New CMOS design, help much appreciated

[WGTP]

You right ETA.  I went back and reduced the gain significantly and then started bumping it back up a stage at a time until it started squealing again.  There is a very definite threshold above which the oscillation goes off.  Below that, nice 4 stages of distotion, with a final compresovertertron stage.  Even doubling the gain in one more stage is too much.  The last 2 stages have 1k between and 10k for the feedback loop.  Second stage is 4.7k/47k.  First stage is like the 22/7th's (with 22k replacing the 33k so I could use 220k at the input and figure the roll offs easy), to get things off to a big start.  I guess 10x10x10= 1000.  Not sure about the gain of the first stage.  Less than 220k at the input has rolled off too much treble in past experiments.   

[earthtonesaudio]

I think the calculation for the gain of the first stage is the same for DC.  Meaning, take the total resistance in the f.b.l. and divide it by the resistance at the input.  For AC signals, it's much higher due to the cap shunting the negative AC feedback.  If you use a large enough cap to pass all audio frequencies, then the whole range gets the boost, but if you use a small cap, it starts working like a treble boost (to a certain point), which can be quite fun...  Too small a cap seems to move the AC gain out of the audio range, but values from 1u to 3n3 seem to do good things. 

Now that I think about it a little more, the AC gain probably depends on the corner frequency set by the resistors AND the capacitor, not just the cap.  Yeah, that makes sense.   

I plan to try the cap blend from the Easy Face in that location to see if it works as a variable bass control.  My circuit is very responsive to input cap changes, but I couldn't get the Easy Face "Blend" control to work at the input.  Probably has something to do with input Z...

[WGTP]

If nothing else, a switch for the caps will work.  I'm using a 4.7n cap with the 22k reistor and the fc should be around 1600Hz which is high, but what I typically use in other distortions.  For sure the gain is less in the low frequencies than with a larger cap.  As you said, a treble boost.  My amp is really dark.  With a 330 resistor and 6.5v decharged 9v battery, it is running at around 5v.  Refinement continues.   

[WGTP]

Another similar question, what value should I use as the impedance when modeling the BMP tonestack with Duncan's stuff?  If the output of the inverters is high impedance, 10K, 100K,    

[gez]

Another similar question, what value should I use as the impedance when modeling the BMP tonestack with Duncan's stuff?  If the output of the inverters is high impedance, 10K, 100K,    

That's a difficult question (for me, at any rate) to answer.  From empirical testing in the past, I'd say a few kilohms.

Not sure if this is correct, but I suppose if you divide supply voltage by current consumption of a single inverter that would give you the effective resistance of the channel...at least in its quiescent state (problem is that the effective resistance changes over the course of a cycle).  Then divide that by 2?

Assuming, say, 3mA per inverter (check data sheet for actual figures) at 9V that gives a channel resistance of around 3k, so each device would have roughly 1k5 effective resistance.  Seems a little low to me: the resistance of each device within the inverter, as it acts as a load, will increases with a signal present.

Short answer: I don't know! 

[gez]

Forgot to mention that, in theory, negative feedback will lower the output resistance.  However, seeing as it's unlikely closed loop gain will be set low, and the open-loop gain of inverters is pitiful (especially at 9V), it's a bit of a non-issue. 

Another pointless post...

[earthtonesaudio]

There's gotta be a way to figure it out... what about sticking a variable load on the output and reading the AC voltage output with a meter?  If you had a 1M pot as the load, you could put a sine wave in, and gradually decrease the resistance.  When the AC output starts to really drop off, I'd say you're well below the output impedance of the inverter.

...Don't know if that would work.  Anyone have an opinion?

[brett]

Hi
Hey Gez, don't be too hard on yourself.

Another pointless post

These discussions are very informative for me, at least.  I suspect that others who half understand feedback, op-amps etc also learn a lot.

There's gotta be a way to figure it out... what about sticking a variable load on the output and reading the AC voltage output with a meter?  Don't know if that would work.  Anyone have an opinion?

Yes, it will definately work.  I've used this method a couple of times.  I think it was described by Mr Keen at geofex.com.  In this case, look for the resistance that causes a 50% voltage drop in the signal (or the resistance that causes a very small reduction in signal loudness ie about 3dB).  That's your output impedance.  Like Gez, I'd be guessing a few Kohms (2? 3? 5?)
cheers

[earthtonesaudio]

Cool, thanks for the affirmative, Brett!  One more problem solved...

[puretube]

IIRC, the output-impedance of those CMOS inverter "linear amps"  heavily depends on the feedback-resistor-value used...


[EDIT]: and the "distortion"/"saturation"/"waveform-deformation" as well as the max. output-voltage also heavily depends
on the loading-impedance of the output (...in relation to the feedback-impedance).

A similar analogous phenomenon has been utilized in tube-circuits for distortion ~70 yrs ago...
(variable loading of the plate-circuit for waveform-distortion, i.e.).

[Nasse]

I think I have a diagram plot somewhere, z out and g/bw of a stage without f/b, or something. It is on same paper I found the gain formula.

