(Diode) Bias Tremolo

[BetterOffShred]

I like it man.  It does the fender swamp thing pretty nice.  Thanks again for the layout. 

[Rob Strand]

You can in principle add a steep low-cut after the thumper to take the sub-sonic off.

Also, you should place any high-pass filtering after the last differential amp not as part of it.   Imbalances in the cutoff frequencies on each diff-amp arm can cause more thump that it removes.

Sometimes it's worth adding an LPF after the LFO to take out any glitches.  Usually not required for a sine LFO.

[samhay]

You can in principle add a steep low-cut after the thumper to take the sub-sonic off.

Also, you should place any high-pass filtering after the last differential amp not as part of it.   Imbalances in the cutoff frequencies on each diff-amp arm can cause more thump that it removes.

Sometimes it's worth adding an LPF after the LFO to take out any glitches.  Usually not required for an sine LFO.

I don't see how to do that well without more chips.

I thought about adding more filtering, but as Paul also figured, this makes the circuit significantly bigger.

The HPF on the diff amp can be nicely balanced if you use fairly high-precision R's and C's.
I did lift one of the caps and it didn't make a massive difference to the hiss. The corner frequency is quite high, so mostly there to ensure the circuit is polite to its neighbours.

C9 provides a little LFP'ing to the LFO. Glitches are more likely to be a problem in the audio path when the LFO wiggles D1, D2 in and out of forward conduction.

Ultimately, this is trying to capture some of the quirks of the old output bias tremolos. These are not know for their low distortion or dead-quiet operation. The circuit as drawn does this, and the noise is certainly no worse than a lot of the other DIY tremolos I have tried.

[samhay]

I like it man.  It does the fender swamp thing pretty nice.  Thanks again for the layout.

Thanks. That's what I was shooting for.

[bool]

Any diffamp can be made into a LP/HP/BP filter. If you are careful, even a one-opamp diffamp can present a reasonably well balanced load ("current balance") to the source.

But before diving in with a solder, it's time to check it with a sim.

[samhay]

>Any diffamp can be made into a LP/HP/BP filter. If you are careful, even a one-opamp diffamp can present a reasonably well balanced load ("current balance") to the source.

Right, and that is what I have done here.
D1 sees a load of 22k (R14) from the diff amp.
D2 sees a load of 20k (R15 + R18).
If you want to do better, you can use 20k resistors at R14 and R17, but this won't improve performance much (I tried) and I don't tend to stock 20k resistors in my parts bin.

LFP is formed from C13/R17 and C12/R15. Both give a corner frequency of 7.2 kHz.
HPF can be made with series caps as I said earlier. However, this may make performance worse as filtering is best done post-mixing/nulling.

>But before diving in with a solder, it's time to check it with a sim.

I agree. This is why I was pulling my hair out debugging - the sim shows this should be nicely balanced, yet that's not what the scope was saying.
This also highlights a problem with sims - they don't know about your build errors until you tell them!

[Rob Strand]

But before diving in with a solder, it's time to check it with a sim.

Do a sim with cap tolerances. You will see the common-mode rejection is diminished even at frequencies somewhat above the HPF cut-off frequency.

[samhay]

The common mode rejection isn't perfect in the sim even if you use perfectly matched components.
It varies with the op-amp model, but there are enough subtle differences in the signal out of IC1A and IC2A (other than phase shift) that you will never get perfect cancellation in this circuit.
I agree that it would be better practice to do all filtering after the diff amp, but most DIY tremolos use no more than 4 op-amps, so I figured this was a good design rule. Others are, of course, welcome to tack filter stage(s) onto the output if they want to improve the design.
I also stress again, the thumping in the first audio clip is gone - this was an artifact of a build error that has been fixed. There is still a little high frequency hash in the audio path, but there is no noise on the supply lines, so it does better than most on the noise front.

Also, I haven't blown this trumpet yet, but the circuit shows you can build a perfectly useful tremolo without using any FETs, LDRs or OTAs/VCAs.

