Tri-Vibe LFO for higher frequencies

[knutolai]

So I'm snooping around for a sine wave oscillator to use as the modulator in a ring modulator (too many nouns in one sentence.. sorry   ).
I haven't really found any circuit that easily lets me modify the frequency of the oscillator (over a broad spectrum, like 0,5Hz to 2kHz) without also changing the waveform. Is the Tri-Vibe LFO the solution to this issue? (Read: would this oscillator retain the same waveform at audio frequencies?)

here is the full circuit (from runoffgroove):


U3a and U3b makes a pretty standard square/triangle wave generator, but then there is the more magical stuff happening around U2b... I haven't really seen anything like this before. Could maybe someone point me in the right direction regarding documentation describing this circuit component? I'm eager to understand how it works 

[PRR]

> lets me modify the frequency of the oscillator (over a broad spectrum, like 0,5Hz to 2kHz)

A 4,000:1 range? On one knob? That is tough.

A standard pot covers about 10:1 (linear pot 100%-10%). You can do 50:1 ("audio" pot 100%-10%).

You can put in a range switch. Use one set of caps for 0.5Hz-250Hz, a smaller set of caps for 50Hz-2KHz.

And then there are LOG tricks. Voltage-controlled oscillator, VERY linear to low frequency, and put a LOG converter in front. This is how the classic Moog/ARP synths worked. One knob 20Hz to 20KHz with pitch accuracy. With the relaxed accuracy needed for vibe/trem LFO, 4,000:1 is not impossible; but will be very tricky to null all the offsets.

The OTA approach can-be naturally LOG. I don't think it is good-LOG over 4,000:1 range without lots of complications.

And there's the BFO tricks: beat 100KHz against 100.1KHz to get 100Hz. Easy to swing a "100KHz" oscillator 100KHz to 102KHz. Very difficult to set very small differences; aside from accuracy, the two oscillators tend to "lock" and give zero Hz.

National had an old plan for a somewhat simple function generator with extra wide range. AN-115. If I can find that I will post it. (It is no longer on NatSemi/TI's site.)

[knutolai]

> lets me modify the frequency of the oscillator (over a broad spectrum, like 0,5Hz to 2kHz)

A 4,000:1 range? On one knob? That is tough.

Heh, yeah I guess that was quite unrealistic..

You can put in a range switch. Use one set of caps for 0.5Hz-250Hz, a smaller set of caps for 50Hz-2KHz

Good idea, thx!

And there's the BFO tricks: beat 100KHz against 100.1KHz to get 100Hz. Easy to swing a "100KHz" oscillator 100KHz to 102KHz. Very difficult to set very small differences; aside from accuracy, the two oscillators tend to "lock" and give zero Hz.

This is a interesting approach, however doesn't the amplitude of the beat frequency drop as the frequency-difference between the supersonic oscillators increase? I did some experiments with this a few years back and can't really recall the results.. I'll have to dive into it again I guess.

How about a simple standard triangle wave VCO followed by a VC-LPF where the frequency of the VCO and the cut-off freq of the VCF is controlled by the same voltage (like frequency equal to cut-off frequency)? I guess it would be very hard to make the two parameters follow each other..
EDIT: Doing some research VCF seems to be very component heavy. Ill drop this option completely. Who needs complicated and hifi when you can have simple and dirty 

[PRR]

> modify the frequency of the oscillator ...without also changing the waveform.

The waveform "should not" change. The oscillations follow the same laws at any frequency. We do have practical limits: leakage at the low end and falling gain at the high end. But practical parts and chips should work the same from sub-Hz to the top of the audio band (limited by the ~~1MHz gain-bandwidth of cheap chips).

The tri-wave you show is the basic Schmitt and integrator. The integrator is quite easy (until you get sub-sub-Hz and leakage effects). The Schmitt has to swing FAST-- otherwise the freq runs low and the tri-wave tips go soft. That particular plan uses TL062, not a real fast chip. It also has 100K:10nFd around it which probably hurts speed more. This may be "nice" for music, softening the tri-tips when the freq gets high enough to hear tip-glitch. But that's a design decision for the specific application.

There's also amplitude/shape effects due to unfortunate choice of circuit. In the classic Wein Bridge, if you change just one of two parts in the R-C-R-C network, the gain changes which will ultimately stall or overdrive. However Merlin posted a variant where the change in the one R also changes gain, so the effect is "flat" up until it runs out of gain. Using cheap chips he showed good flatness up to the top octave of audio.

> magical stuff happening around U2b...

The diodes before U2b do a semi-Sine round-off of the harsh triangle wave.

The 5 resistors and two diodes do something else. This plan drives the OTA gain-pin with a *voltage*. But the OTA's gain is exponential to voltage (linear to current). With voltage-drive we'd either get a very lopsided trem or we have to bend the raw trem-wave into its log form. Diode-bridges can do this. The design is a pain. (Personally I wonder if the designer got off the tracks here... on the face of it a current-drive plan seems more straightforward.)