Handheld meter robustness testing

[2N3055]

Anyone who has ever done thermal testing knows it can take a fair amount of time (weeks, months).   With the handhelds, even if I wanted to run a complex test, the data would have to be collected manually or possibly using multiple cameras inside the chamber.  A lot to invest for some cheap meters.   I'll leave that to the next person.   

I agree, am fully aware of it and I said so.. I understand you don't have time for it...

So we are back to what test could be ran in the shortest time that yields the most useful information.   Again, I am open to suggestions and have posted a proposal that Alm has weighed in on.  Feel free to do the same.   

Fair enough, so to be constructive, I agree with alm and as I stated before him, I think it should be done at volts range.. Even 1-2V is much better than millivolts (some would even say thousands times better ).  Complication is with the fact that meters can have 1,2,3,4,5,6V ranges.  So full scale will be different for each one.
That can be mitigated by testing at each full range (or close to full range) and calculating and specifying tempco with percentage instead of absolute values..
So you have  meters with 4, 5 and 6V range, connect them to 4V and fire away.. Or you can test meters with same ranges together....

But simply testing at 1-2V would be enough to make it much better than testing at millivolts range..

As for testing temps, three points would be nice. 0°C,  some room temp and  40°C. 
I also can understand what you wrote about -10°C, I can see that would be interesting, if you simply have to measure something outside in winter time. So maybe that too.. That would be 4 points. And that is time consuming. Maybe 4-5 hours plus setup time..  Pretty much full day of work.
I would say you need to decide how many points you are prepared to spend time for.
And we can maybe setup some kind of voting about temps.
I have my opinion, but it is biased by my needs.. Maybe there are many more people out there that would like to see meters tested at -60°C.. What do I know what other people want...

As for automation, you might not need more than one camera, and small handheld camera is more than good enough. I have small Canon 260SX that you can manual focus, and with CHDK firmware, you can write scripts that will do time lapse and much more..

Best regards

Sinisa

[EEVblog]

Sorry you feel that way Dave.  I would say if anyone was the fool, it was me.  There is little I can do at this point beyond not mentioning the meter and pulling down the videos.  If you like I can remove the last one as well.   Your call.

I never asked you to nor will I ask you to remove any videos. Entirely your call. I sent you an unreleased and unfinished meter and I can't and and don't want to stop you from doing what you want to do with it.
As an aside, I was surprised to see you doing more with it than I thought you wanted it for (i.e. some fun potentially blowing it up). I'm not really a follower of your videos or this thread so had no real idea about the seriousness you take all this stuff.
Sorry if you thought I was sending you a review unit, that wasn't my intention.
Just today some parts have changed again that affect the tempco and stability of the unit.

[EEVblog]

So we are back to what test could be ran in the shortest time that yields the most useful information.   Again, I am open to suggestions and have posted a proposal that Alm has weighed in on.  Feel free to do the same.   

If you are comparing meter tempcos then just a simple delta at two temps (plus room temp) on a fixed DCV range at close to full scale.
Meter data sheet specs are typically given for room temp (23degC) +/- 5 degC. So I'd pick 10degC and 40degC as the other temps.
Room temp doesn't really take any extra time as you do it at the start of the thermal test (unless you don't work in a fairly controlled temp office environment like I do).
If you want to save extra time then just chose a temp in one direction from room temp. I'd suggest at least 20degC delta in any case. Positive or negative direction is a coin toss, doesn't really matter.

[EEVblog]

Fair enough, so to be constructive, I agree with alm and as I stated before him, I think it should be done at volts range.. Even 1-2V is much better than millivolts (some would even say thousands times better ).  Complication is with the fact that meters can have 1,2,3,4,5,6V ranges.  So full scale will be different for each one.
That can be mitigated by testing at each full range (or close to full range) and calculating and specifying tempco with percentage instead of absolute values..
So you have  meters with 4, 5 and 6V range, connect them to 4V and fire away.. Or you can test meters with same ranges together....
But simply testing at 1-2V would be enough to make it much better than testing at millivolts range..

