How do multimeters measure AC voltage?

[Stray Electron]

Thanks!
Actually, I'm not going to have AC measurements at this moment at all, because I don't need it. I just asked, in case I will do it, what efforts I need to put into

For the multimeter which does not exist, please suggest me one like Keithley 179 - direct selection of modes via dedicated button for each function, no auto ranging, no rotary switch, no multi-function buttons, but big, bright LED display, pocket sized, and fast screen update time - at least 15 FPS. There are none. Maybe someone say that no one needs, but I don't care for other's needs, so I'm making one for myself

   It's not pocket sized but go check out the Keithley 194A.  They're FAST and sometimes sell for peanuts on FleaBay. The autoranging can be turned off and I think it meets all of your other specifications.   Download the manual and read it, you can learn a lot from it. 

[Siwastaja]

A pretty credible multimeter could be made now using an inexpensive sampling ADC and microcontroller, but offhand I do not know of any like this.

Indeed, digital multimeters are still 1980's legacy where every component is penny-pinched. It's pretty interesting that the component BOM (not including case, fuse, etc.) for a $50 multimeter is still like $2-3 (5% of retail price), while the component BOM for a $300 oscilloscope is easily $100 (30% of retail price) or so. Talk about bang for buck!

One could assume that in $((2011+11+1)), more advanced multimeters with <1MHz oscillosscope-like features such as waveform display, min/avg/max/rms statistics etc. would be commonplace even in sub-$50 class (given the triviality and low cost of doing that), but nope, we are still discussing about the freaking true-RMS functionality (which was a big thing in early 1990's.) as a decision factor between cheap and expensive multimeters.

[Zero999]

One more point, in stating he is considering a 200 Hz sample rate for 50 or 60 Hz AC I think he is depending far too much on the Nyquist–Shannon sampling theorem.

If true RMS is really needed, just how does four samples provide any accuracy if the waveform is not going to be a sine wave?

SImplfying David Hess' point. It doesn't mattter if the sample rate is below the half the maximum frequency of interest. The waveform is repetative so all that's required is enough samples to give a decent level of accuracy. Obviouly it can't take too many samples, if the rate is low, because the display will refresh slowly..

[dobsonr741]

One could assume that in $((2011+1+1)), more advanced multimeters with <1MHz oscillosscope-like features such as waveform display, min/avg/max/rms statistics etc. would be commonplace even in sub-$50 class

I want to believe in progress. But when I look back how do I use my multimeter:
Is this battery dead?
Is this supply rail OK?
Is there a short?
Is this resistor really 10K?

Then I do not see the need for progress, neither do the manufactures.

[Kleinstein]

The digital way RMS is not necessay with a sampling ADC. One can as well use a relatively fast SD ADC chip. There are plenty of such solutions for power metering, AFAIK usually with an SD ADC.
There are some relatively cheap handheld DMMs that now include RMS from digital calculations as part of the chip, though usually with a rather limited bandwidth (e.g. ~ 1 kHz for the UT133A ) as they tend to use a low power SD ADC.   For practical use this still has the nice point of fast respond, a big step up from a sluggish analog RMS chip.

Of cause the cheapest solution is likely a µC internal ADC. Even a 10 bit ADC can give reasonable acceptable results. I have tried such a solution with an AVR and was surprized how well it works.

In most aspects the DMMs are mature.
A point where one may want improvements is with a DMM is the case of  a mixed DC+ AC signal. It is a shame that some meters (inlcuding some otherwise good ones) may show total nonsense with rectified half waves. So a display with DC , min and max (with a reasonable BW) and maybe RMS would be great.

[LinuxHata]

There's a new trend of multimeters coming, I'm hoping to get sample by September, btw
These look like smartphones, they run android, have large touch screen and a lot of things, like display font size, color, etc. can be customized.
And since they run stock android, you can have almost any app for any kind of statistical analysis or whatever....

[David Hess]

Indeed, digital multimeters are still 1980's legacy where every component is penny-pinched. It's pretty interesting that the component BOM (not including case, fuse, etc.) for a $50 multimeter is still like $2-3 (5% of retail price), while the component BOM for a $300 oscilloscope is easily $100 (30% of retail price) or so. Talk about bang for buck!

The only change I have seen is the replacement of integrator based integrating ADCs with moderate resolution integrating delta-sigma ADCs, but the later can be implemented incredibly inexpensively on a digital CMOS process.

One could assume that in $((2011+11+1)), more advanced multimeters with <1MHz oscillosscope-like features such as waveform display, min/avg/max/rms statistics etc. would be commonplace even in sub-$50 class (given the triviality and low cost of doing that), but nope, we are still discussing about the freaking true-RMS functionality (which was a big thing in early 1990's.) as a decision factor between cheap and expensive multimeters.

The thing which amazes me is that they are still using analog RMS conversion or averaging, but I guess that is still less expensive than a sampling ADC and DSP.

One more point, in stating he is considering a 200 Hz sample rate for 50 or 60 Hz AC I think he is depending far too much on the Nyquist–Shannon sampling theorem.

