Low-distortion Audio-range Oscillator

[dirkwright]

Originally Posted by richiem

IG-18 #4, the IG-339B

Thanks for all your hard work, but I don't want to buy an old Heath oscillator.

[KSTR]

Samuel,

thanks for the detailed comments.

We already studied (and admirered) the various compensation tricks around the 5534's and found it unlikely to make any improvement there. But will try the class-A mod on U1401 and see if it does something.

We swapped the 3 OpAmps in the SV section to 797's in a quick try, did not work out (got much worse). Of course having the comp pins floating...

Also we tried fine-balancing the gate divider and found that an exact 50% remains to be the best operating point for the FET. Further, in a brute-force fashion, we tried to increase (by 5x) the control voltage smoothing caps by switching additional precharged ones after settling which also did not change anything except a tendency to destabilize the control loop.

We basically are after reducing the higher order components where our guess is that they come mostly from injected control signal spikes/ripple. Therefore, your hint about C1502 and C1603 (and nearby resistors) is very handy (those two in fact were a mystery to us), maybe we can make out something there.

Thanks again,
Klaus

[Samuel Groner]

We swapped the 3 OpAmps in the SV section to 797's in a quick try, did not work out (got much worse).

The AD797 won't be stable in this configuration. You'd need to add a 220r resistor right in series with its negative input. An OPA1611 would probably be the easier choice.

But first of all find out whether the dominant distortion is in the oscillator itself, or the output stage (i.e. measure at U1401 output through a small series resistor).

Samuel

[1audio]

Originally Posted by davada

Hi Demian,

I'm interested in looking at everything. If it's not too much trouble can you post it?

I went back and looked at the older solutions. They are complex with lots of parts but discretes are cheap. I am attaching the oscillator circuit from an Optimation AC125 ac calibrator. The lower right hand circuit is the AVC. It can meet a .002% AC voltage accuracy which is not necessary for this application (dispense with the ovenized zener).

The Fluke 510A is a really interesting solution as well. it captures the peak voltage without a quad signal for triggering. The top half of the schematic is the agc circuit.

Oops, the PDF's are too large to post. PM me for them. The 510a is available on the web but the schematic in it has been sliced into pieces and really hard to follow.

[EssB]

Originally Posted by dirkwright

Thanks for all your hard work, but I don't want to buy an old Heath oscillator.

But you get a lot of hardware for relatively little money and save building time and effort for the new circuit.

[dirkwright]

Originally Posted by EssB

But you get a lot of hardware for relatively little money and save building time and effort for the new circuit.

Why are you trying to talk me into doing something I don't want to do? All of that old hardware is not very good, and I would not do it that way anyway. If you want to do that, fine, go for it. I've done similar things in the past with hifi and it's far more work and then you have to deal with a lot of limitations. No thanks. I'd rather start from scratch.

[myhrrhleine]

Originally Posted by Samuel Groner

Of course--I'd not comment at this detail if I didn't had the background.



Definitely. Building low distortion oscillators is a steep learning curve!

The schematic is only half the rent--bad layout can completely ruin performance.



The former pre-trims the Q of the oscillator such that J1 is working with just as much voltage across it as necessary. The latter adjusts the symmetry of the rectifier to minimize ripple.



An ES9018 will not have -130 dB dynamic range at full scale. This figure is measured at -60 dBFS and degrades above ~35 dBFS. Also it would surely not give 1 ppm distortion at 100 kHz!



A reasonable and realistic design goal for a very good DIY oscillator is perhaps:

* Frequency range 10 Hz to 100 kHz, in 4 decades
* Amplitude range -60 dBu to +20 dBu
* Harmonics below -125 dB for 20 Hz-20 kHz
* Harmonics below -110 dB for 10 Hz-40 kHz
* Harmonics below -100 dB for 10 Hz-100 kHz
* THD+N (20 kHz BW) below 110 dB for 20 Hz-20 kHz
* THD+N (80 kHz BW) below 100 dB for 10 Hz-40 kHz
* THD+N (500 kHz BW) below 90 dB for 10 Hz-100 kHz
* Settling below 1 s for 100 Hz-100 kHz
* Settling below 5 s for 10 Hz-100 kHz
* Amplitude flatness below 0.1 dB for 10 Hz-20 kHz (ref. 1 kHz)
* Amplitude flatness below 0.5 dB for 10 Hz-100 kHz (ref. 1 kHz)

Samuel

I believe you can get those specs from the Cordell oscillator.
The article comes with a nice PC layout too.
Building that oscillator and tweaking it is a good starting point.

[1audio]

Originally Posted by dirkwright

Why are you trying to talk me into doing something I don't want to do? All of that old hardware is not very good, and I would not do it that way anyway. If you want to do that, fine, go for it. I've done similar things in the past with hifi and it's far more work and then you have to deal with a lot of limitations. No thanks. I'd rather start from scratch.

If you are making a continuous tuned oscillator the frequency determining components are a real PITA. There are a lot and they need to be of a certain grade and close tolerance (don't forget tempcos and matching the caps to the resistors to improve the system tempco. . .). Also an attenuator that is accurate and frequency constant is not simple.

Replacing the oscillator circuit lends it self to upgrades easily. Separating the AGC from the oscillator will help keep any junk in the AGC from getting into the oscillator (with sample and hold circuits this is an issue).

For a few fixed frequencies building from scratch makes some sense, but Victors oscillator's are so good and cheap just putting them in a box with switching makes a lot of sense.

[richiem]

@dirkwright -- please keep us in the loop and provide construction details when you achieve a low-cost, 10Hz to 100kHz tunable oscillator that has 20V p-p output with 0.00006% THD at 1kHz.

However, a really great alternative is the easily built Cordell SV oscillator which offers even lower distortion and higher output. I also buit that -- with a few mods -- into an old Heath IG-18.

I'm not being snarky, here; just pointing out that you have been offered many suggestions of things that work very well indeed....

[dirkwright]

Originally Posted by myhrrhleine

I believe you can get those specs from the Cordell oscillator.
The article comes with a nice PC layout too.
Building that oscillator and tweaking it is a good starting point.

Getting the rotary switches is a pain in the @ss for that design. Modifying it for PC mount relays would not be fun either. It's a shame that designs like this are not available for purchase as PCBs. It's also a shame that the highly skilled designers here refuse to take a leadership role and make their designs available for purchase as a PCB kit. Oscillators are for the most part really old technology. These kinds of things should be readily available, but they are not, as bare PCBs that anyone can purchase and make into something useful. I don't think it's fun to have to learn all of these stupid details about oscillators just to have a good one of my own.