DIY Variable AC Voltage and Frequency Source for Calibration

[enut11]

https://www.eevblog.com/forum/metrology/an-experimental-ac-voltage-calibrator/msg3840836/#msg3840836

Some time ago, in the Metrology section, I described how I built a stable AC voltage source for calibration purposes.
The resulting meter could supply up to 200v and up to 30KHz (at lower voltages) and was intended to be used with an accurate meter to transfer the AC voltage to another meter.
In the end it worked to my satisfaction but it evolved into a bulky and costly instrument.

Recently I set about building a simpler 'AC Calibrator', so here it is.
enut11

[enut11]

The basic components are:
1) Input attenuator (5K helipot)
2) Audio amplifier (LM1875 kit, including power transformer)
3) Audio line transformer (MM1120 or equivalent)

Optional components:
1) switchable low/high outputs
2) analog AC voltmeter (HUA brand on eBay)*
3) Audio clipping indicator (LED)

*You can buy dedicated HUA brand AC voltmeters on eBay. They have very good frequency response (> 20KHz)

[enut11]

Also needed is a stable AC signal either from an audio generator or function generator. The signal must be sinusoidal and relatively low distortion.

For this instrument I use a Feeltech FY2300 arbitrary waveform generator, here seen in the background.

The stability of this DIY calibrator is almost entirely dependent on the signal source.
I suggest that you characterise your signal source first to see how long it takes to stabilise from a cold start.
The FY2300 can take up to 30min to stabilise.

[enut11]

The inner workings. Everything, bar the signal source, was mounted inside a plastic Jiffy box (Jaycar Australia) 20x11x6cm.

Rear
Left: 30-40v center tapped power transformer.  Only need <10VA as we are feeding high impedances.
Center: LM1875 audio amplifier
Right: 2-20W Audio transformer

Front
Left: AC voltmeter (converted 200uA meter)
Center: Input helipot and peak indicator PCB

[enut11]

This is the audio transformer that I used (Altronics Australia, M1120). We don't need high power as we are feeding high impedance DUTs. Frequency response should be up to 20KHz or better.

The multiple tappings help to program the optimum inputs and outputs. In my case I fed the output of the LM1875 audio amplifier into the 16ohm tap and output the signal to the front panel from the 2.5W tap. Note that the audio transformer is used back-to-front. This is how we get up to 200vAC from mV signal sources.

This makes for a very flexible design as it can match a variety of signal sources.

[enut11]

Below is an analysis of the M1120 audio transformer showing possible 'gains' from the various taps. I found it best to experiment, starting with the 16ohm tap. If this does not provide enough 'gain' (Column E), move to the 8ohm tap.

Important
You need to ensure that the audio amp does not clip the signal into the audio transformer. The resulting 'square wave' will distort the output to the front panel which must be sinusoidal.
You can monitor the output of the audio amp with an oscilloscope or, as I did, with an audio clipping LED indicator.

https://www.instructables.com/Clip-Indicator-LED-for-any-power-amp/

[trobbins]

Good stuff.  I think I will follow suit. 

Did you use a cheapo LM1875 kit with a clone LM1875, or use a bone fida LM1875?  It looks like the clones typically have a smaller die, although that aspect may not be a concern for this application, but may be for your first calibrator effort.

I note the cheapo pcb layout is not too bad, but doesn't quite follow layout guidelines by using the same 0V return trace for speaker and zobel.  You also seem to be using the pcb main filter caps as the rectifier filter caps, as a short cut so to speak.

Did you use 22k and 1k for feedback, as that is circa 27dB gain, and close to the datasheet typical gain-phase plot.  I just noticed you may not have measured phase shift versus frequency in your other thread although it is likely to be quite low. 

[David Hess]

So the accuracy depends on the calibration of the reference source?

This seems more like a transfer standard for when you already have a calibrated way to measure AC voltage.

[Kleinstein]

The unit seems to be "only" an amplifier with some division. There is a chance that the transformer taps give rather accurate voltage ratios. Especially the ratio of different secondary taps with low loading can be quite accurate as it is set by the turns ratio and thus a fixed number. There seem to be quite good 1:2:4 ratios.

[dietert1]

Won't core losses introduce errors on the input side of the transformer? And core losses may depend on iron temperature (drifting). It might be worth to take amplifier negative feedback from the transformer output side, i mean unless transformer isolation is important.

Regards, Dieter

[Kleinstein]

Core losses would be an additional load to the primary. Ideally (epseically not too high a frequency) the core loss is relatively low and not that temperature dependent.
Taking the feedback from the seconday side would be ideal, but can have isssues with the stabilty. One may still want the highest frequency FB from the amplifier.

[David Hess]

Taking the feedback from the seconday side would be ideal, but can have isssues with the stabilty.

Accurate high voltage compensated dividers are not trivial to implement either.

When I looked at the sine wave AC reference problem many years ago, I thought about using analog sampling and high voltage analog sampling to make an accurate resistive divider easier.

[enut11]

Good stuff.  I think I will follow suit. 

Did you use a cheapo LM1875 kit with a clone LM1875, or use a bone fida LM1875?  It looks like the clones typically have a smaller die, although that aspect may not be a concern for this application, but may be for your first calibrator effort.

I note the cheapo pcb layout is not too bad, but doesn't quite follow layout guidelines by using the same 0V return trace for speaker and zobel.  You also seem to be using the pcb main filter caps as the rectifier filter caps, as a short cut so to speak.

Did you use 22k and 1k for feedback, as that is circa 27dB gain, and close to the datasheet typical gain-phase plot.  I just noticed you may not have measured phase shift versus frequency in your other thread although it is likely to be quite low.

Hi @trobbins
I used the cheap Chinese LM1875 kit as it is hardly being stressed in this application. Also, made no changes to the default gain which is about 20x.
Yes, I did use the on-board 100uF caps and put 2 diodes at the PCB power input for half-wave rectification, again we are only needing to supply milliamps.
The power transformer is a 4.5VA 30v CT but I would recommend at least a 40v CT for more overload margin.
enut11

[enut11]

So the accuracy depends on the calibration of the reference source?

This seems more like a transfer standard for when you already have a calibrated way to measure AC voltage.

@David Hess, this is not a true calibrator but a variable AC voltage/frequency generator. Yes, it relies on being able to measure the output with a known meter and transferring this to the DUT.

[enut11]

The unit seems to be "only" an amplifier with some division. There is a chance that the transformer taps give rather accurate voltage ratios. Especially the ratio of different secondary taps with low loading can be quite accurate as it is set by the turns ratio and thus a fixed number. There seem to be quite good 1:2:4 ratios.

@Kleinstein
Providing a calibrated dial output is something I could explore and worth considering. It would at lease negate the need for a level meter.

[enut11]

Won't core losses introduce errors on the input side of the transformer? And core losses may depend on iron temperature (drifting). It might be worth to take amplifier negative feedback from the transformer output side, i mean unless transformer isolation is important.

Regards, Dieter

@deiter1, this is obviously an open-ended system. I have resisted creating a closed loop mainly because I am not confident on how to do it.
Having said that, as long as the instrument is stable for the short time it takes to transfer a reading using a known meter to the DUT, it has done its job.
enut11

[enut11]

Providing sufficient time is given for everything to settle down, this simple ‘AC Calibrator’ source can give surprisingly good results. The plot below is a 1KHz signal from a FY2300 function generator amplified to 100v. A little bit of drift is evident and short term variations are about 2mV PP (0.002%).

Notes on usage:
1) Allow everything, especially the signal source, to stabilise temperature wise.
2) Use an input capacitor, if your amplifier does not have one, to block any DC from getting to the audio transformer.
3) Always start with the attenuator (helipot) set to zero before applying an input signal.
4) Monitor the amplifier output signal for clipping. Use a CRO or audio clipping indicator (LED).
5) Include a means of isolating the high-voltage output as anything over about 50v can be dangerous.
6) Most DMMs on ACv are calibrated at 1KHz and, at this frequency, the instrument can go all the way to about 200v. Mine will be limited to 100v as this is a normal calibration point for instruments.
7) Be wary of high voltages at high frequencies as these can cause unpredictable (catastrophic?) results, both before and after the audio transformer. Eg, you can easily destroy the output zobel filter on the amplifier. I normally test at 1KHz and 10KHz.
enut11

[Doctorandus_P]

Edit / Oops.

I posted this in the wrong thread. I'm tempted to delete it, but leave it because Kleinstein gave a response and deleting it now would break that answer.

The intended thread was:
https://www.eevblog.com/forum/projects/making-an-lvdt/



I've been reading a bit about LVDT's and not really convinced this is a wonderful thing. It uses a complicated way to generate an analog voltage, and then you have to put that into an ADC.

I also bumped into the AD598. This little chip costs EUR122 from mouser, which seems quite ridiculous.

Now you've built your sensor coils and are doing some experiments, I suggest you also do some experiments with a simpler method.
Ever since the '90-ies there are a bunch of circuits floating around the 'net for an LC meter. It's build as a free running LC oscillator and a microcontroller to measure the frequency of the oscillator though some reciprocal math. Years ago I did some experiments with this, and you can do quick measurements with about 0.1Hz resolution, (and could get more resolution by integrating over more oscillation periods). The hardware is very simple, and schematics and software is floating around the 'net.

I guess both methods are sensitive to nearby ferromagnetic materials which influence the coupling or inductance of the coil(s). But LVDT is also sensitive to the amplitude of the input signal, but it may be less sensitive to humidity or temperature. It would be interesting to see with which method you can get the most resolution and stability.

[Kleinstein]

A LVDT is a displacement (could be a rotational version) sensor. It is not made as a stable source.

Measuring a capacitance / inductance from a LC oscillator is tricky as one get a measurement at a frequency that depends on the part. At the fine details the capacitance and inductance is frequency dependent.

The question with an AC calcibrator may be if one wants precision transformers to set ratios. If made right (good core, ideally dual core) a transformer can make a very accurate (can be better 1 ppm) ratio reference. E.g. the Datron 1281 uses a ratio transformer in it's ACAL system.

[enut11]

Update
1) Fitted a proper ACv panel meter (HUA brand from eBay). The meter provides a rough indication of the High Volts output.
2) Limited output to 100ACv, although the circuit can supply over 200ACv.
3) Removed the 20v/200v meter switch and in its place is a x10 attenuator (switched 42K resistor between input socket and 5K pot).
4) Provided an external jumper from Low Volts ground (black terminal) to High Volts green terminal. This eliminated a tendency for the previously floating output to pick up the odd glitch.

It is difficult to specify the upper frequency limit for this device. It depends on which audio amplifier you use - there are many choices on eBay. I limited mine to 10KHz.
Hi volts coupled with high frequencies can burn the amp output filter. It is possible to fiddle with the filter components but beware of HF oscillation (in the MHz!).
The M1120 audio transformer is capable of over 20KHz.
enut11