How to measure RMS current for non sinusoidal wave forms

[AllTheGearNoIdea]

I have been playing recently with a modified sine wave inverter that converters 12 volts DC to 230 volts AC. Note the output is nothing like a sine wave it’s very much a square wav.  As part of my experiment I plugged in a wall wart type energy monitor and found to my surprise it’s gets very hot internally. The wall wart monitor uses a capacitor dropper circuit and current limiting resistor. It turns out that the sharp rate of change of the voltage from the inverter has a peak current of 1.7 amps as opposed to 80mA when fed with a sine wave from the true mains.  So the capacitor doesn’t limit anything when fed with a square wave and that’s no surprise.   Now I measured the current by using a 10 ohm resistor in series with the power supplied to the neutral of the wall wart.  The current being proportional to the voltage developed across the series resistor. I should say that it’s not really a neutral as it was being sourced via a isolation transformer or through a 12v isolated supply when feeding the inverter.

I also tried to measure the currents supplied to the wall wart monitor plug  using several bench meters both analog and digital. Now all these meters gave wildly different current readings. I assumed that that was because some of the meter were not true RMS types and the internal correction factors were wrong because the current were not a sine wave.  However I do have two true RMS meter. These are a fluke 175 and a BK Precision 5191A. Well the fluke indicates 26mA and the BK Precision just 6mA

So I guess my question is how do you measure true rms for non sinusoidal current pulses. I know how to do it on a scope but it’s messy to try and add up all those nasty little pulses over the cycle time. And when is a true RMS meter not a True RMS meter ?

The current spikes are about  100uS wide  peaking at 1.7 amps at each cycle transition. I’m trying to get a feeling for how much current is being dissipated in the wall warts surge resistor. When being fed with the inverter as opposed to it running totally cold when running from the normal sinusoidal mains supply.

It does appear this modified sine wave inverter are horrible!

Regards Chris

[nctnico]

You have to realise that many 'true RMS' multimeters (especially the handhelds) only work well for 50Hz. Outside that frequency range the error goes up enormously. The easier way for you is to buy a cheap current probe which you can connect to an oscilloscope and use the cycle RMS function. Otherwise you have to buy a very expensive DMM (probably a bench DMM because most handhelds I have come across don't have a wide frequency range).

[AllTheGearNoIdea]

Yes I think your correct I believe the BK Precision is 100khz from memory the flukes just a few hundred Hz. I think I have an old Tektronix oscilloscope clamp probe hiding somewhere.  I think I can measure the RMS cycle on my old Tektronix TDS 2022 via the series resistor if I am careful.  Great post thanks

[joeqsmith]

I agree, that squarewave adds a bit of a snag. 

I have looked at some fairly low end meters with fairly high bandwidths that may do a decent job.    Strange the BK would read lower than the Fluke with a higher BW.  Something is going on there.

Not what you want to hear but I tried using my scope to measure power with a LEM sensor and a fairly odd transformer I had.  I use a computer to run the calculations.  Works but as you mention, its a bit messy.     

https://youtu.be/04I7nHA_HxM?list=PLZSS2ajxhiQBcHhIaGpmm9GyZQfrCzqkv

[AllTheGearNoIdea]

Just had a quick try using the rms cycle function on the scope and I think maybe it’s have trouble identifying a cycle or it could be that inverter is constantly modifying its output PWM and confusing the triggering.  It’s funny never tried to measure current for a square wave before it’s proving surprisingly difficult for me.

[Fungus]

Just had a quick try using the rms cycle function on the scope and I think maybe it’s have trouble identifying a cycle or it could be that inverter is constantly modifying its output PWM and confusing the triggering.  It’s funny never tried to measure current for a square wave before it’s proving surprisingly difficult for me.

What "scope"? Often the "RMS" measurement isn't what you think it is.

The one you might want is called Periodic RMS (or PRMS or something like that).

[AllTheGearNoIdea]

Just had a quick try using the rms cycle function on the scope and I think maybe it’s have trouble identifying a cycle or it could be that inverter is constantly modifying its output PWM and confusing the triggering.  It’s funny never tried to measure current for a square wave before it’s proving surprisingly difficult for me.

What "scope"? Often the "RMS" measurement isn't what you think it is.

The one you might want is called Periodic RMS (or PRMS or something like that).

Is an old Tektronix TDS 2022. It only list a cycle RMS option under the measure menu

[nctnico]

Perhaps you can show us the waveform you are trying to analyse/measure.

[AllTheGearNoIdea]

Yes quite right. I will post a video tomorrow. Drank to much grape juice to play with the mains tonight.

[Electro Detective]

AllTheGearNoIdea is not back yet to update us on the mystery frequency that's playing tricks on his current measurements.   

Most likely something way beyond most multimeters AC current bandwidth,
the Fluke 175 probably claps out at 500hz to 1000hz as do most decent RMS meters


OT? I wonder what that grape juice was?        I want some   

[HalFET]

Keithley 194 high speed voltmeters can do this up to about  500 kHz (if I remember correctly) if you give them a suitable series resistor to work with. But they're a rare piece of kit with Keithley style pricing...

An easier way if you can live with 5% error is a scope + differential probe.

[fcb]

If you're doing this professionally (which it doesn't sound like you are) - then use a power analyser (Yokogawa, Newtons4th, ZES Zimmer, Norma, etc..) - expensive pieces of kit, but pretty good.

For hobby, then probably I would build a resistive current shunt, (differential probe) and use a scope (isolation...?). Download the waveform from the scope and work out the area 'under the graph'.

