variable stuttering pedal

[R.G.]

Hey - thanks!

I did get a chance to put your OTA limiter into the simulator last night. It works in simulation. I was getting feedthrough of 20mV peak, which would otherwise be a noticeable thump, but it was all slow ramps up and down, so that may work out fine.

I suspect that the JFET setup may just need the drive waveform DC level and peak to peak size tailored to the JFET to get rid of residual feedthrough.

[R.G.]

I messed with it a bit more on the simulator. It looks like using a series-shunt JFET switch instead of either one alone produces a major improvement of feedthrough regardless of the slowdown on the gate drives. Using two similar JFETs, one being driven on and one being driven off at the same time seems to cancel the charge injection.

One JFET is hooked up as a shunt switch on the input, drain to signal, source to reference voltage. The other is a series switch to the - input, drain to signal and source to - input. The + input of the following opamp is tied to Vref.

I got transients under 50uV. That's way better than real world will be, of course, but it's also way better than you need, as noise under 1mV is going to be pretty insignificant.

It's also better than I could get with OTAs in any hookup.

Slowing down the gate drive seemed to make no difference. What made a HUGE difference was imbalancing the gate drives. Different cap or resistor values on the gate drives made the residual noise jump by 10:1.

[gez]

I did get a chance to put your OTA limiter into the simulator last night. It works in simulation. I was getting feedthrough of 20mV peak, which would otherwise be a noticeable thump, but it was all slow ramps up and down, so that may work out fine.

It's possible to use quite gentle slopes without having a detrimental effect on the stutter, you'd be surprised.  I set the thing up with back of an envelope calculations then just tweaked the damn thing till the clicking went (it's noiseless now but, despite quite a gradual slope, sounds like on/off action).

Had a mixed afternoon.  I had bleed-through from the LFO despite being really careful with layout - and this was without the FET connected!  No amount of rail filtering/shifting of layout solved it; haven't had this problem in a long time...must be losing my touch...or your circuit's cursed!!  Hooking up the FET did make things slightly worse, so that helped a little in deciding whether the trapezoid had some effect.

Ok, next problem was the circuit you linked to.  I was a bit concerned it would require something esoteric like an amp from the ICL76X1 series, so I tried a more common or garden variety to see how well it would fare.  Can't remember its code and I'm not about to check what it was as it now resides at the bottom of my rubbish bin - I think you can guess how well that experiment went!  I then came up with this shaper:



As you can see it's just an integrator set up to clip.  When the amp's output hits one of the rails and stops providing feedback, the diodes prevent the input being pulled to either rail, thus preventing damage.  It actually worked very well, but needs its own divider.  The slope is the max I could get away with for the range of PWM you've set things up for.  Ideally, I'd have preferred to reign in the PWM and make the slope a lot gentler.  The LFO could do with speeding up slightly too, though it's not too far off the mark. 

It made some difference when I hooked the trapezoid up to the FET, but just how much is difficult to tell because of the bleed through I already had (sigh).  I think your intuition about using more than one FET might be right though; in this set up I don't think slowing the ramp up/down is going to make too much difference...though it definitely works with OTAs. 

[puretube]

It's possible to use quite gentle slopes without having a detrimental effect on the stutter, you'd be surprised.

agreed! the hundred percent (>80dB) on/off thing within a µS is near mojo...
(at least when playing 50% duty cycle)

[gez]

(at least when playing 50% duty cycle)

Yeah, that's the clincher; deviation from 50:50 requires a steeper slope which is noisier.

[puretube]

noisier? - well, depends... 

(supposed you mean: CV-noise, aka: bleedthrough).

[gez]

(supposed you mean: CV-noise, aka: bleedthrough).

Yeah, the ramp up/down is no longer slow enough that you can't hear it.

[R.G.]

The ramp up/down needs to be short relative to the perceptible on/off period. The human ear loudness characteristic shortens the "on" period a bit. Too slow and you get noticeably less "on" duty cycle.

[R.G.]

I posted the series/shunt version under the old "variable stuttering pedal" article, page 2.
http://geofex.com/FX_images/stutter.pdf

[gez]

The ramp up/down needs to be short relative to the perceptible on/off period. The human ear loudness characteristic shortens the "on" period a bit. Too slow and you get noticeably less "on" duty cycle.

Depends how far you take things.  The waveform should still be trapezoid and not a triangle, though that was pretty much the case in your circuit with the PWM at one of its extremes due to 'on' time being so miniscule.  However, the effect was unusable in this position, even without the additional sloping I introduced - a barely perceptable signal at output - which is why I mentioned reigning in the range a little.

I'd have to crack open my box to measure what I did (can't remember & didn't make any notes), but with 50:50 duty cycle (I didn't use PWM, not keen on it with trem circuits...all subjective though), it sounds like 50:50.

[gez]

RG, found a spare 20 mins and managed to get rid of the bleedthrough problem I was experiencing.  Ran out of time so haven't tried you fix yet, but will do sometime over the next couple of days.

The trapezoid shaper I posted was still on the board so I did manage to tweak that and can confirm that it gets rid of the tick, but not with the values shown: the slope needs to be gentler and this is done at the expense of PWM.  If one were willing to forego the PWM it would make more sense to use a single op-amp LFO with the Trapezoid shaper, that way a single dual amp could cover it.  Despite what you said, with 50:50 duty cycle it sounds 50:50 and there's a definite on/off action.