If true RMS is really needed, just how does four samples provide any accuracy if the waveform is not going to be a sine wave?

SImplfying David Hess' point. It doesn't mattter if the sample rate is below the half the maximum frequency of interest. The waveform is repetative so all that's required is enough samples to give a decent level of accuracy. Obviouly it can't take too many samples, if the rate is low, because the display will refresh slowly..

Just to be clear, the RMS is calculated from the amplitude histogram, and the amplitude histogram is not affected by undersampling.  A repetitive signal is not required, but enough sample points are required to limit statistical error.

The digital way RMS is not necessay with a sampling ADC. One can as well use a relatively fast SD ADC chip. There are plenty of such solutions for power metering, AFAIK usually with an SD ADC.

There are some relatively cheap handheld DMMs that now include RMS from digital calculations as part of the chip, though usually with a rather limited bandwidth (e.g. ~ 1 kHz for the UT133A ) as they tend to use a low power SD ADC.   For practical use this still has the nice point of fast respond, a big step up from a sluggish analog RMS chip.

I knew some modern multimeters were using moderate resolution delta-sigma converters, but not that any were using digital RMS conversion.

Of cause the cheapest solution is likely a µC internal ADC. Even a 10 bit ADC can give reasonable acceptable results. I have tried such a solution with an AVR and was surprized how well it works.

I have had bad results doing that because of poor linearity so I rarely even consider it.  I would expect a microcontroller's 10 bit ADC to be 8 bits linear, so not even 2-1/2 digits.

A point where one may want improvements is with a DMM is the case of  a mixed DC+ AC signal. It is a shame that some meters (inlcuding some otherwise good ones) may show total nonsense with rectified half waves. So a display with DC , min and max (with a reasonable BW) and maybe RMS would be great.

My old Tektronix DMM916 measures AC average, AC RMS, and AC-DC RMS just fine.  I would love to find a modern multimeter which is just as good.  Its only flaw is an awful backlight.  APPA might have one but they are unavailable in the US.

Currently I am looking for a modern replacement for my Beckman RMS225.  Ampro has some contenders but their accuracy specifications are suspect.

[EPAIII]

Well, if you are going to be that way about it, may I suggest that the easiest way may be to buy a meter with manual controls (rotary switch) and modify it with your features. That will save a lot of work in the actual metering circuits.

Thanks!
Actually, I'm not going to have AC measurements at this moment at all, because I don't need it. I just asked, in case I will do it, what efforts I need to put into

For the multimeter which does not exist, please suggest me one like Keithley 179 - direct selection of modes via dedicated button for each function, no auto ranging, no rotary switch, no multi-function buttons, but big, bright LED display, pocket sized, and fast screen update time - at least 15 FPS. There are none. Maybe someone say that no one needs, but I don't care for other's needs, so I'm making one for myself

[dobsonr741]

I tried to search ADCs on DigiKey, and most that come up, are with that built-in filter. Can you suggest any moderately priced, fast enough ADC, with SPI or I2C interface?

If OP still wants to build a DMM, instead of waiting for the Android-ish one, two options, available on Analog.com:

Sigma-delta, +/-20,000 count:
https://www.analog.com/en/products/max1499.html $15.89

Dual Slope, +/-40,000 count:
https://www.analog.com/en/products/max134.html  $22.16

Might get a sample, too.

[LinuxHata]

Yes sure, I'm still building
my prototype actually works, using built-in ADC of PIC18F46K20...

[LinuxHata]

These maxim chips do look interesting but besides the price, they have slow read speed and there are even some trimpots, which suggests me that these are from the times when Reagan still was the president

My current setup consists of the PIC microcontroller, mentioned above, connected via TM1629A (LED display driver) to LED display, using built-in ADC with internal reference and TP181 as input amplifier. All this is on breadboard with a lot of jumper wires and definitely not worth to show I have ordered some goodies on LCSC, such as more precise voltage reference (ADR5040) and precision, 20 bit external ADC (CS1180), while "normal" PCB is being designed right now.  Regarding the BOM, all these together will cost less than that single Maxim IC

There will be built-in lithium battery, Type C USB port for charging (isolated from multimeter circuitry)

[dobsonr741]

How many digits on the display? My measure for the home built DMM is to have a stable 0 LSB display with the input shorted.

When I did that I had to go to a 32 bit sigma-delta (ADS1263) and several iterations on the auto zero on the frontend. So I ended up with a 5 1/2 DVM, with 150mV lowest range, and a stable LSB of 1uV.

[LinuxHata]

It has 4 digit 7 segment display for value and 2 digit 16 segment display for V, A, Ko, mA, Mo display.
There are also 15 buttons for directly selecting the measurement type/range and 15 leds for displaying what is currently selected.

[dobsonr741]

What analog front-end and protections are you planning to use?

[Kleinstein]

The mentioned maxim ADCs are relatively expensive. If one wants a low cost dual slope there are still some ICL7135 around, but still a bit outdated as this chip is directly display oriented. So for old style analog adjustment and not for digital calibration constants.

