Multiband Compressor Idea

[scott2000]

Some interesting stuff on crossover filter design in this ...... May be relevant to something somewhere although I suppose it was written in the tube era.......

https://archive.org/details/The_Recording_and_Reproduction_of_Sound_Oliver_Read_1952/page/n1

 

[pvision]

I've been thinking along similar lines recently got an HF limiter. I was looking at a Pultec EQ schematic and concluded that using a voltage controlled resistor instead of the HF Cut pot might create an HF limiter. Additional bands might get you a multiband compressor

VCRs exist but the VCA is probably a better-evolved product for this task. More knowledgeable minds than me will already know how to implement a VCA in this role and whether it's an approach worth pursuing

Nick Froome

 

[volker]

No one has anything to say about it?

In case you are referring to your reply #16: it's empty, there is no attachment visible for me.

 

[Gold]

In case you are referring to your reply #16: it's empty, there is no attachment visible for me.

Ahh. It displays on mine. It points to my server. It previewed when I posted it. I can upload it as an  attachment.  It’s a PDF. I’ll change it to a jpeg. I bet that’s it.

 

[Gold]

I've been thinking along similar lines recently got an HF limiter. I was looking at a Pultec EQ schematic and concluded that using a voltage controlled resistor instead of the HF Cut pot might create an HF limiter. Additional bands might get you a multiband compressor

That would be a dynamic EQ. A multiband compressor has a crossover. A dynamic EQ adds compressor functions to an EQ band.

 

[pucho812]

paul,


this is giving me ideas.  Obviously the best solution here is all VCA as the comp/limiters.... 

 

[pvision]

That would be a dynamic EQ. A multiband compressor has a crossover. A dynamic EQ adds compressor functions to an EQ band.

Yes. But multiple bands of EQ with compressor function = multiband compressor. It just depends which angle you look at it from

As I understand the Pultec schematic, each tuned LC section shunts a bandpass filter to ground depending on the cut pot setting

There is no crossover per se, just tuned bands in series

Nick Froome

 

[Gold]

paul,


this is giving me ideas.  Obviously the best solution here is all VCA as the comp/limiters....

Yes, the RMS delta T switch uses the VCA in its natural state to vary attack and release together. No separate variable attack and release settings.

 

[Gold]

I changed the image in post #16 from PDF to jpg. Hopefully it will display for those not seeing it.

 

[pucho812]

Have you checked out an urei la10?

As a compressor/limiter the variable rms vs peak knob might be worth looking.

 

[Gold]

Have you checked out an urei la10?

As a compressor/limiter the variable rms vs peak knob might be worth looking.

I haven't but I want separate control over each detector. Not a "crossfader" between RMS and Peak detection.

 

[pucho812]

Are they in series or parallel?

 

[Gold]

Are they in series or parallel?

Parallel. Sometimes I'd rather have them in series but which order depends on the program. I thought being able to change the order with them in series was one control too many and too tweaky. I'm trying to keep the controls to a minimum.  Looking at the faceplate may scare some with all the knobs. I think with dead simple controls in each section it helps cut down on the "too complicated" factor.

Some of the advantages of having them in series are mitigated by the Threshold Offset controls. The main reason you would want one detector before the other in series is when there is large crest factor. You can't tune the rms detector because it keeps getting tripped by the peak detector. With the Threshold Offset controls this is less of a problem.

 

[john12ax7]

Could you explain what you mean by RMS delta T?

 

[Gold]

With a VCA attack time and release time are naturally tied together. The faster the attack the faster the release. The slower the attack the slower the release. To vary them independently requires extra circuitry in a VCA. Change in the time constants is what I mean by delta T. Maybe Attack/Release is a better label?

 

[john12ax7]

Makes sense.  Thanks.

It seems doable.  The only real issue I see is the R&D time to get everything right.

Personally I like multiband compression,  but have only used plugins for it.  Is there a decent demand for multiband hardware in the mastering world?

 

[pvision]

I haven't but I want separate control over each detector. Not a "crossfader" between RMS and Peak detection.

Again, I've been thinking along the same lines. My thought was to have two sidechains, one for peak and one for RMS, and to use both to feed one VCA. The question is how to combine the control voltages - a comparator, perhaps?

It seems logical to superimpose the faster peak control voltage upon the slower RMS

How is it done in the GML compressor?

Nick Froome

 

[Gold]

Makes sense.  Thanks.

It seems doable.  The only real issue I see is the R&D time to get everything right.

Personally I like multiband compression,  but have only used plugins for it.  Is there a decent demand for multiband hardware in the mastering world?

I can’t do the circuit design but I can tweak the time constants and other stuff to get it feeling good.

I think there is  enough of a call for a hardware unit in mastering. I use the Flux Alchemist plug in and the Maselec hardware. They both have their strengths. I think this design would alleviate my need for two. When I use two one is usually set up for peak stuff and the other for rms stuff.

 

[Gold]

Again, I've been thinking along the same lines. My thought was to have two sidechains, one for peak and one for RMS, and to use both to feed one VCA. The question is how to combine the control voltages - a comparator, perhaps?

That’s how I’ve seen it done. I think you just sum the control voltages. Probably parallel the VCAs to lower distortion.

 

[Rogy]

Hi,

For multi-timing parallel sidechain inspiration, might be worth reading thru the white paper about the Safe Sound Audio - Dynamics Toolbox. Clever design: single band compressor with three different side chains, interlinked (see attach).

About the bandsplitting: How about using a subtractive VCA approach, where the main path is full range unfiltered, and the band-splitted blocks are mixed in out-of-polarity when they are used? This means, no processing in one band = no unwanted artefacts in that band. Of course the gain law is unconventional using subtractive technology, but I believe in mastering applications the amount of needed gain reduction would be rather modest...

I believe we have not yet seen this implemented in a DIY design, would be a nice educative discussion

PS about the (derivative) crossover for the band splitting, there is lots of info in Douglas Self's 'The design of active crossovers'. Most interesting reading!

Best Rogy