The FM Tube Modulator Jimi Loved

[Eb7+9]

thx for the input guys ...

I agree with the "some" part Dave ...

we can't really play with the resistors anyway as it will change the shape/size of the waveform (flattens it ...)
and the circuit won't allow much variation in resistance anyway, due to Birkhauser's criterion being broken quickly

and the current-mode 3-phase oscillator (paper) poses another problem ...
that of offering a bipolar (+/-) output current, which needs to be made unipolar (+/0)
which would require an additional static current to be summed so we end up with a current that goes down to zero and stays positive
(I shoulda noticed that before) ...
big headaches

the PWM approach needs to be done with a high degree of exactness
not something a 555 PWM circuit could do, let alone a 3-phase version ...

which means , Arduino/PIC's would have to be programmed ...
much work in itself ...
and then we would need 5 Arduino's ...

I think switching caps is really the only practical way to go here ...
maybe Arduino's too
but quite a bit of effort ...

[Tek465b]

5 arduino?, 1 arduino has many input and many output, (depend of the chip, i usually work with the atmega328p) wich has 6 analog input and 14 digital I/O 6 of wich provide pwm output(1 bit, 0-255) on 3 timer wich you can change frequency. we can use another chip model with more pwm output/more timers too.
but i agree, switching caps seem more practical for now.

[daveweyer]

Perhaps the DSP route is the best. I personally like to see freedom from high density microchips from a long term repairability and user accessibility perspective, and for that I am willing to suffer larger sizes and more expensive mechanical parts.
You probably already know this but you can add capacitance by lowering resistance. There is some complexity involved in multiple segment circuits like this one just due to numbers.
Also, again, it isn't really necessary to use a three stage RC oscillator in your design. The only reason Wayne did it was to get the gating of the phase delayed signals. If he had obtained access to OTAs, I'm sure he would have just used a multiplex system to get all his frequencies, and it would have given him the option of using other waveforms for gating them. For instance, in those days it was common practice to use a master oscillator, and then use flip-flops to get octave divisions, i.e. 16HZ, 8HZ, 4HZ, 2HZ and 1HZ. You just use a simple integrator to get sine waves, and it can all be done on a tiny circuit board. With some simple timing, you'd have all you need to operate the OTAs. The master oscillators could be tuned to develop any degree of phase shift, and/or variability in the locked AM functions.
These days they use a micro to develop all the frequencies and all the timing. When Arduinos disappear, like so many of the user programming systems of yesteryear, everyone will be locked out. Other than that I'd be fine with some DSP.

[Xplorer]

For instance, in those days it was common practice to use a master oscillator, and then use flip-flops to get octave divisions, i.e. 16HZ, 8HZ, 4HZ, 2HZ and 1HZ. You just use a simple integrator to get sine waves, and it can all be done on a tiny circuit board.

it makes me think of the 1979 Crumar Performer analog synthesizer, it works this way i think, with an oscillator being divided and subdived again, to have every notes. this is how it made one of the first ( false ) polyphonic synth, so you could play more than a few notes at the same time.
and you could affect the whole pitch by affecting the original sine wave.
But ... i won't say much more because i may be a bit out of the subject. i have the synth and the schematics though, if you think it's helpful.

[Eb7+9]

cheers guys,

I think PIC's will be around for a real long time ...

I'm not hung up on the Arduino, any PIC arrangement will do
and, as long as we have 15 PWM signals to play with ... I can do the rest

indeed, the alternatives appear a little daunting at the moment ...

---

I'm checking out this fascinating paper:

http://www.europment.org/library/2014/i ... CSC-11.pdf

"Electronically Tunable Current-Mode Multiphase Oscillator Employing CDCTA-based Allpass Filters"
by Pitaksuttayaprot and Jaikla

relying on the idea of Current-Differentiating Transconductance Amplifiers ...
each amplifier block requiring 41 bipolar transistors to implement,
and we need three of these arranged in a loop to make a 3-phase oscillator (if I'm reading right)
that's a 123 devices ...
and if we do 5x 3-phase oscillators, 615 transistors in all ... (!!)
plus more devices for current sources and mirrors

yikes, ...

scratch that, then ...

and I'm also looking at

"A 7-decades Tunable Translinear SiGe BiCMOS 3-phase Sinusoidal Oscillator"
by Loizos, et al.

we don't need to make this work in BiCMOS technology here ...
so, I'm checking the references to see if I can fandangle something from older ideas
it's a hassle for me since I'm not an ieee member //

---

yeah, at this point were looking at simply switching caps ...

or maybe some day,
a PIC programmer will come along and want to do a 3-phase sinewave oscillator using PWM's ... x5

thanks for the comments ...!

