The FM Tube Modulator Jimi Loved

His guitar slung across his back, his dusty boots is his cadillac.

Moderators: VelvetGeorge, BUG

Post Reply
daveweyer
Senior Member
Posts: 605
Joined: Wed Oct 29, 2014 9:36 pm
Just the numbers in order: 13492

The FM Tube Modulator Jimi Loved

Post by daveweyer » Sat Nov 07, 2015 12:40 pm

I started this thread to converge all the tube modulator comments and exchanges, thinking it might be easier to access this information if it had its own topic heading. The West Coast topic is getting a lot of pages, only a brave few wade through it all. Some kind of directory would be nice, like an index of subjects and pages where info could be found.
For now, a separate heading is the closest thing. Maybe I should do this with Jimi's Guild Quantum amp, the Wahs and the FF units too.
This was JC's idea and it seems to make good sense.
If someone would quote my expose of the tube modulator theory over on to this topic, we could start with questions about all that and go from there. Anyone interested could follow the exchanges.

I'll start by saying that Jimi went nuts when he heard this thing. Many guitar players did because there was absolutely nothing like it under the sun for guitar players. Everyone wanted to be the first to use it, to gain that little advantage in musical exploration that would set them apart. There were simple phase shift units around, operated by LDR chips, that was the Univibe and several others. But the tube modulator was different, so vastly complex of a sound, and so many other possibilities for using it to create new worlds of sonic material. Given the proper regenerative signals, the unit will actually play itself--I'll post some of this stuff to hear as we go along.
I never could part with the unit, but Jimi wanted me to build one especially for using on the road; I said I would do just that, but the thing is complex and would have required a lot of engineering to convert into a solid state unit which would be truly portable but still maintain that incredible randomness that it has in its tube form. Needless to say I never had time to get the job done, what with all the amp building and mods. By the time I did have a chance to get started on it Jimi had gone into the really big tube modulator of infinity.

But I did get a few moments of Jimi playing through the unit, and I have since taken some of Jimi's material and re-recorded it through the unit--just to get an idea of what folks would have heard if I had let him have the unit and made the proper switch boxes he would have needed to use it on the road.
I'll share some of this stuff as well during the course of this topic.

With that all said, fire away!

daveweyer
Senior Member
Posts: 605
Joined: Wed Oct 29, 2014 9:36 pm
Just the numbers in order: 13492

Re: The FM Tube Modulator Jimi Loved

Post by daveweyer » Sat Nov 07, 2015 1:55 pm

JC, the approximate modulation frequencies are basically percentages of the number of HZ in that band, obviously the effect of a one cycle change on a very low frequency is much greater than it would be on a high frequency--therefore it it necessary to use greater modulating frequencies the higher the band of frequency you need to modulate. I have experimented with changing the frequencies, you can get changes in the pulsating pattern, but there is magic in the 1,2,4,8 numbers because it sounds like you can time it out and play along with the pulsations--but you can't because they are ever so slightly off and it always turns around on you. Dropping by half works too, but it lessens the effect--the bands would need to be shifted as well.
For the moment and the first try, it might be best to use the engineering which is all done for you.

Eb7+9
Senior Member
Posts: 104
Joined: Wed Mar 09, 2011 1:30 pm
Just the numbers in order: 13492
Contact:

Re: The FM Tube Modulator Jimi Loved

Post by Eb7+9 » Sat Nov 07, 2015 11:09 pm

thx Dave,
I'll start by saying that Jimi went nuts when he heard this thing.
:jimi:

well, I can't wait to check this thing out myself ...

For my first "experiment" I need to verify that the modulator replacement circuit will have comparable effect on signal ... to that end I'm still not sure I understand totally how the tube version works ... as you know, if we alter the input bias on a triode or pentode the small-signal gain will drop (gradually) as the transfer curvature provides a gradual shut-off // the principle behind many tube based Limiter circuits ... and so, the oscillations might be simply modulating the gain levels in each of the three legs of the modulator, and then summed into a common load (albeit, in a non-linear manner) ... OTOH, the circuit might simply be "gating" the injected signals // you might have mentioned this already ... but, for some reason, I don't see how that would cause a time shift ... maybe it's just me, but I have a problem with the later explanation

for sake of argument, if the process relies on gradual gain modulation we might be able to get away with using jFET's (as they exhibit somewhat gradual transfer slope variation as well) ...

