Quantcast
Limiters and Bass Intermodulation Distortion - Page 2 - Gearspace.com
The No.1 Website for Pro Audio
Limiters and Bass Intermodulation Distortion
Old 12th February 2009 | Show parent
  #31
Lives for gear
 
Alexey Lukin's Avatar
 
Verified Member
🎧 10 years
Quote:
Originally Posted by fader8 ➑️
10mS release times, 1.6dB gain reductions.
On some of your graphs you can even see that some limiters are driven much harder than others (e.g. Flux against Kjaerhus). Doesn't look like a fair comparison to me.
Old 12th February 2009 | Show parent
  #32
Gear Maniac
 
fader8's Avatar
 
🎧 10 years
Quote:
Originally Posted by Alexey Lukin ➑️
On some of your graphs you can even see that some limiters are driven much harder than others (e.g. Flux against Kjaerhus). Doesn't look like a fair comparison to me.
Hey Alexey!
I am TOTALLY open to suggestions on making these plots fair. Please feel free to critique the the test conditions. Keep in mind that the spirit of this isn't a "performance" test. I'm not advocating any one limiter or anything, ie it's not a marketing ploy!. It's just an evaluation of the flavor of these things. But I think some additional disclaimers on that page might be prudent, eh?

Anecdotally, here's an example of what I mean. I'm working a video mix right now that has some pretty wild stuff on the LFE channel. It's not music, it's effects. Normally I'd just throw an LA2A on it to keep it under control, but I'm trying out some of the "noisier" limiters in this bunch and I gotta tell ya, some of these are going to be in my permanent collection of LFE sweetening tools! But I digress. Having not been much of a limiter user in the past, I'm really just trying to figure out what makes these things tick. This inter-relationship of release time and distortion is fascinating, not just from a very occasional mastering need but from a sound design standpoint as well.

As I mentioned previously, I'm not clear exactly what signal levels some of these are using to initiate their release phase. I believe that's why some are showing an increase in the close sidebands, but I haven't confirmed that yet. Perhaps you or others have some insight on this. If I get some more downtime, I may try modeling this up in Kyma, as then I'd be able to control this release threshold parameter and watch what happens.

So any and all suggestions are more than welcome! Cheers.
Old 12th February 2009 | Show parent
  #33
Lives for gear
 
Alexey Lukin's Avatar
 
Verified Member
🎧 10 years
I already outlined the conditions that I consider fair: equal peak and RMS levels of all signals (in your test they are not equal) and equal threshold levels in limiters. If you upload the waveforms, I can show this.
Old 12th February 2009 | Show parent
  #34
Lives for gear
 
John Suitcase's Avatar
 
🎧 15 years
It seems to me that a 40 hz sine wave would require a release time of at least 13-15 ms. A 40 hz sine would take 25 ms to complete a regular cycle, if rectified, it still would take 12.5 ms. So, to be safe, you'd want an extra ms or so, right?

My math may be completely wrong here, so...

I often hi-pass my mixes so they have a little bump around 50-60 hz, and everything below 45 is cut pretty severely. That wouldn't work well for those sub effects, though!
Old 12th February 2009 | Show parent
  #35
Gear Maniac
 
fader8's Avatar
 
🎧 10 years
Quote:
Originally Posted by Alexey Lukin ➑️
I already outlined the conditions that I consider fair: equal peak and RMS levels of all signals (in your test they are not equal) and equal threshold levels in limiters. If you upload the waveforms, I can show this.
I'm not quite following you here, Alexey. If you're referring to the two tones in the source signal, the attached pic is the instructions I gave the sig gen, and you can get the file here:

Here's the source file.

