EEVblog #853 - How A Multimeter Works

[EEVblog]

This was supposed to be a repair video of a Fluke 17B multimeter.
But embarrassingly turned into an impromptu How a Mulitmeter Works video.
Dave goes through the Fluke 17B schematic and explains how stuff works.



Teardown video:

Thermocouple Tutorial:

[Paul Moir]

Please let me be the first to thank you for your use of the word "tragic" in this video.  It's been a long time since I've heard it in the right context.  Way to break those Engineer stereotypes!

Cmon, that chip has got to be a Hycon, right?  The date code, the H20x, the Hycon LCD driver.  The crystal is coming off the wrong pins though, so that's where I gave up.

[max666]

This video made me think many times of Futurama:


That cold junction compensation can be used to measure temperature, without having a probe connected, can't it? My meter shows me the temperature inside the meter, when I have no probe connected.

[HAL-42b]

 

[EEVblog]

 

I sooooo missed the opportunity to put that in!
And I tweeted it the other day too 

[zapta]

I have attached the schematics of the Fluke 17B ...

It's interesting how the big rotary switch is actually a set of many, independent, smaller switches.

[Zbig]

I'm a bit surprised that Dave considers having a cold junction compensation in a thermocouple-based meter anything fancy. Unless I'm having my understanding of how a thermocouple works totally wrong, I'd think all you could possibly get off a non-compensated meter would be "your DUT is X degrees beyond/below the ambient temperature"?  Even the cheapest, nastiest meters with temperature range I came across were displaying absolute temperature. As max666 mentioned, you can tell that by the meter showing its internal temperature with no leads connected but it's not only for that - that's more of a "side-effect", really. Without that, you'd always have to sum the meter's reading and the temperature in the room you're sitting in in order to be able to tell how hot the FET you're poking really is, anyway.

[f4eru]

I alway appreciate a good video on "moldy meters"

[Fungus]

a) The non-standard spacing between the two "amps" connectors might be deliberate.

You'd never want to plug something into both of those at the same time.

b) The newer version of the 17B (the 17B+) is closed-case calibration - no pots!

Still... you've got one with twiddle pots and a schematic. Family fun!

[Fungus]

Question: How much do the switch contacts affect the readings on the current ranges?

There's all sorts of precision resistors and trimmers in there but at the end of the day it still goes through the switch.

eg. When switch 18 is closed (at 13:20 in the video) it's in series with a one Ohm resistor, surely the switch resistance will be noticeable. 

[retrolefty]

I'm a bit surprised that Dave considers having a cold junction compensation in a thermocouple-based meter anything fancy. Unless I'm having my understanding of how a thermocouple works totally wrong, I'd think all you could possibly get off a non-compensated meter would be "your DUT is X degrees beyond/below the ambient temperature"?  Even the cheapest, nastiest meters with temperature range I came across were displaying absolute temperature. As max666 mentioned, you can tell that by the meter showing its internal temperature with no leads connected but it's not only for that - that's more of a "side-effect", really. Without that, you'd always have to sum the meter's reading and the temperature in the room you're sitting in in order to be able to tell how hot the FET you're poking really is, anyway.

 Right on. 

That is also why to measure temperature accurately one must NOT try and extend the temperature sensor using copper extension leads as then the temperature compensation measurement must be made at the transition of TC wire to copper wire, not at the meter's input jacks.

[ModemHead]

Question: How much do the switch contacts affect the readings on the current ranges?

The voltage sense line picks off below the switch, so the measurement only includes the voltage drop from the top of the shunt to analog ground.  Voltage drop on the switch is not included in the measurement, but it will contribute to the burden voltage.

[Huluvu]

Exactly - but I'm sure the mA Fuse will contribute much more.

[Fungus]

Question: How much do the switch contacts affect the readings on the current ranges?

The voltage sense line picks off below the switch, so the measurement only includes the voltage drop from the top of the shunt to analog ground.  Voltage drop on the switch is not included in the measurement, but it will contribute to the burden voltage.

Oh, duh!   

It's measuring current. It only has to measure across part of the circuit.

[scopeman]

Hi Dave,

From one that has made a few shunts in his time:

Multi-meter shunts are normally made from Manganin an alloy that has an excellent TC.

