Fender type tremolo schem. Opinions?

[Morocotopo]

Hi people, I just drew this. It´s a tremolo, imitation of the one found on some Fender amps. In basic, the signal is divided in two, one half lowpassed and the other hipassed, then both branches are LFO tremoloed, but out of phase. 
This idea comes from a JFET circuit posted by R.G. on his site, that I loved but couldn´t make to work well (ticking, low volume...). Then I found a similar one (Tremolo-matic X) in the Stompboxology issues found at Moosapotamus site, that is great sounding but too hissy. Sigh. So I tried to cook my own... Yes, I really want this FX.
This is based on the VCA applications article in R.G.´s site, using a LM13700. What I did was take the LFO and lopass/highpass filters from the Stompboxology one and grafted it to one of R.G.´s schems in the mentioned article (application 3: pan pedal).

The schem:


A little description: IC1a is a buffer, IC2a and b and associated parts are the lopass and hipass filters going to each half of the LM13700. IC4a and b are the LFO, IC5a and b buffer and split the control signal and invert one of them to generate the two out of phase control signals. IC6a and b mix the resulting signals.

Well, I´d like your opinions on this. Some values I guessed, mostly I copied and pasted but some parts required a bit more than that. I wanted to present this to you people before refining and prototyping. Will it work? suggestions? Any commentary will be welcome.

Thanks for reading.

[Morocotopo]

Sorry for the too big image, but I wanted it to be legible.

[Morocotopo]

If you read this, R.G., I´m interested in your opinion since I implemented some of your ideas presented in Geofex in this.

Thanks

[Morocotopo]

Somebody? I´m itching to try this. Maybe I´ll go ahead and see if it works.

[petemoore]

  Kudos !
  I've been observing and reading, but can't purport to be of much help, you've probably got all the things I could help with figured out and have advanced beyond that anyway.
  I'm not familiar with the big bug chip in there 13700.
  Fender type tremolo does seem like a good sound to persue.
  I have a feeling someone who can read through the big bug portion of the schematic will pipe up soon...

[R.G.]

The only issues I can think of have to do with the quirkiness of OTAs in general.

In this setup you've used the predistortion to allow bigger signals, which is good. Be ready to adjust R9/R10 to get the signal down to where it doesn't distort, but use as big a signal there as possible. This helps a lot with the hiss. For most OTAs, use as big an input signal as you can without distortion. Use as big an Iabc as possible, then adjust the load resistors (R16 and R22) to get the size of output signal you want. All of this is to avoid running in the lowest current regions of the device where it's going to be hissy.

Also be prepared to fuss with the LFO DC levels. Most OTAs have big changes in feedthrough (unwanted control signal noise appearing on the output) at very low bias current levels. My experience is that IC4 and IC5 will need to be able to have both inputs and outputs work down to V-. This lets the TL072 out, because neither the inputs nor outputs can go closer to V- than about 1.4-1.8V. The LM2904 dual and the LM324 quad work well here. You'll probably want a different bias voltage for the LFO from the signal part of the circuit. The signal part wants Vb = 1/2 V+, but on the LFO, you need to be able to turn the DC level of the LFO down so that you're getting the right static level on Iabc with the LFO depth turned all the way down. From there, you need to set the maximum depth to make sure that the bottom of the LFO voltage (a) does not turn the OTAs completely off - the Iabc never goes to zero, which means the voltage on the LFO never gets down to about 1.4V and (b) never gets low enough to cause offensive feedthrough problems. All of that says to me to make the Vb for the LFO section different from the Vb for the signal side. That's a bit of subtlety that you don't find just floating around everywhere on the internet. 

Aside from that, you're using one opamp more than you need, possibly two. Obviously you have one spare section, but you're using duals, so if one section more could be dispensed with, you could remove an opamp package and make it cheaper and smaller. IF you were to move the input signal to the inverting inputs of the OTA (R9 to pin 13 and R10 to pin 4) then the output signal would be inverted at the OTA output, re-inverted by IC6b, and the output would nominally be noninverted without using IC6a, R38 and R39.

However, talking about noninverting phase when filters are involved is kind of simplistic. The fact is, only signals far from the filter frequency keep their original "phase" in the sense of true or inverted. Filters always kick in a variable, frequency dependent phase. In this case the lowpass and high pass filters will both have a phase shift of 45 degrees at their critical frequency, gradually becoming 0 or 90 degrees at 10x above and below the critical frequency. The biggest thing to do is not to invert bass frequencies.

