An Experimental AC Voltage Calibrator

[trobbins]

Perhaps push the zobel corner out one decade to 480kHz, given you have a somewhat controlled 'speaker' loading on the amp.

As an aside, many commercial ss 'brute' PA amps sell for peanuts and have flat frequency/phase response out past 50kHz (I use my American DJ Proformer V1000 for output transformer testing as it gets to a moderate Vrms per channel and can be configured as a balanced amp for PP output transformer testing, and is effectively flat response to 90kHz).

[Kleinstein]

The 15 Ohms and 220 nF are part of the compensation to make the amplifier stable, so it would not oscillate, especially not with a capacitove load. The LR combination at the ouput serves a similar purpose. One may be able to get away with less RC load and larger RL seires impedance. How much is actually needed may needs tests with the actual harware. Quite often the RC element was more like a thing added because it helped in the past with no dailed calculation. The transformer usually has some series inductance anyway. So the transformer is often not such a critical load and may get away with less than a normal cross over circuit.

It could also help to look at the amplifier itself to make it better behaved even with capacitive load. A first step would be to get a model for the simulation to see the more critical points. There may be options with different transistors or maybe more current to the VAS stage to speed up the amplifier.
Besides the load, there could also be a slew rate limit that could be a problem.

Usually audio amplifiers don't use extra voltage regulators for there main supply.

[enut11]

Perhaps push the zobel corner out one decade to 480kHz, given you have a somewhat controlled 'speaker' loading on the amp.

As an aside, many commercial ss 'brute' PA amps sell for peanuts and have flat frequency/phase response out past 50kHz (I use my American DJ Proformer V1000 for output transformer testing as it gets to a moderate Vrms per channel and can be configured as a balanced amp for PP output transformer testing, and is effectively flat response to 90kHz).

@trobbins
I interpret this as substituting a 22nF for the 220nF. This would certainly be a lower load, approx 160 ohm total, at 50KHz. Is there a downside to this approach?

@Kleinstein
Agreed, regulating the rails of an audio power is even frowned upon but in this case I am hoping for better amplitude stability. In all probability,  up to 30KHz would be more than enough to test most meters for AC response.

[Kleinstein]

reducing the capacitor in the zobel element from 220 nF to 22 nF could effect the loop stability. It is not sure how much. For a relatively fast amplifier the 22 nF may still be enough. It really depends on the details (e.g. transistor type, layout,...) of the amplifier and the load how much is needed. Some amplifiers get away without that RC element. The suitable size is more a thing to determine from experiment or maybe a careful simulation.

A flat frequency response to 90 kHz does not guarantee the amplifier could do that at full amplitude. There could still be a slew rate limit and the RC element at the output may overheat. It not so rare to find a burned resistor in the zobel element due to some oscillation of either the amplifier itself or some other stage before.

[trobbins]

You may be able to gain some confidence on HF stability performance of just the ss amp by applying a square wave test with the nominal zobel values in to your known load, and then modifying the zobel cap value - if there is no substantial ringing (and any ringing is say a single much higher frequency) then you should be fine given you have a known load scenario (your transformer likely presents a quite low capacitive impedance above its main resonance around 1kHz, up to when the first HF resonance kicks in).

[enut11]

reducing the capacitor in the zobel element from 220 nF to 22 nF could effect the loop stability. It is not sure how much. For a relatively fast amplifier the 22 nF may still be enough. It really depends on the details (e.g. transistor type, layout,...) of the amplifier and the load how much is needed. Some amplifiers get away without that RC element. The suitable size is more a thing to determine from experiment or maybe a careful simulation.

A flat frequency response to 90 kHz does not guarantee the amplifier could do that at full amplitude. There could still be a slew rate limit and the RC element at the output may overheat. It not so rare to find a burned resistor in the zobel element due to some oscillation of either the amplifier itself or some other stage before.

The Humingbird designer quotes amplifier response (-3dB) to 150KHz. I do not know how to test for HF instability. Would this be at rest or with an input signal?
EDIT: I think @trobbins answered my question.

[David Hess]

The Humingbird designer quotes amplifier response (-3dB) to 150KHz. I do not know how to test for HF instability. Would this be at rest or with an input signal?

The small signal response bandwidth is responsible for stability and for a properly designed audio amplifier it is 100s of kHz.  (1) The large signal response bandwidth is usually but not always just sufficient for audio because of slew rate limitations.  It actually takes some design finesse to get enough slew rate for power audio in a conventional design.

