True rms converters

[Vgkid]

Since I do nothave a stable ac voltage source I thought that I would build one using the schematic from the LTC1968. Since they are only a few offerings from LT/AD/Maxim. This would increase the odds that others would have dealt with them. My own multimeters(Fluke 87/34401) have these chips inside. There fore, they should be pretty good.
That is not what I'm after. What I would like to know is how stable are these chips(day,months,yearly) drift rates, and what is the tempco of these units.  I will be using a fixed frequency, and voltage.

[tszaboo]

I've used them. They are some 0.1% accurate. Not the most accurate things in the world, but for AC it is very good in fact. I definately prefer them over any heat transfer RMS or those pseudo RMS things from Analog devices. They do some delta-sigma magic in them, but that is OK. Very important, use a good quality foil capacitor on the output.

[TiN]

I definately prefer them over any heat transfer RMS or those pseudo RMS things from Analog devices.

Why so both ways? I think more detailed reasoning could help others to understand better too.

[e61_phil]

I definately prefer them over any heat transfer RMS or those pseudo RMS things from Analog devices.

Why so both ways? I think more detailed reasoning could help others to understand better too.

I'm also interested in any answer to this question. I've bought a Fluke 540B thermal transfer standard for AC measurements some time ago..

[ebclr]

Another interesting circuit


http://www.linear.com/solutions/1594

[tszaboo]

I definately prefer them over any heat transfer RMS or those pseudo RMS things from Analog devices.

Why so both ways? I think more detailed reasoning could help others to understand better too.

Well, they are simple to use. You only need to select one capacitor and scale the input voltage to 100-1000mV. For AD's solution you need a bunch of high value, stable capacitor which themselves could cost more than the IC.
It is more accurate, i would say 2-5 times. And it is tiny. Price was also better. The old AD chips are like 20 euros. And then there is the fact, that it is differential input and output. Fits better to a system, and you can possibly avoid a level shifting, decreasing the errors. And it is linear technology.

[MosherIV]

Hi

I am very new to the forum and I am also well out of my depth when it comes to precision stuff but......

Surely any oscillator, now matter how precision, low noise, low thermal drift etc the op-amps and resistors are, the capacitors are the weakest things in the circuit as far as stability goes?

Any oscillator based on RC oscillations is going to drift horribly over time and with temperature - or am I missing something?

Surely, an Xtal or a silicone based oscillator is going to give better temperature and aging stability?

What I am not sure about when using an Xtal is how much the aging of the Xtal is going to affect the amplitude of the oscillation.
As far as I have seen, most silicone oscillators (or DDS) do not quote stability for the parts, so how good or bad are they for creating stable AC waveforms?

I do not understand why 'heat transfer' is relevant to RMS voltage measurement. Heat transfer is a measure of power, not a measure of root mean squared of amplitude.

Could someone explain please?

I have been thinking about how to make an AC voltage standard for hobby use. So knowing the answer to these will be helpful.
Thanks.

[Kleinstein]

The heat transfer units compare the power of an AC voltage to the power of an DC voltage. In the application to stabilize an generator it does not matter if you keep the voltage or the power constant.

There are also quite stable capacitors, so it is possible to build a relatively stable RC oscillator. Not as stable as a crystal one, but still reasonably stable (e.g. 0.1 %) in frequency. In the shown circuit the caps are only responsible for the frequency - the amplitude is set by the extra loop that uses the RMS converter. So the amplitude can stay stable not matter how much the caps are drifting. If you really need a stable frequency too, one could add PLL stabilization of the frequency. As the wien bridge type oscillator gives a good quality sine, there is no real need for a true RMS - simple full wave rectification could do the job as well.

DDS chips might have a specification on how stable the DAC is. Together with the filter, this is what mainly sets the amplitude stability at low frequencies. For low frequencies like 50/60 Hz one might choose a separate DAC and µC/FPGA instead of an integrated DDS chip, and can thus chose a DAC with specified drift and maybe more resolution than typical low end DDS solutions.  Still the filter could be a challenge. So even with a DDS generator one might chose the way of amplitude measurement and control loop.

Crystal drift is usually quite small and thus less of an issue with such an reference voltage source.

[barry14]

The RMS voltage of a waveform, by definition, generates the same power in a resistor as a DC voltage of the same magnitude.  Thus one way to build an RMS voltmeter is to compare the power dissipated in equal resistors by the input voltage and a variable DC voltage. When the power is equal (as measured by the temperature rise for example), then the DC voltage is equal to the RMS voltage of the input waveform.  This method, which was used by Hewlett-Packard, is capable of wide bandwidths and is completely independent of the input waveshape.

