Simple DIY DVM Voltage Reference

[Nick]

My pet hate with DMMs being misleading is the variable interpretation of "RMS".

To most folk, the concept of RMS voltage should be straight forward, but most DMMs are AC coupled on RMS ranges, so ignore any DC component.

Tektronix used the term "True RMS" to indicate a) that they were computing an RMS value by integrating the waveform, not by just assuming that its a sine wave, and b) they allowed you to either include or ignore any DC offset.

Almost all cheap DMMs do not compute a REAL RMS value - they ignore any DC and make assumptions about the waveform. The Tektronix DMM 916 (which Mark was referring to) does all of these correctly, as I suspect will decent Flukes. However, you pay for that

[ppppenguin]

My Fluke 87 computes true RMS but I don't think it measures the DC component. It would be useful to have an AC/DC coupling switch as on a scope.

Then there's form factor and frequency to consider. All RMS instruments will have an upper frequency limit which is typically in the audio range. For the same reason they will tend to underread spiky waveforms (high form factor).

Traditionally, meters were calibrated in RMS but measured the mean. Hence they were only correct for a sine wave. A square wave would overread by 11%. Most lowe cost DMMs are also like this.

[Mark Hennessy]

Yes, AC is difficult.

Cheap multimeters use an average-responding circuit; the posh ones ("TrueRMS") use what is effectively a miniature analogue computer to calculate RMS. The classic AD536 was the IC of choice for many years: http://www.analog.com/media/en/technical...AD536A.pdf

Detection is separate from AC+DC measurements. The Fluke 187/8 and 287/9 can do this - most don't (I'll check the 87V at the weekend). Bandwidth is important - but knowing the bandwidth of your instrument is the critical thing.

My best AC meter is a Fluke 8920A - this uses a thermal converter, and has a bandwidth of 20MHz. The crest factor is up to 7 - double most good non-thermal DVMs. It does AC+DC, and dBm - so it's ideal for me. It's a complex beast, but most of the complexity is there to prevent the thermal sensor being overloaded. Funnily enough, despite all the goodness, the best accuracy at AC is only 0.5%. It's only a 2000-count meter, but no point having more with that accuracy. BTW, I recommend these. Because they are old and crusty, and a bench meter form factor, they don't seem to fetch huge amounts - I paid just over £30 for mine.

[Nick]

Thermal converters are OK, but have two serious drawbacks:

1. They are fairly inaccurate (powerin->heater->thermocouple/thermistor->reading).
   
2. They draw significant power from the DUT
   

On the plus side:

1. They are simple
   
2. Handle any waveform
   
3. Very wide bandwidth
   

So they are ok for power measurement, but of little use for small signal work.

[pwdrive]

Out of interest his was before semiconductors (I think)

http://conradhoffman.com/stdcell.htm

Lawrence.

[Mark Hennessy]

Thermal converters are OK, but have two serious drawbacks:

1. They are fairly inaccurate (powerin->heater->thermocouple/thermistor->reading).
   
2. They draw significant power from the DUT
   

Perhaps in isolation, but not in the Fluke 8920A I mentioned. The input impedance is 10MΩ, and the accuracy is 0.5%.

The Fluke 87V is 0.7%. Even the 287 only manages 0.4%.

On the plus side:

1. They are simple
   
2. Handle any waveform
   
3. Very wide bandwidth
   

So they are ok for power measurement, but of little use for small signal work.

The lowest range on the 8920A is 2mV FSD. Admittedly, in this range, the BW falls from 20 to 2MHz, but if this matters, you can stop it auto-ranging into that range.

As I said before, highly recommended

[Nick]

All true  I went and read up on the 8920A - my comments were related to the traditional thermal RMS converters - the 8920A is far more modern - very interesting write-up in its instruction manual (http://assets.fluke.com/manuals/8920A_21imeng0000.pdf) - see sections 3.17 onwards, 

As Mark says, nice bit of kit - in this case, there is a whole bunch of auto-ranging & conditioning, hence the high input impedance and accuracy. Lot of work in there.

Quite tempted, I am

EDIT: Has to be the absolute ugliest bench DMM I have ever seen ! Simply for aesthetic reasons, it's not going into my shack !

[ppppenguin]

I remember using a tradtional thermal RMS meter to standardise PAL subcarrier (4.43MHz) amplitude. You measured the result at 4.43MHz. Then you applied DC instead and adjusted to get the same reading. Tedious and prone to burnout. The 8920A uses a pair of resistive elements and a pair of transistors as temperature sensors, nicely matched on a single chip. Then a feedback network drives the temp of one resistor to match the other one. Thus non-linearity (and there's plenty of that) is cancelled out.

[Mark Hennessy]

EDIT: Has to be the absolute ugliest bench DMM I have ever seen ! Simply for aesthetic reasons, it's not going into my shack !

Yeah, I can't argue that one

And they normally fade to a pretty disgusting shade of beige to complete the look!

Being a bit of a "Fluke Fanboy", I enjoyed an interview with Dave Taylor, who worked at Fluke at the time:

http://www.eevblog.com/forum/amphour/drt...this-week/

His crowning achievement while there was the 8060A (a lovely meter for an audio geek), but he also discusses that case style. For a while, it was the "big thing" at Fluke, but funnily enough, they had terrible thermal problems with them. They did prototype a bench PSU, but there was no way it was going to work (remembering that this was before cheap fans, etc). There is a picture of the prototype somewhere on the EEVBlog forum.

In fairness, the front panel of the 8920 is quite attractive, and that's all I can see of it because it sits in a rack!

In case (bad pun!) you're wondering about the weird shape, it's partly explained by the way they are designed to stack. There is a locking mechanism that works very well. A friend has a frequency counter in that style case, and I keep trying to persuade him to sell it to me. The 8920 has a frequency counter output on the rear (optional, but usually present) which is very convenient...

[EinsteinAGoGo]

I came across this interesting article by Jim Williams and some other guy, some time ago.
http://www.edn.com/design/test-and-measu...voltmeters
Different AC measuring topologies are compared.

The 8920 fares extremely well, and yes, I agree, its a hideous looking thing!
It's a pity they didn't 'name and shame' all the meters.

8060As are very useful, if rather fiddly to use. Certainly pretty flat to 100kHz or so, IIRC. However, I much prefer my Tektronix TX3 to any Fluke I've ever used; certainly better made, well, that's bygone Tek for you.