American Sine Language - How to adapt triangle LFOs to sine waves?

[Mark Hammer]

The old Boss CE-1 has a certain reputation.  One of the distinctive features of the CE-1 was its use of different LFOs, or at least modified waveforms, for Vibrato and Chorus settings.  In particular, it used standard triangle for chorus, and a sinusoidal LFO for vibrato.

Looking at a drawing of the Little Angel PT2399-based chorus this morning, I was prompted into wondering how one might tack on a few parts to the same LFO one sees on the LIttle Angel, MXR Phase 90, and so many others, to transform the triangle LFO into something closer to sinusoidal, so that vibrato settings might be more pleasing.

That could probably be generalized to a great many modulated effects whereas "rounder" turnaround at the ends of the sweep comes closer to the desired sound/feel.

So how does one do it, in a manner that doesn't cost too much in LFO output amplitude?  And I'll say up front that the modulation does not necessarily have to be pure sine.  I'm fine with a barebones closer-to-sine-than-to-triangle compromise if it requires fewer parts, less intervention, simpler switching schemes, etc. 

[R.G.]

The concept is easy - one does soft clipping on the peaks of the triangle. This was at one time the bread and butter of almost all waveform generators that made sines. There are a number of schemes to do this, including diode-resistor or transistor-resistor ladders, JFET clippers, and at least one MOSFET-as-a-diode clipper.

There are three practical problems with this. One is that the amplitude relationship of the triangle to the clipper circuits was critical in getting a decent sine. For your purposes, that's not too bad since a pure sine isn't the issue. But it's a recurring problem, in that each design to be retrofitted will probably need a different clipper design to get near-sine, as the LFOs' peak points are all different.

The next is that the design of the clipper is a bit tricky in standard pedal practice with a unipolar power supply. Most of the standard waveform shaping circuits used a center-ground to simplify the clippers.

Finally, you'll have to either make the triangle larger before clipping so the sine shaper restores the peak values to the LFO for the good of the circuit it drives, or you'll have to amplify the shaped sine after the shaper to get the peaks back to the same voltage. Amplification is more trouble, but it's also more flexible, as the opamp stage likely to be used can also restore DC position as well as size.

Bottom line - not too difficult in concept, but nit-picky design to be done for each pedal to make it work right.

[slacker]

The small clone has a simple RC lowpass filter after the LFO, this rounds the waveform up a bit at faster speeds and reduces the amplitude a bit, whilst keeping it triangular at slower speeds.
Think it's the small clone anyway, I've seen it somewhere, might even have been you who mentioned it at some point.

[Jazznoise]

Why not a low pass filter? You could use a dual gang pot and have the cutoff "track" the LFO speed.

[R.G.]

Lowpass works somewhat. You do have to track at least approximately the triangle frequency, otherwise the amplitude changes. The need to keep the amplitude on the final LFO you use for the circuit makes a lot of this sticky.

[Mark Hammer]

The small clone has a simple RC lowpass filter after the LFO, this rounds the waveform up a bit at faster speeds and reduces the amplitude a bit, whilst keeping it triangular at slower speeds.
Think it's the small clone anyway, I've seen it somewhere, might even have been you who mentioned it at some point.

It was used on the first-issue (6-chip) Small Stones.  If you look at the schematic, you'll see a simple 10k/50uf RC filter on the output of the LFO.  That forms a lowpass filter with a corner frequency of just over .3hz.  For very slow sweeps, the LFO output is "hypertriangular" (sinusoidal at one extreme, and "pointy" at the other, like this: http://userdisk.webry.biglobe.ne.jp/000/024/65/1/200510_img_4.gif ), but as the sweep speeds up, the RC filter adds some lag to the turnaround at the "pointy" end, making the sweep more suitable for fast bubbly sounds.  The same trick is not used on later-issue Small Stones, but you can always add it.

The interesting thing about it is that it is useful precisely BECAUSE it does not track the LFO frequency.  Very slow sweeps benefit by being nonlinear and moving through different parts of the spectrum at different rates, while faster sweeps benefit by being more even and gentle.  The Small Stone trick gradually changes the form of the sweep with LFO rate.  So simple.  So smart.

[slacker]

Ah yeah, that's what I was thinking of, knew Small Clone didn't sound right.

[Mark Hammer]

So let's say one wanted to sine-o-fy the LFO on a Phase 90, to make it more suitable for bubbly vibrato.  Would one split the 3M3 (3M9 in some issues) resistor from the LFO to the JFET gates into several series resistors of roughly equivalent total value, and stick a pair of diodes to ground to soft clip the LFO?  Is that "all it would take"?

[Fender3D]

This is what I added to my custom ADA flanger:



With a simple 2P4T you'll obtain 4 different LFO waves:
Sine, triangle, Hyper positive and hyper negative.

The out op-amp will raise LFO level (R7 gain resistor as per ADA flanger)

[R.G.]

So let's say one wanted to sine-o-fy the LFO on a Phase 90, to make it more suitable for bubbly vibrato.  Would one split the 3M3 (3M9 in some issues) resistor from the LFO to the JFET gates into several series resistors of roughly equivalent total value, and stick a pair of diodes to ground to soft clip the LFO?  Is that "all it would take"?