(1) If the designer naively implemented bandwidth control with the feedback network, then the measured small and large signal bandwidth is lower but this causes other performance problems so it is not favored.

[enut11]

I have completed most of the construction for calibrator #2.

Learnings from this build:
1) A bigger box was better but still not big enough. This one is a re-purposed modem box measuring 26x19x8cm.
2) The larger audio output transformer is a tight fit and probably too close to the font panel. There is an option of driving this externally from the LV output.
3) The small fan transformer was getting too hot so was relocated outside rear panel.
4) Some minor mods were made to the Hummingbird audio amp - 10uF input cap now 1uF and 1nF input cap removed.
5) 500 ohm input trim pot added to top of 220K amplifier input resistor

Next step is to test the signal output stability and noise.

[enut11]

Initial 1KHz tests of the new calibrator did not show any significant improvement over the the prototype wrt noise. See 10v and 100v tests below.

The new modules and transformers do, however, provide extended LV and HV ranges of 20v and 200v RMS.

I have yet to implement any form of feedback around the audio transformer as per previous suggestions (replies #42 and #44). I am not a circuit designer and would only be guessing what to do so I will leave that alone for the time being.

I am now going to concentrate on the input signal source. Previous tests have shown this to be the component that most affects the warm-up time.
I may also look at a PC based audio generator which is supposed to be very stable wrt amplitude over a wide frequency range (thanks @mzzj and @trobbins).
enut11

[enut11]

There are a few USB soundcards that have adequate audio bandwidth out to 80-90kHz before amplitude droop becomes noticeable.  The advantage of that type of oscillator source is that software can be used to null out harmonics from the 'output'.  As such, a very low distortion output can be sourced from the soundcard, but also extend further out to include any following buffer amp, as well as any step-up audio transformer.  Nulling harmonics in that manner can achieve low distortion levels that are intrinsic in just the ADC of the soundcard, and of course if the frequency is not too high in that bandwidth (eg. circa 20kHz fundamental for first few harmonics to be nulled), and the same applies for distortion introduced by a step-up transformer at the low-frequency end.  Certainly a cheap and easy way to start testing without the need for a low distortion oscillator.

Using a step-up audio transformer may benefit from some judicious loading of the high voltage output winding to maximise high frequency response flatness.  The output transformer likely needs to be 'hi-fi' in design to both push out amplitude variations from the first resonance, and to constrain the 'Q' of that resonance, given that resonance could well be below 50kHz depending on the transformer (a good transformer can push that resonance out to circa 100kHz).  The soundcard technique with suitable software (like REW) can even test the impedance of the transformer to confirm where the first resonance occurs (and hence the likely limit to frequency response for high voltage calibration efforts).

Hi @trobbins
I have an ASUS XONA7 U7 soundcard which should be suitable as an audio frequency generator. I have not used it in many years so I need to learn to drive it all over again. I originally bought it to test the distortion of amps and tape decks. You mentioned REW software as being suitable. Any how-to references would be appreciated.

[trobbins]

REW has a good help section.  REW has many functions and options, so it can take some effort to read all the relevant descriptions, and defer reading about features that you initially won't be interested in trying.  You would typically use the Generator and RTA functions for manual observation, and use the Measure function for automated sweeps.  The Preferences function is the starting point for selecting the soundcard and driver (hopefully there is an ASIO driver available) and the soundcard channels you want to use as output and input.   The https://www.avnirvana.com/ website has an REW forum, and is also where the software is available for download.  Some forums have a strong use of REW and threads such as https://www.diyaudio.com/community/forums/software-tools.123/.  Googling well known soundcards (like EMU0404, and Focusrite 2i2) brings up posts/threads where the soundcard has been tested using REW (and others) and indicates how they have been used and setup.  Googling tutorial may also be worthwhile - I don't look at the youtube posts but I can imagine there are a few helpful ones.

The 2019 review link (https://www.audiosciencereview.com/forum/index.php?threads/review-and-measurements-of-asus-xonar-u7-mkii-adc-dac-hp.8165/) indicates your card does have an ASIO driver available, but it sounds like there are issues, and the ASUS site only shows software up to 2016.  Hopefully you have a Mk2 as that has later software.  If you can get ASIO with operation at 192kHz at 24bit then that is great start.  Some soundcards are just fair performers, in which case you may need to spend time to characterise their performance for DAC and ADC paths to appreciate how to get a suitable performance for your application and to appreciate any limitations (eg. https://www.avnirvana.com/threads/measuring-frequency-response-and-distortion-of-amplifiers-%EF%BC%884x150w-rms-and-4x80w-rms%EF%BC%89.9764/).

