaiting some some more TDR goodness.

Rather soon. We are fine tuning last details and writing the docu. Beta testing is almost complete, the products are stable and already "in a relation" with several people in the test team.

Here's a small appetizer:



We plan to release it in sync with 4 other products over the next days (surfing the BF madness). Incl one for free.

Can't wait. TDR plugins are my most used 3rd party plugins.