PWM compressor besides Mantis 2?

[nilhartman]

Isn't Kotelnikov relying on PWM / similar topology ?

[10bd01]

True PWM is hard to do in digital. For it to make sense the control path needs to operate at a frequency much faster than the audio sample rate.

Interesting, melco_go! I wasn't aware of this. It sounds like a significant difficulty, making me wonder how good these emulations may be.

No, I don't think melco_go's remark is absolutely accurate:

from the Dave Hill paper linked to above:

The modern PWM gain control circuit design has the advantage of being able to use current technology, very high speed parts. They were not available 20 years ago or were very costly and power hungry. Components that can switch between off an on in less than 1nS, that are suitable for build a switched gain control element are now available

(take note kids: language skills are not the only path to a successful career )

So yes, it might prove to be very difficult to emulate this Crane Song PWM compressor as software (if possible at all ) but on the other hand the Pye is age old (a lot more than 20 years or 30 years old (no idea when that paper actually has been written)), so it's absolutely safe to say that its switching works at a significantly lower speed (with the resulting artifacts propably being part of its desired character) and therefor will also be a lot easier to emulate as software.

Btw: I assume that a FPGA might come to the rescue in regards to solving potential speed limits with "software" emulations.

Thanks for chiming in, jens! I'm intending on working with the PYE/PIE and Mantis2 at the moment. I was wondering that too about the PYE - was just watching Eddie Kramer praising it and was using it decades ago, so I can't imagine the older PWM compressors couldn't be modeled now.

Mantis2 appears to be sampled off of a PWM comp designed specifically to be sampled from, so it's a bit more opaque in terms of what's going on there and what I'll be getting.

[meloco_go]

Well, I did some googling and people say the frequency of the PWM oscillator in PYE compressor was 250 kHz.
https://gearspace.com/board/showpost.ph ... tcount=154

As for the plugins, I don't think it is necessary to copy the operational principle to produce an emulation.

Also, yeah, it is probably possible to make a PWM circuit operating at a lower frequency and make it output a control voltage for something like VCA. The compression range and attack/release will be limited.

[jens]

So that would be 4000nS/p then... i.e. 1/4000 the speed of the Crane Song STC8 - and still sub-sample speed even @192khz...
but am I wrong in thinking that 6xOS would nontheless make it possible even @44.1khz?
Perhaps the Waves Pie already uses OS for this part of the algorithm?

[aMUSEd]

Dave Hill is the man.
http://www.cranesong.com/pwm_comp.pdf

I thought PWM stood for pulse width modulation not 'pulse with modulator'?

[jens]

that's certainly a typo worth discussing at length!

What do you think would be worse: "with" instead of "width", or "widt"? And what is more reprehensible in your opinion:
lack of education or being dyslectic?

Let's discuss this instead of the boring topic at hand!

[aMUSEd]

that's certainly a typo worth discussing at length!

What do you think would be worse: "with" instead of "width", or "widt"? And what is more reprehensible in your opinion:
lack of education or being dyslectic?

Let's discuss this instead of the boring topic at hand!

Huh? I was just asking for clarification - I hardly think he's dyslexic or lacking education (obviously not the case) but did wonder if there was more than one meaning for the term that I wasn't aware of

[jens]

Like there's even just a hint of chance that this is anything other than a mere typo... "pulse with modulator" doesn't even make any sense... besides the paper is full of both spelling and grammar mistakes...

from the same paper btw.:

"This circuit is known as a pulse width modulator"

edit: either way my apologies if you genuinely assumed he might have meant "pulse with modulation" - I will also do my best to avoid making fun of you in the future....

[aMUSEd]

Maybe a translation error then?

I just assumed there was some esoteric alternative DSP guru interpretation of PWM I wasn't aware of

[jens]

I see - I think he may really be dyslectic - the kind of mistakes you can find in the paper won't be caught by correction tools...
with geniusses something like this often kind of comes with the territory (not everyone can be a "jack of all trades, master of none" such as me ).

[meloco_go]

So that would be 4000nS/p then... i.e. 1/4000 the speed of the Crane Song STC8 - and still sub-sample speed even @192khz...
but am I wrong in thinking that 6xOS would nontheless make it possible even @44.1khz?
Perhaps the Waves Pie already uses OS for this part of the algorithm?

Typically OS goes in the powers of two, so the closest necessary rate should be 8x. I don't know about Pie, OS should yield some latency, does it have it?

[bmanic]

OS does not require any significant latency if you don't care about using linear phase filters.

[jens]

So that would be 4000nS/p then... i.e. 1/4000 the speed of the Crane Song STC8 - and still sub-sample speed even @192khz...
but am I wrong in thinking that 6xOS would nontheless make it possible even @44.1khz?
Perhaps the Waves Pie already uses OS for this part of the algorithm?

Typically OS goes in the powers of two, so the closest necessary rate should be 8x. I don't know about Pie, OS should yield some latency, does it have it?

No, it is entirely irrelevant what is "typical". It's perfectly possible to have 6x OS. Probably the main reason why it's normally doubling up is because that means shifting the Nyquist limit one octave up, so in the case of using OS to reduce aliasing it probably makes sense - but that doesn't apply here at all.

[meloco_go]

No, it is entirely irrelevant what is "typical". It's perfectly possible to have 6x OS. Probably the main reason why it's normally doubling up is because that means shifting the Nyquist limit one octave up, so in the case of using OS to reduce aliasing it probably makes sense - but that doesn't apply here at all.

Hmm, you're probably right. However, AFAIK at least sometimes oversampling is done in stages, e.g. twice, then twice to get quad, etc.

[jens]

I am not a coder at all (I couldn't program "hello world" in any language if my life depended upon it), however if I understood it correctly it can yield a relative effiency advantage in certain cases to do it that way...