How do old (25+ year) digital multimeters measure RMS voltages?

[e100]

... and can the measurements be trusted when the signal has a bit of odd harmonic distortion?

I have an ancient "Maplin Precision Gold M-4510" multimeter which is similar to the 5010 shown here
http://www.petervis.com/meters/maplin-precision-gold-m-5010ec/maplin-precision-gold.html

It has been working great for all these years when doing DC measurements but now I'm trying to use it to verify the output from a microcontroller which is doing true RMS measurements from an isolation transformer stepped down mains voltage. The measurements are broadly in agreement, but the multimeter reading is suspiciously stable (1 part in a 1000) over a period of 30 seconds. Is that to be expected for mains voltage measurements in an urban environment? The micro is giving readings that fluctuate by about 2 parts in a 1000 over a 1 second period so I don't know which one is telling the truth.

The web page for the 5010 references an Intersil chip
http://www.intersil.com/content/dam/Intersil/documents/icl7/icl7106-07-07s.pdf
which only mentions "RMS" once.

Do these things simply multiply the measured peak by 0.707 to get an approximate RMS reading, or are they doing something else?


 

[nidlaX]

I could be wrong, but unless the multimeter included a true RMS converter, then it looks like the Intersil chip will just use a conversion factor as you suspect.

[Macbeth]

It's not actually calculating an RMS from the peak x 0.707. It's rectifying and smoothing the DC, much like a power supplies bridge rectifier and RC filter. This average DC value is what the meter will display but with an adjustment to scale it to RMS. This will be calibrated with a trimmer so a typical mains derived 50/60Hz sinusoidal voltage will read correctly.

[TiN]

e100

Suggest to be more specific, as you may know that most of bench system DMMs are 25+ year old as well. Fluke 8506A is using thermal convertors to measure ACV. 30+ year old 3458A using fast digitizing + AC conversion. Keithley 2001/2002's use AD637 + separate floating full-wave rectifier circuitry to enhance accuracy and frequency range.

[Dr. Frank]

Look at this AD536, True RMS Converter, salvaged from my first DIY kit 4 1/2 digit DMM, built when I just left school in 1980, obviously.
These chips really square the voltage input, average, and draw take the root, so it's a real analog calculator for the integral of the input signal.



Geeze, that's been 36 years already! I'm vintage also.. 

True RMS DMMs could be recognized by a big "TRMS" sign on them, so yours obviously measures averaged AC volt only.


(The ICL 7106 is the 3 1/2 digits A/D converter and display driver only) 

When I worked in this Airforce calibration lab, also around 1980, we had many of these Fluke 6 1/2 digits state-of-the-art DMMs, I think, already the Fluke 8506A with their proprietary thermal RMS converter inside.
 
Frank

PS: To answer your question, even if your DMM only measures the average, and will not give the correct absolute RMS value, it will anyhow display any changes of the line voltage.
Maybe your µP TRMS device is not stable; I would test that with a stable sinus signal, as higher harmonics may make a difference.

[Kleinstein]

The method of sampling an digital RMS calculation works really well. So even with an 12 Bit ADC it might rival the analog RMS converters. So I am not so sure blaming the µC solution. It could be the waveform changing or the averaging time not exactly at one cycle with either the DMM or the µC.

Even the poor quality boards dielectric will be pretty linear - the problem is that they are frequency dependent and can also change with humidity and not all boards will behave the same, as the material is not that homogeneous.

[Jay_Diddy_B]

Hi group,

In the Fluke 8050A and the 8010A, maybe others, Fluke has a circuit built on a ceramic substrate using a quad op-amp and a transistor array LM3046. The circuit is an analog computer that computes the RMS value.



Here is a picture of the hybrid:




I am trying to find the circuit ...


Regards,

Jay_Diddy_B

[Dr. Frank]

After my military service, in 1982, I visited the PTB department in Braunschweig, where my Airforce calibration unit had close connections to.

