An Experimental AC Voltage Calibrator

[enut11]

Thanks @bdunham7. I did notice a little VHF noise dancing on top of the 30v square square wave. Don't know if this is normal for an audio amp or what I can do about it though.

[enut11]

Hi @trobbins. My scope is a 100MHz Tekway. The 10KHz 30v squarewave output (LVout) is shown below. The 20KHz test was little different.

[trobbins]

Ta, just for clarity, did the generator output squarewave have a much faster risetime than circa 2.8us (as shown for the power amp LV output signal)?

[enut11]

Wavetek Model 80 function generator wave specs attached:

[TiN]

Just for reference, 3458A can do ACV measurements in multiple ways, with RMS to DC conversion (ACV ANALOG mode), by accurate sampling (ACV SYNC mode) and also by fast sampling (SSDC and SSAC modes). I would highly recommend Rado Lapuh's book showcasing all the various operation modes and benchmarks of 3458A in terms of AC voltage.

There is also very nice appnote showing methods to measure ACV to very high precision from designer of 3458A ACV function, Robert Swerlein.

[enut11]

Hi @TiN. It is interesting to see that the HP3458A is fast enough to be able to accurately sample an AC waveform in DC mode. Amazing.

[1audio]

Setting up an amp with an output transformer works OK. Ideally feedback around the transformer would help. Even if its just a leveling loop since the transformer/amp will have a rolloff and will be sensitive to the load. I have done that in the past with LM3886's and transformers to make a 120V AC supply. Stability is a little tricky but worth the effort.

Is there any info on the amplitude stability of the DDS sources? That stability is critical to the utility of the project. If the voltage is changing (like the 1/2 hour warmup) then it can all be very frustrating.

[David Hess]

Is there any info on the amplitude stability of the DDS sources? That stability is critical to the utility of the project. If the voltage is changing (like the 1/2 hour warmup) then it can all be very frustrating.

Stability of the analog output from a DDS depends on the reference, digital-to-analog converter, and signal conditioning circuitry between the DAC and output.  They are not usually designed for DC precision, but could be.

Hi @TiN. It is interesting to see that the HP3458A is fast enough to be able to accurately sample an AC waveform in DC mode. Amazing.

I have used many different high resolution bench multimeters with triggering capability to make low frequency AC measurements in DC mode.  The low frequency limit depends on how long a time period the measurements are made, and the high frequency limit usually depends on the integration time which is typically an integer number of power line cycles, but can be much shorter at lower resolution.  This is a great way to make 0.1 to 10 Hz noise measurements.

I used this method to design some incredibly low noise DC amplifiers.

[TiN]

I've got ManateeMafia's Holt HCS-1AF on a bench as well and can do some measurements with them. If anyone have particular interest, feel free to drop test suggestions. Sadly 20A shunt from a set is blown open, but other seem to be ok (5A, 2A, 0.5A, 0.2A and 0.02A).

[enut11]

Setting up an amp with an output transformer works OK. Ideally feedback around the transformer would help. Even if its just a leveling loop since the transformer/amp will have a rolloff and will be sensitive to the load. I have done that in the past with LM3886's and transformers to make a 120V AC supply. Stability is a little tricky but worth the effort.

Is there any info on the amplitude stability of the DDS sources? That stability is critical to the utility of the project. If the voltage is changing (like the 1/2 hour warmup) then it can all be very frustrating.

Hi @1audio. Do you have a circuit diagram to illustrate your point? At the moment, the audio transformer output in this project is fully floating.

[enut11]

I have been gathering parts for an upgraded AC Voltage Calibrator, including a larger box. Picture shows the main components to be incorporated. Still waiting for the toroidal power transformer with two 25v windings. This will extend the calibrator low voltage range to just over 20v RMS.

The LV  output will be 0-20v RMS, 20Hz-50KHz (direct power amp output)
The HV output will be up to 200v RMS and 50Hz-25KHz (power amp + audio transformer)

I have a HUA 0-300v AC analog voltmeter which was to go into my 50Hz variac. I tested the meter frequency response and was surprised that it still registered 95% of the true value up to about 50KHz! This will now be used to give a rough indication of voltage on the output terminals. Important from a safety point of view when dealing with voltages over 50vAC. I am also thinking of other ways to indicate the presence of dangerous voltages.

[enut11]

This is the new AC voltage amplifier module. Based on the Hummingbird audio amp design from Silicon Chip (Aust) magazine December 2021 - see reply #14. The only change I made was  smaller, lower power output transistors (TIP41C and TIP42C).

The first question that may come to mind is, why such a big heatsink? The short answer is to control heat - the nemesis of all precision instruments. Class AB low distortion amplifiers draw a significant standing current. In this case 50mA, and this generates approximately 3 watts with no output signal. The 12v fan will be run at 8v to reduce noise.

The module has a gain of ~18 and it will amplify the audio oscillator signal  to 0-20vRMS for frequencies 10Hz-50KHz. This should be suitable for calibrating the lower AC voltage ranges of most multimeters.