[Rob Strand]

agree that it would be better practice to do all filtering after the diff amp, but most DIY tremolos use no more than 4 op-amps, so I figured this was a good design rule. Others are, of course, welcome to tack filter stage(s) onto the output if they want to improve the design.

Yes, but the reason your design works better than other diode based units is *because* it uses that differential stage.  Unfortunately that means using at least three opamps.

I also stress again, the thumping in the first audio clip is gone - this was an artifact of a build error that has been fixed. There is still a little high frequency hash in the audio path, but there is no noise on the supply lines, so it does better than most on the noise front.

The reason you get more noise is because there are three opamps all processing the same signal level.  In fact the last stage contributes twice as much noise, so it's like having four opamps adding to the noise.

You could boost the level coming in and drop it going out but that will make the diode distortion worse than they already are.

The way around the noise is to make the first opamp a pre-emphasis stage.  The de-emphasis network is perhaps best implemented as a passive RC filter on the output of the last stage.  You can use low values to keep the output impedance down.  (You can't put the de-emphasis in the feedback path of the last opamp as it will skew the frequency response of each differential input differently.)    Look at how Boss do it on the Flanger and Chorus.
The de-emphasis network in that position also filters distortion products from the diodes.

[bool]

If you have any desire to revisit the cap-coupling to diffamp idea, you can try to do it a "C-R-C" way to minimize the influence of the diffamp input current imbalance: simply cap-couple, and just insert a resistor after the caps AT the diffamp inputs.

This resistor would act as a corner-freq-setting part of a balanced hpf, but will also affect the input differential impedance.

My suggestion would be to use 330nF caps, and resistor in range from 20k-10k. (Do an impedance sweep sim of the diffamp to make it behave). There will be some interaction with diode-biasing resistors as well.

[samhay]

The way around the noise is to make the first opamp a pre-emphasis stage.  The de-emphasis network is perhaps best implemented as a passive RC filter on the output of the last stage.  You can use low values to keep the output impedance down.  (You can't put the de-emphasis in the feedback path of the last opamp as it will skew the frequency response of each differential input differently.)    Look at how Boss do it on the Flanger and Chorus.
The de-emphasis network in that position also filters distortion products from the diodes.

Having played a little more with the circuit, it looks like some more aggressive LPF'ing would be an improvement - it is lovely and quiet if you role the guitar tone all the way down.
The pre/de-emphasis is a good compromise and will only cost an extra handful of passives. It would require some moderate changes to the layout, so if I can muster the motivation to build another one I will give this a try.

[samhay]

If you have any desire to revisit the cap-coupling to diffamp idea, you can try to do it a "C-R-C" way to minimize the influence of the diffamp input current imbalance: simply cap-couple, and just insert a resistor after the caps AT the diffamp inputs.

This resistor would act as a corner-freq-setting part of a balanced hpf, but will also affect the input differential impedance.

My suggestion would be to use 330nF caps, and resistor in range from 20k-10k. (Do an impedance sweep sim of the diffamp to make it behave). There will be some interaction with diode-biasing resistors as well.

I think this could just make it worse. The input current imbalance is not that bad and adding HPF'ing is more likely to cause problems with the nulling of the low-frequency component of the LFO - thus bringing back the thump (which is nowhere near as much fun as bring back the funk).

[Rob Strand]

Having played a little more with the circuit, it looks like some more aggressive LPF'ing would be an improvement - it is lovely and quiet if you role the guitar tone all the way down.

That's a bit weird (and perhaps a little bit more weird because you are using a JFET input opamp).   That means a lot of noise is coming from guitar and first-stage region.   Past the first stage the opamp noise is independent of the tone control and the guitar.   Maybe that first stage opamp is damaged?

Does the noise go down when you set the guitar volume(s) to zero?

[samhay]

I puzzled this at first too.
The guitar is pretty noisy, and my house isn't that quiet either, so I don't think there is anything wrong with the input stage - it is just high impedance.

I am beginning to agree that the diff amp probably doesn't do high frequency rejection very well, so removing hiss  anywhere before this stage is helpful - you may be right that I should remove the LPF from this stage, which would become a must if I add pre/de-emphasis.