This guy at PTB was designing by hand (the first ever) multi junction thermo couples. These looked relatively simple, and it seemed not so difficult to build.

So, having the right equipment, it should be possible to copy that DIY.

Frank

[coppice]

The method of sampling an digital RMS calculation works really well. So even with an 12 Bit ADC it might rival the analog RMS converters. So I am not so sure blaming the µC solution. It could be the waveform changing or the averaging time not exactly at one cycle with either the DMM or the µC.

Even the poor quality boards dielectric will be pretty linear - the problem is that they are frequency dependent and can also change with humidity and not all boards will behave the same, as the material is not that homogeneous.

Using an ADC to sample the waveform and computing the RMS digitally will generally beat the accuracy of analogue methods. Just make sure you sample fast enough to meet your bandwidth requirements. That's no problem if you are only measuring mains power, but the true RMS modes of most multimeters claim 50kHz or 100kHz bandwidth.

[Dr. Frank]

The method of sampling an digital RMS calculation works really well. So even with an 12 Bit ADC it might rival the analog RMS converters. So I am not so sure blaming the µC solution. It could be the waveform changing or the averaging time not exactly at one cycle with either the DMM or the µC.

Even the poor quality boards dielectric will be pretty linear - the problem is that they are frequency dependent and can also change with humidity and not all boards will behave the same, as the material is not that homogeneous.

Using an ADC to sample the waveform and computing the RMS digitally will generally beat the accuracy of analogue methods. Just make sure you sample fast enough to meet your bandwidth requirements. That's no problem if you are only measuring mains power, but the true RMS modes of most multimeters claim 50kHz or 100kHz bandwidth.

The agilent 34410/411, and the Keysight 3446x/470A DMMs all use a digitizing technique for AC measurement, working up to 300kHz, and 0.02% best accuracy.
The 1989 vintage HP3458A, using its ACV SYNC mode, already implemented this, up to 10MHz (15%), and best accuracy of 0.007% between 40Hz and 1kHz.
Using Swerleins ACV Algorithm further improves uncertainty to less than 0.002%, at frequencies up to 1kHz.

Frank

[David Hess]

The big advantages of the integrated RMS converters like from Analog Devices over discrete implementations with matched transistors are specified "no-adjustment" accuracy and ultimately greater economy; I doubt the performance of the integrated design can be duplicated with a discrete design for the same price.  A discrete design using integrated matched transistors can achieve the same result but is more complicated and requires adjustment.  A thermister compensated design is almost as good.  You can find examples in the old National Semiconductor (and Analog Devices?) application notes.

It is too bad Linear Technology discontinued their integrated thermal RMS chip.  That thing was magic.

Even the poor quality boards dielectric will be pretty linear - the problem is that they are frequency dependent and can also change with humidity and not all boards will behave the same, as the material is not that homogeneous.

Back in 1978, Tektronix called this "hook":

https://www.eevblog.com/forum/projects/getting-rid-of-'hook'-and-hidden-capacitance/

Using an ADC to sample the waveform and computing the RMS digitally will generally beat the accuracy of analogue methods. Just make sure you sample fast enough to meet your bandwidth requirements. That's no problem if you are only measuring mains power, but the true RMS modes of most multimeters claim 50kHz or 100kHz bandwidth.

There is no need for the Nyquist bandwidth to be higher than the signal bandwidth although this is easy to do with power line measurements; aliasing will not affect the result and undersampling has the effect of lowering the certainty for a given measurement time.  The sampling bandwidth is very important of course.  RF sampling RMS voltmeters take advantage of this to produce good accuracy into the GHz range while sampling at 10s of KHz.

[coppice]

The agilent 34410/411, and the Keysight 3446x/470A DMMs all use a digitizing technique for AC measurement, working up to 300kHz, and 0.02% best accuracy.
The 1989 vintage HP3458A, using its ACV SYNC mode, already implemented this, up to 10MHz (15%), and best accuracy of 0.007% between 40Hz and 1kHz.
Using Swerleins ACV Algorithm further improves uncertainty to less than 0.002%, at frequencies up to 1kHz.