Yes, the different counts are annoying.
1V might be the best all-round choice as it gets almost all the meters including the 10,000 count ones that usually go a little bit over.
10V reference would include another resistor in the divider chain, and that could be good or bad depending upon your viewpoint.
You don't want low ranges like 50mV that some meters have, because they will be typically using the internal chipset x10 chopper amplifier and external resistors different from the main divider chain. Although if your plan is to characterise those aspects in particular then of course it's what you'd do.

[SeanB]

Saying 60C is not a real world thing is not true, many lift motor rooms here will regularly reach that and higher in summer, as they typically are brick and concrete slab buildings on top with a lovely reflective silver concentrator of the rest of the building cooking the walls all day. Sitting with almost zero cooling aside from air vents typically closed with rodent proofing and roach proofing mesh, and with 10kW of motor heat being dissipated in the room with it as well, and all the brake, controller and shaft heat rising up into there as well. If you are unlucky you get the older ones that are basically a tin shack, walls, roof and with a tin sheet door, basically a solar oven. You go in there and it might be 60C easily, and you work fast and take regular breaks out in the up to 40C cool outside. Your meter should work there though, you will be leaving it there while you go out for the break and the look for some water to drink.

You tend to get the hotter non room temp conditions a lot more than the cold ones, though you can also be working in a cold store with it sitting at -30C as well, though you probably are really only going to be concerned with rough resistance and continuity, and if voltage is present in about the right range in these conditions.

[alm]

I think picking a range that all meters share (e.g. the range that includes 2 V), and doing a short and a voltage close to full scale (95%) would be the best. If you present the change with shorted inputs in percent of full scale, and the change near full scale in percent of value, you would have something that is easy to compare against accuracy specs, which are typically X% of value + X% of range (or counts). It would also give you the advantage of maximizing the resolution for the gain measurement.

I do not think that using 2 V or 10 V for different meters would make any difference, assuming you have a variable precision voltage source. No worse than comparing a 2000 count meter to a 20000 count meter already does (low values may disappear in quantization noise for the 2000 count meter).

If you are comparing meter tempcos then just a simple delta at two temps (plus room temp) on a fixed DCV range at close to full scale.
Meter data sheet specs are typically given for room temp (23degC) +/- 5 degC. So I'd pick 10degC and 40degC as the other temps.
Room temp doesn't really take any extra time as you do it at the start of the thermal test (unless you don't work in a fairly controlled temp office environment like I do).
If you want to save extra time then just chose a temp in one direction from room temp. I'd suggest at least 20degC delta in any case. Positive or negative direction is a coin toss, doesn't really matter.

The question, however, is will there be anything to measure? Even for -10°C to 40°C, the deviation for several meters was down to 1 LSD (i.e. down in quantization noise), see attached frame from Joe's video that I linked to in my previous post. If you limit the temperature range even more, pretty much all meters might score < 1 LSD, which makes the test not very sensitive (mediocre meter in this regard performs the same as an excellent meter).

I would consider averaging multiple readings to increase resolution, but that would be a total pain to set up, especially in a thermal chamber, and would not be feasible for meters without computer interface (camera + OCR ).

[2N3055]

Saying 60C is not a real world thing is not true, many lift motor rooms here will regularly reach that and higher in summer, as they typically are brick and concrete slab buildings on top with a lovely reflective silver concentrator of the rest of the building cooking the walls all day. Sitting with almost zero cooling aside from air vents typically closed with rodent proofing and roach proofing mesh, and with 10kW of motor heat being dissipated in the room with it as well, and all the brake, controller and shaft heat rising up into there as well. If you are unlucky you get the older ones that are basically a tin shack, walls, roof and with a tin sheet door, basically a solar oven. You go in there and it might be 60C easily, and you work fast and take regular breaks out in the up to 40C cool outside. Your meter should work there though, you will be leaving it there while you go out for the break and the look for some water to drink.

You tend to get the hotter non room temp conditions a lot more than the cold ones, though you can also be working in a cold store with it sitting at -30C as well, though you probably are really only going to be concerned with rough resistance and continuity, and if voltage is present in about the right range in these conditions.

I understand your point. I'm from Croatia, it's 37°C in my home town now. Products I make are usually installed in metal boxes in full sun, and everything I make is specified to work to at least 65°C...

But....