If you are lucky enough to have a soundcard and ASIO driver that operates 'norminally' with REW, then that can open up a lot of measurement functions that imho far exceed legacy test equipment like scopes and distortion analysers and signal generators and impedance analysers/LCR meters, within the bandwidth constraints of typically 2Hz to 96kHz.

As an aside, I recently did a quick impedance check on a 1-to-9999 ohm 1930's 0.1% tolerance resistance decade box that I just restored, and due to the bifilar wind of the manganin coils and layering it seems like phase shift is fairly low out to 100kHz (circa 10deg at 30kHz).  So I hope to do some more investigation on how practical it may be to normalise for effectively zero phase shift up to 100kHz - all with the help of my soundcard and REW.

[enut11]

I installed REW on my Win10 PC and set the sinewave gen to 1KHz 1.0v out on the ASUS Xonar U7 soundcard and monitored the signal. Ambient temp was 28C.

The amplitude-time stability was poor compared to even my worst analog sig gen. It dropped by about 0.6mV per minute and showed no sign of leveling out even after 30 min.

Unless I am doing something drastically wrong, a PC soundcard source is not good enough for this project.

[Kleinstein]

Different sound- cards can be quite different in there performance. A super stable amplitude is normally not the main traget there, more like low distortion and stable frequency.

[David Hess]

Audio DACs can be very drifty as well.  Gain and offset precision is not a requirement for audio.

[1audio]

You would need to dive into the DAC implementation etc. to get a good sense of the causes of amplitude stability. I'll check a few later today (I have a number to check). Most audio DAC's derive their reference from the main supply. They may have an internal reference but its not super precise and may well be temperature sensitive and not compensated.  E.G. https://www.es.co.th/Schemetic/PDF/AK5394A.PDF  If you look the reference is sort of accessible but only for bypass caps. There may be an option for an external reference in the DAC on the Xonar card. Really good AC source stability may require a different approach (like the anal0g solution). My Boonton 1120 uses a state variable oscillator + sample and hold + Ref02 to get a stable source. The Fluke 510 also has a pretty carefully designed reference. It simpler and may be something that lends itself to the eBay oscillator board. https://xdevs.com/doc/Fluke/510A/510A_AA_imeng0000.pdf

[enut11]

I have completed the build of the AC Voltage Calibrator Ver2 and tested the LV at 10v and HV at 100v output. This final (?) version has a built-in 1KHz LD Audio Range Signal Generator mentioned in Reply #7 and has provision for an external signal generator.

Learnings from this project:
1) It is possible to build a stable 0-200vAC signal source at 100Hz to >30KHz frequency using readily available audio modules, voltage regulators and transformers.
2) Amplitude stability is almost entirely reliant on the signal source.
3) Low signal distortion is good but far more important is signal noise as this is amplified by up to x400.
4) Fan cooling of the audio amplifier is recommended and speeds up the settling time.
5) Separating the LV and HV output terminals allows the HV output to float above ground.

One application would be to calibrate up to 5.5 digit AC meters using a 6.5 digit or better reference meter.

[enut11]

As a follow up to this project, I found a cheaper solution for the power amplifier module.
A class AB MOSFET audio amp kit for less than $31AUD at:

https://www.ebay.com.au/itm/124926624992

This amp can be powered by up to +/-70v rails and deliver up to 40vRMS at the output.

As before, regulated rails improve amplitude stability. Also, for frequencies over 20KHz, the output stability capacitor should be reduced in value accordingly otherwise the 10R/1W series resistor will go up in smoke!

A heatsink is essential to assist with thermal tracking.

L7 mono MOSFET class AB audio amplifier board kit.
Specs:
Output power of 150W 8R DC +/- 56V rails (this applications only needs a couple of watts)
Frequency response 4Hz-350KHz, -3dB
THD <0.01% 100W 1KHz
Slew Rate = 38V/uS
SNR> 99dB
Requires AC 50V(or less)x2 transformer, or DC ±70V(or less) power supply.
PCB size = 78 * 63 MM