http://www.lynx.net/~jc/transferCurvature12ax7.jpg

http://www.lynx.net/~jc/transferCurvature2n5457.jpg

but not BJT's as they do not exhibit a concave transfer curvature like triodes and jFET's do (compare below with two cases above) ...

http://www.lynx.net/~jc/transferCurvature2n5210.jpg

OTOH, we could well use OTA's if gain modulation is the player at hand here, since it would be relatively easy to modulate individual bias current on each OTA ... come to think of it, an OTA based modulator would be preferable as it is quite the task to properly characterize jFET devices for sake of grouping them into tightly matched trios ...

as well, the transfer curve of jFET's is not as spread out in the cutoff region as it is with triodes (from my sims), and if the process relies on that effect then jFET's might not have enough gain reduction/headroom range ... we'll see about that when I've got things happening on the bench ... but, of course, an OTA approach would make the circuit build way more accessible to DIY'ers ... with carefully matched jFET's, not so much ... so this might help me make a choice towards using OTA's

for now I'm trying to rule out the possibility that I'm missing something (else) before I jump to conclusions ... just for sake of proofing the conversion I'm wondering if I can simply build myself a single oscillator and modulator block and test that alone, or do I have to build the whole thing with tank circuits, etc etc ... again, we'll see when I get there

right now, I want to try each kind of modulator (jFET and OTA) with a separate oscillator and single modulator block // and see what that produces ... maybe the non-linear aspect of jFET's (like triodes) is crucial, maybe not - in which case, a linear OTA approach would work ... regardless of whether gating is or is not the underlying mechanism, as both can be achieved in these two cases as well

(yeah ... much guess-work at this point)

I know that eventually I will have to get much closer to the complete architecture in order to have taste-able pudding
but at this point I need to come up with a simple test platform just to confirm basic jFET/OTA modulator operation ...

more shortly ...

~jcm

---

(ed.)

btw, here's the link to pictures of Dave's "Chora-Tone" unit and the Baldwin schematic ...
http://imgur.com/a/mOQr2 ...

thanks for posting that Adrien
Last edited by Eb7+9 on Mon Dec 07, 2015 9:45 pm, edited 7 times in total.
modern VT circuit analysis and modeling: https://viva-analog.com/product/ifmta-book-pdf/

Eb7+9
Senior Member
Posts: 104
Joined: Wed Mar 09, 2011 1:30 pm
Just the numbers in order: 13492
Contact:

Re: The FM Tube Modulator Jimi Loved

Post by Eb7+9 » Sat Nov 07, 2015 11:20 pm

I may was well point out I'm thinking of using a basic op-amp based 3-phase oscillator here ...
a very common circuit

having op-amp outputs to play with might be an advantage in the new architecture
if need be, oscillation amplitude can be set by adding multiple didoes, or simply using Zeners ...
once we know what the modulator requirements are we can set these accordingly

http://www.lynx.net/~jc/ChoratoneOscillatorLF411a.jpg
http://www.lynx.net/~jc/ChoratoneOscillatorLF411b.jpg
Last edited by Eb7+9 on Mon Dec 07, 2015 9:44 pm, edited 2 times in total.
modern VT circuit analysis and modeling: https://viva-analog.com/product/ifmta-book-pdf/

Tek465b
Senior Member
Posts: 200
Joined: Sat Jan 10, 2015 9:43 am
Just the numbers in order: 13492
Contact:

Re: The FM Tube Modulator Jimi Loved

Post by Tek465b » Sun Nov 08, 2015 9:42 am

Instead of a diode or zener, could you use a vbe multiplier.
Reverse breakdown voltage is pretty low, but your at about 2/-2 volt from the graph its not very high.

daveweyer
Senior Member
Posts: 605
Joined: Wed Oct 29, 2014 9:36 pm
Just the numbers in order: 13492

Re: The FM Tube Modulator Jimi Loved

Post by daveweyer » Sun Nov 08, 2015 11:30 am

I think the main concept to wrap the brain around is that in the phase shift oscillator, each triode stage conducts for 1/3 of the cycle, in other words it becomes a gate for exactly 120 degrees. The real genius of the circuit is that this gating action is also used to impress another signal on the oscillator output--the oscillator gives the designer the modulating frequency and also gives him the gated signal from the matrix network.
So first of all, it is only important that whatever you use in the three stage RC oscillator as an amplifier, is able to be switched off smoothly during the 2/3 of the cycle when it is not conducting. Any device should work here, but it has to be configured in such a way that it will provide oscillation and gate at the same time. The designer of the tube modulator used a high perveance triode, the 12AU7, which is low gain, about 6 to 10 in the circuit, has some inverse feedback, the 270 ohm cathode resistor, was readily available and cheap, and would conduct with extremely low plate voltage, the high perveance characteristic.
No particular attention was paid to the cutoff characteristics except that is fairly sharp in that tube. In other words it didn't need a long linear ramp, just that the point at which it cut off was equal to the point where the next stage turned on so that the oscillation was of low distortion.

