Nova: Uneven frequency response in the compressor

Official support for: tokyodawn.net/tokyo-dawn-labs
Post Reply New Topic
RELATED
PRODUCTS

Post

Hi,

I am a very happy user of Nova GE, but recently I found that I was not getting the gain reduction in the bass frequencies, I was expecting. I did a test, and it seems that responds very differently to the same level and threshold across the 20 Hz - 20 kHz frequency range.

The best way to show it, is with a video. I apologize in advance. It is super boring, but shows what I think is a real problem for Nova.

https://www.youtube.com/watch?v=KECE1cWHwCU

The video shows that using only the wideband compressor, there is a slight difference in the compressor response across the frequency range, but (and this is worse) using bands with the exact same settings as the wideband, there is a much larger difference in the frequency response. In this example, there is 12 dB of gain reduction around 1 kHz, but only 5 dB of gain reduction around 50 Hz.

Why is thi?

Post

In Nova, all dynamics have a 3dB/oct high pass filter in the sidechain, at 200Hz if I remember correctly.
Fabien from Tokyo Dawn Records

Check out my audio processors over at the Tokyo Dawn Labs!

Post

OK, it would be great, if that could be turned off, at least in GE.

It still does not explain the big difference in compressing the same signal using the wide band compared to using a low shelf band.

Post

In the band example, you still had the W band compression active. So you see twice the effect of the 3dB/oct energy equalization filter sitting in the SC, in total, something close to a 6dB/oct slope.

Wband compression in combination with parallel bands that are also compressing will of course do weird things. This explains the weird curve you are getting.


About the slope, technically speaking, the average frequency distribution expected by an audio device is not flat, it's very much like pink noise. The SC filter NOVA uses takes this into account. From the musical point of view, it's perfectly fine and IMHO much more correct than a flat SC. ;)

We'll consider the idea to add a flat option, though. Any good use case for it?


BTW, check out PluginDoctor instead for a much better way to analyze processors.
Fabien from Tokyo Dawn Records

Check out my audio processors over at the Tokyo Dawn Labs!

Post

Hi Fabien, first of all - thank you for replying so fast to questions - this is great. Thanks for the heads up on plugindoctor - I did not know about that one. In the past, I have used VST analyzer, but it does not work very well.

In a nutshell, I would like all bands in Nova (including the wideband) to respond equally to a pink noise signal.

Here are a few screenshots showing that it does not

First - here is the wideband response to pink noise - no other bands active
Skærmbillede 2018-04-12 07.50.25.png
Next, I activate the low band - with the same threshold as the wideband
Skærmbillede 2018-04-12 07.50.52.png
If I look at the analyzer and the threshold line in the UI, the pink noise signal is clearly above the threshold, but there is no gain reduction. If, however, I look at the Threshold knob, it seems the signal level is not above the threshold.

If I deactivate the wideband compression, nothing changes in the low band, so now there is no compression although the graph indicates, the pink noise signal exceeds the threshold.
Skærmbillede 2018-04-12 07.51.18.png
As I said before, I love Nova, and I use it for a lot of different things. But I would like to add mastering multiband compressor to my list of use cases. For this purpose, my default setting would be a high shelf and a low shelf and let the midband be handled by the wideband. I would use the same compression settings in all bands (It would be great, if it were possible to link the threshold in the bands as well). Right now, I cannot really do that due to the hidden side chain setting for low frequencies.
You do not have the required permissions to view the files attached to this post.

Post

The FFT graph crossing the threshold line is just a crude indicator. It doesn't know (=it ignores) anything about knee, the band's internal sidechain and timing. The big display generally doesn't show the sidechain signal, it shows the FFT of the input or output signal! Hence the mismatch between display and the circular SC meter around threshold (the latter is far more precise). The fact that you can see the FFT plot cross the threshold is a nice coincidence, but it never meant to show the SC signal anyway! :)

I'll make sure to add this to the manual.


To summarize and explain your observations:

- It's not exactly easy to set up nova for multi-band, because there is no crossover.
- All dynamics use a 3dB/oct in the sidechain.
- All dynamics have knee and max GR, the threshold is never crossed instantly.
- Dynamics are highly automated over several angles.
- Emulation of multi-band compressors is not intended in this design. Quite the opposite! Do not try to approach Nova like you'd do with an MB compressor, it won't work.

These aren't bugs, they are intentional and likely a reason why Nova feels nice than most technically pitiless MB compressors. :)

Maybe Vlad will pass by and drop a word or two on your observations, I'll ask him.
Fabien from Tokyo Dawn Records

Check out my audio processors over at the Tokyo Dawn Labs!

Post

Hi Fabien,

Thanks. It is all a bit clearer now.

I don't agree that Nova is not suited to be used as a multiband compressor. Your genius wideband takes care of that. A low shelf for the low frequencies and a high shelf for the high frequencies, and the wideband takes care of all the rest. That's a nice 3-band MBC - I agree that it is difficult to get more bands to work. The cross-over between a low-shelf and the wideband may not be a perfect equal loudness cross-over. But the only other MBC I have (ReaXcomp) doesn't have perfect cross-overs either. In my measurements, Nova is actually closer til equal loudness at the crossovers.

It's true that there is a 3 dB/Oct high pass in the side chain in the dynamics if you use low- and high-shelves (and bell/peak curves) - I guess the lo-cut starts around 3-400 Hz. But there is no sidechain filter on the wideband dynamics.

If only I could switch off the high pass filter on the low shelf, Nova would fulfill all my multiband compressor needs.

Post

OK this time with graphs instead of videos.

I tested a single peaking band, the wide band and a low shelf with a very high corner frequency, and a high shelf with a very low corner frequency.

Here's my setup - I use a tone generator with an output of -24 dB, and I set Nova to give 8 dB gain reduction, so the output from Nova is -32 dB
ReaperSetupForGraph.jpg
Here I plot the Threshold setting that gives 8 dB gain reduction in the four different scenarios.
Graph.jpeg
It is easy to see that there is a 3 dB/oct lo-cut in the side-chain starting at ~350 Hz for the bell curve, and the high and low shelves. But there is no lo-cut for the wideband compressor - That's a strange design choice to me. Lo-cut on the wideband but not in the 'narrow' bands would make more sense.

Please please can't we get an option to turn the sidechain filter off. It would eliminate all my FF ProMB desires.

Another feature request is to make it possible to link the thresholds between bands (including the wideband). It is already great that we can link all the other compressor settings, but if you make the option to include the threshold, it would be fantastic.
You do not have the required permissions to view the files attached to this post.

Post

FabienTDR wrote: Tue Apr 10, 2018 8:48 pm In Nova, all dynamics have a 3dB/oct high pass filter in the sidechain, at 200Hz if I remember correctly.
Is it possible to avoid this sidechain high pass filter in the latest version of Nova GE?
Last edited by omiroad on Fri Feb 14, 2020 8:05 am, edited 1 time in total.

Post

Yes, check the SC drop down for a "flat" option
Fabien from Tokyo Dawn Records

Check out my audio processors over at the Tokyo Dawn Labs!

Post

Awesome, awesome

Post Reply

Return to “Tokyo Dawn Labs”