For 4 digits the µC internal ADC is not really enough for DC, but there are plenty of cheaper more µC oriented SD ADC chips, like the mentioned CS1180.

[uliano]

For 4 digits the µC internal ADC is not really enough for DC, but there are plenty of cheaper more µC oriented SD ADC chips, like the mentioned CS1180.

Can you expand on this?

I would also be interested in cheap DACs

[Kleinstein]

The µC internal ADCs are usual 10 or 12 bit (1024 or 4096 counts) and this is just not enough to get 4.5 digit (e.g. +-20000 counts), even with oversampling. Oversampling can help with noise but hardly with linearity. For AC use one gets usually averaging over multiple codes and linearity is less of an issue. The alternative analog RMS converters have a somewhat limited linearity.
There are relatively low cost SD ADC chips like the mentioned CS1180,  MCP3421, ADS1112 or some LTC24xx variants that are good enough for 4.5  digits in combination with a µC to do the scale factors in software. This is different from the old style display oriented chips that directly map ADC counts to display counts and are made for adjustment on the analog side (e.g. trimmers).
The SD ADC chips may however need some extra input amplifier / buffer as a 10 M resistance divider can be quite high in impedance. The old DMM chips like ICL7135 usually also inlude the amplifier/buffer and not just the ADC.

With 1 switch per range one would likely do the range switching old style with the mechanical switches, which quite a bit simpler than µC controlled switching for autoranging. This is at least if one does not need a higher CAT rating / high voltage rating.

I have build a similar DMM circuit with MCP3421 ADC for DC and µC internal ADC for AC and 2 rotary switches for the ranges (1x voltage, 1 x current) - if interested I could share more details / pictures.

[uliano]

Any "low cost" DAC?

I'm trying to build a "measuring" PSU, so I need  also a couple of 14-16bit DAC for setpoints, but they all seem to be in a different price class than the CS1180

[MrAl]

Hello.

I'm building my own multimeter with some specific to myself features.
I'm not exactly sure, whenever I need AC measurements at all, but I'd like to ask about it, to determine, whenever I'm moving into correct direction.

To measure AC, we have two ways - rectify input voltage/current and measure resulting DC.
Another way is to sample input voltage with high resolution (say, 200 samples/sec), and, by measuring and comparing the lowest and highest peaks, calculate AC voltage based on that.

The first method, rectification, will introduce significant errors, due to non-linearity of diode volt-ampere characteristics. But as I googled, there are some special ICs, like AD737, which do TRMS to DC conversion.

2nd method, which is more modern, should work better, but I see a little issue here with existing ADCs - since most time, we well be measuring either 50hz or 60hz AC voltage, most ADCs have built-in filter, which does 50/60hz rejection, so this method will not work.

So my question is, if I decide to do AC measurements in my DVM, which path should I follow? which will more price-performance effective?

Hi,

Yes, i remember the old days of purely analog meters that used diodes to rectify the AC into DC and then average it.  They had a factor built in to convert to RMS, but either it was a sine wave, or you had to know the waveform and apply a "crest factor" for that particular waveform in order to get the RMS value.
To make up for the diode nonlinearity they used a nonlinear scale on the face of the meter so it would work down to lower voltages.  Nothing like we have today though, but then that is the world of pure analog movements.

For what it is worth, RMS stands for the "ROOT of the MEAN of the SQUARE" which in the continuous math domain simply means first you take the square of the voltage, then take the mean of that, then take the square root of that, and that is the RMS value regardless of what the waveform is assuming you have the required measurement accuracy.

Mathematically it looks like this:
Vrms=sqrt(integrate(v^2,t,0,Tp)/Tp)
The v^2 is taking the square of the voltage which is really v(t), the integration is to calculate the mean, and the square root is just taking the square root of the result.  The integration is performed over the interval 0 to Tp with Tp being the period.
That is the theoretical calculation.

[David Hess]

The SD ADC chips may however need some extra input amplifier / buffer as a 10 M resistance divider can be quite high in impedance. The old DMM chips like ICL7135 usually also inlude the amplifier/buffer and not just the ADC.

The difficulty of adding a 10 megohm buffer to a high resolution delta-sigma ADC should not be underestimated.  Precision low input bias current operational amplifiers with sufficient common mode rejection to preserve the linearity of a high resolution delta-sigma converter are rare to non-existent. (1)

The ICL7135 got away with it by correcting the CMOS buffer's common mode error as part of the automatic zero cycle.  The same thing can be done with a delta-sigma converter by measuring the input offset voltage of the buffer *at the common mode voltage* which is the input voltage instead of ground or some other voltage, which is what the ICL7135 does.

I think this problem could also be solved by bootstrapping the input buffer to follow the input voltage, thereby increasing the common mode rejection, but I have only seen this done in electrometer circuits where the common mode voltage is way beyond the input range of the high impedance buffer.

(1) There are some really good modern parts now though like the OPA140.