Page 3 of 7

Re: A Discussion About True Power Summing for Stereo Compressors

Posted: Sun Mar 22, 2009 4:11 pm
by mediatechnology
I still want to make a fully variable attack/release circuit on that timing pin. What do you think?...
Go for it!

Re: A Discussion About True Power Summing for Stereo Compressors

Posted: Sun Mar 22, 2009 5:33 pm
by Crusty
Go for it?! Need to understand it first, LOL The non-linear capacitor dealy is too 'snappy', but a cap multiplier is an obvious idea. Or maybe 'flying' the timing cap?
At the moment, I'm trying to make 1510's distort nicely - so a new A/R would be a future project anyway for a dedicated mastering comp.

Re: A Discussion About True Power Summing for Stereo Compressors

Posted: Sun Mar 22, 2009 6:46 pm
by mediatechnology
Go for it?! Need to understand it first, LOL The non-linear capacitor dealy is too 'snappy', but a cap multiplier is an obvious idea. Or maybe 'flying' the timing cap?
Well, for RMS, you could consider averaging it with a resistor, a bucket of water and a thermister. Use input to heat the resistor. Varying the amount of water is your variable A/R. Take the output from the thermister and viola: Variable A/R! :ugeek:

Seriously though the non-linear C has two values. Could it be made to have more via diode steering/clamping at multiple levels? It would seem that multiple op amps could also be used and their outputs summed into the same node.
At the moment, I'm trying to make 1510's distort nicely
Clipping behavior or musicality?

Re: A Discussion About True Power Summing for Stereo Compressors

Posted: Sun Mar 22, 2009 10:23 pm
by Crusty
Multiple-leveled NLC approach sort of seems too 'zippery' to me (if I can use another technical term). I was thinking something fully variable, but that's down the road for me. I'm a little burnt out on comps right now.

Re:1500 chip - musically distort, not clipping, ha ha. Probably not user-adjustable, more of a 'signature' sort of thing... Don't tell your boss though - I know a great deal of effort has been put into making these things not distort, but...

Re: A Discussion About True Power Summing for Stereo Compressors

Posted: Mon Mar 23, 2009 9:13 am
by JR.
I don't know how useful my experience is for getting a "sound". The vast majority of my effort with dynamics processing was to make it less audible in action.

Once you start using asymmetrical attack/release the difference between Peak/Ave/RMS is kind of moot.

I would add there are additional tools in our control voltage kit besides just attack and release to mitigate the audibility of dynamics processing. "Hold" time, keeps control voltage unchanged within a window, this is a little like hysteresis. A variant on hold that is a little more complex but IMO even more transparent is a variant on an active capacitance multiplier configured as a ripple filter. i.e. small level changes see an apparent long/slow time constant, while larger level changes up or down revert to faster time constants. Additionally wrapping pre-/de- emphasis around the gain element can reduce HF artifacts in the audio path caused by rapid side chain modulation voltage changes.

I have used all of these tricks in the past with some degree of success, but I repeat perhaps only of interest if the goal is to make a comp you can't hear working.

JR

Re: A Discussion About True Power Summing for Stereo Compressors

Posted: Mon Mar 23, 2009 9:34 am
by mediatechnology
Once you start using asymmetrical attack/release the difference between Peak/Ave/RMS is kind of moot.
Definitely moot. Add A/R to an RMS detector and it becomes just a log-based level detector.

Re: A Discussion About True Power Summing for Stereo Compressors

Posted: Mon Mar 23, 2009 6:34 pm
by JR.
I designed one comp at Peavey where I used a 3 position switch for slow- med- fast, so you could keep the popular RMS and just switch in different Cap values for faster or slower general response, while still RMS. (note: mine wasn't RMS but somewhat level adaptive. I don't recall many circuit details, that was the comp using a class A VCA I patented 20 years ago...

I'd still like to do an uber-comp with tens of adjustments, but would probably have to craft a PC based interface since the hardware is too awkward with that much control.

JR

Re: A Discussion About True Power Summing for Stereo Compressors

Posted: Tue Mar 24, 2009 11:29 am
by JR.
raf wrote:
The Class A VCA you patented... Can you share the details?


rf
You can find more than you want to know from Patent office website: US04818951 Roberts.

