Help: Need low noise stable amplitude audio signal generator or design

[enut11]

I am working on an AC Voltage Calibrator project. See reference below from reply #82 for the latest version.
https://www.eevblog.com/forum/metrology/an-experimental-ac-voltage-calibrator/75/
Am now in need of a very stable amplitude low noise signal source, ~1vRMS 20-20KHz, to drive the calibrator.

I have tested several signal sources, commercial and home-made, and most have fallen short wrt either amplitude stability or noise.
Wavetek Model 80 function generator
Leader LAG126S audio generator
Chinese low distortion audio range generator kit
ASUS Xonar U7 PC USB sound card

Obviously there are dozens of good designs out there that can generate stable audio frequencies. What I am looking for is an affordable one to work with.
Ideally, for this project, a generator that stabilises in amplitude (<60uV/min) within ~30 min and has less than 30uVpp noise on a 1v/1KHz signal would be OK.

Looking for suggestions on equipment or designs. I will post more info and the results of my tests soon.

How good is your audio frequency generator for amplitude and noise in a metrology role?

[enut11]

@evb149
I have only tested one PC sound card - an ASUS Xonar U7 and it had a very drifty amplitude. See Reply #86 at https://www.eevblog.com/forum/metrology/an-experimental-ac-voltage-calibrator/msg3911450/#new.

Are you saying this can be compensated for in software?

[enut11]

I may look into that later on, ie letting the sound card measure its own output to determine amplitude drift.
Meanwhile, I am looking for an AF signal source that stabilises in a reasonable time with minimum interference from me - ie 30- 60min.
One thing I did like about the sound card, there was no measurable noise at the output.

[Messtechniker]

Soundcard output is normally pretty stable.
For example, for my old and trusted M-Audio Delta 192
1+ hour stability in loopback is below 0.01 dB. 

Long term drift after a month or so when switching on is around -0.04 dB staying there being stable again at this level below 0.01 dB.

Maybe a long term stability of -0.05 dB
is not good enough for your purpose. You probably
cannot expect more from even the best sound cards.

[enut11]

@Messtechniker
Thanks for your input. This is useful information and puts a number on what is possible from a sound card. No doubt my inexpereince with these devices does not help.
On the negative side, high performance sound cards are very expensive and could not be justified simply to run this AC Cal project.

[Kleinstein]

Sondcards have additional adjustements for the ouput amplitude. The stability may depend on the amplitude setting and maybe the output (line out, headphone) used. Another point can be the supply voltage and thus the PC it is used in. Amlitude stability is not one of the main targets, the qualtity in this respect may not correlate with price and there are rarely good data in the describtions. Trying another type may be worth it.

[Messtechniker]

You might even try on-board audio.
Line inputs and outputs can be quite
good on some more modern motherboards.
Or absolutely horrible especially on older boards.
Note: on-board MIC inputs are always basically
bad to horrible when it comes to moderately
demanding specs.
Lots of penny pinching going on here.

[moffy]

If you are able to measure the output of the sine wave precisely would that solve your issue? A simple synchronous demodulator(switch based) filtered and fed into a semi decent volt meter will give you a good measure of the average value of the sine wave, from that you can compute the RMS. I don't know if that suits your situation though?

[2N3055]

This has been discussed before. Best way is to use DDS chip with a reference. No audio DAC will be PPM stable in amplitude.

[David Hess]

The old solution is to make a precision measurements of the AC output, like with a precision rectifier, and use that to control a leveling loop, which is just part of an analog oscillator anyway so that is easy enough.  Offset and drift can be removed from the rectifier by AC coupling the gain stage, but that would not be necessary today.

[djsb]

Have you looked at the QuantAsylum equipment

https://quantasylum.com/

[moffy]

The old solution is to make a precision measurements of the AC output, like with a precision rectifier, and use that to control a leveling loop, which is just part of an analog oscillator anyway so that is easy enough.  Offset and drift can be removed from the rectifier by AC coupling the gain stage, but that would not be necessary today.

That would work well, as you only need to adjust the gain to compensate for the instability of the sine wave variability. So if you have +/-1% amplitude variation then you only need say +/- 2% gain variation. A small value LDR optocoupler as the controlling element(ensure the gain variation is small) works very well.