That's probably more information than you wanted. Ask questions about the parts I've made more confusing rather than less.

[Morocotopo]

R.G. thanks for the answer. Very, very helpful and informative. And educative.
A couple of quick questions:
- By Iabc I suppose you mean what you call in your article Ictl, right (pins 1 and 16 of the OTA)?
- I don´t quite get what you mean by predistortion...

Give me some time to redraw according to your suggestions and re-post.

Regarding LFO DC level, could something as simple as making the LFO Vb resistors a trimpot work, to get variable Vb? Or am i missing something? I´m just thinking as I type...

[R.G.]

- By Iabc I suppose you mean what you call in your article Ictl, right (pins 1 and 16 of the OTA)?

Yes. I_control is the same as I_{Amplifier Bias Current}.

- I don´t quite get what you mean by predistortion...

The nature of the input differential amplifier is that it begins to distort gently, then harder at differential voltages from about +/-10mV on up. The predistortion setup corrects for this to a great extent by distorting the signal in the opposite way from the diffamp before the signal goes through the diffamp. The predistortion almost exactly cancels the distortion, allowing signals up to a couple of volts at the differential input without huge distortion.

Regarding LFO DC level, could something as simple as making the LFO Vb resistors a trimpot work, to get variable Vb? Or am i missing something?

You could do that, but the better thing to do would be to either make a second Vb for the LFO and adjust it, then replace the pot with fixed resistors, or to do the math ahead of time so that the LFO plus LFO Vb did not get into the 1.4V (or so...) of the I_ABC input.

[Morocotopo]

OK, second attempt:



I implemented the mods suggested. Now, Let´s see if I got your ideas right, R.G. What I have to do besides the mods already in the schem is:
- Adjust R9/10 for maximum signal into the 13700 without distortion
- Use as big Iabc as possible (I assume you mean DC V bias, or LFO V swing?) and adjust R16/20 to get the wanted out level from the 13700
- Adjust LFO DC so that at min depth the Iabc into the 13700 is correct and at max depth the LFO V negative part never goes below  around 1.4V. All this with the LFO Vb trimpot, and possibly limiting the LFO V swing.

What do you think? I believe I have to start breadboarding...

Thanks for the help.

[R.G.]

... What I have to do besides the mods already in the schem is:
- Adjust R9/10 for maximum signal into the 13700 without distortion

Right - keeps signal to noise as high as possible.

- Use as big Iabc as possible (I assume you mean DC V bias, or LFO V swing?)

In the case of a changing Iabc, keep it all as big as you practically can. In your circuit, the Iabc is the sum of bias voltage C and the LFO swing. So keep the C voltage as high as possible without (a) exceeding 2ma; at 2ma, the chip dies, so probably 1ma is a good place to shoot for and (2) without ever letting the Iabc go to zero by keeping the LFO depth down compared to the C voltage.

This can be tricky. In fact, I realize that I have made an oversimplification in telling you about it.  The Schmitt trigger part of the LFO (IC4a) has some peculiarities when running from a single power supply. I neglected those, but crawling through it now made me remember. The presumption of the Schmitt trigger circuit is that the circuit is biased at some voltage (C does nicely) on the inverting input, but that its output is symmetrical around voltage C. That means that the voltage fed back to the + input through R23 and to the integrator through R24 and R25 from the output of IC4a should be symmetrical around voltage C. That does happen if (a) you're using an opamp for IC4 which has an output voltage that goes equally close to the V+ as the V- output and (b) voltage C is right in the middle. Neither of these are true here. The LM324 output pin goes much closer to V- than V+, and voltage C is variable. They ...might... cancel out, but that's counting on luck, not design. A sharper thing to do is to force the output to be symmetrical around C by tying it to C with a pair of series anode-to-anode zeners. This forces the output pin to be be symmetrical around C whatever C is, which gives you symmetry on both the input of the Schmitt and the integrator. A 1k in series with IC4a's output pin and a pair of 3V zeners would work well down to C = 3.0V.

Sorry - I haven't dug through a LFO Schmitt for a while, and this all came back to me when I thought about it.

and adjust R16/20 to get the wanted out level from the 13700

Correct. The 13700 output is a current. It doesn't much care what voltage it puts out, as long as it's within the range its output drivers can provide. The output signal voltage will be the current times the output resistor.

- Adjust LFO DC so that at min depth the Iabc into the 13700 is correct and at max depth the LFO V negative part never goes below  around 1.4V. All this with the LFO Vb trimpot, and possibly limiting the LFO V swing.

