Why can't I make my Frobnicator quiet?

[Mark Hammer]

I'm using RG's 2nd (Oct, 2001) layout.  I've gone over the PCB, gone over the parts placement and orientation, gone over the solder joints, and used a 10-turn trimmer for precision nulling, in addition to an LM358.  I can make the carrier somewhat quieter (maybe 3db), but it is still annoyingly obvious, and well above what any gentle gating might be able to remove.  While it is quite present in high-frequency mode, it is also present when in low-frequency/tremolo mode, if I turn the frequency up.

I'm beginning to think it may have something to do with the biasing, and the manner in which the bias voltage is derived, using the pair of LEDs.  With a 9.6V supply, the Vb reading that Hollis states as 3.2V reads 3.47V.  Is that somehow problematic, or within he margin of normal variation?

Have any of you who built the unit been able to tame the carrier/modulator to your satisfaction?
And should the pair of LEDs used for biasing pulse at the rate of the LFO?

[anotherjim]

If you used an LM358, it can swing closer to its neg rail than the 062, so it might be that IN914 in the OTA control also wants to be an LED.
Does the layout keep the neg rail (ground symbol) separate from the pedal supply/audio 0v?

[willienillie]

What's the advantage of using the LEDs in the bias supply?  To regulate the Vref?

[Mark Hammer]

I dropped the 10M resistor feeding the trimmer voltage to the 3080 down to 1M and that helped things quite a bit, though it's still not acceptably quiet at higher frequencies.

In answer to your question, Jim, as near as I can see, yes.

As for the use of two LEDs for biasing, I suspect it was used in place of a zener value that doesn't seem to exist.  3.3V is a standard zener value, but 3.2V isn't.

[anotherjim]

I sometimes have to use mixtures of LED and diode on breadboard to get particular min and max ranges that won't vary with supply - the actual voltage depends on a few things. However, if it's more convenient, I'll build it with those same parts and not even bother to see if I have a Zener that would do it.
The OTA control pin needs a minimum voltage to bias although the control is actually current dependant. It would probably cut off at less than 1 diode drop at the OTA control.
The LFO swing, for a decent range of depths, must (usually) never go so negative that the OTA cuts off, so there's another parameter that the supply reference has to be sized for.
I doubt there is something in there that demands an exact reference voltage. LED Vf isn't always consistent.

I'm wondering that if the breakthrough is audible even with the depth at minimum, the problem could have more to do with the LFO shaking the reference when it should stay rock solid.

[Mark Hammer]

So are you suggesting that decoupling the dual op-amp used for the LFO from the reference might fix things?  I'll have to try that out tomorrow.

[anotherjim]

So are you suggesting that decoupling the dual op-amp used for the LFO from the reference might fix things?  I'll have to try that out tomorrow.

I think that as designed, the reference should be pretty well stable. I would check the supply filter caps to make sure they are good. The only way from the oscillator into the signal path should be via the OTA control. If the Depth pot at minimum doesn't stop the breakthrough, it's getting through another way.
Original TL062 op-amp may have been chosen for a good reason, but any type ought to be possible there. TL062 is a very low-power type and that might be the charm.
Like anything with an oscillator that can run at audio frequency, there's always the chance of getting radiated crosstalk from the pot wiring into the signal path.

[ElectricDruid]

The reference might be solid. It could be the actual 9V rail that's getting dragged around. If that's moving but the reference isn't, you'll get horrible feedthrough.

Swapping for a TL062 low current op-amp sounds like a good idea. Especially if the reference LEDs flash with the LFO.

[Mark Hammer]

I actually had a TL062 in there at the outset.  The whining prompted me to try out another low-power op-amp, being the 358.  Didn't make a pinch of difference.  On the other hand reducing the current-limiting resistance for the biasing trimmer did.

But the rate-blinking of the LEDs used for biasing, suggests that the decoupling of the power to the LFO from the basic biasing may be in order.  Hollis's designs tend to opt for the leanest referencing arrangement possible.  You'll recall that many have had to decouple the LFO in the Zombie Chorus builds in order to eliminate audible ticking.

[anotherjim]

You could add some signal filtering on the input, since any inaudible noise could modulate and become audible.
Reducing the input cap to reduce rumble/hum - 5.6nF for guitar
Add a cap to ground after the 330k input - 100pF for guitar

[Eb7+9]

has nothing to do with rails, decoupling, etc ... that approach tries to chop dc output offset current plus noise / which cannot be perfectly nulled in either case ... the only way to get zero output is by multiplying chopper signal by zero / which can't happen with a real-world OTA ... you can swap other OTAs and hope that one naturally comes close to having no offset, but still getting zero output with no input is a practical impossibility ... also, notice the futile attempt at reducing the clock noise by softening clock edges via 10k/100nF low-pass filter ...

[jonnyeye]

I'm not happy with the feedback circuit around the OTA; the 3080 biases with ~4.5V at its output, which will be divided down by the 470k/10k to about 27mV of offset... might be enough to cause a problem. Try putting a cap (10-47n) in series with the 470k. You can also try running it open loop (remove the 470k entirely).

[ElectricDruid]

has nothing to do with rails, decoupling, etc ... that approach tries to chop dc output offset current plus noise / which cannot be perfectly nulled in either case ... the only way to get zero output is by multiplying chopper signal by zero / which can't happen with a real-world OTA ... you can swap other OTAs and hope that one naturally comes close to having no offset, but still getting zero output with no input is a practical impossibility ... also, notice the futile attempt at reducing the clock noise by softening clock edges via 10k/100nF low-pass filter ...

can you explain...this in whole sentences...finding this a bit sketchy...more details...zero output with no input can be done with trimming...or not? please explain..."futile" attempt reduces higher harmonics and therefore softens thump...less perceptible if not actually less...

Ta!
Tom

[anotherjim]

I was kind of thinking everyone assumed this a minimal design that won't achieve perfection.
If its a dirty kind of effect, you might try series back to back diodes in the output for the easy zero-crossing noise gate trick. Or add a simple envelope detector and arrange it to stop the mod oscillator on note-off. Or redesign for LM13700 and have a proper VCA envelope follower on the output.

[Mark Hammer]

This is getting more complicated than the circuit warrants.  I think I'll omit the higher/audio-frequency LFO and simply use the sub-audio modulation as a "normal" tremolo where the modulation frequency can't be heard.  There ARE better ring modulator circuits.