But if you're referring to matching the rms levels of the outputs of two different limiters, that's tricky, since not only the distortion varies that, but the stability of the rms does too. Anyway, I'm game to follow your advice, I just need you to be more specific. Thanks!
Attached Thumbnails
Limiters and Bass Intermodulation Distortion-siggen_settings.jpg  
Old 12th February 2009 | Show parent
  #36
Lives for gear
 
Taurean's Avatar
 
🎧 15 years
I had a compression test on here a while ago. In my opinion, these plugin tests should be two fold. One, have a default set of parameters and set each limiter to those parameters and then observe how each one reacts with those parameters; how each behaves: gain reduction, distortion, etc. Then two, have a set goal of a sonic result or such as an rms you are aiming for and adjust each limiter's settings (regardless of how they end up) to achieve such results and then observe that.

oh, and thumbsup for elephant

-bob
Old 12th February 2009 | Show parent
  #37
Gear Maniac
 
fader8's Avatar
 
🎧 10 years
Quote:
Originally Posted by John Suitcase ➑️
It seems to me that a 40 hz sine wave would require a release time of at least 13-15 ms. A 40 hz sine would take 25 ms to complete a regular cycle, if rectified, it still would take 12.5 ms. So, to be safe, you'd want an extra ms or so, right?

My math may be completely wrong here, so...
No, your math's not wrong, but take a look at the envelope/power history attached. That's a 25mSec slice. For the history, blue shading is rms (based on some sample period) and the trace is peak. So, if my logic is correct, the limiter is initiating an attack and release phase 4 times during this period.

Edit: The image shows the source file used in these tests.
Attached Thumbnails
Limiters and Bass Intermodulation Distortion-25ms_history.jpg  

Last edited by fader8; 12th February 2009 at 09:53 PM.. Reason: Needed to clarify
Old 12th February 2009 | Show parent
  #38
Lives for gear
 
John Suitcase's Avatar
 
🎧 15 years
What settings do this image represent? Just curious, I'm interested in this subject, too!
Old 12th February 2009 | Show parent
  #39
Gear Maniac
 
fader8's Avatar
 
🎧 10 years
Quote:
Originally Posted by TranscendingM ➑️
I had a compression test on here a while ago. In my opinion, these plugin tests should be two fold. One, have a default set of parameters and set each limiter to those parameters and then observe how each one reacts with those parameters; how each behaves: gain reduction, distortion, etc. Then two, have a set goal of a sonic result or such as an rms you are aiming for and adjust each limiter's settings (regardless of how they end up) to achieve such results and then observe that.
LOL, and you're volunteering me to do all that work, eh?!?!

OK, but at the rate we're moving, it could take a while! As it stands, this thread is specifically about the effects of limiting relatively steady-state bass signals. But it holds for compression too, just the timing and detection parameters, etc., are different. This isn't a "shootout", just a way to explore what may work well under certain conditions. I may start another thread about transient handling with these, but for now, one thing at a time!
Old 12th February 2009 | Show parent
  #40
Gear Maniac
 
fader8's Avatar
 
🎧 10 years
Quote:
Originally Posted by John Suitcase ➑️
What settings do this image represent? Just curious, I'm interested in this subject, too!
That's the source file. Sorry, I should have mentioned that.
Old 12th February 2009 | Show parent
  #41
Lives for gear
 
Taurean's Avatar
 
🎧 15 years
Quote:
Originally Posted by fader8 ➑️
LOL, and you're volunteering me to do all that work, eh?!?!

OK, but at the rate we're moving, it could take a while! As it stands, this thread is specifically about the effects of limiting relatively steady-state bass signals. But it holds for compression too, just the timing and detection parameters, etc., are different. This isn't a "shootout", just a way to explore what may work well under certain conditions. I may start another thread about transient handling with these, but for now, one thing at a time!
Ahah nah man, was just generally throwing it out there. I do however appreciate you got the ball rolling with these.
Old 12th February 2009 | Show parent
  #42
Lives for gear
 
