Group Design: PWM Limiter Project

[QUEEF BAG]

this is a great read, even though so little has become of it.

the info on PWM comps, the related tangents people go off on discussing options...
fascinating. 
and i hope this is picked up again, it could, should be a winner.

the only PWM comp i have used much is the fearn,  and i dig it a lot.
mainly the texture.  well, everything about it is cool

i am glad i heard one before i realized how it works.

i mean, if i had seen sausage being made before i ever tried it, i probably would
have never tried it.

 

[Rob Flinn]

Did anyone test the attached PWM signal-generator?
This is build around a SG3525A PWM-control-chip and runs up to 400kHz. http://www.onsemi.com/pub_link/Collateral/SG3525A-D.PDF
The original design comes from Clive Sinclair and Mark Case. http://robots.freehostia.com/Circuits/PwmGenerators/PwmGenerators.html
Extremely simple circuit with few components, the supply voltage can have a wide range (need not be +12V)
The smart part is the way the 2 outputs are tied to the 10k pull-up resistor, delivering the full duty-cycle control (0-100%).
Theo

I built a gain cell that worked quite well using that chip a few years ago.  It worked fine.  I never got round to adding a sidechain, but the chip will work.

 

[Balijon]

Thanks Rob.
Can you share what you build?

grT

 

[topcat]

For Info, the nice folks at Inovonics have been making PWM based broadcast processors for a while and have complete manuals including schematics at their website http://www.inovonicsbroadcast.com/downloads.

I used them for a number of years in both the on-air chain and for production, they have a very sweet sound....IIRC the peak control was not 'absolute' and they struggle with 'sledgehammer' compression.

Thanks to the posts here I'm going to look at PWM for the compressor/limiter in a 4 channel broadcast mic amp idea I'm kicking arround.

Thanks guys!

Cheers

tc

 

[Rob Flinn]

Thanks Rob.
Can you share what you build?
grT

If I can find the stuff, yes I can help, but it was a few years ago, so may take som etracking down.

 

[Rob Flinn]

I have been looking through some stuff, but haven't found any diagrams of what I did.

However I found some scope traces that show the signal while it's being chopped & then after the low pass fileter where the chopped frequency has been removed.

 

[Rob Flinn]

Here is the circuit for the low pass filter I used.  I designed using a circuit & formula from the active filter cookbook.  I'm not sure what the values of the Rf resistors were...... I'm not very good at making notes ... sorry

 

[Balijon]

Thanks Rob.
What did you use as a switching element?

I am working through the Invotronics documentation from the link that Topcat provided, great stuff.
Most interesting are the curves they use to align the ratio to the PWM duty-cycle generation.

I will try some old contacts next week to get a hold on the BBS schematics.
It would be great if we could come to a simple, 'few-chip' and easy to build, module design.

Did anyone use a PIC-microcontroller, with a A/D to generate the PWM signal, in a design before?
This way you could create a table for the desired ratio and attack/release behavior.
It would reduce the amount of parts a lot.
Adding a noise-gate function would come at almost no cost.
For instance a PIC18F1220 (+/- 2$) would have enough A/D-ports, PWM-ports and I/O-ports to control threshold pots, switches and GainReduction-led-bar.

Theo

 

[Rob Flinn]

Switching element was pretty much the same as the 1176 fet arrangement.

This gives you some values for the filter.

 

[Rob Flinn]

This is how I configured the PWM controller chip I think

 

[Rob Flinn]

I may have the strip board prototype around somewhere.  Difficult to know where.  There's piles of junk here.

 

[gfr]

Hello. I've been lurking a long time, this is my first post.

I designed and built this with a friend when we were in college.
Ramp is generated by transistor current source + 1/3 339 to reset.
The gain control uses a pair of 4066 switches as a "L", in the feedback loop of an opamp. Envelope extracted from input.
We found the envelope detector to be overkill - it needn't to be so fast for bass, could have used something simpler..
I don't know how much gain reduction, but it sounded nice with my friend's fender bass.
I remember it needed a very good power supply or else strange noises would leak to audio.

 

[Balijon]

Ramp is generated by transistor current source + 1/3 339 to reset.
The gain control uses a pair of 4066 switches as a "L", in the feedback loop of an opamp. Envelope extracted from input.