If you need precision, can't afford a power analyser then probably look at the MCP3903 (we used this in a clients project, when we needed to build a basic power analyser into their product).

[Alex Nikitin]

when is a true RMS meter not a True RMS meter ?

The current spikes are about  100uS wide  peaking at 1.7 amps at each cycle transition. I’m trying to get a feeling for how much current is being dissipated in the wall warts surge resistor. When being fed with the inverter as opposed to it running totally cold when running from the normal sinusoidal mains supply.

In your case the crest factor is too large for your RMS meters (about 100, most "True RMS" meters can not handle anything over about 5 , at best 10) so these can not measure the current accurately. From your data my estimate would be about 130mA RMS (for 50Hz inverter frequency). To measure this kind of current reasonably accurately you may try to use a scope or a proper RMS meter with a thermal converter. Or you can stick a thermocouple on the resistor in your wall wart monitor and measure it's temperature rise when connected to your inverter and then try to get the same temperature rise with a DC current through the resistor (assuming the circuit allows it).

Cheers

Alex

[capt bullshot]

In your case the crest factor is too large for your RMS meters (about 100, most "True RMS" meters can not handle anything over about 5 , at best 10) so these can not measure the current accurately.

Excactly this is the OP's problem. I wouldn't expect a reasonable result from any True RMS meter at all with this kind of waveform. RMS measurement with a scope may work, a suitable thermal converter would work. The LT1088 comes to my mind, but you'd still have to bring up some circuitry or a wide BW current clamp to feed the signal into it.

[David Hess]

In your case the crest factor is too large for your RMS meters (about 100, most "True RMS" meters can not handle anything over about 5 , at best 10) so these can not measure the current accurately.

Excactly this is the OP's problem. I wouldn't expect a reasonable result from any True RMS meter at all with this kind of waveform. RMS measurement with a scope may work, a suitable thermal converter would work. The LT1088 comes to my mind, but you'd still have to bring up some circuitry or a wide BW current clamp to feed the signal into it.

Unfortunately Linear Technology stopped producing the LT1088 many years ago.

[dj831]

Hi, in my opinion, as said earlier, you have only two choices:

• Use a THERMAL RMS meter. Unfortunately, they are mainly found on auction sites, and may need repair. Use for example a Fluke 8506a, 8920 - or something else from HP.
• Use a digital oscilloscope, and try to find one that computes RMS.

Just FYI, using an LT1088 is not that simple, and may be easily blown is not correctly protected. I purchased a pair from a German seller a year ago on that famous aution site (c.a. 25€ each). I think he still has some to sell, since he had a bunch of them. These circuits seemed genuine, since resistance of sensors matched datasheet. I can't tell you more since I have not used them.

[David Hess]

If the crest factor is high, then a higher range may be used sacrificing resolution and settling time.  One way to detect errors do to high crest factor is inconsistency between the nominal and higher input ranges.

Most handheld multimeters still use an integrated analog computing RMS converter like those from Analog Devices.  But even back in the 1990s, some did it through sampling which may have advantages including higher performance and detection of excessive crest factor.

For high crest factor and high bandwidth, a DSO that can calculate gated or windowed RMS and not only cycle-RMS is the ideal way to go.  The calculation is trivial, it is just the standard deviation, however be warned that some DSOs including the Rigol DS1000Z series do not calculate RMS correctly because they operate on the display record which lacks the needed information.

Back when I was calibrating oscilloscopes without a calibrated peak-to-peak source, I used my best AC voltmeters to calibrate the level of a low frequency, roughly 200 Hz, square wave.  I was very pleased when my Tektronix DMM916 returned identical RMS AC and corrected average AC measurements to within 1 count.  I would not mind finding another multimeter which can measure average AC and RMS separately.

[nctnico]

For high crest factor and high bandwidth, a DSO that can calculate gated or windowed RMS and not only cycle-RMS is the ideal way to go.  The calculation is trivial, it is just the standard deviation, however be warned that some DSOs including the Rigol DS1000Z series do not calculate RMS correctly because they operate on the display record which lacks the needed information.

As long as a cycle is clearly visibly then you should get the proper RMS value. Some have been making a lot of fuss about doing measurements on the on-screen data only but for the majority (if not all) of the measurements this is not really an issue because if you can't see the shape of the waveform you have no idea what the measurement result is based on. Taking numbers at face value isn't a good approach.

[David Hess]

For high crest factor and high bandwidth, a DSO that can calculate gated or windowed RMS and not only cycle-RMS is the ideal way to go.  The calculation is trivial, it is just the standard deviation, however be warned that some DSOs including the Rigol DS1000Z series do not calculate RMS correctly because they operate on the display record which lacks the needed information.

As long as a cycle is clearly visibly then you should get the proper RMS value. Some have been making a lot of fuss about doing measurements on the on-screen data only but for the majority (if not all) of the measurements this is not really an issue because if you can't see the shape of the waveform you have no idea what the measurement result is based on. Taking numbers at face value isn't a good approach.

I would not take it for granted that a DSO correctly selects a cycle on a noisy waveform.  Some DSOs can automatically show the cycle endpoints they are using for the calculations but having the ability to make a gated or windowed RMS measurement always works.

The other way is simply to use a slower time/div setting so that the measurement is made over multiple cycles.  This attenuates the error from including a partial cycle so the cycle endpoints, knowing the shape of the waveform, or even triggering becomes irrelevant.  Unfortunately, this is exactly the type of RMS measurement which will not work on a display record because the processing corrupts the standard deviation (except in the trivial case of straight decimation which "advanced" oscilloscopes do not do) so a lot of modern DSOs cannot make an accurate RMS measurement this way.