Alexey Lukin's Avatar
 
Verified Member
🎧 10 years
Quote:
Originally Posted by fader8 ➑️
But if you're referring to matching the rms levels of the outputs of two different limiters, that's tricky, since not only the distortion varies that, but the stability of the rms does too.
Yes, that's what I meant: RMS and peak levels of the resulting files should be equal. You are trying to put limiters into equal conditions by setting the "same" release time: 10 ms. However all limiters interpret these 10 ms differently: some of them release faster, some - slower; the shape of the release varies as well, not to mention PDR/ARC/IRC features. As a result - they all limit with different speeds, hence different RMS of the resulting files. See e.g. your graph of Kjaerhus vs. UAD: they measure -5.58 dB vs. -5.02 dB. Although seemingly small, this difference in RMS does a lot to the level fo distortion.
The test that I'm proposing would require equal peak and RMS levels from all limiters, down to +/-0.01 dB. This should be achieved by varying the aggressiveness of limiters, not the threshold level. Aggressiveness can be varied by the Attack/Release/Hold/Character/PDR controls. And only after all limiters achieve the same speed (read "same RMS", "same loudness"), it's fair to plot distortion spectra.
Old 13th February 2009 | Show parent
  #43
Gear Maniac
 
fader8's Avatar
 
🎧 10 years
Quote:
Originally Posted by fader8 ➑️
So, if my logic is correct, the limiter is initiating an attack and release phase 4 times during this period.
OK, I take that back, John. Having eaten something, and looking more closely, and based on a -2.6 threshold giving it 1.6dB gain reduction, (because the input signal peaks at -1dBFS) I don't think it is happening 4 times. It will only trigger the threshold when the waveforms are additive by some amount. So my guess now is that it's likely averaging 3 times during any 25mSec period. The 2 frequencies are not harmonically related, so there's a bit of randomness there.

Alexey,
OK, understood. I'm going to chew on that and return with some thoughts on it. Thanks.
Old 13th February 2009 | Show parent
  #44
Gear Maniac
 
fader8's Avatar
 
🎧 10 years
Quote:
Originally Posted by Alexey Lukin ➑️
You are trying to put limiters into equal conditions by setting the "same" release time: 10 ms. However all limiters interpret these 10 ms differently: some of them release faster, some - slower; the shape of the release varies as well, not to mention PDR/ARC/IRC features. As a result - they all limit with different speeds, hence different RMS of the resulting files.
Absolutely agreed. To minimize this to an extent, I've kept any ARC or PDR functions disabled. This only looks at the devices in their manual modes. There's nothing fundamentally unfair about loading up a limiter, setting it to manual, choosing a desired gain reduction and release time, then listening or measuring. It's just one data point, and a common one from a typical user perspective. But I am by no means diminishing your point, only saying that it's a valid next step in further refining results.

As you say, limiters will interpret various settings differently. Since we don't have absolute control over many of those parameters, we can only measure what they provide. From an engineering perspective, we'd ideally like to have all things equal. But considering the limitations of the control interfaces and the different implementations, I'm not sure it's going to be practical. But I'm keeping an open mind.

Why don't you take a crack at this yourself? Pick any two of the limiters in this test to try and we can work through them together. (Just don't pick the Flux or Oxford as I'm only running time limited demos of those!) No rush, just as your time permits.

Quote:
Originally Posted by Alexey Lukin ➑️
See e.g. your graph of Kjaerhus vs. UAD: they measure -5.58 dB vs. -5.02 dB. Although seemingly small, this difference in RMS does a lot to the level fo distortion.

The test that I'm proposing would require equal peak and RMS levels from all limiters, down to +/-0.01 dB. This should be achieved by varying the aggressiveness of limiters, not the threshold level. Aggressiveness can be varied by the Attack/Release/Hold/Character/PDR controls. And only after all limiters achieve the same speed (read "same RMS", "same loudness"), it's fair to plot distortion spectra.
Right, so, pursuant to that, would you agree that setting them up with a single pure sine wave first, matching rms out levels, would be the correct calibration procedure? Just an FYI, I tried this with the Kjaerhus and Flux using the same settings as the last test and they do match rms and peak output levels within .01dB. Of course the crest factor was 3dB for both as it was a single sine wave. Obviously the intermod test signal cannot be used for calibration as the power spectral density of the distortion itself impacts the rms measurement too much.

Anyway, food for thought.
Old 13th February 2009 | Show parent
  #45
Lives for gear
 
Alexey Lukin's Avatar
 
Verified Member
🎧 10 years
What I did is I matched peak and RMS leves right on your intermodulation waveform. I used your exact settings in Kjaerhus, but to match Kjaerhus'es 10 ms, I had to set release of L2 to 60 ms, Ozone Brickwall to 56 ms, Elephant 3 to 26 ms (the rest of settings match yours), Ozone Intelligent II Character to 7.6.