In general Class A is not really a good thing in VCAs.. It means you have worst case noise, all the time. The new generation VCAs are class B so this noise floor is a modulation noise below the signal and well masked.

My particular patent was a variant on a fairly conventional class A VCA where I wrapped a DC feedback loop around the VCA to automatically trim out DC offsets so I could use unselected transistors to make my VCA... A few $.05 transistors and a couple opamps and presto I had my VCA. We abandoned that approach when DBX VCAs became widely available. And I used THAT VCAs in later products at Peavey.

JR

PS: Some more food for thought, I suspect similar to the way opamps combine with small caps to make them look bigger in DC servos, a clever combination of small caps and opamps (and a pot) could synthesize a variable capacitor at that integration node, so you would have effectively an adjustable speed RMS. It would retain the att/release characteristic of RMS but be variable generally faster and slower. Of course this assumes this is even useful or not, which could be proved with a mutli-position switch.

PPS: I think I could do it in two opamps, but it's is not worth drawing up, unless of determined to be of value or interest.

Re: A Discussion About True Power Summing for Stereo Compressors

Posted: Tue Mar 24, 2009 11:42 am
by mediatechnology
Some more food for thought, I suspect similar to the way opamps combine with small caps to make them look bigger in DC servos, a clever combination of small caps and opamps (and a pot) could synthesize a variable capacitor at that integration node, so you would have effectively an adjustable speed RMS. It would retain the att/release characteristic of RMS but be variable generally faster and slower. Of course this assumes this is even useful or not, which could be proved with a mutli-position switch.
That sounds very clever John.
You can use different caps on THAT's RMS detectors for variations but I am not sure if affects both attack and release in a usable way. Maybe Wayne can illuminate this?
Roger - You can vary the C and the timing current but the attack/release characteristics move in lock-step. Their relationship is fixed.

The variable A/R, which "unlocks" the two, is indeed a useful tool. I haven't had a chance to build my board. From looking at it, it appears that the response can go to very fast attack/long release. There's always room for peak in the dynamics control toolkit.

John - You had earlier mentioned the windowed approach and I did some similar thing back in the 80s using a S/H. Aphex patented one. IIRC mine looked like figure 3: http://www.ka-electronics.com/Images/pd ... pdf#page=3

I did an auto-threshold gate in the 80s for the Cochlear Implant that worked pretty well discriminating speech from slowly varying background noise.

I think Dugan patented an adaptive auto-threshold didn't he?

The point is we use a lot of adaptive tricks. The log-domain makes those tricks even easier.

Re: A Discussion About True Power Summing for Stereo Compressors

Posted: Tue Mar 24, 2009 2:45 pm
by JR.
That Aphex track and hold, that is gated on off, by the delta between input and output looks like it should do something useful but looks complicated. Perhaps intentionally busy in the patent description to conceal best implementation. In practice you'd really want the hold to actually be a slow/long time constant so you don't have it jumping between close just beyond threshold levels.

The classic hold "time" circuits are usually a fixed time or amount of hold, after attacks occur, before CV starts releasing again. Certainly much simpler and not patented AFAIK. I recall doing one design with a variable hold time, but that product got killed before it was released to production. I don't recall an adaptive threshold from Dugan, his classic automatic mixer is threshold independent, so it compares inputs to sum of inputs, perhaps something he did later?

The (IMO) best of my work in this area was my later tape NR where I had a combination of fast att/rel for very large changes, slow att/rel, for small changes, with a long term slow overall restoring response, so short term wave shape variations didn't cause long term tracking errors (especially useful for tape NR). As you can appreciate when you have assymetrical att/release the DC result varies with waveshape because of that. In my approack any such wave shape related tracking errors were kept short term, so much less audible when dealing with the tendency of tape media to scramble phase response. Probably more complexity than is useful for a comp/limiter. In fact there were even more bells and whistles in it than I described.

I think a variable integration time RMS detector could be somewhat useful, and would probably be something unique to talk about from a marketing/merchandising perspective. ;) If there is a difference between the characteristic sound of different RMS comps, this seems a likely candidate (integration time).

My hypothetical circuit requires two opamps (assuming it actually works), but I haven't really researched variable capacitance circuits. Most AFAIK use small C voltage biased diode junctions or whatever, so there may be some simpler approach in the art I am not aware of.

JR