TDR Kotelnikov, Limiter6, Nova, lookahead?

[bbaasszz]

I’ve been reading a lot of discussion on TDR plugins recently. Including these very informative ones:

https://www.gearslutz.com/board/masteri ... ssing.html
https://vladgsound.wordpress.com/2014/1 ... a-version/
https://www.gearslutz.com/board/product ... ds-10.html

No doubt TDR plugins are high-quality in many technical aspects, they deal with oversampling, alasing, IMD etc. in greatest detail. And I’m readily happy to use them as my default mixing tools as I trust the developer.

But some question comes to mind. Firstly, I found that there’s no lookahead option in either Kotelnikov GE or Limiter 6 GE, and with Kotelnikov’s fastest attack time = 0.02ms, I might still have a little transient leakage between 0 – 0.02ms? While by contrast, Nova GE has auto-lookahead. What’s the design concept behind this? Isn’t lookahead necessary if I wish to kill those transient completely?
Cannot find a suitable analyzer to test this, so I think I would post on KVR.
@3:17 He shows that compressor A - F all leave a spike before the GR go full downward.

Also I do see an anti-aliasing filter when turning on each of the modules in Limiter 6. But didn’t see the same behavior in Kotelnikov and/or Nova. I presumed these two are also non-linearities and should have oversampling (followed by a LPF above 20 kHz). But they display a flat curve in the EQ analyzer, making me wonder where the OS and anti-aliasing filters are gone.

regards,
Seb

[Telematix]

and with Kotelnikov’s fastest attack time = 20ms

Actually, Kotelnikov’s attack time = 0,02ms. But you should read about Peak Crest parameter in the manual.

[imrae]

Also I do see an anti-aliasing filter when turning on each of the modules in Limiter 6. But didn’t see the same behavior in Kotelnikov and/or Nova. I presumed these two are also non-linearities and should have oversampling (followed by a LPF above 20 kHz). But they display a flat curve in the EQ analyzer, making me wonder where the OS and anti-aliasing filters are gone.

Both of these plugins have a multi-rate design, combining an oversampled delta path with an untouched dry signal. So there are no oversampling artefacts if the plugin isn't doing anything.

[bbaasszz]

and with Kotelnikov’s fastest attack time = 20ms

Actually, Kotelnikov’s attack time = 0,02ms. But you should read about Peak Crest parameter in the manual.

I misread it as 0.02s, my bad. But still 0.02ms is not zero (which might require lookahead) to kill all the transients.

[imrae]

Kotelnikov is primarily intended for transparent processing of mixed material. "Killing all the transients" isn't really in that spirit.

As Dan Worrall demonstrated in a recent video, you could introduce lookahead by adding a negative-delay plugin to the sidechain - but this may have unintended consequences for the release portion of the envelope.

[Telematix]

I misread it as 0.02s, my bad. But still 0.02ms is not zero (which might require lookahead) to kill all the transients.

0.02ms is smaller than one sample at 96000
Also the Attack Time is NOT a pre-delay before gain reduction. This is the time it takes for the signal to become fully compressed after exceeding the threshold.

[heavymetalmixer]

Kotelnikov is primarily intended for transparent processing of mixed material. "Killing all the transients" isn't really in that spirit.

As Dan Worrall demonstrated in a recent video, you could introduce lookahead by adding a negative-delay plugin to the sidechain - but this may have unintended consequences for the release portion of the envelope.

Do you know of any free negative delay plugins?

[bbaasszz]

Kotelnikov is primarily intended for transparent processing of mixed material. "Killing all the transients" isn't really in that spirit.

As Dan Worrall demonstrated in a recent video, you could introduce lookahead by adding a negative-delay plugin to the sidechain - but this may have unintended consequences for the release portion of the envelope.

Yes I also watched that video, and it's him who makes me want to further research in this topic
He also mentioned that lookahead is one of the features that's better be built into the plugin, cos it cannot be recreated manually. So I wish my "all-in-one" compressor would have this feature. But I agree negative delay sidechain should be pretty close.

[bbaasszz]

0.02ms is smaller than one sample at 96000
Also the Attack Time is NOT a pre-delay before gain reduction. This is the time it takes for the signal to become fully compressed after exceeding the threshold.

That makes a lot of sense to me. So in other words, you're saying Kotelnikov is already very fast?

[imrae]

JS: time_adjustment in REAPER is what I use.

https://www.kvraudio.com/product/latenc ... t_sleepers looks like an alternative?

I think the ideal reduction behaviour can be achieved by

• Summing the sidechain to mono, creating identical L, R channels
• applying negative delay to the left channel only of the sidechain
• setting stereo linking to 100%

It looks complicated but would be easy in REAPER.

Stereo linking usually applies the same "use whichever reduction is largest" logic as suggested in the video.

[Telematix]

So in other words, you're saying Kotelnikov is already very fast?

Yeah, but it's also worth remembering that very small attack times can lead to unwanted distortions.

[Kr3eM]

Do you know of any free negative delay plugins?

https://www.voxengo.com/product/latencydelay/

[Recourt]

You would use Limiter6 for limiting instead of Kotelnikov or Nova. It will even catch intersample peaks.

[MogwaiBoy]

0.02ms is an incredibly fast attack time. With a high ratio and heavy reduction (like a limiter), I doubt it would let any spikes through ...