Forcing crest factor to be the same also forces the amount of distortion to be similar. However the spectral shape of the distortion is different, see the graphs.

Old 14th February 2009 | Show parent
  #46
Gear Maniac
 
fader8's Avatar
 
🎧 10 years
Quote:
Originally Posted by Alexey Lukin ➑️
Forcing crest factor to be the same also forces the amount of distortion to be similar. However the spectral shape of the distortion is different, see the graphs.
Well, this is revealing. Mostly due to the extraordinary similarity in the close sidebands, and that makes sense as their being the largest magnitude they make up the bulk of the rms difference. I'm trying this with the Kjaerhus and L2 at the moment and I've attached a plot using your method. As you can see, I've matched crest factors exactly, but had to increase the L2's release to 42.5 mSec's. (L2 in blue).

Tell me if I'm thinking straight here. If we view the close sidebands as a common intermodulation product of the gain compression, ie I mean common to all these dynamic processes, then the additional artifacts are what make each limiter unique.

If that's true, then we have to make a fundamental paradigm shift in what we may consider "common practice" for limiter use. If we look at the snippets in this thread from a couple of our esteemed colleagues:

Quote:
Originally Posted by kjg ➑️
1.6 dB at 10 ms release seems to me like more realistic real world "limiting for loudness" use,
Quote:
Originally Posted by csl ➑️
I second the suggestion of testing the limiters at 10ms -- more of a real-world value, perhaps.
It would seem that our notion of what "works" in certain applications, (our typical go-to settings) needs to change as we try one limiter or another. If this is the case, how is the user to determine this when going about limiting a kick drum, for example. It's clear that these products of the release cycle are going to be present in any signal, steady-state, transient or otherwise, as long as long as there is more than one attack and release cycle during the event.

Assuming that a look-ahead control allows us to change how soon the limiters attack occurs before the onset of the waveform, we have no such control, like a hysteresis, that allows us to control when the release starts on the downward slope of the waveform. Or is that what, for example, the "smoothing" control does on the Sonalksis?

One question just begets another!
Attached Thumbnails
Limiters and Bass Intermodulation Distortion-k-l2_cm.jpg  
Old 5th March 2009 | Show parent
  #47
Gear Head
 
🎧 10 years
Kjaerhus...is anybody there....?

After reading this thread I was about to purchase the Kjaerhus limiter. But looking at the website and the forum support, it seems that there's nobody there anymore....
everybody is complaining that the support doesn't answer neither through the forum nor emails....
People can't get their authorization codes it seems....I'm afraid to buy the limiter and not get the codes to have it work.....
anybody bought something recently from Kjaerhus ?

Juan
Old 6th March 2009 | Show parent
  #48
Gear Head
 
BertilAlving's Avatar
 
🎧 10 years
Hi fader8,

Thanks for your excellent investigation! Please notice that you've made a setup mistake at your 2nd webpage: The release time of UAD Precision Limiter is 0.10 ms, NOT 10 ms (therefore giving more distortion).

Regards,
Bertil
Old 6th March 2009 | Show parent
  #49
Gear Maniac
 
fader8's Avatar
 
🎧 10 years
Quote:
Originally Posted by BertilAlving ➑️
Thanks for your excellent investigation! Please notice that you've made a setup mistake at your 2nd webpage: The release time of UAD Precision Limiter is 0.10 ms, NOT 10 ms (therefore giving more distortion).
Thanks for catching that, Bertil! I'll fix that soon. I've been meaning to get back to this little study, provide better explanation, incorporate some of the stuff discussed with Alexey here, etc. I got waylaid by real work! Anyway, I'll get back to it as time permits.