The point of all this is that the gated signals from the matrix appear on the output of the oscillator.
So to answer the question about what happens with these gated signals: Since the signals from the matrix are actually three versions of the input signal, 120 degrees lagging each other, they can be introduced into each stage, and gated by the action of the LF oscillator operating in its normal pattern. The oscillator is just an oscillator, it doesn't know anything unusual is going on; it's just doing its thing.

So lets say the input signal is 100HZ, and the oscillator is operating at 1HZ, Three versions of the 100 HZ signal appear at the output of the matrix, 120 degrees lagging each other. The first version of the signal is applied to the first stage of the oscillator, and is gated through the stage for 120 degrees of the 1HZ frequency. The next version of the 100HZ signal, which is 120 degrees behind the first signal, is gated through the second stage of the LF oscillator for the next 120 degrees of the 1HZ frequency, and the third version of the input signal which is 120 degrees behind the second version, is gated through the third stage of the LF oscillator for the last 120 degrees of the 1HZ cycle.

So what just happened? A signal appears on the output of the 1HZ oscillator which equals the original input signal shifted by 120 degrees times 3, or 360 degrees in the lagging direction, meaning it is now a 99 HZ signal. One HZ has been subtracted from the original signal by repetitive shifts backwards of the original signal, you might describe it as removing time, or time delay. The reason the whole thing works is that the 120 degree shifts are perfectly timed by the 120 degree conductions in the oscillator to equal a smooth low distortion composite sine wave of 1HZ less than the original.

This is the most important concept of the circuit. I have to admit, it took me quite a while of thinking about it to grok the idea and get a decent mental picture of what was happening.
It is quite conceivable to make the circuit using other methods, but nothing I have thought of does it as simply as Mr. Wayne's idea.

Eb7+9
Senior Member
Posts: 104
Joined: Wed Mar 09, 2011 1:30 pm
Just the numbers in order: 13492
Contact:

Re: The FM Tube Modulator Jimi Loved

Post by Eb7+9 » Sun Nov 08, 2015 4:36 pm

alrighty then ...

I spent the last hour drawing a circuit that will do both ... it will do a sine wave modulation or a gated modulation ... simply a matter of altering the voltage-to-current conversion going into the trio of OTA's ... I think I'm gonna go with that first as that's what the design process is leading me to at the moment ... if I fail there I will head back towards a more class-A implementation using matched jFET's ...

if all that fails, then we'll simply have to go back to tubes ... ha!

:P

but I have hopes, and feel that it will be revealing either way ...

I first need to double-check the values in the filter lattice ...

and I also ordered a bunch of sealed inductors in Ebay ...
or I might just wind my own (tho time is a factor for me ... we'll see)

once I know what inductors I'll be using I can draw a pcb and send it off for CNC'ing
(usually takes me a couple of days to get it back)

I think, I'm gonna go with the 5-bank design for the full-meal deal and can always leave one out anyway
... and also modify operating frequencies like you did in your unit Dave,

I'm taking a chance that this simpler design will work (ie., create THE effect, in an equivalent manner)

I appreciate re-iterating your point of view on all this ...
point taken!

definitely gives me something concrete to shoot for at this point

~jcm

---
by Tek465b » Sun Nov 08, 2015 9:42 am
Instead of a diode or zener, could you use a vbe multiplier.
Reverse breakdown voltage is pretty low, but your at about 2/-2 volt from the graph its not very high.
thanks for the input Tek

that's a good idea ... and I may use it

but, just to be clear,
my 2v/-2v output swing(s) is enough to drive a degenerated differential pair for linear V-I conversion ...
providing a linearly modulated bias current on the OTA's ...

in fact, the exact level is not crucial as I can adjust the degeneration resistance accordingly ...