[daveweyer]

JC just thinking out loud here, but if you did use a master oscillator and subdivided it, you could use a modification of Wayne's circuit to get three 120 degree lagging sine waves from each subdivided frequency. Then you could just feed those to your OTAs to gate the input signals.
There is no reason why the three phase principle wouldn't work on the LF system too, especially now that we have huge capacitors in tiny packages.
Then if you wanted to change the frequencies, you would just change the values of one oscillator and it would change all the other frequencies together. Also, the master oscillator would not have to be a three stage affair, just one gain stage with a gain of at least 29 and 3 RC networks (that is if you wanted to use a phase shift oscillator).
But your mind is probably already racing with the possibilities and the problems they present.

[Eb7+9]

Dave,

a good idea, but let me play devil's advocate for a second here ...

---

what you're saying (I think) involves using square pulses and doing this using divider chips
in other words, using logic gates and related circuitry

but we need sinusoidal-like output, and that is a problem here
you can't go from square to sine, not directly anyway

(unless you use a high order tracking filter to cut out the square wave's higher harmonics = quite complicated)

the other problem is all the sub-phases would be phase-locked to the master clock
and (again, I think) part of the magic comes from 5 oscillators NOT being in that (synchro) state

ever shifting by un-periodically

so, really ... (I think) we want 5 independent 3-phase oscillators
NOT in any state of synchronicity

and, providing sinusoidal outputs ...

---

I thought of using sine-based analog PLL's (phase-locked loop) and make it so that they track to a master oscillator with a constant (120/240 deg) phase offset ... but that gets complicated fast also ...

[daveweyer]

Yeah, that would lock them all together, but you could use two oscillators which would solve that, and still save an awful lot of circuitry. Turning square to sine is no problem, and it really doesn't have to be perfect sine as it only affects the LF amplitude modulation. The constant phase networks don't need much bandwidth for the LF section, you'd only need about a 3 to 1 ratio, for instance on the 1HZ section you would go from 1/2 HZ to 2 HZ or thereabouts.
You could use a resistor and capacitor for the integrator to get the sine function back, and you could probably design those networks so you got your perfect 120 degree phase shift in the same block.
It just sounds like a lot less circuitry to me, but please advise me if I'm wrong.

[Tek465b]

Arduino is just an AVR microcontroller like the others..
Its just that it has a bootloader, so we can program it over usb instead of using one more piece of hardware(programmer). but we can still program it using a propgrammer over the icsp headers without a bootloader and without usb(we use the icsp header, because it allow programming of surface mount component but you can also use a DIP socket programmer)... they are just microcontroller chip. and arduino(the compiler) has a good community with lots of library wich make it alot easier to code something. but you can program the arduino using avrdude or using other program too..
Its just that PIC chip are made by Microchip and AVR by atmel
Its just a matter of requirement, AVR vs PIC.
http://www.electricstuff.co.uk/picvsavr.html
we can have 15 pwm signal.

[daveweyer]

It looks like Arduino is going to be a possible solution.

Not to sound like broken record, but the reason Wayne used the three stage RC oscillators is because they would do the gating for him; this was the economy of the circuit. If you really like that idea, it seems to me that just recreating his idea with FETs or Mosfets would offer the same advantages of economy, plus you could just resistively sum the outputs without worrying about dual rails, negative signal swings, and all the rest of that.
It really is a simple solution to all the circuit complexity of separate gating.

[Eb7+9]

gentlemen,

my only peeve with the PIC/AVR approach is that "serious" amounts of programming
would be required every time some change is needed or wanted

which means asking someone else (not me) to do it ...
plus the output is perfectly timed, and almost (perhaps) sterile compared to an imperfect/quirky analogue circuit

I know, because I built a stereo-vibe unit with aTap-Tempo PIC in it
and, well ... it's perfect, as digital circuits always do ...

but, a little boring (unlike say, a Univibe ... with its quirky LFO)

just saying, it's not my first choice here ...
workable no doubt

thx Tek ...!!

---

the approach you're alluding too Dave seems cumbersome to me
would require much hardware

let's put that idea on the back burner for now if you don't mind

---

well,, ... as usual I was up late and couldn't sleep

so I sat down and carefully read over the ieee paper by Loizos et al.
(of course, the aim of their paper is WAY different than ours)

the authors describe in passing a simple circuit block that wouldn't work in a BiCMOS environment
well, that's not a problem for us here ...
whereas modern academics are looking for circuits operating in the hundreds of Mhz
and at 3.3v supply voltages ...