Correct.

[Morocotopo]

Let´s see if I understood you well. Like this?

[R.G.]

Let´s see if I understood you well. Like this?

Yes, except that R23 and R24 come off the junction of R39 and D2 cathode, not the output of IC4a. IC4a then pulls R23 and R24 higher or lower than Vc, but the zeners limit how far they get pulled from C.

[Morocotopo]

OK. I think I´m gonna start breadboarding this and see how it works in reality. R.G, I can´t begin to tell you how helpful you are... thanks for educating me in this addicting stuff! I hope this will end up working to my satisfaction. The Keen Tremolo?   

[Morocotopo]

Just to be sure, like this?

[soggybag]

Great idea, I was thinking about doing something similar myself after reading R.G.s article on OTAs. Then I found your layout for the Tremolo X and thought I build that instead.

I was wondering about what R.G. was saying about the LFO voltage range. I wondered if, rather than biasing the LFO op-amps, if it would be possible to bias the output?

[soggybag]

Much of the talk about biasing the LFO voltage to work properly reminds of John Hollis Frobnicator. This was a trem based around a 3080 OTA. You might take a look at to get some ideas for strange biasing. There are a few LEDs used to bias the Frobnicator. I remember these needed to all be the same or it would not work.

[R.G.]

Just to be sure, like this?

Yes, that's what I meant.

I was wondering about what R.G. was saying about the LFO voltage range. I wondered if, rather than biasing the LFO op-amps, if it would be possible to bias the output?

If by that you mean adding an offset voltage into the two opamps that make the true/inverted LFO for the OTAs, yes it's possible. But it gets complicated too. These things are really doing what amounts to analog computation. Not bad actually, since that's what opamps were originally intended to do. The guns on WWII battleships and the bombsights on WWII bombers were aimed with analog computers - literally vacuum tube opamps - doing the computations. But it is a computation and you have to manipulate the things like terms in an equation. I prefer to use the "factored" form of the equations, adding a constant offset all across the board by moving the reference rather than adding a constant term into each part of the equation. But you can get to it both ways.

Much of the talk about biasing the LFO voltage to work properly reminds of John Hollis Frobnicator. This was a trem based around a 3080 OTA. You might take a look at to get some ideas for strange biasing. There are a few LEDs used to bias the Frobnicator. I remember these needed to all be the same or it would not work.

For the same reasons as above. Diodes, including LEDs, have an only approximately-constant forward voltage, and then only when looked at from a distance. Different types of diodes, different diodes of the same type, and the same diode at different currents have slightly different voltages. When doing sum-and-difference, especially differences, small offsets can add up to big errors.

[Morocotopo]

OK people, just finished breadboarding and it works!

I have ticking, now I have to determine if this is because of layout (messy breadboard), ground distribution, feedthrough of the control signal in the 13600, or what. I still haven´t tested if the output is unity gain. No distortion so far, at least with single coils. Noise/hiss besides the ticking seems to be fine (low). Moving the feedtrough trimpots doesn´t seem to vary the ticking, only effect is the corresponding filter section stops "working". Moving the LFO Vb trim only seems to make it go slower and eventually stop oscillating. I´ve changed some values to what´s available: 7K5 for 6K8, 3V zener for 3V3 ones, 4K3 for 4K7.  Now I have to start taking measurements: voltages, looking at waveforms with my "new" borrowed oscilloscope, etc.

Some sound samples:

Noise:

Noise.mp3 - 0.09MB

6th string, to hear the lows tremoloed:

Lo.mp3 - 0.35MB

Idem with 1st string:

Hi.mp3 - 0.20MB

Some chords to hear the full effect, different depths and speeds:

full test.mp3 - 1.71MB

I´m happier than a dog with two tails...

EDIT: with depth pot at minimum, there´s practically no ticking.

[Morocotopo]

Update: I was using a 13600 OTA, now I tried a 13700 and the ticking is gone! The 13700 is better for audio applications, says so on the datasheet, if I didn´t misunderstood that the Iabc is independent of the Id bias current, and that minimizes pops when the signal changes abruptly (audio, and maybe control?). Or something like that. 
Now I have to optimize gain in all the stages, to minimize distortion and maximize signal to noise ratio. And also, I´m gonna try making the lo and hi "branches" switchable in and out, so I can get only-high-tremolo and only-low-tremolo. Madness!

[mr.adambeck]

Wow!  from listening to your mp3's that sounds GREAT!  So much more lush than most tremolo's.  Glad you were able to get rid of the ticking!