Low-distortion Audio-range Oscillator

Yes I took it from pin 7. I see that the main output has 2 x 1.2k in parallel as output R. Could that also be reduced to a few 100 ohms, in the interest of not loosing too much signal in lower loads?
Can you reveal the type of resistors you used for the series Rs?
Jan
Yes, also the main output can run with the same lower series resistor at least 200 ohm. All the resistors in the sine signal chains are metal film type. The 0805 cases are mainly Susumu (25ppm/C). MELFs in older boards may be from any manufacturer, but also metal film.

Vic.
 
Disabled Account
Joined 2012
BTW, previous post: left is Vics osc. into the analyzer, right is AP gen into the analyzer....

This is Vics osc. measured with a dedicated passive notch. This osc. is clean down to -160dB. An heroic achievement!

Jan

I can measure directly down to -160 with my gear. So, I would like to back up this test by doing a measurment also. I have to order a 1KHz, now.

It wasnt too long ago, I was asking for lower thd generator and could we go to -160dBv. No Way! But now, it appears to have been done with encouragement and dedication.

:)


Thanks,
Richard Marsh
 
Last edited:
Dear Victor,

very interesting to get insight in your measurement setup, excellent craftsmanship BTW!

This gives me more confidence in my own measurements, since the basic setup is very comparable (except of the opamp-type and using 10n caps in the passive Twin-T Notch instead of 22n).

[see post #6870]: https://www.diyaudio.com/forums/equ...n-audio-range-oscillator-687.html#post5134739

Thanks!

Nice measurements.
Maybe the OPA627 is little more quiet.
Also the ADC linearity region is different, and may ask for deeper notch.
I am using two time higher FFT lengh: 524288, and sometimes small averaging.

Vic.
 
I am sorry for the confusion. Maybe I did a mistake? I'll try to explain:

Referring to the second plot:
Just like Victor, I used a 40dB gain-stage after the notch. The Attenuation of the passive notch at 1kHz is around -47.6dB and therefore the fundamental level is at -7dBV. I have also stated the Twin-T-loss at 2nd and 3rd harmonic in the referenced post. They are less than 10dB. Therefore I assume the harmonic levels of the Oscillator are "magnified" by at least 30dB. I must admit, the scaling was not adjusted....might be misleading. Anything wrong with that assumption?

The first plot was a reference measurement without notch and gain-stage in order to get an idea the distortion levels achievable with the QA401. Of course, the settings of the device were not changed.
 
I can measure directly down to -160 with my gear. So, I would like to back up this test by doing a measurment also. I have to order a 1KHz, now.
It wasnt too long ago, I was asking for lower thd generator and could we go to -160dBv. No Way! But now, it appears to have been done with encouragement and dedication.
:)

Thanks,
Richard Marsh
The modified Jan's board has the second harmonic around -158dB (as I mentioned, needs to add +3dB at 2kHz, and -1dB at 3kHz on this plot).
Richard, I don't forget about you. As I remember, you wish the oscillator finished in box with PSU, but this is not so easy for me. Also I don't wish to send you the similar board as you already have. Probably you have not so old 1kHz board. Can you measure this board with your equipment? Probably your oscillator board must have the harmonics lower than -150dB at max output. I can modify your board, too. But I can't give the warranty for -160dB. Some boards may can run so low, but I have no statistics. If I can get the stable performance under -160dB, then I immediately offer you the solution.
You can see how around 5-years old board looks before the modification (1) and after modification (2,3). Last plot (3) was done with three times averaging.

Vic.
 

Attachments

  • 1kHz1.jpg
    1kHz1.jpg
    124.2 KB · Views: 1,452
  • 1kHz2.jpg
    1kHz2.jpg
    123.5 KB · Views: 1,419
  • 1kHz2A.jpg
    1kHz2A.jpg
    121.9 KB · Views: 1,264
Thank you, Bob.
The real levels of the harmonics can be easy calculated. There is app. 4dB level difference between 2nd and 3rd when this type of notch is used. Every time we know the fundamental level at the notch input. Then we can adjust the proper scale (calibrated injection from other source) at 3rd or 2nd and then calculate the other harmonic level. In my picture, which Jan previously post, needs to add +3dB for the second and -1dB for the third. And it is not so important how deep is notch rejection, only needs to know that the signal after the notch is in the linear region of the ADC. The notch fine tuning practically not affects the second and the third harmonic levels. When the oscillator is adjusted with +/-0,5% precision via the manufacturing process, then in practice usage not needs to tune the notch, and I am using fixed ones.
I don't like active notches. The distortions will depend from the opamp. This is not so good when extremely low harmonics needs to measure.