either way, diodes are easy and convenient to use here

since this general "op-amp based oscillator >> pot >> V/I converter >> OTA" approach requires no special extra-matching of components, as OTA's come already "matched" intrinsically, it makes it easier for others to build afterwards (assuming it works) ... more DIY friendly

(I'm sure you know all this, just pointing it out to the general population here ...)
modern VT circuit analysis and modeling: https://viva-analog.com/product/ifmta-book-pdf/

daveweyer
Senior Member
Posts: 605
Joined: Wed Oct 29, 2014 9:36 pm
Just the numbers in order: 13492

Re: The FM Tube Modulator Jimi Loved

Post by daveweyer » Sun Nov 08, 2015 4:55 pm

I thought maybe I should add an expansion to the above post. If you have ever worked in ProTools or other audio editing program that does to-the-sample editing, you will probably have noticed that any two waveforms can be glued together without an audio artifact if the point at which they are joined is at the exact DC level component of the AC signal, and that point does not leave a spike in the waveform. (Such as when the waveforms are steep and heading in the opposite direction)
The same thing happens with the joining of the three lagging phases of the input signal. If you graph it out on paper you can see the 120 degree points on the falling waveforms, lined up at lagging intervals of 120 degrees, fall at the exact DC level component of the AC waveform, only shifted in time. When the first version of the input signal reaches the 120 degree point, the oscillator switches to the second version of the input waveform, which is lagging the first by 120 degrees--this transfer from one to the next only takes less than a microsecond, and since there is no DC level shift, there is no audio artifact other than the pitch is lower.
The first gate subtracts 1/3 of a cycle, the second subtracts another third of a cycle, and the third subtracts another third of a cycle, that is, if the oscillator is running at 1HZ. All this by continually jumping from one lagging phase to the next ad infinitum.

More clear, or more muddy?

daveweyer
Senior Member
Posts: 605
Joined: Wed Oct 29, 2014 9:36 pm
Just the numbers in order: 13492

Re: The FM Tube Modulator Jimi Loved

Post by daveweyer » Sun Nov 08, 2015 5:08 pm

JC, I think I'd get the 1HZ phase shift oscillator going, and find a point to inject the input waveform so that it gets gated at exactly 120 degree intervals. The input matrix should be easy enough to do to get the three 120 degree phase lagging signals for injecting into the oscillator amps.
As far as needing the LC output network, you could easily accomplish that with active filters just using your OP amps.
It ends up just being a non-adjustable parametric EQ.

It seems like getting one band going first would save a lot of time in the long run. Then you could measure the output frequency and see if you really got x- (1HZ) and it was a nice sine wave which would keep making your scope retrigger.

daveweyer
Senior Member
Posts: 605
Joined: Wed Oct 29, 2014 9:36 pm
Just the numbers in order: 13492

Re: The FM Tube Modulator Jimi Loved

Post by daveweyer » Sun Nov 08, 2015 6:45 pm

JC, you know, you don't have to use the RC oscillator system to get these signals gated; it was extremely handy in the days when four quadrant multipliers were out of the question. Gates are ubiquitous now, so you are not restricted to using an oscillator to get accurate gate times.
You could just as well use gates, and then set up any kind of counting system to trigger the gates to open at the exact intervals needed to sequence the three versions of the input signal, that is, at every 120 degrees of the LF oscillator frequency. I can think of a number of ways to do this as I am sure you can too.
Maybe it would make it a little more logical from today's perspective, and it might make it more straightforward to design. Just thinking out loud here.........

Eb7+9
Senior Member
Posts: 104
Joined: Wed Mar 09, 2011 1:30 pm
Just the numbers in order: 13492
Contact:

Re: The FM Tube Modulator Jimi Loved

Post by Eb7+9 » Sun Nov 08, 2015 9:14 pm

Thanks Dave,

I will take all this info into consideration ... very helpful !!

I think I also understand when you say the following:

... it is also a feature of this system that the four oscillators are approximately at frequency, but vary just enough that the resultant amplitude modulations are not perfectly timed with each other--this provides a random but rhythmic quality which sounds like you can lock up with it, but never will. As soon as you think you have it counted out it is somewhere else ...
this, I'm guessing, is because the LF-oscillators are not phase-locked ... and so, they independently interfere/re-inforce each other in irregular and unpredictable patterns over time, regardless of frequency relationhips ... I can see how this would provide an interesting alternative to the more-mundane (and so dominant) sine-wave style effect, that is pretty much all Phasors and Chorus/Flanger boxes, ...