You drew a lucky one there.

Dumped the P90 LFO into the simulator. In the case of the P90, the peak-to-peak output of the LFO at the LFO integrating cap is about 1.4V. Hey, wait! That's two diode drops! And there's a fixed, firm reference in the form of the zener voltage to tie them to.

OK, two diodes back to back to reference voltage ... um, about 150K to the integrating cap to not mess with the Schmitt trigger unduly, and - nothing. The thing doesn't put out enough to trip the diodes.

OK, change the Schmitt trigger trip points by dinking with the 470K/150K reference. About 330K instead of 150K gets some bigger output. Now it puts out enough to trip the diodes over, but it's too flat.

Inserting a 47K smooths things out a bit, but it's still a bit small. Inserting another pair of diodes in series with the existing two gives a pretty nice waveform of about 1.4V peak to peak.

Well, that wasn't too hard. All done. It's definitely a lot more sine-ish than it was.

Completed mod:
With reference to the Phase 90 schematic at GGG: http://www.generalguitargadgets.com/pdf/ggg_p90_sc.pdf
1. Break the connection between the junction of C7, R19, and R21.
2. Insert in series from the junction of C7, R19 and R21 a 150K, a 47K and two pairs of silicon clipping diodes (1N4148 works nicely) to the zener reference voltage at the top of D2 (5.1V zener).
3. At the junction of the new 47K and the new 150K, re-connect the 3.9M (3.3M) resistor disconnected above.

At least Mr. Simulator thinks that works nicely. This will need some verification in practice. If you'd like to tell me how that works...  

[Mark Hammer]

I'd LOVE to tell you how it works, but most of my evenings this week will be taken up, and they frown on me bringing my soldering iron to work, so it will have to wait until next week.  But I will try it.

fender3D, I see someone was reading that POLYPHONY article I posted!    I was thinking of that too, but was hoping for something that was simple enough to only require tacking a few parts to the copper side of a board.  On the other hand, if working from perf or self-designed layout, pretty frickin' cool!

[R.G.]

I'd LOVE to tell you how it works, but most of my evenings this week will be taken up, and they frown on me bringing my soldering iron to work, so it will have to wait until next week.  But I will try it.

No biggie - I was just poking you.  Your time for doing actual hardware is about as limited as mine is.

I'll probably write/draw that up and post it when I get the chance.

[tca]

Actually there is a trig-to-sine converter in some vol. of the Polyphony Publishing magazine. I don't recall the number of that particular issue but I think I have the PDF in this computer, have to look. Mean while have a look here, it only takes 4 resistors two diodes and a jfet.



It goes from some signal, square, triangle (sawtooth), and sine. I was thinking in something else when I simulated that circuit but the the trig-to-sine works.

Cheers.

P.S. (edit)
I forgot to tell, for the trig-to-sine to work it needs a +-5V trig signal.

[tca]

Just saw the image posted by fender3D... big figure with small screen laptop,  could not see the big picture!  

[R.G.]

That's the one I was referring to as "JFET clipper" in my first reply.

The choice of JFET is critical, as is the amplitude of the triangle you feed it. It can be tweaked in, and was used in some waveform generator circuits, back in the day.

[cloudscapes]

XR2206 is an easy-to-find LFO-on-a-chip, it does sine and does all the nitpicky stuff within.

[Mark Hammer]

If you want to go all out, there's always this oldie-but-goodie:
http://hammer.ampage.org/files/SEWAR.PDF

Addendum: This one uses the aforementioned XR2206 to derive sine, triangle, ramp, sawtooth, and square waves, in addition to providing a filtered noise source for "random".  The stock circuit also includes an XR2207 for producing a HF clock, suitable for driving a BBD, but you don't have to use that.  It is relatively easy to adapt to use just the LFO output.

Stupid me, I HAVE one of those boards, stuffed and sitting around, plus an uncommitted (and also populated) voltage-controlled delay board sitting around.  Man, I really have to take some vacation time and use these things!

[Devius]

This is slightly off topic but watching you guys talk is like being a fly on the wall in a Jedi council meeting and you guys are talking about the force.
Mad respect for you boys.

[R.G.]

Stupid me, I HAVE one of those boards, stuffed and sitting around, plus an uncommitted (and also populated) voltage-controlled delay board sitting around.  Man, I really have to take some vacation time and use these things!

The Shoemaker's Chlildren come to mind.

[Fender3D]

fender3D, I see someone was reading that POLYPHONY article I posted!   

I thank you, you're an invaluable material source...

I was thinking of that too, but was hoping for something that was simple enough to only require tacking a few parts to the copper side of a board.  On the other hand, if working from perf or self-designed layout, pretty frickin' cool!

please note that you'll just need 4 resistors, FET and 2 diodes, they might happily rest under a 2P4T  LFO and recovery stage are those on ADA board, just tinker with R7 (moosapotamus' R66) for a proper LFO level...
pretty easy to try out