[Kleinstein]

With smaller power transistors, the emitter resistors should also be larger. This makes the adjustment of the standing current more stable and may allow to run the amplifier with less current.  A little more current flow still helps to make the amplifier fast. So some power loss is normal and with an instrument not used 24/7 should not be so bad.
If speed of the amplifier gets a problem, I would consider the slightly faster D44H / D45H instead of the TIP41/42.

[enut11]

Thanks @Kleinstein. I have replace the output resistors (0.22ohm) with 1 ohm 1W and reset the bias current to ~25mA, a little lower than spec but the waveform is clean and the amp runs much cooler. Tested to 15vRMS running off a power supply. The full output of 20vRMS will have to wait until the power transformer arrives.
enut11

[1audio]

RE feedback around the transformer. You want to lower the output impedance so that you can be confident in the measured output voltage not changing once set.  Here is feedback around the XFMR. I set up around 6-10 dB of feedback so stability is not really hard to achieve. I have used this to drive a turntable motor with success. I was using LM3886 amp chips which are easy to work with.

The other option, leveling, is more involved.

[Kleinstein]

For the feedback around the transformer the very high frequency part should usually bypass the transformer. This makes it easier to get stability. Obviously the DC part also needs to be from before the transformer and the DC offset may need some care.

[enut11]

For the feedback around the transformer the very high frequency part should usually bypass the transformer. This makes it easier to get stability. Obviously the DC part also needs to be from before the transformer and the DC offset may need some care.

More information please or where I can find some. Working with signal transformers is new to me.
enut11

[1audio]

Unfortunately transformer coupled linear audio amps without tubes are reare so there is not a lot to refer to. Here is some that can help:
https://www.jensen-transformers.com/wp-content/uploads/2014/08/Audio-Transformers-Chapter.pdf
https://www.jensen-transformers.com/wp-content/uploads/2014/08/an001.pdf
https://www.lundahltransformers.com/wp-content/uploads/datasheets/feedbck.pdf     Interesting but tricky to implement. Audio precision has something similar. Details are in their patent: https://patents.google.com/patent/US4614914A/en

You can look at the tube stuff but its not all that detailed in how to stabilize, except for stuff from the 1950s and 1960s.
Using HF feedback to reduce the gain around the transformer is important for stability. Cap coupling to keep DC out may also be important. Even a little DC can saturate the transformer.

[trobbins]

The soundcard/software technique can measure HV output magnitude/phase response before feedback is applied to the output of the step-up transformer, to get an 'open-loop' response to guide whether the HF response is smooth or prone to resonance that may limit or make closed loop operation unstable when applying too much feedback, especially if your HV loading varies a bit.  Or you can do some spot checks with a scope and use X-Y for phase assessment.  As indicated, the feedback signal from after the transformer needs to be rolled-off to reduce HF content level that may cause unstable operation (ie. phase and gain margins need to be kept to respectable levels and not go too close to unstable conditions).

[enut11]

You can look at the tube stuff but its not all that detailed in how to stabilize, except for stuff from the 1950s and 1960s.
Using HF feedback to reduce the gain around the transformer is important for stability. Cap coupling to keep DC out may also be important. Even a little DC can saturate the transformer.

The new Hummingbird amp module I am working with has ~60mV DC at the output. Is this significant wrt transformer saturation or not?

[trobbins]

That is 60mVdc across an 8 ohm winding with a 5W rating that could be used for up to 6.3Vrms across it (ie. 9Vpk).  So no, 60mVdc is not going to bias the core much at all.

[1audio]

The dcr of an 8 Ohm winding is more like .5 Ohms so the current could be significant. Is there an offset trim on the amp?

[enut11]

No output offset is provided on the Hummingbird amp. See reply #63

[1audio]

A basic fix would be a pot across the emitters of the input pair tied to the input current source. 1K should be enough to get adequate adjustment.

[enut11]

The power transformer has arrived and I can now get into part 2 of the AC Calibrator project.

Aim: 0-20vRMS directly out of the power amp 'LV' module for frequencies from 50Hz up to around 50KHz.

This will require a little over 60Vpp or a minimum +/- 30v rails for the power amp. As I intend to regulate the power rails for additional stability, I chose a power transformer with two 25v secondary windings. Rectified and filtered these will give about +/-36v DC.

The voltage regulator kit I used had LM317/LM337 chips. I had to change the 5K pots as they did not allow enough adjustment. I set the regulated outputs to +/- 31v DC. This allowed about 5v (36-31) for the regulators to work.

At lower frequencies (<20KHz) there was no problem generating 20vRMS with this setup. However, as frequency was increased, the amp output capacitor to ground (220nF) has lower and lower reactance and presents a heavier load. At around 48KHz the reactance is ~15 ohm and this, along with the 15 ohm output resistor, puts a load of about 30 ohms on the amp output.

The resulting current of 670mA (20v/30 ohms) is enough to drop the 317/337 pair out of regulation with significant ripple breaking through.

Looking for feedback here. How much leeway do I have to fiddle with the 15 ohm resistor and 220nF cap? Looking to a least halve the amp output loading at 50KHz.