WE don't have those limitations imposed on us here in audio-gizmo land

so, I checked out this circuit block presented in Fig.3 where the caption reads "Possible implementation of the Gm - R module. The architecture is fully translinear and avoids PNP devices, however requires arge voltage headroom"

well, PNP devices are not an issue for us as we're looking to have a circuit operate between, say 1/3 Hz and maybe 16 Hz tops ... as in, 1/3, 2/3, 4/3, 8/3. 16/3 Hz ... up to 1, 2, 4, 8, 16 Hz ...
(PNP devices are slower than NPN in IC environments, preventing ultra high speed operation)

the circuit is basically a differential pair with a string of three diodes per collector load ... a very simple circuit

the author does not mention how that circuit block would be used in a ring arrangement ... I tried in vain a couple of times, simply by stringing them in a loop ... but I didn't get an action (on the simulator) as much as I tried

then decided to hit the sack ...

but my mind kept at it, and then realized there was an issue with DC conditions ...
and then the idea popped in my head of using complimentary (PNP) diff pair to produce some level shifting

jumped out of the sack, hit the simulator a couple of times, and ...

EUREKA baby ...!!!

I won't spill the details until I have something that looks cooked ...
but for now, I'll say that I have something that seems to work and won't require huge amounts of circuitry to govern
each 3-phase LFO would require only 12 transistors, 15 if we include the current sources

the idea here is that the bias current on the diff pairs can be used to increase and decrease the oscillator frequency
and it does so in a linear/proportional manner

in other words, we can use DC currents to set the speeds ...

and, if we use scaled current mirrors of a master current we can make this master current vary and have all other currents scale in proportion ... what this means is that a master current reference can be used to scale all groups of 3 so that the whole works wobbles up and down at proportional rates

almost hard to believe, considering the complexity ... but that's current-mode thinking for ya

from this circuitry I can arrange things so that I get current outputs to run my OTA's directly ... the only downside to all this is that it would make it cumbersome to introduce some kind of "DEPTH" control ... my initial thoughts about this are that we would have to convert to voltage first, go thru a triple-ganged pot as before, and then go from there ... but this requires an op-amp circuit per output phase, and n extra mirror ... if I go directly with currents, much simpler (only an extra diff pair and current source per phase)

upon thinking about it, the effect does not seem very strong to begin with (unless my sims are off in that sense) ... and so I wonder if an INTENSITY control is really required here ...

so, I've decided to go down this route and have a master RATE control and forgo the INTENSITY side of things ... I got some more sim'ing to do before I understand how this circuit swings, so-to-speak ...

I'll chime back once I've got a working example on the breadboard ...
~jcm

[daveweyer]

JC it looks like we're starting to get in the way of your thinking process. I think you have a fairly solid concept of how you would like this thing to look, so we'll just support you in that.
I sent Xplorer a file of a simple guitar played alone and through the modulator, that as you requested a few days ago. You may be able to use it to compare to your simulations.
Does anybody get any sleep?

[Tek465b]

JC it looks like we're starting to get in the way of your thinking process. I think you have a fairly solid concept of how you would like this thing to look, so we'll just support you in that.
I sent Xplorer a file of a simple guitar played alone and through the modulator, that as you requested a few days ago. You may be able to use it to compare to your simulations.
Does anybody get any sleep?

I get good sleep, i just don't focus to much on the technical side .

[Xplorer]

No .... slept just two hours "to be up all night", as usual. Big fail.

congrats on your EUREKA time JC, here is a clip from JC, two speeds apparently. it really has a univibe aspect in this chorus effect, preserving the signal, i understand that it could have been a smooth evolution for Jimi, very nice :

https://soundcloud.com/xplorer80/deniz- ... -tone-demo

from Dave :

Here is a short demo of the modulator done like JC wanted it. Just Deniz noodling on guitar.

The first cut is straight guitar through a '48 Fender Deluxe, nothing in the signal path.
The second cut is the straight Choratone signal, driven from the Deluxe speaker output.
The third cut is guitar on left, modulator on right, all levels identical.

[Eb7+9]

Wicked ...!!



thanks for preparing and posting the clips Dave & Adrien,
I can say I feel like I'm in the same ballpark at this point ...


... and, the addition of a Master SPEED control might just be the cherry on the cake
... plus, the stereo option ... firecracker

ok, ... back to the Current-Controlled 3-phase Sine-wave LFO design
the end is finally in sight ...

~jcm