Of top of my head, Fluke 87V (–20°C to + 55°C specified), Fluke 27/28 II (-15°C to +55 °C , -40 °C for up to 20 minutes) ,  Keysight U1273AX (-40°C to 55°C ),
Brymen 839 (-10°C to +50 °C), Brymen 235 (-10°C to +50 °C).......
I presume if you look further, there will be some other.  Those are specified and guaranteed to work at those temp extremes..

I wouldn't trust any meters operated outside their operating specs for serious (professional) work, no matter how well they fare in Joe's testing.
Especially if I worked on elevator motors and such installations, that are CAT III best case, and some are even CAT IV as far as overvoltage goes..

It's not that anything is wrong with his measurements, but product is not guaranteed to work out of specs.  On specific one that Joe tests, all  might even work fine and be safe, but you might go out and buy same instrument from different batch/revision... That will be assured and tested by manufacturers according their specs...But might have different behaviour outside, at the extremes....

My main handheld meters now are Brymen 869S and 525S.. But I have my trusty old Fluke 77 III  (0°C-50°C) in my carry on toolbox for when I go where it's hot...

What I'm trying to say that meters should not be tested outside specs, and then expect that because someone measured it, it is now characterised and safe to use outside manufacturer's warranty. It's not, and only manufacturer can respecify instrument for extended temperature range or whatever..

Most of the meters out there are specified for 0-40°C... For work where it's enough.

If your work requires extended operating range, you're out of luck, you're gonna have to go buy extended specification meters from known brands and pay for it...
Thinking about it, this is one of occasions, where it is much better to buy, for instance, used Fluke 87V in good shape than brand new Brymen 869..
Or maybe for motor work on line voltage, one of electrical testers like FLUKE T150 ..

Regards..

[joeqsmith]

Sorry you feel that way Dave.  I would say if anyone was the fool, it was me.  There is little I can do at this point beyond not mentioning the meter and pulling down the videos.  If you like I can remove the last one as well.   Your call.

I never asked you to nor will I ask you to remove any videos. Entirely your call. I sent you an unreleased and unfinished meter and I can't and and don't want to stop you from doing what you want to do with it.
As an aside, I was surprised to see you doing more with it than I thought you wanted it for (i.e. some fun potentially blowing it up). I'm not really a follower of your videos or this thread so had no real idea about the seriousness you take all this stuff.
Sorry if you thought I was sending you a review unit, that wasn't my intention.
Just today some parts have changed again that affect the tempco and stability of the unit.

I should have ran my test at 10V and then looked at the mV as a percent change and really make those numbers look good.  But instead, I ran the same test as before.  1mV, highest sensitivity range, -20 to 60C then calculate a TC from that by looking at the change in voltage / change in temp. 

Want to know how the pre-production 121GW and Gossen M248B compare against my most stable meter, watch and find out. 

Joe, this test is completely invalid. The meter you have is not the current pre-production unit, it's a prototype that does not have the current voltage reference or divider resistors. I mentioned this in the emails to you when I sent it, so I'm not sure why you tested this.
It is most definitely not representative of the final unit.
I'd appreciate if you don't do any more further testing on this unit as it will only confuse people.
Thanks.

Again, to avoid any confusion, I have no intent to do more with it. 

[joeqsmith]

The question, however, is will there be anything to measure? Even for -10°C to 40°C, the deviation for several meters was down to 1 LSD (i.e. down in quantization noise), see attached frame from Joe's video that I linked to in my previous post. If you limit the temperature range even more, pretty much all meters might score < 1 LSD, which makes the test not very sensitive (mediocre meter in this regard performs the same as an excellent meter).

And now we get to the root of the problem and why I changed the test method.  I can go back to looking at 1V and run at an even less temperature swing but from the little bit I looked at it, I don't think you will learn much from it. 

[alm]

I do not see away around that. You can come up with all kinds of tricks like introducing noise that you remove by averaging (Datron used this trick in some of their high-resolution bench meters), but absolutely none of that is useful if all you have is a human looking at a display. Maybe the conclusion would just be that the temperature coefficient is negligible for many 3.5 digit meters, at least as far as gain error is concerned?