Thanks again!
Old 18th March 2009 | Show parent
  #50
Lives for gear
 
Table Of Tone's Avatar
 
Verified Member
8 Reviews written
🎧 10 years
I did buy the Xenon thinking it was gonna be good but after a couple of records, I realized I didn't like it.
The two lookahead modes can be real snare vampires!
Old 18th March 2009 | Show parent
  #51
jdg
Lives for gear
 
jdg's Avatar
 
Verified Member
24 Reviews written
🎧 15 years
Quote:
Originally Posted by Table Of Tone ➑️
snare vampires!
Old 19th March 2009 | Show parent
  #52
Lives for gear
 
Table Of Tone's Avatar
 
Verified Member
8 Reviews written
🎧 10 years
Yeah OK!
You like that?
I'm not measuring s**t.
I don't wanna hang waveforms up on the wall and analyze em!
I'm just listening to it.
It either sucks or it rocks!

Cheers
Old 22nd March 2009 | Show parent
  #53
Lives for gear
 
🎧 10 years
Hi, fader8

interesting thread, should be in the gear shoot outs!
anyway...
i think that limiters should be judged by the Ear,
with a DACs based in AKM chips that can reproduce inter-sample peaks.
or with free SSL X-ISM.
www.solid-state-logic-com

THD like choresterol, thers good and bad ones.
omega3&6 vs. trans fats.

done diferent listening tests, sadly i didnt recorded the files,
tests at 15dB of Gain or -15dB of Limiting, near the Limiter Limit.
that gives a nice idea of the sound of each plugin.
also done tests at -3dB / - 6dB
with diferent source files,
also tests over a mastered song. over-mastering.
that gives another level of understanding of the distortion & look ahead capabilities of each plugin.

ive tryed to test all limiters ive could,
KVR: Virtual Instruments, Virtual Effects, VST Plugins, Audio Units (AU), DirectX (DX), Universal Binary Compatibility - Audio Plugin News, Reviews and Community
i only miss UAD and powercore MD3 brickwall, and voxengo elephant.

have found that some sound verry analog, some are verry transparent sound, some are colored sound, some are very loud, some are limited, some are free, some create clipped waves, some create intersample peaks.
but all comes down to personal taste..

also, soft-sat is another piece of the pie...
and thats another story.

its intersting to see&hear that most million dollar selling records latelly, have too much inter-sample peaks, and/or clipped square waveforms.
and thats a big issue in big sound systems, and also with DACs that are sensitive to intersample-peaks.

THD in the Bass region is nothing compared to Intersample Peaks in the mid range.
Old 30th May 2009 | Show parent
  #54
kjg
Lives for gear
 
🎧 10 years
50% off at Kjaerhus

For whoever doesn't own the MPL-1 yet and is longing for more flavors of IMD to choose from, this excellent plugin is currently on sale for $59 for either mac (au) or win (vst), while $74 buys you a license for both platforms (au/vst/rtas).

Kjaerhus Audio - Professional Audio Plug-ins
Old 31st May 2009 | Show parent
  #55
Lives for gear
 
Jesse Graffam's Avatar
 
🎧 10 years
Quote:
Originally Posted by Table Of Tone ➑️
Yeah OK!
You like that?
I'm not measuring s**t.
I don't wanna hang waveforms up on the wall and analyze em!
I'm just listening to it.
It either sucks or it rocks!

Cheers
Well then good, because this thread never was about subjective research, but objective research, so you are correct: this is not the thread for you.

My clipper gets totally pwnd in this test before 1.2kHz and above (in comparison to no distortion, it fairs ok against the limiters although I only bothered with 16bits), but feed it some normal program material and it's a whole different story.



And also this was with 1.6db of reduction straight up, so the RMS increase is possibly more than double that of the limiters. However advanced, it is a clipper, especially for simple test tones like this.
πŸ“ Reply

Similar Threads

Thread / Thread Starter Replies / Views Last Post
replies: 217 views: 47601
Avatar for wavemechanic.net
wavemechanic.net 8th November 2016
replies: 323 views: 105347
Avatar for VSTSlut
VSTSlut 11th February 2015
replies: 152 views: 38103
Avatar for samzilla
samzilla 1st March 2016
replies: 75 views: 18035
Avatar for Robb Robinson
Robb Robinson 6th June 2021
Post Reply

Welcome to the Gearspace Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…

Forum Jump
Forum Jump