[amptramp]

There is a similar circuit with additional series diodes connected cathode-to-cathode as the squelch in my Heathkit AJ-1510A so you have to be careful that you get the amplitudes right.  Otherwise you may have a fuzz that varies in completeness of clipping at the tremolo frequency.  This could be a useful effect but maybe not what you were looking for.

[Rob Strand]

The guitar is pretty noisy, and my house isn't that quiet either, so I don't think there is anything wrong with the input stage - it is just high impedance.

My place is pretty noisy but I usually get more  buzz (mains derived) rather than hiss.

I am beginning to agree that the diff amp probably doesn't do high frequency rejection very well, so removing hiss  anywhere before this stage is helpful - you may be right that I should remove the LPF from this stage, which would become a must if I add pre/de-emphasis.

Yes, there's plenty to go wrong putting filters around a diffamp.  It often looks OK from an idealized theory perspective but when you have part tolerances and finite opamp gain-bandwidths the performance can be quite different to "what you thought".

[R.G.]

Just as a lead on prior art, using the varying resistances of diodes was more common back when opamps were not yet thought of.

One variant I always liked was the tremolo in the Thomas Organ Vox and MKii and MKiii UK Vox amps. They used a full diode bridge modulator, with the bias currents fed to four diodes from top and bottom sides differentially, and the signal used to "cancel" against an inverted version of the same signal after amplification.

The problem with diode resistance modulators is that they are highly non-linear for signals bigger than about 20-50mv. More than that and the diode nonlinearity shows up grossly in the signal. Which may or may not be OK, depending on how you like your sound. Using a diode modulator bridge helps because you have a chance at least at cancelling some of the distortion.

[samhay]

>My place is pretty noisy but I usually get more  buzz (mains derived) rather than hiss.
You're right, I need to be more careful with my semantics. My guitar buzzes - there is hiss too, but the buzz is what I was thinking about last night.

[samhay]

>Just as a lead on prior art, using the varying resistances of diodes was more common back when opamps were not yet thought of.

Indeed, although I have also seen them used in a few more contemporary op-amp filter circuits, like Escobedo's 'Phuncgnosis'.
I haven't seen them used quite like this before, but I wouldn't be surprised if it's been done.

>One variant I always liked was the tremolo in the Thomas Organ Vox and MKii and MKiii UK Vox amps. They used a full diode bridge modulator, with the bias currents fed to four diodes from top and bottom sides differentially, and the signal used to "cancel" against an inverted version of the same signal after amplification.

I can see how that would work if one was careful. Do you rate the tremolo effect in these designs (I know you are fond of the Thomas Organ in general...)?

>The problem with diode resistance modulators is that they are highly non-linear for signals bigger than about 20-50mv. More than that and the diode nonlinearity shows up grossly in the signal. Which may or may not be OK, depending on how you like your sound. Using a diode modulator bridge helps because you have a chance at least at cancelling some of the distortion.

I am aware of the limitations of using diodes like this. Frankly, I'm surprised how well my circuit worked straight off the drawing board.
The diff amp can (almost) deal with a lot of gross behaviour from the diodes.
Also, the design goal was to design a tremolo that had inherent distortion at the heart of the tremolo effect. The non-linearity of the diodes is very much exploited to achieve this!

[samhay]

>There is a similar circuit with additional series diodes connected cathode-to-cathode as the squelch in my Heathkit AJ-1510A so you have to be careful that you get the amplitudes right.  Otherwise you may have a fuzz that varies in completeness of clipping at the tremolo frequency.  This could be a useful effect but maybe not what you were looking for.

The LFO swing and the bias voltage range have been chosen so that (at least) some of the time the diodes are not biased into conduction. The LFO can conk out at low speeds, so if you do this and then turn the bias up (lower voltage) you get a pretty gross fuzz.
I did try to tweak it got get the varying fuzz effect (i.e. not clean anywhere through the LFO cycle), but this was quite difficult to dial in.