Super Low Distortion Ultra Pure Audio Oscillators Revisited

[terkio]

Thanks for your answers.
I forgot I do have a microchip PIC developement kit Unused for so many years.....
It might very well support what I need for audio projets.

[mediatechnology]

The low distortion oscillator Protoboard has been sitting at the end of my work bench for several years now.
I decided it was time to finish it.

Lately I've been testing gain helper stages to increase the 5532 open loop gain and revisited this circuit: http://www.proaudiodesignforum.com/foru ... ?f=6&t=926



I lowered R3 and R4 to 10K and can get 1 kHz open loop gains of about 130-135 dB. (C1 was removed.)
The 5532 open loop gain measured 90 dB at 1 kHz so I'm getting a little better than 40 dB from the discrete stage.

The inverting closed loop test circuit shown above requires a 33 or 47 pF in parallel with R2 to maintain stability.

While there are better op amps than the 5532 I want to see what performance I can get from it.
"Flip Flop World's" oscillator web page, which now 404s, claimed a DC open loop gain of aboout 170 dB with this circuit.
170 dB is about 30 dB better than an LME49720 at DC.
At 1 kHz this stage has more open loop gain than the LME49720.

My goal from the oscillator is to have second and third harmonics near -140 dBc.
The added gain may help.

[mediatechnology]

I moved the oscillator to a new Protoboard and used the "super" op amp topology provided by "Flip Flop World."

I haven't moved the notch filter to see the actual distortion but this is the lowest THD number from the Focusrite 2i2 A/D I've ever seen.


Super Low Distortion Oscillator Using Gain-Assisted "Super" Op Amps.

The 135 dB 1 kHz open loop gain of the "super" op amp appears to make a huge difference.

[JR.]

Cute... have you characterized the ein, and PSRR? I ASSume the THAT array is low noise.

JR

[mediatechnology]

The rbb of the THAT340 NPN is 32Ω typ. The PNP rbb is 25Ω.
I could use discrete transistors and match them but I have 340s and they make it simpler.
I played with the higher gain low rbb Zetex parts and they seemed to work well but have higher output capacitance.
The 340 NPN and PNP don't have a very high hFE so the bias current is high - the Zetex provide an advantage.
I didn't measure the ein or PSRR.

I need to move the notch filter to the new board, run some tests and then post a schematic.

[JR.]

It may be useful to add simple RC filters in series with the voltage feeding R3 and R4 if they appear to be a vector for PS corruption.

JR

[mediatechnology]

JR: I agree that some filtering would be a good idea.

The following are the notch filter output.
There is +40 dB gain post-filter so -100 dBu displayed on the graphs is actually -140 dBu.

With a +4 dBu single-ended output into a 1KΩ load the distortion is in the noise floor.


Low Distortion Oscillator With Composite Op Amps Notch Out +4 dBu 1K Load

At +10 dBu output some second and harmonic becomes visible.
The filter insertion loss is about -9.5 dB at 2 kHz and about 5.5 dB at 3 kHz.
Second harmonic is about -131 dBc or 0.3 ppm.
Third is about is about -135 dBu.


Low Distortion Oscillator With Composite Op Amps Notch Out +10 dBu 1K Load

I think the second harmonic distortion is from the highly-degenerated VCA running about -50 dB attenuation.
Third harmonic is load current.

Second harmonic rises out of the noise floor around +7 dBu single-ended.
With the anti-polarity output also available the actual differential voltage is about +20 dBu maximum into 2KΩ.

[mediatechnology]

I've been able to lower second harmonic distortion by identifying three sources.

1) The VCA. I lowered the input current by 14 dB. It no longer contributes.

2) The half-wave detector. With the VCA now being distortionless I could see detector residual ripple contributing. I added a passive filter pole at the VCA Ec- pin. The detector no longer contributes.

3) The third source is the output stage when heavily loaded at high levels. Working on it. Unloaded the second order is in the noise floor.

[terkio]

I've been able to lower second harmonic distortion by identifying three sources.

1) The VCA. I lowered the input current by 14 dB. It no longer contributes.

Is this the VCA from THAT Corp ?

Have you figures about the VCA distortions at these low levels. And comments about noise when going to low levels for lower distortion ?

I was going to suggest using a 16bit MDAC as a VCA for excellent gain resolution and stability, but this pushes to digital at the gain control.

[mediatechnology]

I am using a THAT2180 pre-trimmed VCA.
I prefer them over the 2181 because they eliminate a trim and the on-chip Zener-zap trim of the 2180 has better THD tempco.

The VCA is in parallel with a non-inverting path and subtracts a tiny fraction of the signal with the VCA typically operating with about 50-60 dB attenuation.
The reduction in VCA input current, by making Rin 100KΩ instead of the typical 20K, doesn't significantly affect output noise which is low.
By utilizing the current output of the VCA to subtract the VCA doesn't add an I-V op amp providing additional noise savings.

The VCA's even-order distortion becomes so low it's beneath the noise floor of the oscillator's op amps.

When the output is unloaded I'm not seeing any THD popping out of the noise floor line at -140 to -145 dBu.

This is a 132 kpt 20 sample avg FFT of the notch filter output.
There is +40 dB post-filter gain; -100 dBu is -140 dBu.
The output level is +12 dBu (single-ended) and there is no load other than the notch filter.
With a +12 dBu carrier the noise floor line is actually -152 dBc at 2 kHz. (Includes post-filter amp noise.)

There is no THD-2 or THD-3 visible.
The notch filter insertion loss is about -9 dB at 2 kHz and -5.5 dB at 3 kHz.
The 1 kHz is notched by about -92 dB.

It's reasonably safe to say that THD-2 is no more than -143 dBc and that THD-3 is no more than -147 dBc.


Low Distortion Oscillator with Composite Op Amps Notch Out +12dBu Out No Load

Loaded outputs are a different story...