[Video] FOSS Limiter Shootout

[unfa]

Here's my new video, where I test a bunch of limiter plug-ins, old and new to see which is the cleanest, most trustworthy of them all.

https://youtu.be/xinY-4eusYM


Enjoy!

[finotti]

That was a nice and helpful video, thanks for it!

I do wonder if one could get similar results to the x42 limiter with LSP limiter, though. I certainly see the point that x42 is easier, so for most there would be not much of a point, but, if for nothing else, it would be nice to learn more about the details of how a limiter (or the LSP limiter in particular) works.

[Basslint]

Great! I think it'd be awesome if you made a shootout series which covers all types of plugins

[milo]

Wow, maybe I should stop using LSP limiter. That would explain some weird behavior I have gotten from it sometimes.

[finotti]

Wow, maybe I should stop using LSP limiter. That would explain some weird behavior I have gotten from it sometimes.

If you do get something weird with LSP limiter, I'd recommend filing a bug report at GitHub. sadko4u is usually very fast in addressing issues.

[milo]

Mostly the weird behavior is what @unfa found in the video -- the transient peak is not actually limited very well by the plugin. When I put a limiter on a track hoping to prevent a clip, the clip happens anyway. That has happened to me so many times in the past year. But I just dropped in x42 limiter on the track and it solved the problem with fewer knobs to worry about. I learned a lot by having to tweak all of those knobs, but I think I'm ready for a solution that just works.

[finotti]

That is totally understandable. Go with what works best for you!

I just wanted to point out that reporting bugs is quite helpful to the developer and the community. Someone might need more fine tuning than x42's limiter allows and having a better LSP limiter would be quite helpful in that situation.

Maybe I will do some experiments and report it myself (if I can reproduce) when I find the time. (Or maybe just point out to unfa's video.)

[milo]

I agree. It would be better to have an improved LSP. I always thought it was some kind of user error on my part, but it's pretty easy to see why the behavior happens when you put the output of the plugin on the oscilloscope. Unfortunately the video wasn't very kind or fair to LSP, and I expect @sadko4u's head to explode when he sees it. Hopefully this video will help him debug/improve the DSP code, or maybe improve the documentation if it is a user error.

[sadko4u]

I've already described my position to @unfa:

1. You also definitely need to try all knobs in ALR section since ALR works as a compressor with the infinite compression ratio.
2. The LSP does not pretend to be a 'clean' limiter. It does the job it was designed for: prevent short peaks in the audio mix.
3. Trying the limiter on a single instrument is not the good comparison. I would like to recommend to make a 'crank-up' test where you try to get the maximum loudness (RMS/LU) from the final mix. I didn't try the x42 limiter but the Calf one completely lose the 'crank up' game to the LSP limiter in the heavy metal music style. Anyway, such comparison would be more informative for me since limiter is in many cases used as a last device in the mastering section.

I think I need to review a bit the limiter, maybe implement yet another additional algorithm.
But this won't be earlier 1.2.0 since I'm very deep inside of the 1.2.0 release process now.

[unfa]

If you do get something weird with LSP limiter, I'd recommend filing a bug report at GitHub. sadko4u is usually very fast in addressing issues.

I've got in touch with @sadko4u even before publishing the video to figure this out.

I've already described my position to @unfa:
1. You also definitely need to try all knobs in ALR section since ALR works as a compressor with the infinite compression ratio.
2. The LSP does not pretend to be a 'clean' limiter. It does the job it was designed for: prevent short peaks in the audio mix.
3. Trying the limiter on a single instrument is not the good comparison. I would like to recommend to make a 'crank-up' test where you try to get the maximum loudness (RMS/LU) from the final mix. I didn't try the x42 limiter but the Calf one completely lose the 'crank up' game to the LSP limiter in the heavy metal music style. Anyway, such comparison would be more informative for me since limiter is in many cases used as a last device in the mastering section.

And for the sake of completeness, here is how I replied:

I see that a single kick might not be the common use case, and a representative test fro limiters (as they usually go on master buses), though I use limiters on single drums or drum buses to cleanly control the transients as well. The single kick seemed like a very clean and easy sound to test with - it has everything that reveals problems: sharp transient, high frequencies and very low frequencies that often crunch on short release times.
One thing I saw is that LSP Limiter Mono had been applying ~4dB of gain reduction on a signal peaking at 0 dBFS, with the threshold being at 0 dB. This is very strange. And also - the attenuation did not actually limit the transient, it has let that through - so it seems to me there's a problem and it sometimes doesn't catch short peaks - at least it did not on default settings in my test.

screenshot.jpg (237.07 KiB) Viewed 7202 times

There's two things going wrong in here:
1. Attenuation happens even though signal shouldn't have crossed the threshold yet.
2. Tail of the drum being attenuated instead of the transient.

I am rewatching the video and it seems I did not touch the ALR Attack (5ms by default).
However since there's 5ms of lookahead delay I would expect the attenuation to come in time to catch the transient, and it did not.
I assume it's a bug. As you said - it's made to stop short loud bursts. And even after quite some time spent trying to achieve that by tweaking settings - I still had the loud bursts come through unaffected, while the other parts of the signal were altered in different ways.
Would you like to try and reproduce this on your end?
Here's the kick sample - it's normalized to 0 dBFS so with default settings LSP Limiter should not attenuate that, right?

True_Peak_Kick.zip

(6.2 KiB) Downloaded 78 times