[johansen]

The RMS voltage of a waveform, by definition, generates the same power in a resistor as a DC voltage of the same magnitude.  Thus one way to build an RMS voltmeter is to compare the power dissipated in equal resistors by the input voltage and a variable DC voltage. When the power is equal (as measured by the temperature rise for example), then the DC voltage is equal to the RMS voltage of the input waveform.  This method, which was used by Hewlett-Packard, is capable of wide bandwidths and is completely independent of the input waveshape.

almost sounds like something that might arrive in my email, given enough time.

[MosherIV]

Hi

A big thank you to Kleinstein and barry14 for explaining things to me.
I am still not sure that using power to measure RMS voltage is valid but in thinking about it, real world physics mean that even the integral (capacitor reservoir with bleed capacitor) method still takes current and therefore power.

I had a look at how commercial units do it (Fluke 5200a).
https://www.eevblog.com/forum/repair/fluke-5200a-ac-voltage-calibrator-teardown-and-repair/
Sounds like they too use some kind of RC oscillator. Then amplify it and buffer through transformers. Finally rectify back to DC to compare with stable DC reference and use offset to adjust freq and voltage of original AC oscillator.
This is more complicated than I would like to do in build my own AC reference.

My concern with using Xtal is a) thermal and aging drift, even though I have found 10ppm parts
b) how pure is xtal sine wave
c) xtal p-p voltage is small - around 0.6Vp-p - amplifying this up 20 times - will this introduce noise

Will using a silicone based solution (eg the old 8038 oscillator) be as stable as an Xtal?
It will produce larger output voltages. However, it is now obsolete. Not seen any replacement for it.
DDS devices are too complex to use for my liking.
Silicone oscillators all seem to be replacements for Xtal
Looks like it will have to be RC oscillator around an Op-Amp unless anyone has a better idea?

[Kleinstein]

An RC oscillator will never be as stable as a XTAL when it come to frequency, but xtals are only fixed frequencies at about 32 kHz and up. So they are good to get a stable frequency for something like an DDS or PLL or a frequency counter. It depends on the oscillator circuit how pure or not the sine will be. In principle the Xtal can give a pure sine at that one frequency. The low amplitude and thus amplitude noise is not a problem, it's not that small.

The ICL8038 or XR2206  RC oscillators give about a 1% distortion on the sine, so not very pure. Also stability in frequency and amplitude is not that great. The good thing is that it is easy to adjust the frequency over a large range. So for a reference source one would need stabilization in amplitude and maybe the frequency.

The Wien bridge type oscillators are relatively difficult to adjust in frequency and usually need a amplitude stabilization even for basic operation. The sine quality can be really good (usually < 0.1 % distortion). So one would need amplitude stabilization and maybe frequency stabilization (with a PLL referenced to a XTAL).

DDS is very stable in frequency (about as good as the XTAL / clock), good purity (better than 8038 but usually not as good as a good Wien bridge) and the amplitude can be quite stable (depending on the DAC). So one does not need a PLL and may get away without an amplitude regulation.

For higher frequencies (e.g > 10 kHz)  one may need the amplitude feedback or measurement, as it's difficult to impossible to make the amplifier the essentially zero output impedance at higher frequencies - so one would need some way to take the output impedance into account.

For not so high frequencies, there is also the option to do RMS conversion with an ADC sampling the AC waveform and doing the RMS conversion in software. This is how high end DMMs line the 3458 do it. This way can be more stable than analog RMS converts. One might even use relatively cheap chips made for electricity meters for this purpose.

[timb]

Hi

A big thank you to Kleinstein and barry14 for explaining things to me.
I am still not sure that using power to measure RMS voltage is valid but in thinking about it, real world physics mean that even the integral (capacitor reservoir with bleed capacitor) method still takes current and therefore power.

I had a look at how commercial units do it (Fluke 5200a).
https://www.eevblog.com/forum/repair/fluke-5200a-ac-voltage-calibrator-teardown-and-repair/
Sounds like they too use some kind of RC oscillator. Then amplify it and buffer through transformers. Finally rectify back to DC to compare with stable DC reference and use offset to adjust freq and voltage of original AC oscillator.
This is more complicated than I would like to do in build my own AC reference.

My concern with using Xtal is a) thermal and aging drift, even though I have found 10ppm parts
b) how pure is xtal sine wave
c) xtal p-p voltage is small - around 0.6Vp-p - amplifying this up 20 times - will this introduce noise

Will using a silicone based solution (eg the old 8038 oscillator) be as stable as an Xtal?
It will produce larger output voltages. However, it is now obsolete. Not seen any replacement for it.
DDS devices are too complex to use for my liking.
Silicone oscillators all seem to be replacements for Xtal
Looks like it will have to be RC oscillator around an Op-Amp unless anyone has a better idea?

