Wideband dynamic compression bug, or misunderstanding TDR Nova

[macmelon]

Hi, TDR Nova seems to me fantastic, but I came across something strange, maybe the error is on my side, I have tried many different scenarios, ratios, even with with noise, but here is an example of the issue.
I have generated a test signal, 5 sec intervals of 880 Hz Sine waves from -72dB to -1dB with 1 dB steps. I have compressed this signal with:
Wideband dynamic, 0 dB Gain, -30 dB Threshold, 10:1 ratio, fastest attack&release, no EQ Gain, no Out Gaing, 0% Dry.

What I expected from the math:
-40 dB -> -40 dB
-30 dB -> -30 dB (with a softer knee less)
-20 dB -> -29 dB (with softer knee slightly less)
-10 dB -> -28 dB
-5 dB -> -28,5 dB
-1 dB -> -28,9 dB

What I got:
-40 dB -> -40.04 dB
-30 dB -> -31.20 dB (OK, not hard knee)
-20 dB -> -23,87 dB (huge difference from the expected -29)
-10 dB -> -14,07 dB (seems almost linear from -20 dB)
-5 dB -> -9,07 dB
-4 dB -> -8,07 dB
-3 dB -> -7,07 dB
-2 dB -> -6,07 dB
-1 dB -> -5,07 dB (on the last few 1dB/1dB, not 10db/1db as the ratio would suggest)

So, this not a compressor? What am I missing?

Thanks in advance!

[Gamma-UT]

> 880 Hz Sine

> fastest attack&release

I can’t help feeling this combination is the source of the problem.

[macmelon]

Any attack (eg. slowest/slowest, 50/200 ms) produce the same. (And eg. Dead Duck work as expected.)

[heavymetalmixer]

First: No compressor out there is perfect, not evne NOVA, and they all work differently.
Second: Not many people knows this, but NOVA uses RMS compression with a very small RMS window, it will let go very fast peaks.

[macmelon]

I still don't get it: I have tried a -1 dB 40 Hz square wave, with the same 10:1 ratio, and it is -5,07 dB again. In RMS it is from circa +2 dB to circa -2 dB. And with 2:1 ratio is the same. What does the 10:1 ratio even mean?

My best bet is still a misunderstanding on my side, but please help me understand what it is!

[macmelon]

Maybe it is something like what we can see on the left, a range for the compression, but a surprisingly small range: https://www.junger-audio.com/uploads/im ... ressor.jpg

[cron]

Nova's wideband is unusual in that it can still produce strongly frequency dependent behaviour. Are all bands other than the W-band actually disabled? Parts of the spectrum covered by active bands (even if they don't appear to be doing anything) won't be included in the W-band signal IIRC.

Nova's knee is ratio-dependent incidentally - higher ratio = harder knee.

[heavymetalmixer]

Nova's wideband is unusual in that it can still produce strongly frequency dependent behaviour. Are all bands other than the W-band actually disabled? Parts of the spectrum covered by active bands (even if they don't appear to be doing anything) won't be included in the W-band signal IIRC.

This, and the fact that all bands (except the Wideband IIRC) have a 3 db per octave HPF in their side-chain, so they don't go absolutely crazy with low frequencies.

[macmelon]

Yes, I'm aware of the ratio dependent knee.

nova.png

I guess we can even see the -4,07 dB from the yellow line.

[ScrLk]

(double post)

[ScrLk]

Have you ever found an answer to this? Maybe you can contact TDR directly...

Hi, TDR Nova seems to me fantastic, but I came across something strange, maybe the error is on my side, I have tried many different scenarios, ratios, even with with noise, but here is an example of the issue.

[macmelon]

No, yes, maybe if I will have time to do so. In the meantime I have purchased the GE version, will check it too, but not yet. Thank you for your question!

[ScrLk]

Oh, I remembered something.

Are you aware that Nova has a maximum range of gain reduction?

No, yes, maybe if I will have time to do so. In the meantime I have purchased the GE version, will check it too, but not yet. Thank you for your question!

[macmelon]

Oh! This must be it.

[ScrLk]

In PluginDoctor with Limiter 6 GE's Nova compressor setting, the dynamics graph rises diagonally 1:1... then is compressed by 1:x from the threshold onwards... then reverts back to rising 1:1.

(I don't have a screenshot handy...)

Kotelnikov GE doesn't do this, unless you enable the GR Range parameter.

Whether any of this is desirable... I'm not sure. For cleaning up signals yes, for creative effects like sidechain ducking not so much.