Swerlein's Algorithm only works for sine waves, or close approximations to since waves, so it doesn't really measure RMS.

[acbern]

Dr Frank:
You mentioned somewhere you use the Swerleins algorithm. Is there a simple way to do this (I do not want to reinvent the wheel and do programming or so on this one but rather buy old computer gear if need be if there is a readily available code somwehere). Any hints appreciated.
And what about the CREST factor / THD acceptable to the algorithm?
Thanks

[coppice]

Dr Frank:
You mentioned somewhere you use the Swerleins algorithm. Is there a simple way to do this (I do not want to reinvent the wheel and do programming or so on this one but rather buy old computer gear if need be if there is a readily available code somwehere). Any hints appreciated.
And what about the CREST factor / THD acceptable to the algorithm?
Thanks

If you are interested in crest factor and THD, Swerlein's algorithm is not for you. You need a pure sine wave for it to work. If signal exceeds 1% THD or so it starts to break down. This is spelled out in the documentation for the algorithm.

[coppice]

Using an ADC to sample the waveform and computing the RMS digitally will generally beat the accuracy of analogue methods. Just make sure you sample fast enough to meet your bandwidth requirements. That's no problem if you are only measuring mains power, but the true RMS modes of most multimeters claim 50kHz or 100kHz bandwidth.

There is no need for the Nyquist bandwidth to be higher than the signal bandwidth although this is easy to do with power line measurements; aliasing will not affect the result and undersampling has the effect of lowering the certainty for a given measurement time.  The sampling bandwidth is very important of course.  RF sampling RMS voltmeters take advantage of this to produce good accuracy into the GHz range while sampling at 10s of KHz.

Note that although you only need to sample at greater than twice the bandwidth to capture the content of the signal precise, you need to use a sufficiently narrow sampling aperture. Of course, the sampling theorem says you always need to use a Dirac delta sampler, but in practice we can only approximate that. To sub-sample you need to use a sampling aperture consistent with being a fair approximation of a Dirac sampler at the actual frequency of the signal.

[Dr. Frank]

Dr Frank:
You mentioned somewhere you use the Swerleins algorithm. Is there a simple way to do this (I do not want to reinvent the wheel and do programming or so on this one but rather buy old computer gear if need be if there is a readily available code somwehere). Any hints appreciated.
And what about the CREST factor / THD acceptable to the algorithm?
Thanks

Correct, I forgot, that the original Swerlein Algorithm is based on a pure sine wave only; higher harmonics will give errors, specified in the description of the algorithm, for up to 1% distortion.
In fact, I use this in conjunction with my Fluke 5200A, which has very low distortion (max. 0.041%).

I also found some other publications, which were extensions to Swerlein, where harmonics were included, but you have to program this on your own.

The source of the Swerlein Algorithm for the 3458A is available on Keysight site, and there's also an executable which runs directly on modern  PCs (even on W10), with N.I. compatible GPIB cards.
That's currently not visible on Keysight.com, but if you are interested, I can send it to you.

Frank

[TiN]

I can see related page on Keysight site.
If you have no access to the page and linked files by some reason, I mirrored it here.

[Dr. Frank]

I can see related page on Keysight site.
If you have no access to the page and linked files by some reason, I mirrored it here.

Nope, there's an additional zip, which consists of an installation file. This sets up the Swerlein program as an executable VB program (.exe)

Frank

[branadic]

After my military service, in 1982, I visited the PTB department in Braunschweig, where my Airforce calibration unit had close connections to.

This guy at PTB was designing by hand (the first ever) multi junction thermo couples. These looked relatively simple, and it seemed not so difficult to build.

So, having the right equipment, it should be possible to copy that DIY.

Are you refering to something like this? http://www.nist.gov/pml/div684/grp06/acdc.cfm

There are other converter beside LT available like: http://www.nikkohm.com/resistor_thermopile.htm