I am not convinced that the larger differences in the mV range are indicative of true differences between DMMs, or if you are just magnifying noise (e.g. thermoelectric). Assuming the temperature is constant throughout the meter, the only explanations I can come up with why a full scale reading on the lowest range would have substantially more drift than on a higher range is either offset in the amplifier that is switched in for the mV ranges (is it really that interesting to focus on this detail?) or thermoelectric offset voltages between the meter and voltage source outside the thermal chamber.

It might be interesting to see how the results compare for a shorted input on the higher range. It could be that the temperature coefficient mostly affects the offset (e.g. input buffer), rather than the gain (voltage reference).

For the Fluke 189, temperature coefficient is specified as 0.05 x specified accuracy/K, which for 30K temperature difference (the tempco only applies for temperatures < 18°C or > 28°C) would +/- 30 µV for 0 V on the 50 mV range, +/- 70 µV for 50 mV on the 50 mV range, +/- 1.5 mV for 0 V on the 5 V range, and +/- 2.25 mV for 5 V range. That suggests that the error should only increase as you get closer to full scale (as makes sense). Based on that, a value close to full scale of whatever range would be optimal.

Obviously these are maximum tolerances, so real world performance will likely be much better. Is there a handheld that has the tempco specification split it in offset and gain to give an indication of expected drift?

[joeqsmith]

is it really that interesting to focus on this detail?

For the normal person, I would hope not.  But we could say the same thing about the transient testing.  Is that interesting to the normal person?  I would again say no but it does seem there is a small group of us who are interested in the subject.  Maybe myself along with the other subscribers should seek help on a professional level.   
     

Maybe the conclusion would just be that the temperature coefficient is negligible for many 3.5 digit meters, at least as far as gain error is concerned?

I believe this to be the case but again my sample size is small.

With the chatter, it sounds like the temperature testing is a source of confusion.  If people want to point the finger at me or my videos, I don't mind shouldering the blame for some of it.   However, it does seem like if people are going to present thermal data for handheld meters, we should have some clear understanding about how to go about benchmarking them and what does "wow, that's pretty good" really mean.  Otherwise it seems we are just adding to the confusion. 

If I said I hooked my Fluke 101 to a 100,000VDC power supply and it lived, that would cause confusion and I could see people hooking their 101s up to their MOTs any wondering why they cooked.   This is why I try to provide as much detail as I can about the tests I run.  My tests may be flawed or deemed to have little to no value but at least you will know that my 100,000V supply was limited to 1fA.   

[alm]

is it really that interesting to focus on this detail?

For the normal person, I would hope not.  But we could say the same thing about the transient testing.  Is that interesting to the normal person?  I would again say no but it does seem there is a small group of us who are interested in the subject.  Maybe myself along with the other subscribers should seek help on a professional level.   

Just to be clear, that statement referred to the x10 chopper amplifier that is used only for the mV ranges. To me it seems that if you want to pick one range, it is more interesting to study the components that are used for all voltage, current and resistance measurements (ADC and voltage reference). I make no claims about being (more) sane.

[joeqsmith]

If only there was something interesting to look at with a volt signal.  Seeing a few counts of change is not much fun.  Your idea about testing them with condensation is sounding better and better...   

Maybe the two members who asked to see the 8002/8 tested over temp have an opinion on what they would like to see.  Feel free to add to the mix.

Am I the only one confused about why a meter that is checked against two higher cost units and deemed to be pretty good, would go through yet another round of mods that effect the temperature performance?   My guess is it must have been a side effect from another change and not that they are directly changing the reference circuit or attenuator network for lower drift.  It would not make sense.   

[evava]

If only there was something interesting to look at with a volt signal.  Seeing a few counts of change is not much fun.  Your idea about testing them with condensation is sounding better and better...   

Maybe the two members who asked to see the 8002/8 tested over temp have an opinion on what they would like to see.  Feel free to add to the mix.

I think I was one of the two, I asked and still I ask you to test AN8008 - but in the original rig!
I'm not convinced that original test is completely invalid - on the contrary!
Even from previous posts is evident, and I agree with that, that test with narrower range of temperatures ( and 1V level) would probably mean "nothing to write home about".

If I were you, I would continue in that original test (-20 Deg C to +60 Deg C and 1mV), and when new 121GW is available to public, then test it too, please.
When UT181a is stable enough, let others be stable too.