Jimi ...?!!

ahead of his time, would likely appreciate that aspect of it right away ...

---

For reference sake, it would be cool if you (Dave) could process a simple E chord on a live guitar (not taped)
and have us hear it going thru this box alone ...

I'm sure the dry/wet stereo is to die for ...
but just so I have some kind of reference to go by ... 100% wet reference effect sound would be useful to me

thanks for extra info Dave,

~jcm
Last edited by Eb7+9 on Sun Nov 08, 2015 9:39 pm, edited 2 times in total.
modern VT circuit analysis and modeling: https://viva-analog.com/product/ifmta-book-pdf/

Eb7+9
Senior Member
Posts: 104
Joined: Wed Mar 09, 2011 1:30 pm
Just the numbers in order: 13492
Contact:

Re: The FM Tube Modulator Jimi Loved

Post by Eb7+9 » Sun Nov 08, 2015 9:34 pm

daveweyer wrote:JC, you know, you don't have to use the RC oscillator system to get these signals gated; it was extremely handy in the days when four quadrant multipliers were out of the question. Gates are ubiquitous now, so you are not restricted to using an oscillator to get accurate gate times.

You could just as well use gates, and then set up any kind of counting system to trigger the gates to open at the exact intervals needed to sequence the three versions of the input signal, that is, at every 120 degrees of the LF oscillator frequency. I can think of a number of ways to do this as I am sure you can too.
Maybe it would make it a little more logical from today's perspective, and it might make it more straightforward to design. Just thinking out loud here.........
using a daisy-chain counter and analog gates would not be too hard to implement, I agree ...
not sure about the glitching at the boundaries tho (... a 2nd'ary issue)

ok, I will try that if my "linear" approach fails ... miserably ...

;)
modern VT circuit analysis and modeling: https://viva-analog.com/product/ifmta-book-pdf/

Eb7+9
Senior Member
Posts: 104
Joined: Wed Mar 09, 2011 1:30 pm
Just the numbers in order: 13492
Contact:

Re: The FM Tube Modulator Jimi Loved

Post by Eb7+9 » Sun Nov 08, 2015 10:23 pm

I'm adding a useful Wiki page link on Analogue Filtering
for anyone wishing to check out some background info on all this frequency shifting business, and its history (electronically) ...

... mention of "resonance" and the work of Sydney Darlington are noteworthy here

https://en.wikipedia.org/wiki/Analogue_filter
modern VT circuit analysis and modeling: https://viva-analog.com/product/ifmta-book-pdf/

daveweyer
Senior Member
Posts: 605
Joined: Wed Oct 29, 2014 9:36 pm
Just the numbers in order: 13492

Re: The FM Tube Modulator Jimi Loved

Post by daveweyer » Sun Nov 08, 2015 11:00 pm

For a minute I thought you wrote the Wiki article on filters! Wow, pretty thorough romp. The link is good too, if there are any Marshall lovers with a strong left brain.

The gates would have to be matched obviously so that the DC trigger proportionate level would be nearly identical, say out to 16 bits or so, perhaps a microvolt or so.
Wayne just used the LC oscillator because it forced the tubes to match their conduction angles, saving him the trouble of little feedback circuits which would guarantee the gates open at the identical voltage level. It's not hard to envision such circuits these days though--we have a lot more precision to work with.

daveweyer
Senior Member
Posts: 605
Joined: Wed Oct 29, 2014 9:36 pm
Just the numbers in order: 13492

Re: The FM Tube Modulator Jimi Loved

Post by daveweyer » Mon Nov 09, 2015 12:16 pm

JC, here's another possibility for the modulator: Use the Supertex LND 150 N3-G depletion mode mosfet. These are normally on, and biased exactly as a tube, even working at the same supply voltages.
With these devices you could recreate the circuit nearly part for part, the drain voltage would be in the 100 volt range just like the tubes in the original circuit. These devices are no bigger than a 2N2222 signal voltage transistor. With the RC oscillator, all the old bulky capacitors could be replaced with tiny monolithic caps so the whole circuit could conceivably be stuffed in a pedal size box; there is no need for fancy tubular caps in the oscillator section, and even if there were, the size of these has shrunk dramatically as well.
Just food for thought.

Post Reply