Vic.

Hi Vic,

Thanks for your answers. I think what you are saying about taking account of the loss of the passive twin-T at 2H and 3H is essentially what I was referring to as manual correction, which is fine.

Thanks for clarifying that you are not depending on a known notch depth of, say, 40 dB.

I understand your reluctance to use an active notch for such extremely low distortion measurements. I was just pondering how good the active notch would be in these situations if you use the same very good op amps you use in your oscillator. It could be an interesting experiment, and in fact quite easy to switch a twin-T from being passive to active, assuming you have to buffer the output of it with subsequent gain anyway. Just a thought.

Cheers,
Bob
 
Hi Vic,
Thanks for your answers. I think what you are saying about taking account of the loss of the passive twin-T at 2H and 3H is essentially what I was referring to as manual correction, which is fine.
Thanks for clarifying that you are not depending on a known notch depth of, say, 40 dB.
I understand your reluctance to use an active notch for such extremely low distortion measurements. I was just pondering how good the active notch would be in these situations if you use the same very good op amps you use in y our oscillator. It could be an interesting experiment, and in fact quite easy to switch a twin-T from being passive to active, assuming you have to buffer the output of it with subsequent gain anyway. Just a thought.
Cheers,
Bob
Hi Bob,
Thank you. I understand your idea. Probably the active notch can be easy made, and can have acceptable performance, but now my interest is to go deep as possible for to minimize the distortions. One of the main problems is the noise. Now I have the new discrete buffer on my table. Needs to check it. Next needs to build the new oscillator.

Vic.
 
One thing about noise in oscillators that is a tradeoff in my experience is noise vs distortion from the AGC element, usually a JFET.

My experience here is mostly with low-distortion state variable oscillators and notch filters, but I think it applies pretty much in general. Assume that the main source of distortion in the oscillator is the JFET agc element (this will not always be the case, of course). We can reduce the voltage across the JFET to reduce the distortion that it introduces, but if we only do that we reduce the amount of agc range that we have. To get back the lost control range so that it is enough to control over worst-case tolerances, etc, we must increase the strength of its control injection. But the JFET and its associated circuitry already has noise, so doing so introduces more noise into the oscillator.

This noise generally shows up as a combination of amplitude and phase noise around the fundamental. We care about this a lot if we are making a traditional THD analyzer because this noise will come through in a THD+N measurement. It is a little less of a problem if we are using a spectrum analyzer to look at the distortion components coming out of the THD analyzer.

Cheers,
Bob
 
The FET distortion problem was discussed some time ago, and the radical method is the balanced FET regulator. As example in the attachment. Other simple method may be paralleling or cascoding of the FETs, if we can't find the FET with needed specification. As example, the MMBF4391 runs in my oscillator with around 40 ohms resistance. The fully open resistance of this FET is around 20 ohms. So, I have only 20 ohms for to change, but other 20 ohms as unwanted ballast. When I use the matched FET in parallel, I can reduce the resistance and also the voltage across the FETs without losing the same control.

Also other serious problems we can find under -150dB - capacitors, opamps and the PCB board quality. Some PCB boards don't allow to run so low. Maybe chemical residues, moisture, bad mask or PCB material. The oscillator under -160dB probably must be mounted on teflon board without the mask, or air mounted - point to point soldered.

Vic.
 

Attachments

  • FEtregBal2.PNG
    FEtregBal2.PNG
    19.9 KB · Views: 537
Member
Joined 2004
Paid Member
Tektronix identified the PCB issues in the early 1970's as an issue in high impedance circuits. Its like a poor cap in the circuit. its also not consistent, some boards happen to be worse from the same batch. Tek introduced a new PCB material at the time. it was clear with no fiberglass and pretty fragile. I used it in some Spectral projects until Tek stopped making PCB's for outside customers and discontinued the stuff.

Teflon should be good (or some of the other exotic microwave materials) but quite expensive. Radiometer did use glass epoxy and got to as low as -170 but the impedances are lower.

One question I have is whether adding the LME49600 buffer inside the loop on the output of the opamp would degrade anything. It would enable lower impedances with its extra current capability.
 
I can measure directly down to -160 with my gear. So, I would like to back up this test by doing a measurment also. I have to order a 1KHz, now.

It wasnt too long ago, I was asking for lower thd generator and could we go to -160dBv. No Way! But now, it appears to have been done with encouragement and dedication.

:)


Thanks,
Richard Marsh

Hi Richard,

Being able to measure distortion down to -160 dB is truly impressive. I'm not quite sure what you mean by "directly". Can you describe in detail the setup you are using to measure down to -160?

Cheers,
Bob