[WGTP]

Another part of the design mystery.  If I understand correctly, input resistance/feeback resistance, 1k/10k, 10k/100k, 100k/1M, etc. all result in a gain of 10, but what "differences" result.  Lower value resistors are desireable from a noise stand point, but most designs seem to use larger value resistors.  In my original TSF/Mutant, etc. it seemed a lower resistor in the FML was brighter.   

[WGTP]

Latest Proposal:  http://www.aronnelson.com/gallery/main.php/v/WGTP/Proposal.GIF.html?g2_imageViewsIndex=1

[earthtonesaudio]

Cool, WGTP.  Thanks for sharing. 

If you wanted to add another control (there are already a lot if  you include trimmers), you might try something to do with the cap to ground in the feedback loop of the first stage.  A switch to select between different cap (or resistor/cap combos), a 1-10k pot in place of the 22k resistor, or a variation on the Easy Face "Blend" control to vary the gain boost from full-range to mid/treble to treble.  I've not tried something like this, and I don't know how much it would work for CMOS, but it works alright for discrete components and opamps, so it might be worth it to give it a try.  Actually, a spdt switch might be good, you could call it a "fat" switch or something.

Nice use of the cmospressor stage on the end (or whatever people are calling it now).  I still haven't breadboarded it yet!

[zencafe]

Latest Proposal:  http://www.aronnelson.com/gallery/main.php/v/WGTP/Proposal.GIF.html?g2_imageViewsIndex=1

Looks interesting, I've been designing some stuff with CMOS lately, I'm on my way to finish a high gain preamp, 5150 sound type.
And based on that design I'll try ti do some more stuff, a bass and guitar overdrive, and maybe a cool fuzz octaver.

But well, Looking at your design I see some trouble (I didn't get what's de "G" point? Let's guess is ground ) The first stage looks pretty ok, but I think you'll have some trouble with the later ones.
One thing I'd recomend is addign coupling caps, If you want good and fuzzy lo-end 100nF is a good value to start, if you want more tight and compressed, 22nF should do.
Another thing is that I'd add feedback caps of at least 100pF on the stages that lack them, and the third, the buffer after de 5th stage won't work... your signal will just go through the wire instead of the inverting gate... for it working you must think of the buffer as a unity gain stage, (with an input and equal feedback resistor) 10k and 10k should work fine for a ncie buffer.

Good luck with the design.

Regards from Chile!

Douglas.

[puretube]

Hi Zencafe:

to understand what WGTP is intending with that "5th buffer",
it helps to fully read the first linked-to thread in reply#13,
where that little different type of circuit was discovered & posted first by Brett...

[WGTP]

Thanks for the comments.  Yes, see Brett's post above.  With all the trim pots, may as well add one to replace the 22k resistor. 

Revised:  http://www.aronnelson.com/gallery/main.php/v/WGTP/Proposal_001.GIF.html?g2_imageViewsIndex=1

I'm not sure if the caps between stages are required.  Right now it is working with 1k resistors only between stages, may not need them!  May be needed as a design consideration.  Let me know if I'm wrong.   

I don't want to roll off the highs or lows too fast.  By using 2 caps for each, it should produce a 12db/oct. roll off and I set the fc as low or high as possible without loosing too much tone.  The G is ground and the 4.7n cap with 22k of resistance, should produce a 6db/oct. roll off starting at 1.6kHz to reduce bass thru the distortion which the BMP Tone control (properly tweaked) will restore after the distortion.  Also the input cap can be used to further "contour" the bass.   

This looks a lot like STM's 22/7th's and earthtonesaudio's postings. 

[zencafe]

Great!

Didn't know about that  .

It was puretube from whom I read that that kind of buffer coudn't be done... ironic that he corrected me hehe.

Well, you may not need the caps, that would depend on how much gain you want in everystage.
On my designes, I've been fiorced to cut the higher harmonics (very loud and higher than 10khz) when clipping the gates at very high gain levels... at lower gain (overdrive, or moderate distortion) I don't think so much filtering would be needed.
I'm still experimenting with filters with the Cmoster (I'll call like that probably), but probably a 1 pole lowpass after the EQ (which is gonna be active as well) is the best choice on my case.

[puretube]

hehe: we had to find out together, why it does work, although it looks like it can`t... 
(you checked the 2nd link in reply#13 ?...)

[earthtonesaudio]

I was messing around with WGTP's "proposal," and I noticed that the sounds from the stages with the 100k trimpots can get quite unruly when you turn the wiper all the way left (as shown in schem).  If you wanted to make it "user proof" you might want some limiting resistors between stages, maybe 1k-10k.

Also, an update: I was messing around with some 4049B (yes, the buffered kind) chips and I got some good sounds.  I have always read that you should avoid the buffered chips for audio, but so far no problems.  The chip seems to act just like a 4069UB as far as I can tell with my ears. 
Note: I was using them in 2-stage, "tube sound fuzz" type of configurations, so I don't know yet how the buffered chips would work for 3+ stages.
Do people say to avoid the buffered versions because of current draw or for some tonal reason?  I imagine they would be more power hungry than the unbuffered, but at first listen they sound the same.   

Anyway, back to experimenting!  I think these multi-stage CMOS circuits sound good for medium overdrive tones.