It is a test, it can not be invalid (unless you compare it with manufacturer data - which you are not going to do)

Thank you in advance!

[kalel]

I was probably the other (unless there was someone else too). Personally I only cared about how they compare with each other (the meters). It didn't necessarily need to be real world usage. So I'd be fine with any test that can compare meters (i.e. multiple meters being tested under the same conditions), but being a beginner to electronics, I don't have a deep enough understanding to suggest some high quality methodology. I understand that the more points measured, the more valuable the data, but it would be fun to watch either way. It's not data to draw conclusions from (e.g. how much the drift would be at specific setting at specific temperature range), but it still gives an idea on what different meters display under the same or at least very similar working conditions.

But, if most people would find that running these tests is not useful, and especially if you find it needing more time than the benefit, then I will gladly support that choice.

[evava]

Am I the only one confused about why a meter that is checked against two higher cost units and deemed to be pretty good, would go through yet another round of mods that effect the temperature performance?   My guess is it must have been a side effect from another change and not that they are directly changing the reference circuit or attenuator network for lower drift.  It would not make sense.   

And if I understand it well, your original test (which caused all that fuss) was about "the x10 chopper amplifier that is used only for the mV ranges" 

[alm]

If only there was something interesting to look at with a volt signal.  Seeing a few counts of change is not much fun.  Your idea about testing them with condensation is sounding better and better...   

I guess the ultimate choice is between interesting and entertaining or accurate .

It is a test, it can not be invalid (unless you compare it with manufacturer data - which you are not going to do)

The test may not be invalid, but is it testing the meter or the setup? Is it showing differences between meters or small differences in test setups?

Joe, do you have any idea how reproducible the 1 mV tests were? I.e. if you go back the next day and reconnect a meter, will it show the same results? If thermoelectrics indeed play a role, than I would expect a fair variation between runs due to how the cables are run and how temperature gradients across the cables are.

[EEVblog]

And now we get to the root of the problem and why I changed the test method.  I can go back to looking at 1V and run at an even less temperature swing but from the little bit I looked at it, I don't think you will learn much from it.

On the contrary, you learn the most important thing you want to know about a meter in terms of tempco, that the meter reading basically does not change with temperature.
A null result of a test does not mean there is no value in that test, far from it.

[EEVblog]

Am I the only one confused about why a meter that is checked against two higher cost units and deemed to be pretty good, would go through yet another round of mods that effect the temperature performance?

We missed something on the 50mV range.

[joeqsmith]

Am I the only one confused about why a meter that is checked against two higher cost units and deemed to be pretty good, would go through yet another round of mods that effect the temperature performance?

We missed something on the 50mV range.

Strange.  How did you find it?  Running temperature tests in the low range?  If so, how did you run them?

[joeqsmith]

Joe, do you have any idea how reproducible the 1 mV tests were? I.e. if you go back the next day and reconnect a meter, will it show the same results? If thermoelectrics indeed play a role, than I would expect a fair variation between runs due to how the cables are run and how temperature gradients across the cables are.

When I compensated that UT-61E, I made several sweeps.   Granted, I was making small changes to the meter between runs to bring it in but sure, it was stable enough setup to setup to make those measurements.   

[joeqsmith]

If only there was something interesting to look at with a volt signal.  Seeing a few counts of change is not much fun.  Your idea about testing them with condensation is sounding better and better...   

Maybe the two members who asked to see the 8002/8 tested over temp have an opinion on what they would like to see.  Feel free to add to the mix.

I think I was one of the two, I asked and still I ask you to test AN8008 - but in the original rig!
I'm not convinced that original test is completely invalid - on the contrary!
Even from previous posts is evident, and I agree with that, that test with narrower range of temperatures ( and 1V level) would probably mean "nothing to write home about".

If I were you, I would continue in that original test (-20 Deg C to +60 Deg C and 1mV), and when new 121GW is available to public, then test it too, please.
When UT181a is stable enough, let others be stable too.

It is a test, it can not be invalid (unless you compare it with manufacturer data - which you are not going to do)

Thank you in advance!