Silicon is a metal/semiconductor, while "silicone" is a compound that contains silicon.  Silicone is used to make oils and rubbers [WD-40 oil, silicone bathtub calk, etc.] and other things, but as far as I know there is no such thing as a "silicone oscillator".  In engineering, words mean things.  If you use a word incorrectly [worst case] you can end up killing someone.  Same thing goes for illegible writing and misplaced decimals, or mixing up the imperial measurement system with the metric system without specifying units.  Please learn to be more precise in your communications.

Now, that said, there are MEMS-based oscillators on silicon chips that rival the stability of crystal oscillators, and they are very rugged.  Crystals can be very delicate and can be damaged by rough handling.  I guarantee you the UPS driver in my area can break a crystal that is mounted on a PC board that is carefully packaged without breaking a sweat.

As for an AC source, the best way I know of at this time in history is to use a DDS chip.  These are made by multiple manufacturers, but Analog Devices seems to have the best ones.  They have a range of devices available.  If the voltage reference used for the output DAC is very good, then the amplitude or your AC sine wave is going to be very stable.  This is followed by a filter to smooth out the waveform and remove harmonics.  Some of these DDS chips can generate 14-bit accurate sine waves, and if followed by a good filter they can have extremely good amplitude linearity [10ppm would be extremely good, 1ppm is achievable, but not by a hobbyist].  These require programming [usually via a serial "SPI" port], so you would need a microcontroller of some sort-- an Arduino could probably handle this just fine.  Analog devices has some evaluation kits available for some of their DDS parts-- and this would help you get started without having to design a PC board.

While the output of the DDS chip+filter will be very stable, any follow-on amplification can and will introduce non-linearity.  This is where the output monitoring and feedback [to the microcontroller] comes in.  While you could use an analog RMS-to-DC converter chip, these are not very accurate-- you would be better off building a AC/DC thermal converter.

DDS is not complicated at all.  In fact it is easier to comprehend and implement a good DDS based waveform generator than it would an analog one with similar specs.  Just get on the Internet, use your favorite search engine, and LEARN.

I have been mulling over an idea for a low-cost multi-junction AC/DC thermal converter that a hobbyist could build.  I just have not had the time to try out the idea.  If I ever do and if it works correctly, then I promise to post the design on this forum.  This would allow sub-ppm AC/DC differences-- on par with NIST.

There's a few good Jim Williams app notes on this subject. He was really into thermal RMS conversion for awhile. In fact, LT actually made a thermal RMS chip! It contained the heater coil/NTC pairs and everything! Shame they no longer make the chip.

[Gyro]

There's a recent thread on thermal converters:

https://www.eevblog.com/forum/metrology/diy-precision-ac-rms-to-dc-transfer-standard/msg920360/#msg920360

[Vgkid]

Since it appears that a lot of meters use the AD637. How well is there short term stability?
I plan on making a pretty stable source using one, a stable V-ref(possibly divided down, or ovenized), and  the classic WBO*.**
That is, if I don't get lucky with a 510A, currently 0 for 2 .
* Other techniques might be implemented, it is the easiest.
** A poor mans fluke 510A...

[David Hess]

You do not need RMS conversion to stabilized the amplitude of a relatively pure sine wave source; rectification and averaging is enough and a bunch of the Linear Technology application notes show examples.  RMS conversion is needed where the source waveform is undefined.

Thermal converters have the advantage of dealing with wide bandwidths and high crest factors and it is a shame Linear Technology discontinued their LT1088 integrated thermal RMS converter.

As far as the accuracy and drift of integrated RMS converters, check the datasheets.

[ManateeMafia]

 I know it defeats the idea of DIY, but if you want to buy a calibrated MJTC for a project, it can be bought from NIST.

https://www-s.nist.gov/srmors/view_detail.cfm?srm=6002a

https://www-s.nist.gov/srmors/view_detail.cfm?srm=6002b

[Vgkid]

I will most likely revisit this thread in the future. So anyone is welcome to post any related info in this thread.
Why is this on hold...
A 510A is in my future.

[Vgkid]

Looking into the spec sheets of both my 3456A, and 34401A. I noticed one thing, that the tempco's of the ac converters are rather large. 0.008%+60Counts (3456A) , and 0.005%+.004%(34401A). Interestingly enough the 34401A uses the ad637J, while the B/K variants are actually specced better.* Will need to look into the 34401A's manual more, I haven't had to fix that meter yet...
Currently both are hooked up to my 510A...

[Kleinstein]

There is more than just the RMS converter chip that influences uncertainty in the AC ranges.  It's also the amplifier / range switching part with frequency compensation parts. Further the specs are more like upper limits, not best estimates or even real estimates on how much the meters are really drifting. Sometimes the engineers are more conservative, as it takes more testing to give confidence to tighter specs, but more testing alone does nothing to improve the meters itself.