Simple Pulse Width Modulation Compressor Design

[PRR]

Inovonics was/is a strong player in broadcast dynamics, using PWM.

http://www.inovonicsbroadcast.com/productline?cat=Broadcast+Processors

The (now Legacy) 260 is a classic. Full manual with theory and schematics.

The 261 is the same general idea BUT minimal analog, all the work is done in a DSP.

The 264 is the same general idea, is still analog, and cleverly simpler than the 260. All power is reduced to +/-9V rails, which means stealing chunks for stompbox use is less problematic.

[rring]

Its funny - My first thought was "well it should be easy to use PWM because all of the spurious products would filtered by the ear"

It seems so simple... but based on my tinkering here, what you have described , and the lack of product offerings, it becomes clear the design of such equipment is not so trivial - if you need studio grade features, specs and very wide dynamic range limiting .

The major brick wall I am seeing in my design approach is a limit in compression range to about 20 to 23 dB - which is plenty I guess for stomp boxes. I just can't get the resolution in duty cycle to do better than that.

[PRR]

> all of the spurious products would filtered by the ear

Stuff going in will beat against the carrier and possibly alias into the audio band.

Carrier coming out may distress later amplifier stages.

It is "easier" in this application (guitar compressor) because the source (e-guitar) is strongly band-limited, and guitar amplifiers tend to be forgiving, while the speaker is a strong high-cut.

Still in all, I'm not shocked you can find birdies and other oddities in the output. Take a 20 pound bologna, slice it very thin, throw-away half the slices... it's a brutal process.

> a limit in compression range to about 20 to 23 dB

Duty-cycle 10% to 7%.

The Inovonics text suggests 40dB is hard. These guys wanted to make money in the lucrative broadcast limiter market, so they spent days tinkering the ramp and comparator. Even so I don't see much evidence that the Program limiters were ever supposed to be wound-up past 26dB GR, with 12dB as more-typical. However the Utility 264 does do double-PWM and 80(!)dB on paper (certainly much less in practice-- the input couldn't stand 80dB-over). (Actually the double-PWM may be to get a steeper control law and thus avoid the complex log-function used in the 260.)

The reason we "want" 40dB GR, when we only use 12dB-26dB, is: what happens when things go WAY WRONG?

Some VCAs overload semi-gracefully. They distort but the output rises very little. FET/LDR designs bottom-out and the output continues to rise. But a simple PWM runs off the tip of the ramp and goes OFF. Punches a hole in the audio. That's so unexpected that, as you did, the designer puts a stop somewhere before 0% duty, but that's back to the FET/LDR case of rising output (and overload in a later stage, perhaps a 25,000 Watt transmitter stage).

Note that the "cascade PWM" tossed-off by that guy in the other forum *must* have a hi-cut filter between PWM stages. (Otherwise the 2nd PWM is making nothing out of nothing, and doing nothing.)

The key to pushing 1% duty cycle is a really sharp tip on the ramp, really fast comparator and switches.

Note that if ramp frequency is 200KHz, and we want 1%, this spike is similar to 200KHz/0.01 or 20MHz pulse. That's pushing the limits of cheap parts and easy layouts. If you are running 1MHz PWM, then indeed a 10MHz-20MHz switch-speed limit is 20dB-26dB at best.

I'm afraid, after too many decades in broadcast and recording limiting, I'm liking the improved 260, the 261. Dump the program into a digital processor, hammer it, dump back to analog (optional). Right away we can incorporate 4mS of transparent latency, so we can slam sudden peaks "before they happen". All sorts of complex logic become code instead of analog-computations. All the parameters of that decision-making are potentially user-modifiable. True, a 4mS latency may "matter" in live performance (not alone, but on top of 25mS air-path amp-artist and misc latencies in other digi-effects).

You want a whole nother approach? Inovonics 201 uses the same shunt-transistor trick used in $13 cassette recorders' AGC (and more brutally as a mute in almost everything). There was even a patent, but assigned to another company, so Inovonics potted this part until they could acquire the patent. Despite being clever for the time, it isn't much different from a '3080. I can see why James turned to PWM when the chips became commodity.

[MicrosiM]

Hello rring.

Very nice work indeed.

I noticed in your video that the duty cycle is not 50% fixed when the circuit is in idle. and it changes with audio, smooth and cool

I am not sure if I am right, But I was planing to test this circuit with Class-d amplifier, Can you provide help if this circuit will provide 50% fixed duty cycle? in idle. I mean when there is no sound?


Or how to adjust the duty cycle with 50% fixed.

Also, does the frequency changes with the sound? or its fixed?


Best Regards

[rring]

Frequency is constant and is adjustable over a huge range by one resistor. Yes you can set the duty cycle at 50% - there are a number of ways to accomplish this. It will matter why you need to be @ 50% is this a top limit, bottom limit or do you want to go lower and higher with respect to an AC signal? Not sure what you are doing. Have you looked at the data sheet for the PWM chip? if not check it out- it is very straightforward and may give you the answer you need.

If this isn't sufficient - tell me a little more about what you are doing.

Just a side note I found this circuit which really works great as a full wave rectifier, has very few parts and is single supply.

http://diodes.com/_files/products_appnote_pdfs/zetex/an55.pdf

Doing a new version of this compressor using this for the side chain

[MicrosiM]

Frequency is constant and is adjustable over a huge range by one resistor. Yes you can set the duty cycle at 50% - there are a number of ways to accomplish this. It will matter why you need to be @ 50% is this a top limit, bottom limit or do you want to go lower and higher with respect to an AC signal? Not sure what you are doing. Have you looked at the data sheet for the PWM chip? if not check it out- it is very straightforward and may give you the answer you need.

If this isn't sufficient - tell me a little more about what you are doing.

Just a side note I found this circuit which really works great as a full wave rectifier, has very few parts and is single supply.

http://diodes.com/_files/products_appnote_pdfs/zetex/an55.pdf

Doing a new version of this compressor using this for the side chain

Hello


I am making an CLASS-D Audio amplifier, and as you know you need 50% duty cycle, and the duty cycle will vary and change with the audio signal you apply


In other word, modulation of analogue audio into digital signal

[rring]

So In my circuit I have configured the input to have less than unity gain and 0 -100% duty cycle is over a range of 0-3V (not actually 0 -100% but more like 5% to 90%). You will have issues with obtaining any more range than this, because the pulse jitters and generates horrible noise the limits duty cycle. At unity gain the range is more like .7 to 1.3 volts( from the data sheet). So all you need to do is just DC couple in your audio with your bias point set accordingly to put you at 50%. In my case around 1.5 volts...so you will have limits with the valid range of input signal etc.... Does this make sense?