Read all about it here:

https://en.wikipedia.org/wiki/Manganin

Sam
W3OHM

[sbprojects]

The ohms function in multimeters doesn't usually use constant currents. What they basically do is put a voltage across the voltage divider resistors (known resistor) in series with the unknown resistor. Then they use the voltage across the known resistor as reference voltage for the ADC and the voltage across the unknown resistor as normal input voltage. The result is the ratio between the two input voltages, scaled to ohms. The actual voltage across the the two series resistors doesn't affect the accuracy, as long as it is high enough for the ADC.
On older Fluke meters (8010, 8012, 8020 etc) they used a little trick to swap the known and unknown resistors to get the infamous nano Siemens range. To be exact, they swapped the reference voltage and the input voltage of the ADC.

[Muxr]

Nice video Dave. Honestly I am torn on the whole trimpot vs. eeprom calibration. Fluke 87V has eeprom calibration but you need some pretty specialized gear in order to calibrate it, whereas trimpots don't require anything special (not sure about Brymen's DMMs). As for the drift, I can say that my 30+ year old 8060A is still in calibration and it's all trimpot based. Eeprom by its nature should provide for less drift I agree but I wonder how big of a deal it really is considering these are low count meters.

[Kleinstein]

If done right, the trimmer based calibration can also work well, but it takes quite some time and good trimmers are not cheap. In this low count meter they use precise resistors for some places to keep the number low. The software / EEPROM calibration can usually do the adjustment for more points than the trimmers used here - so also a few more precision (0.1%) resistors could be saved. Adjustment of the pots usually have to be made by hand, that adds costs even in china and is more error prone. With special gear, software calibration is fast, precise and low cost. So I very much understand why they choose software calibration today.

It depends in the software in the meter if you need special gear to do the calibration - with these cheap meters this is likely , as it speeds up calibration and no need to add software for manual calibration. The bigger problem is that you are lost without instructions for adjustments even if the instruments allows it in principle. Cheap meters could even have PROM for adjustment.

Doing the adjustment with trimmers is also very hard if you don't have instructions or at least a schematics.

[Fungus]

If done right, the trimmer based calibration can also work well, but it takes quite some time and good trimmers are not cheap.

Yep. I've got nothing against it in principle so long as it's done well.

I wish they'd publish the procedure/protocols for the closed-case calibrations though. How can I adjust my friend's meter to be 20% off without trimpots and without knowing that?

[sync]

I wish they'd publish the procedure/protocols for the closed-case calibrations though. How can I adjust my friend's meter to be 20% off without trimpots and without knowing that?

There are DMMs with published calibration procedures. If this is important for you buy one of them. If a meter is 20% off then it's probably broken and needs repaired.

[Neilm]

@4:00 - where is the fail WaWaWaaaaa?

At least it was easy to spot and get at.

[drtaylor]

I was going to comment on how ohms are measured, but SBprojects beat me to it. In integrating DMMs, it is a simple task to integrate with the unknown input resistor, and de-integrate with the known resistor. So the only trick is to make sure that the input common mode range of the A/D is not exceeded by the voltage across either the reference or unknown resistor. SBProjects also correctly pointed out that you can get the Siemens ranges just by reversing this process. This allowed measurement to very high resistances (inverted) but resolution went down as resistance went up (or Siemens went down). The 200/2000megohm ranges in the 8060 just made a Siemens reading and reciprocated it to get it back into ohms.

The protection afforded by the PTCs can only be used in Ohms circuits, because the very noisy voltage across the PTCs is not measured. An advantage of ratiometric ohms conversion is that it only requires the reference resistor to be precise. Any other voltage, including the VRef is not at play.

The main down side to radiometric ohms conversion is that it does not lend itself well to four terminal ohms measurements. Although theoretically possible, it would pick up a lot of noise. All meters I am aware of with four terminal ohms connections, use constant current measurement techniques.

As far as the ohms protection clamp goes, the old way of using back to back zener connected transistors (usually 2N3904) dates back a long way with Fluke. I first saw it used on the 8020 meter, but I suspect that older Flukes used it too. Bob Pease had an article on this. The main reason for using transistors instead of Zeners is that it makes an extremely low leakage clamp, usually less than 2pA.  This was required for the Siemens ranges. Fluke used selected 2N3904s for this way back when. I touched on this in an earlier post.

I also concur with other posters that the switch resistance on the current inputs is negligible. It only very slightly contributes to burden voltage, and the fuse would be a far larger contributor to that. Note that the high current range does not go through a switch. Certainly the sense voltage does, but any series resistance in the sense lines could be very large and not affect the accuracy until you hit a point that leakage current becomes an error.

The 8060 used dual trimmers, a fine and coarse trim. These trimmers were selected for Temp-co and were highly evaluated for long term stability. It also had a primitive constant current source, but that was only used for the diode check.