AKG - Orban novel Compressor sidechain design

[mr coffee]

Has anyone worked with or heard a compressor design based on the pilot-VCA sidechain described in Robert Orban's US patent # 5,444,788, entitled, "Audio Compressor combining Feedback and Feedforward sidechain processing." :?:

I know the Neve SSL sidechain has some aspects of being a feedback design due to the VCA at the beginning of the sidechain, but this adds a few twists.

Check out the block diagram (assuming you already have the free .tiff viewer) at the US Patent Office server:

5444788

Here's the abstract if you haven't seen it - (hey, I know the patent is 10 years old - I've been busy, OK? )

"An audio compressor having both a feedback compressor and a feedforward compressor. The feedback compressor operates so as to provide envelope detection. A sidechain processor which receives an input from the feedback path includes a non-linear low-pass filter and a non-linear amplifier. The output of this processor provides the gain-control signal for the feedforward compressor. The main audio path is through the feedforward compressor. "

Has a nice soft-knee, and includes a "release-time linked to make-up gain" feature, too.

Sounds good on paper. Anybody heard one?

What do y'all think?

BTW-If you don't altready have the freebie altera .tiff viewer, it's a small plugin that works well. The patent office has a link or you can google it.

 

[clintrubber]

Try this, as found by NYD:

http://www.groupdiy.com/index.php?topic=11049&highlight=patent

Directly to pdf, very nice.

OK, more specific: http://www.pat2pdf.org/ & put the number in.

 

[mr coffee]

Hi Clintrubber,
Thanks! That is a GREAT link for downloading whole patent images, and they are clearer than the .tiff images, too, especially for the text. That Patent2pdf site is much better for getting nice clear figures\drawings off patent pages quickly.

Here's a link straight to the pdf version of the drawings and text for the Orban patent:

http://www.pat2pdf.org/patents/pat5444788.pdf

I sure find the US Patent Server plain vanilla large font text in full-screen width columns easier to read and study than the pdf images of tiny font text in two columns, magazine-style, like the .tiff images themselves from the patent office. 'Course my vision ain't what it used to be... :wink:

Link to USPTO text page for patent 5444788

Check out this quote from the patent:

In any compressor, the dynamic characteristics of the envelope detector or integrator (such as attack and release time constants) must be matched to the psychoacoustic requirements of the ear to minimize audible artifacts. Because in a feedforward compressor it is easy to process the output of the envelope detector without compromising the stability of the control loop, such designs have become very popular. However, the available envelope detectors (such as average or r.m.s. detectors) have tended to introduce obvious artifacts.

Some designs of feedback compressors have a dynamic response smoother than most commercial feedforward compressors. A significant amount of this smoothness results from the unique dynamic characteristics created by the feedback control.. However, these feedback designs have two important limitations. First, it is difficult to smooth the output of the integrator well enough to prevent some audible distortion. This distortion is caused by modulation effects when the gain-control signal is multiplied by the audio. Second, it is inconvenient to create a compression ratio of less than infinity because the feedback always attempts to hold the output level of the compressor equal to the threshold level of compression.

Caught my interest!

Does everyone agree with Orban that feedback compressors inherently track better?

And has anyone seen\heard a compressor with this type of sidechain envelope detector? (He adds some other goodies like a non-linear filter and soft-knee) It sounds good on paper anyway!

The GSSL clone seems to incorporate *some* aspects of this sidechain approach, and that seems to draw a lot of positive comments around here.

Or maybe this has been discussed before and I just missed it searching the archives. :?

 

[PRR]

"...it is inconvenient to create a compression ratio of less than infinity because the feedback always attempts to hold the output level of the compressor equal to the threshold level of compression."

Huh? It is not easy to get infinite compression with feedback: you need infinite loop-gain. This even contradicts the previous bit about "stability of the control loop".

It's a patent. The big trick in writing a patent application is to get legal control of an idea without actually saying what it is good for or the best way to use it. The general text is just that: general. It talks all around the problems without nailing any of them. The Claims are the actual legal points, and these are written very legally, not technically.

Note that we've bought two VCAs. Until THAT Corp cranked-up good VCA chips, this was extravagant.

Note the name Orban. They don't care about an extra VCA. They've made some VERY sophisticated Broadcast Processors. Expensive; but in a big-city broadcast market a louder signal can get you 5% more listeners and $100K more ads per year, so it is worth it. And some of the things the "best" Broadcast Processors do to audio should NOT be done to music except for mass-market air-play. (In fact if you pre-process your master brutally, and then run it through a radio processor, it may freak-out and sound awful.)

> Does everyone agree with Orban that feedback compressors inherently track better?

Rather the opposite view of the same result: if you don't have accurate VCAs, tracking a feedforward across more than a few dB is very hard. Very sloppy VCAs, like VaryMu and Photoresistors, can make great feedback limiters.

> The GSSL clone seems to incorporate *some* aspects of this sidechain

Yes. But a topology is not a sound.

Best way to build a compressor, the way most of the pre-DBX ones were designed, is to build a big breadboard, try many different things, and LISTEN.

 

[kruz]

[quote author="PRR"](In fact if you pre-process your master brutally, and then run it through a radio processor, it may freak-out and sound awful.)[/quote]

There is an excellent (imho) white paper, co-wrote by Robert Orban too, about this same issue here:
http://www.omniaaudio.com/tech/mastering.htm
highly recommended!

 

[Viitalahde]

[quote author="PRR"]In fact if you pre-process your master brutally, and then run it through a radio processor, it may freak-out and sound awful.[/quote]

I don't even want to imagine what a certain EP I was forced to make STUPID LOUD will sound like on the air..

 

[Larrchild]

I don't even want to imagine what a certain EP I was forced to make STUPID LOUD will sound like on the air.. Mad

Well....picture any portion of your waveform that is in any way flat-topped being zoomed in on by a huge magnifying glass.

Those processor designs operate on the assumption of a complete waveform arriving.=)

 

[Viitalahde]

[quote author="Larrchild"]Those processor designs operate on the assumption of a complete waveform arriving.=)[/quote]

I'm glad I don't listen to the radio much. :razz:

 

[mr coffee]

Hi Kruz,

Thanks for the link to the excellent article. That scope shot of the clipped waveform makes me cringe just imagining how harsh that must have sounded. :shock: The discussion brought up several interesting points (new to me anyway) about the assymetry of the waveforms of male vocalists and the use of phase-shift filters to even out peak amplitudes for broadcasting. I learn lots of interesting tid-bits in the oddest places sometimes. Thanks again :thumb:

Hi prr,

I don't think or believe this would be a great approach for a compressor for a whole mix by any means, and I agree with you and Viitalahde totally in that respect.

But I'm not sure if you are on target here in assuming this sidechain detection method can only be used in, or is only good for, the type of broadcast program compressors Orban is admittedly most noted for.

The whole patent BS can be a turn off, though, that's for sure :roll:

I found it useful conceptually and thought others might as well. FWIW.

 

[topcat]

IIRC the Focusrite ISA 430 (Octo pre ii dynamic is supposed to be similar) and dBX 2XX with 'density' controls use this scheme...the dBX link is not surprising because Bob Orban was involved/resposible for the dBX 286(?) voice processor design in the days when Sid Harman owned a good chunk of pro audio.

I don't recall seeing this in any of the Orban analog broadcast processors or compressors.....as they were IIRC (age) all feedback type.....

tc