I was probably the other (unless there was someone else too). Personally I only cared about how they compare with each other (the meters). It didn't necessarily need to be real world usage. So I'd be fine with any test that can compare meters (i.e. multiple meters being tested under the same conditions), but being a beginner to electronics, I don't have a deep enough understanding to suggest some high quality methodology. I understand that the more points measured, the more valuable the data, but it would be fun to watch either way. It's not data to draw conclusions from (e.g. how much the drift would be at specific setting at specific temperature range), but it still gives an idea on what different meters display under the same or at least very similar working conditions.

But, if most people would find that running these tests is not useful, and especially if you find it needing more time than the benefit, then I will gladly support that choice.

I finally have both meters.  Hope to have some time this weekend to play with them.  If there are other non-destructive tests you or anyone else would like to see ran to compare the two, feel free to ask.  Otherwise, it's temperature then ESD. 

   

[joeqsmith]

Just for kicks I hooked my 121GW up to my 5kV insulation tester yesterday and it survived.
I might do some more testing to see if this can kill other meters.
Don't have a 5kV scope probe to see the waveform though, but assume there will be a small initial energy burst and then clamping down.

Hard to say much about the post.  You may have had the meter connected and in the off position for all I know.  Maybe the insulation tester was off or set to 500V.   Let's assume you actually programmed the insulation tester for 5KV and you connected between the 121GW's  V & Com inputs and checked it with the meter set to every mode and it survived.  I still have no idea what insulation tester was used.   Looking at the Hioki 3455  it looks like it has a short circuit current of 2mA or less.  It may be enough to damage the meter but I would expect the 121GW's front end to clamp that down easily, maybe. 

Added.  You are aware I had changed out the HFE part.  This is one of the weak points of the design.  If you wanted to try it with the meter set to Hz with the insulation tester putting out a positive voltage on  V/Ohm relative to GND, it may do something.  You need to somehow get the insulation tester to put out the voltage with an open, get the 5KV then discharge it across the meter.  The capacitor that AC couples the grounds together may be enough to exceed the HFE's absolute maximum supply voltage.   

A member here, Scott was playing around with electronic fly swatters and actually damaged a meter with one.   TI bought one and tried to damage the UT90A with it.  The UNI-T UT90A has had more abuse and damaged more times than any meter I have.   They flyswatter did not have enough energy to get the job done.  The UT90A's clamp would just load it down.  So I added a little external capacitance, let it charge, then discharged that into the meter.  None of this is useful data and it's not something that is recorded in my spreadsheet.     

Dave,
I watched your last video and give you an A for your efforts.   Your original post makes more sense now. 

I've mentioned that I had someone post how I was directly discharging the capacitors into the meters with the generators I made.  While I have done some tests like this, the data is not recorded.   There is a coupling network that shapes the wave.  There is no magic or top secrete information to be had.  Its all explained in the IEC standards.  The wave shape is with an open circuit and the majority of the energy when I test meters is absorbed by this network, not the meters.  That is unless a meter happens to go low impedance.   This is partly why I monitor the wave shape with the scope during the tests.  Its not because I am concerned the generator looses alignment or fails, it's because I can detect when a meter is starting to breakdown.   

The big difference with the combo generators is there are two basic waveforms for surge.  The open circuit voltage and the short circuit current.  As I have explained, I am not concerned with the short circuit current waveform and limit to about 20 joules.

If you were interested in running these sort of tests for the fun of it, I would not recommend anything like my setup.  First, it will never generate the drama.  The energy levels are so far down in the mud, at best I get a few sparks emitting from a case from time to time.   My goal has never been to have them explode like some of the Fluke videos show.   If you want drama, or even to get a better idea how safe the meters are, I would recommend getting a real generator and build a small containment chamber. 

Again thanks for the video and at least putting in an effort to show what you were doing.

[floobydust]

Wait a sec, the AN8008 rotary switch wiper has less clearance than the PCB. I think that is the first breakdown point
I measured 0.6mm and the PCB has 1mm I faintly recall.
Also the voltage switch-position is furthest out and somewhat isolated, so a bit better clearance there.

[joeqsmith]

The 8002 is the same way but just because they are closer does not make them instantly the breakdown point.   Without tracing the whole thing out, I would be guessing and I'm not sure it would matter.