Anyway, I'll get back to you on your amended schematic, stay tuned.

[R.G.]

Hey, thanks. I have not had a chance to get to the workshop yet.

Despite what you said, with 50:50 duty cycle it sounds 50:50 and there's a definite on/off action.

I guess that the ear is just shifting the threshold in time, giving back on the trailing edge what it lops on the leading edge. Ah, well. Sometimes thought experiments turn out wrong, just like real ones.

[gez]

Had a few moments to check the trapezoid shaper with the scope.  I think your solution is going to work out better, but for the sake of completeness this is what I ended up with:



I took the lead from one of your posts above and incorporated a trim, which made a huge difference.  The PWM went out of the window (rise/fall time too slow) so I replaced the 1M pot, diodes and 10k with a 470k resistor.  Still a little slow for my liking.

This circuit was not easy to tame.  The rise/fall time ended up as 4mS!  This would vary from amp to amp though (the one I used didn't have rail-to-rail swing).  Despite the slowness, it still sounds like on/off stutter.  In the interests of science I kept increasing the value and even 20mS sounded ok (very little in it).  Any more than this and it starts sounding more like choppy trem that stutter.

Anyway, I'm looking forward to trying your idea (nifty by the way).

[gez]

Unfortunately it seems to make things worse RG. 

I think in order to get this idea off the ground both FETs would have to be incredibly well matched.  Might be an application for a 4007?  With the gate of the FET in parallel with the 2M2 grounded, the bleedthrough is about as the same as it was in version 1.  With this FET dialed in, the background click goes way up in level, to the point where it's louder than the signal.

I'll keep it on the board for a day or two.  If you want me to try anything, just give me a shout (though I might not be able to get round to it till Thursday).

[gez]

What made a HUGE difference was imbalancing the gate drives. Different cap or resistor values on the gate drives made the residual noise jump by 10:1.

I'm wondering if one of those 100Ks were replaced by a trimpot it might help to tune things in...or out in this case?

[Gus]

I read most of this thread.

  what about a power mosfet?

  Power mosfets have a lot of gate cap that and a resistor in series to the gate might be something to try.
  or even one of the photovoltaic IC Mosfet gate drive chips?

  You could do high and low side switching because you can "float" the mosfet and photo cell side of the drive circuit and adjust the gate charge time with a series resistor to make it turn on even slower

[gaussmarkov]

R.G.:  thanks for the new variable stuttering pedal.  i'm looking forward to building and using it. 

anyone:  in R.G.'s version 2 of the variable stuttering pedal he says,

There are two bias voltages for isolation. Note that there are current spikes of about 20ma on the grounds of the three LFO opamps. If you don't isolate these spikes from the other analog ground by careful wire routing and decoupling , it will still tick.

i searched a little to try to understand what this means for a layout of his circuit.  i understand keeping the ground line for the LFO opamps away from the ground for the rest on the circuit board.  i suppose one might even run separate ground wires from the board to the ultimate ground on the output jack.  maybe even use shielded wire for those two connections.

would someone please clarify whether this is right and what "decoupling" the grounds means.  i have only seen decoupling using capacitors as a way to isolate the dc supply of different parts of a circuit.

thanks in advance, gm

[gez]

i suppose one might even run separate ground wires from the board to the ultimate ground on the output jack.  maybe even use shielded wire for those two connections.

More or less,  use separate grounds for the audio and the LFO and join them at the -ve connection for the battery/PSU.  Most cases you can run it to the -ve connection on the board, not litterally to the -ve terminal, ie the PSU jack socket.

I still had a problem (see above) on the breadboard as the wires I used were too long and were coupling tick into neighbouring ones.  Keep things short and if you still have a problem use screened wire for audio in/out.

[gaussmarkov]

i suppose one might even run separate ground wires from the board to the ultimate ground on the output jack.  maybe even use shielded wire for those two connections.

More or less,  use separate grounds for the audio and the LFO and join them at the -ve connection for the battery/PSU.  Most cases you can run it to the -ve connection on the board, not litterally to the -ve terminal, ie the PSU jack socket.

I still had a problem (see above) on the breadboard as the wires I used were too long and were coupling tick into neighbouring ones.  Keep things short and if you still have a problem use screened wire for audio in/out.

got it!  thanks, gez.  and thanks for the whole thread.  you and R.G. are sharing cool stuff.

[gaussmarkov]

here's a draft layout of R.G.'s variable stutter, version 2.  i think i am on my way to a reasonable layout after switching the connections around on the tl072s for lfo a bit.  before investing anymore effort in this layout, i was hoping to get some feedback on the layout of the grounds.  (edit:  for example, i just noticed that it is possible to move ic2 down and make the bias d net more direct and shorten the lfo ground net to boot.)  is this the sort of layout that you all are talking about?

the red net between ic2 and ic3 is the lfo ground net.  light green is ground for the rest of the circuit.  i did  this in eagle, so if there are problems they will appear in my transcription of R.G.'s schem (without part values) which is here.

(2nd edit:  i couldn't resist.  i changed the layout to make several nets a little shorter and to move the output further away from the lfo ground.  i take it this is a good idea.)