[David Hess]

The old solution is to make a precision measurements of the AC output, like with a precision rectifier, and use that to control a leveling loop, which is just part of an analog oscillator anyway so that is easy enough.  Offset and drift can be removed from the rectifier by AC coupling the gain stage, but that would not be necessary today.

That would work well, as you only need to adjust the gain to compensate for the instability of the sine wave variability. So if you have +/-1% amplitude variation then you only need say +/- 2% gain variation. A small value LDR optocoupler as the controlling element(ensure the gain variation is small) works very well.

An operational transconductance amplifier can also be used as the variable gain element, even in a low distortion design.

[SiliconWizard]

This has been discussed before. Best way is to use DDS chip with a reference. No audio DAC will be PPM stable in amplitude.

Yeah, I agree. But, depending on the load, they are likely to need an output amplifier after the DDS chip, and the challenge will be to make this amplifier stable at this level.

We need to read the requirements: "a generator that stabilises in amplitude (<60uV/min) within ~30 min and has less than 30uVpp noise on a 1v/1KHz signal".

[2N3055]

This has been discussed before. Best way is to use DDS chip with a reference. No audio DAC will be PPM stable in amplitude.

Yeah, I agree. But, depending on the load, they are likely to need an output amplifier after the DDS chip, and the challenge will be to make this amplifier stable at this level.

We need to read the requirements: "a generator that stabilises in amplitude (<60uV/min) within ~30 min and has less than 30uVpp noise on a 1v/1KHz signal".

Yeah I read that, stability better than 60 ppm and noise less than 30 ppm...

I agree with others, a leveling loop from the output is probably the most realistic way to achieve something respectable on output connector.... And by that I don't necessarily mean analog negative feedback. It can be running digital calibration cycles in the background.

[enut11]

Shown below is typically what I find when I test a low distortion audio signal generator.
From a cold start, amplitude drops steadily with stability setting in after about 60min.
A close-up of the flat part of the curve shows short-term variations, the worst in this case being about 60uV/min droop at around data point #1230.

[Ground_Loop]

Audio distortion analyzers like a Sound Technology 1700 have very stable and low distortion 20-20K sources.

[ Specified attachment is not available ]

[enut11]

Thanks. I Like the old world analog audio test gear. HP also made some good ones. I downloaded the 1700 manual but there were no specs included.
What I need is better than 0.1dB amplitude stability around 1v, at least at mid audio frequencies.

[Conrad Hoffman]

AFAIK, amplitude stability isn't a requirement for THD measurement. That said, and I've never measured it for stability, the oscillator used in the Bob Cordell THD project has always done very well for me. It's put away right now, but maybe I can do a stability measurement in the near future. http://www.cordellaudio.com/papers/build_a_thd_analyzer.shtml

[Messtechniker]

50 minutes soundcard startup drift test:

Test equipment: Keysight 34465A and M-Audio Delta
calibrated to 0.000 dBu approx. 1 year ago. Is now off by approx. -0.036 dB
Measurements started approx. 2 minutes after switching the 34465A on and the soundcard output on. I.e no long warmup. Logging programme: Test controller.

[jonpaul]

Tektronix SG505, Audio Precision were Bruce Hofer and Rich Cabot classic designs.

N+ THD 0.0018% frew 10..100 kHz.

You cant do better.

Jon

[Messtechniker]

2 hour drift measurement with using the "High definition audio"
soundcard on motherboard  Asus M5A78L-M.
All equipment warmed up this time.

[Terry Bites]

Use a PCM1794 DAC with a passive attenuator on the output. https://www.ti.com/product/PCM1794A-Q1

[enut11]

2 hour drift measurement with using the "High definition audio"
soundcard on motherboard  Asus M5A78L-M.
All equipment warmed up this time.

(Attachment Link)

@Messtechniker. More info on how you did this test please.
enut11

[Messtechniker]

My measurement setup is as follows:

Computer with motherboard Asus M5A78L-M with its "High Definition" onboard sound chip. Signal generating software: "ASIO4All"
https://www.asio4all.org/
and "Audio Tester"
https://www.audiotester.de/
Set to produce 1 kHz 0.00 dB at the green output socket.

Connect green output socket to Keysight 34465A set to AC and use math to display dBs.

Connect its LAN to a second computer running "Test Controller"
https://lygte-info.dk/project/TestControllerInstalling%20UK.html
for logging.

Do logging as long as required. Screenshoot
https://app.prntscr.com/en/index.html
the results of "Test Controller" "Current values" tab as required.