Problem with SVF oscillator at high frequencies

I built this Rod Elliott SVF Oscillator:-
ESP - Sinewaves

An externally hosted image should be here but it was not working when we last tested it.


The op amps are NE5532.
Supply rails are +/- 12V.
I used a J111 FET (Vgsoff = -7v) instead of 2N5484.
It works fine at lower frequencies, but I want it to operate up to 150kHz.

I made the following changes:-
VR1 = 10K (I happen to have a quality wirewound Colvern 10K dual gang pot).
R4 = R5 = 1K
C1 = C2 = 1000pF

What happens is that as the frequency approaches around 100kHz, it suddenly jumps into a state in which it oscillates between the rails at approximately 80kHz. The output is clipped. This is despite the AGC fully pinching off the FET.

It's as if there is excess positive feedback and/or unwanted phase shift at high frequencies, presumably peaking at 80kHz.

The gain-bandwidth product for an NE5532 is 10 MHz, so I did not expect problems at 150kHz.

There is plenty of supply decoupling around the op amps.
I don't think there is excessive stray capacitance or inductance with the layout.

I tried carefully matching the values for C1 and C2 using a bridge.
I tried replacing the pot with matched fixed resistors in case stray capacitance or inductance was affecting things.

I tried a different FET (an old 2N3819, Vgsoff = -5v), thinking it might be capacitance between source and drain. Initially, this seemed to help: the maximum frequency almost reaches 150kHz as the control pot is advanced. However, if you switch off and on at that setting, the oscillator still goes back to hitting the rails at around 80kHz.

I'm running out of ideas. Any suggestions?
 
There may be another problem that's worth eliminating.

Notice that D1 connects C4 pretty directly to the output of U2A. Now, I'll assume your output level is 2 volts peak. That would require a current peak from U2A of about 2*2*pi*150000*1e-6=1.88A until the cap charges. Of course, the diode probably limits it some, but its dynamic resistance isn't much.

That capacitive load probably makes U2A unstable, as well as driving it into current limiting at start-up. So, the test would be to put about 100 Ohms in series with each diode to limit the loading on the opamps, and the charging current taken at start-up. (note that U2B faces a similarly thankless task).

Even if it doesn't fix your 150 kHz issue, I think it is one of a couple of fixes that you may need.
 
I would use the original RC values, short out the pot or set to min resistance and look @ the high frequency waveform. If stability is good, then get a new 100k dual pot.

I think that by scaling the RC (10x lower to fit the dual pot) the amount of current drive is a problem.

The control voltage for the filter & FET should have a buffer amplifier. The peak pulse current into the rectifier will distort the output waveform.
Duke
 
A couple of good suggestions from djoffe and Audio1Man.
I will give them a try while waiting for the LM4562s to arrive.

I tried adding a small cap in parallel with R1 as suggested on page 446 in the diyaudio thread that discusses Q enhancement. In my case it didn't seem to make much difference. However, I'm still reasonably convinced that this is the problem.
BTW: That thread seems to have got way out of hand - 697 pages!!!
 
I built this Rod Elliott SVF Oscillator:-
ESP - Sinewaves

An externally hosted image should be here but it was not working when we last tested it.


The op amps are NE5532.
Supply rails are +/- 12V.
I used a J111 FET (Vgsoff = -7v) instead of 2N5484.
It works fine at lower frequencies, but I want it to operate up to 150kHz.

I made the following changes:-
VR1 = 10K (I happen to have a quality wirewound Colvern 10K dual gang pot).
R4 = R5 = 1K
C1 = C2 = 1000pF

What happens is that as the frequency approaches around 100kHz, it suddenly jumps into a state in which it oscillates between the rails at approximately 80kHz. The output is clipped. This is despite the AGC fully pinching off the FET.

It's as if there is excess positive feedback and/or unwanted phase shift at high frequencies, presumably peaking at 80kHz.

The gain-bandwidth product for an NE5532 is 10 MHz, so I did not expect problems at 150kHz.

There is plenty of supply decoupling around the op amps.
I don't think there is excessive stray capacitance or inductance with the layout.

I tried carefully matching the values for C1 and C2 using a bridge.
I tried replacing the pot with matched fixed resistors in case stray capacitance or inductance was affecting things.

I tried a different FET (an old 2N3819, Vgsoff = -5v), thinking it might be capacitance between source and drain. Initially, this seemed to help: the maximum frequency almost reaches 150kHz as the control pot is advanced. However, if you switch off and on at that setting, the oscillator still goes back to hitting the rails at around 80kHz.

I'm running out of ideas. Any suggestions?
URL is dead. go to
ESP - Sinewaves (overview resp. TOC)
ESP - Sinewaves (SVF oscillator)
or all at one page
ESP - Sinewaves
 
Last edited: