DIY Precision AC-RMS to DC Transfer Standard

[amspire]

I have been working on a simple precision DC to Sinewave converter - so if you enter 5V DC in, you get a very precise and very stable 5V AC out. Enter 10mV DC in and you get 10mV AC out.

But then I started to think - How do you verify it is accurate?

I have a Fluke 540B that can thermally compare the heat produced in a 100 ohm resistor by AC to the heat produces by DC and it has a 0.01% accuracy. The great thing about this solution is you only need to have an accurate DC standard, and you can use it for AC calibration as well.

http://bama.edebris.com/manuals/fluke/540b/

Can common parts get similar results?

Thinking about it, I went for a 100 ohm 402 SMD load resistor glued to diode (a C-B junction of a SMD NPN transistor). The use of a diode as a temperature sensor allows two options - forward biased where you get about -2mV/C coefficient. This is good as we need to know the resistor temperature. My initial target is 100 degC so that I have a safe margin. The diode could also be used in reverse mode using the diode leakage current.

The wires are 0.1mm enamelled copper - soldering wasn't much fun. Finding very fine wire is easy - old phone earpieces, small speakers, etc. Mine came from a dud 40mm speaker from a transistor radio.

The initial results surprised me. Got the 100 Deg C target temperature at 3V - that is 12 times worse then the Fluke, but the settling time looked useable. I was getting the diode voltage going up and down by 50mV every few seconds. Then I realised that this was just air movement in a room. Stuck it in a jar and it became very stable - no fluctuations at all.

Based on my very quick first tests, 0.01% should be achievable. The sensitivity means that 0.01% difference between the AC and DC corresponds to a 20uV DC voltage change.  I am not concerned with the high power requirements of my sensor as we have something that Fluke did not have in the 1960's - we can easily get 100MHz+ opamps that should have a flat enough frequency response to get DC to 1kHz accuracies of 0.01% easily.

[alanambrose]

Wow very neat

[Mickle T.]

Nice work!

There is another way to make a stable and predictable AC voltage standard via the calculable AC voltage reference.

The Solutions : Calculable AC Voltage Reference https://www.ncsli.org/c/f/p13/REG_2013.PRE.1147.1876.ppt
Digitally Generated AC Reference Source http://www.transmille.net/Presentations/AC%20Reference/Digital%20AC%20Source.pdf
Guildline Model 7410 AC Voltage Reference http://www.guildline.com/Datasheet/Guildline7410Datasheet.pdf
Sine Wave Generation Techniques http://my.ece.ucsb.edu/York/Bobsclass/2C/Tutorials/App%20notes/an-263.pdf

This is my first try: simplified version of the 50 Hz - 1 MHz fixed voltages AC calibrator.

[quarks]

Looks very interesting
thanks for sharing

[Vgkid]

Very interesting, I know in a LT app note they use a dual matched thermistor. That was for higher frequencies, many Mhz.
 Subscribing to this topic.

[Gyro]

I tried making one of these a few years back. I've just had a look and I can't see it, I'll post photos if I can find it.

Dual heaters and sensors are the way to go, you want excellent isolation between the two and slightly lower isolation between each and ambient.

I used a sandwich of two aluminium plates and two sheets of expanded polystyrene with the heater/sensors squeezed between them. I used 1206 resistors (200R) glued to the back of BAS16s (SOT-23). Connections were very fine 'magnet' wire (loosely twisted) inputs from one side, outputs from the other.

Although I was only using an LM358, results were pretty good. Zero and tracking were ok, but I got horrible overshoot (hadn't worked out the opamp time constants correctly). I dropped it in favor of some other project and never got back to it.

I've been thinking of having another go (your post is well timed) and have purchased some 1k2 0.1% 1206 resistors and BAV99s. The idea is to sandwich the BAV99 between two paralleled 1k2s to make 600R in. The BAV99 gives two series diodes, so twice the temperature coefficient, and connections on the same side of the SOT23 so better isolation (snip the center pin).

Two 100R resistors would make a 50R input, not sure about frequency performance, that would depend on the terminations (as I said, twisted pair in and out).

You probably want to look at Jim Williams's App note Lin Tech AN22, on the subject, he was using it for LCD CCFL backlight inverter evaluation and was using the LT1088 thermal converter. (edit:There's also) a rather good write-up somewhere comparing it with the technology used in the the HP3400A RMS meter.

http://cds.linear.com/docs/en/application-note/an22.pdf

[Gyro]

Ah, here's the other App note, LT AN61:

http://cds.linear.com/docs/en/application-note/an61fa.pdf

The Thermal RMS bit starts on page 16, and the HP3400a comparison in Appendix A, page 28.

It's quite a catch-all app note... It's also got his avalanche pulse generator on page 21. 

[mmagin]

I see the LT1088 is discontinued and nearly $40 on ebay from sellers with not particularly excellent feedback

[Gyro]

Sadly yes, but most of the app note info is equally applicable to home made heater-diode pairs.

Edit: ...and SO much cheaper to repair if you burn it out, you wouldn't want to say goodbye to $40 each time!   

[Gyro]

I've found my original and grubby 'thermal module'. It's looking pretty sad, I realized that the 'few years' are more like 20 odd  . The wires have all got broken in the mean time but as I said, I'm planning to re-do the elements anyway.

I hope it might be of some help anyway. Pictures attached.

P.S. Looking at it now, it could probably be made rather more compact!   

[timb]

There's an older Jim Williams app note before the LT1088 that discusses the construction of a thermal RMS meter with a Yellow Springs dual thermistor package. The idea being use two of them, thermally insulated. Since it's a dual thermistor, one heats and the other senses the heat. That part is no longer made either, but the principle is sound.

I'll see if I can find the app note. Jim Williams had an obsession with thermal RMS stuff, as it appears in several app notes (and VF/FV circuits, but that goes without saying). It always puzzled me why, after the LT1088 died (which was his baby) that he wrote a whole app note on a new non-thermal responding RMS chip LT made, and never once mentioned thermal responding meters. It was like the LT1088 never existed. Sad.

[Vgkid]

http://www.linear.com/solutions/1378
Not the app note per se, but the relevant circuit.

[amspire]

Matched dual resistor-sensor pairs are great at getting a real time output and eliminating 99% of the hassles of a 540B-type transfer standard, but it is at a big hit to the accuracy. Metrology is all about confirming everything. You cannot just trust in data sheets and hope the design has no unknown flaws.

The transfer standard approach is slow, and you need very stable DC and AC sources. You need the sources stable to better then 0.001% over a period of several minutes. The Fluke procedure requires that you repeat the measurement cycle 3 times and all of the measurements have to agree within the intended tolerence. Also at least one of these sources has to be adjustable in increments of at least 0.001%. My original DC to AC design was actually intended for this purpose. So if this sensor is a go, it will make a useful pair with the AC/DC source design.

Additionally, with the Fluke sensors, you can start to get thermocouple-type EMF voltage problems. They have a resistor reversing switch, and basically if your 540B starts to give different readings for different sensor directions, your 540B is cooked. Probably have to include a reversing switch in any thermal-based design.

There is also the quality of the resistor to consider. There might be a significant voltage coefficient, or at low frequencies, the resistor temperature may be varying within a cycle. With the resistor's temp-coefficient, this could mean the resistance varies with the cycle. Best way I can think of to detect these problems is to attach the resistor in series with a stable resistor (like a vishay foil resistor) to a low distortion AC source (0.0001% distortion or lower) and check the distortion across the resistor.

I am going to try and do a test today with my prototype sensor and match it to the Fluke 540B.

[amspire]

This is my first try: simplified version of the 50 Hz - 1 MHz fixed voltages AC calibrator.

That looks like a neat effort for a first try. I want one!

My design design for an AC source is much less ambitious. It is a calculable AC source, but the maximum frequency will probably be 10KHz initially. I can go to 10Khz very simply, and above that, the project escalates.

I hadn't thought about using an oven, but it makes sense as a way to get extra stability out of standard parts.

Did you calibrate it to get the 10.0000V AC and 0.9999V AC on the meters, or did you calibrate the DC reference only, and it correctly generated the right AC volts?

Also, do you have any numbers on output stability? I can see a number of metal packages on your boards and I am guessing you have used hermetic Vishay foil resistors. Very nice!

Richard.

[Mickle T.]

I did the calibration on the AC meter only. Short-term stability of the output voltage is about 0.01% and limited by the temperature fluctuation in the oven (~0.3C). 10 V frequency response linearity is ~0.05% (100 Hz to 100 kHz), ~0.16% (100 kHz to 1 MHz) - measured via Datron 1071. All critical feedback resistors are hermetic S5-61 metal-foil.

[amspire]

I have some initial results. I have a very jury rigged setup, but I have exactly the same signal going to the Fluke 540B and my RMS sensor.

The sinewave source signal was 1Khz at close to 3V RMS. The DC supply was also 3V.

I couldn't precisely adjust the DC to the match the AC voltage. The Fluke indicated that the DC was lower by 0.021%.

Using the forward biased junction of the diode at 1mA on my sensor, I got:

Cold temperature:  0.731195V
AC volts: 0.637657V (The higher the voltage, the lower the reading)
DC Volts: 0.637684V

It seems that a 0.01% difference causes approximately a 20uV change. So with a 37uV change, that means that my sensor indicates the AC source was about 0.018% higher then the DC voltage. That is only a difference of about 0.003%.

I will test some more frequencies, and I will add a 0.01% switch to the AC source (it will lower the source by 0.01%) along with a vernier adjustment, so I can get a better match between the DC and AC sources.

So far, it is looking good. Interesting thing is my sensor was only running at about 47 deg C above ambient which is a bit low. I will boost it up to about 60 deg above ambient next time and it should respond a bit quicker.

Richard.

[amspire]

Did some more test at different frequencies. Ran at a higher temperature, but didn't make it any better.

I found that for the initial measurement, I really had to leave it for about 30 minutes. I think the air in the jar containing the sensor has to stabilize. The sensor only warms the air in the jar very slightly, but until the air warms the right amount, none of the readings are stable. A smaller jar would probably be quicker. After it stabilizes, it is not to bad.

I had a lot of problems with my Sig gen stability (down at the 0.01% level) but these are the results for the difference between my sensor and the Fluke 540B:

100 Hz	1KHz	10KHz	100KHz	1MHz
+0.003%	+0.0006%	-0.0005%	+0.01%	+0.01%

It will only ever be of use as a calibration tool, and unless the sensor can be put in a vacuum tube, it only works as long as you follow a strict procedure. Might put the procedure into a micro-controller and so it will force you to wait until it is ready to measure. Starting off with the DC and AC as correct as possible will make the measurement much faster.

Before I go further, I definitely need to set up stable DC and AC sources

[Kleinstein]

For temperature measurement there is no need to use a 1 mA current through the diode - that is already a lot of current and thus heating. So I would suggest more like 10 µA - this still gives an 2.5 K output impedance. As a side effect the TK also gets a little higher at lower current.

To make the thing react faster you would need less isolation, e.g shorter wires. So smaller is faster.  A higher temperature does not help very much. If it is only 45 C this means you could use even higher voltage, so a larger range.

The design that Gyro showed actually look quite good: two cells to compare and a reasonably thin (and thus fast reacting) isolation layer. The time it takes to stabilize goes up with the square of the thickness - so it can easily get to much isolation and thus to slow. Also the aluminum frame gives similar conditions for both sides.

[amspire]

10 V frequency response linearity is ~0.05% (100 Hz to 100 kHz), ~0.16% (100 kHz to 1 MHz) - measured via Datron 1071.

That sounds like you are hitting the limits of the Datron accuracy. Sounds like you don't actually know how good your AC calibrator is. It is as good as the Datron at least.

[amspire]

For temperature measurement there is no need to use a 1 mA current through the diode - that is already a lot of current and thus heating. So I would suggest more like 10 µA - this still gives an 2.5 K output impedance. As a side effect the TK also gets a little higher at lower current.

I tried lower currents, but they make the diode impedance higher and so there is more AC noise that gets though to the diode voltage detector circuit. Also, the time constant of the 1mA diode current heating is faster then other time constants, so it makes little difference.

To make the thing react faster you would need less isolation, e.g shorter wires. So smaller is faster.  A higher temperature does not help very much. If it is only 45 C this means you could use even higher voltage, so a larger range.

The lengths of the wires make no difference, but the sensor has to be kept away from any thermal mass. The wire length may make a difference for RF.

The way the Fluke 540B-type sensors work is they rely on the sensor radiating heat. If I could put my sensor in a vacuum inside a black aluminium tube, it would work brilliantly. I cannot do the vacuum, but the sensor may end up in a cheap torch tube.

The design that Gyro showed actually look quite good...

It is a very interesting design, but it achieves a complete different goal. It makes a useable real-time wideband RMS voltage measuring device, that possibly could get to 0.1% accuracy with a lot of work. 0.1% would be extremely impressive. But very hard to use it as a precision calibration tool.

The sensor I am testing looks like it can probably manage 0.001% accuracy that makes it a candidate for calibrating other devices to 0.01%. The thing about this kind of calibration is you only use it, say, once a year. For a hobbyist, taking 30 minutes per voltage range calibrated is not bad. It is not like you have to be watching it the whole time. Once you have calibrated your best meter, you use that meter to calibrate everything else. So I don't want to make this transfer standard complicated - the simpler the better.

[Gyro]

Your results look very impressive!

I can understand that you want to keep it as simple as possible, but wouldn't you benefit from some ambient temperature compensation, ie. another unheated transistor operating at the same current, to compensate ambient changes during your run?

[amspire]

I can understand that you want to keep it as simple as possible, but wouldn't you benefit from some ambient temperature compensation, ie. another unheated transistor operating at the same current, to compensate ambient changes during your run?

Absolutely! For the next stage, I need to make a new sensor - and it will have a second unpowered transistor/diode. I will probably even glue a resistor on the top so that it has an identical thermal mass. I think that Fluke do exactly that in the 540B as well - I seem to remember that there are actually two thermocouple sensors, but only one is powered.

[Gyro]

I will probably even glue a resistor on the top so that it has an identical thermal mass. I think that Fluke do exactly that in the 540B as well - I seem to remember that there are actually two thermocouple sensors, but only one is powered.

That sounds like a good idea - I must study the 540B documentation and try the same myself. I plan to re-make both resistor-diode elements, but there's no reason that I have to apply power to the second one.

[mmagin]

I can understand that you want to keep it as simple as possible, but wouldn't you benefit from some ambient temperature compensation, ie. another unheated transistor operating at the same current, to compensate ambient changes during your run?

Absolutely! For the next stage, I need to make a new sensor - and it will have a second unpowered transistor/diode. I will probably even glue a resistor on the top so that it has an identical thermal mass. I think that Fluke do exactly that in the 540B as well - I seem to remember that there are actually two thermocouple sensors, but only one is powered.

Is this essentially the same thing as a bolometer?

[amspire]

I think a bolometer is for measuring electromagnetic
radiation.