MClass Multband Compressor with a steeper slope

[Ahornberg]

I compared several Multiband Compressor REs to that Multiband Compressor Combinators that can be found in the MClass Mastering Patches folder and I found that the MClass Combinators work phase linear and totally cancel out to a phase-inverted parallel channel if the compressors are bypassed.

None of the Multiband REs do cancel out completely (except the Splex in BSL mode with the same flat slope as the crossover provided by the MClass Stereo Imager).

Then I found out that at the 8-Band Compressor in the MClass Mastering Patches folder some crossover slopes are steeper than the others (about 12 dB). So I tried out patching cables to build a 3-Band Compressor with a steeper slope that works phase-linear ... an here it is

3 Band Compression.zip

(1.8 KiB) Downloaded 67 times

Beside the MClass devices I used the free RE CV8X4 https://shop.propellerheads.se/product/ ... generator/ to extend the limited count of knobs provided by the combinator (hey Props, my 10 years old Kore controller has 8 knobs and 8 buttons).

So feel free to use it and tell me how you like it!

3 band compressor.png (1.45 MiB) Viewed 1816 times

[Marco Raaphorst]

Cool! You need to open the Combi to make adjustments per band of course.

[Abstrax]

This is cool. By any chance, can you explain it? Are the multiple imagers making the steeper slopes? I can't figure it out.

[Ahornberg]

A brief explanation how it works:

multiband explanation.png (32.64 KiB) Viewed 1725 times

The blue and red lines are showing how the crossover in the MClass Stereo Imager works (upper picture).

In the lower picture the crossover is applied on the red line of the upper picture. The blue line (with the steeper slope) is sent to the compressor.

The red line (marked with "uncompressed" is sent to the output. Each crossover produces two of this artefacts, so there is some parallel copression happening in my multiband compressor combinator. Because the MClass doesn't introduce phase shifts, applying parallel compression works fine. If you want more parallel compression, you can use a parallel channel (try this on the multiband REs to hear the difference).

Compared to other mulipand compressors my version works more in the center of the band and less near the crossover frequencies. This is because auf the uncompressed artefact mentioned above.

multiband comparsion.png (17.28 KiB) Viewed 1725 times

[orthodox]

Because the MClass doesn't introduce phase shifts...

You mean the Stereo Imager? AFAIK, it does. The crossover consists of two 2-pole LP and HP filters, which sum to identity. Would they be 4-pole bi-quad filters, they could be zero phase.
Btw, they are not even symmetric, they don't mirror to each other on the graph.

[Ahornberg]

Because the MClass doesn't introduce phase shifts...

You mean the Stereo Imager? AFAIK, it does. The crossover consists of two 2-pole LP and HP filters, which sum to identity. Would they be 4-pole bi-quad filters, they could be zero phase.
Btw, they are not even symmetric, they don't mirror to each other on the graph.

I did a lot of tests using a parallel channel with inverted phase and if I turn the compressors in my combinator to bypass, it cancels out to silence completely. I'm not an expert in audio theory but I guess the Stereo Imager uses a Bessel filter.

[orthodox]

if I turn the compressors in my combinator to bypass, it cancels out to silence completely

This only proves once again that the two sides of crossover with equal gain sum up to identity. Phase shifts cancel out in this case. When they are different, the phase shift can be seen.
I made a test and drew a graph about two years ago. There's a chance I could make mistake, but they were not Bessel.

[Ahornberg]

here's my simplyfied test setup

[orthodox]

here's my simplyfied test setup

Yes high + low = 1, but this has nothing to do with phase. Both bands introduce phase shifts, but they are annihilated in this sum.

[Ahornberg]

here's my simplyfied test setup

Yes high + low = 1, but this has nothing to do with phase. Both bands introduce phase shifts, but they are annihilated in this sum.

Oh, so my testing setup is wrong .... so I have to go back to the start

[orthodox]

Oh, so my testing setup is wrong .... so I have to go back to the start

I am not sure if this ruins your design. I only wanted to make a remark about phase. Phase shifts are ok, they are on the slopes where the signal level is suppressed.

[Ahornberg]

Oh, so my testing setup is wrong .... so I have to go back to the start

I am not sure if this ruins your design. I only wanted to make a remark about phase. Phase shifts are ok, they are on the slopes where the signal level is suppressed.

Yes, phase shifts are not a problem until I do parallel processing.

[orthodox]

Btw, the M-Class Stereo Imager crossover response graphs, just plotted on real data:

[Ahornberg]

Interesting! The blue line goes down to -0.9! That explains some strange notches I encountered using Reason's spectrum graph.

Btw. what software did you use for that plot?

[orthodox]

Btw. what software did you use for that plot?

Wolfram Mathematica

I prepared an audio clip with a sine wave of gradually increasing frequency and fed it to the device. A cosine in the left channel and a sine in the right channel, so that I could measure phase shift on the output. And it is more convenient with level measuring, too, as in this case the level remains constant at every instant and never gets 0.