Thanks for your post.
What frequency is the ramp generator running?
Nice concept to use the 'chopped' signal in a feedback loop. Did you measure the phase shift on high-frequencies between the output from U8A and U13A?
Did you try the output on U13A instead of feedback to U9A?

Theo

 

[gfr]

If I remember correctly, it was around 200kHz.

I didn't measure the phase shift, but I remember if we added another pole on the RC filtering or it would oscillate (as one would expect from theory).

I don't remember trying the output from the other node.

This was back in 1990 or 1991, so I don't remember everything about the project...

I hope there's no error in the schematic, I redraw it from the hand draw schematic because it was starting to fall apart

The waveforms of the chopped signal were not "perfect", in practice this translated to a limitation on the amount of gain reduction but there was no (audible) distortion, and it looked nice on the scope after filtered too.

 

[Balijon]

Thanks for the info gfr.
200kHz is common for this style of limiter, the 4066 will do 4MHz according to the specs.

I hope there's no error in the schematic, I redraw it from the hand draw schematic because it was starting to fall apart

I have a remark:
I think that the +/- inputs from U9A should be reversed, if it should become an inverting opamp.

The amplification around U9A for the input to output is +/- +25db (20log(820k/47k), for the chopped 'feedback-leg' this is +/-32db (20log(820k/22k), so at maximum attenuation (at full limit / 100% duty-cycle) would be -7db output level. Is this a correct assumption?

Theo

 

[gfr]

I have a remark:
I think that the +/- inputs from U9A should be reversed, if it should become an inverting opamp.

I think you're right.

The amplification around U9A for the input to output is +/- +25db (20log(820k/47k), for the chopped 'feedback-leg' this is +/-32db (20log(820k/22k), so at maximum attenuation (at full limit / 100% duty-cycle) would be -7db output level. Is this a correct assumption?

Theo

Hmmm... When there's no signal being fed back through the "chopped path", gain is 820k/47k or 25 dB. When the  "chopped path" is "on" 100% of the time, the gain is  (820k//22k)/47k or -11dB, a little lower than -7dB. That's 36 dB of gain variation, with a distorted "knee" near the extremes because the chopping isn't perfect near 0% and 100% duty cycles. There's the input stage before it which can be set with a good amount of gain.

Perhaps it can be designed with a better gain distribution, we designed this as a final task for a course on college, and since we didn't have much time, we had end of semester exams, and spice took years to run on a PC Jr (hacked to boot with a floppy and with a memory upgrade - credits go to my friend), etc., we were happy when we breadboarded something that worked

IIRC, the "bandwith" of the 4066 is 4Mhz, meaning that you can pass a signal with that bandwith through the switch (I've seen NTSC to PAL video transcoders that used a 4066 for the switching). But it can't turn on and off that fast, if you try you end turning on the switch before it's fully open, or vice-versa. In the video transcoder circuit it passes a 4Mhz video signal but turns on and off at 16KHz or so (scan).

 

[Balijon]

IIRC, the "bandwith" of the 4066 is 4Mhz, meaning that you can pass a signal with that bandwith through the switch (I've seen NTSC to PAL video transcoders that used a 4066 for the switching). But it can't turn on and off that fast, if you try you end turning on the switch before it's fully open, or vice-versa. In the video transcoder circuit it passes a 4Mhz video signal but turns on and off at 16KHz or so (scan).

The signal pass-through bandwidth is something like 40-60MHz. Turn on/off frequency can be much higher than 16kHz (if not used for video frame switching), I recall it was somewhere around 50-60ns to turn on/off (depending on the voltage that you run).
I understood you were running the 4066 around 200kHz?

Theo

 

[gfr]

I tried simulating it with the values in the schematic and the ramp in the sim is at ~400kHz, but from memory I could swear it was around 200kHz. Anyway, another slowing factor (perhaps more important than the 4066) is the speed of the 339 comparator, used to reset the ramp.

 

[Balijon]

Anyway, another slowing factor (perhaps more important than the 4066) is the speed of the 339 comparator, used to reset the ramp.

Good point there. This was the reason I was looking at the SG3524 that Rob Flinn used, it runs 400kHz with a dc input, and it has a nice duty-cycle range. I am pretty sure that the 4066 will deal with 400kHz.

grT