Handheld meter robustness testing

[CDaniel]

Probably the voltage reference is shitty ... pretty obvious

[dcac]

That temperature drift from -10 to 60C is very concerning, and what the hell is causing it.
...
I just don't get it.

I allow it to sit at the two extremes for a half hour to stabilize.  There's not a lot of room in that box and there is a decent sized fan.  At -10 we had 1.003mV, 20ish 0.997mV and 60C 0.928mV.   

It’s difficult to tell exactly as this part of your video is so time compressed, and is perhaps the overlayed chamber camera not really in sync, but:

121gw showing 0.899mV at -8.9c, control program (chamber temp) is shoving about 20c:




121gw showing 0.871mV at 20.9c, control program (chamber temp) is shoving about 43c:




121gw showing 0.997mV after 2h in 20.2c “room temperature”




so I get 0.899mV when chamber temp is shoving about 20c and then 0.997mV at 20c “room temperature”, why the big difference?

[e0ne199]

i just watched joeq's recent video about 121GW and now i kind of regret of buying it....

[rsjsouza]

I just finished watching the video and was truly dismayed by the results. It is hard to see a great featured meter failing so spectacularly under public scrutiny. Old or new design, the absence of a robust input protection takes it back to the ultra-cheap field restricted to low power electronics, which can be had by a fraction of the price - my UT61E works wonderfully for that and it is not a very good place to be.

I couldn't stop thinking what triggered UEi to stop selling handheld DMMs for a while before the 121GW was released. Did they disband their previously experienced dev team, saw the opportunity to reenter this market but couldn't hire them back and did Ctrl+C/Ctrl+V on older designs but without careful consideration? Or was it a cost cutting measure (the explanation that everyone loves to believe) and replaced greybeards with new hires at a fraction of the cost? We can only speculate, but I have seen both scenarios way too many times in the industry.
 
Oh well... Thank you for this christmas bonanza. I wish you a merry christmas as well.

[joeqsmith]

It’s difficult to tell exactly as this part of your video is so time compressed, and is perhaps the overlayed chamber camera not really in sync, but:

121gw showing 0.899mV at -8.9c, control program (chamber temp) is shoving about 20c:

121gw showing 0.871mV at 20.9c, control program (chamber temp) is shoving about 43c:

121gw showing 0.997mV after 2h in 20.2c “room temperature”

so I get 0.899mV when chamber temp is shoving about 20c and then 0.997mV at 20c “room temperature”, why the big difference?

That temperature drift from -10 to 60C is very concerning, and what the hell is causing it.
...
I just don't get it.

Different parts inside the meter are changing at different rates when I ramp the box, even at this slow rate.  With this meter, it causes a fair bit of error.      I allow it to sit at the two extremes for a half hour to stabilize.  There's not a lot of room in that box and there is a decent sized fan.  At -10 we had 1.003mV, 20ish 0.997mV and 60C 0.928mV.   

But if you leave it in your car overnight in the dead of winter (much colder here than -10C) and bring it inside to use it, expect there to be a fair bit of error until it settles.  Normally we would test something like this with a shock chamber but in my home lab, you have to settle for my cardboard box.

Yes it's VERY compressed.  There's about 6 hours of data, maybe more.  It takes several hours for that peltier setup to cool down to -10.  You get to see the last half hour, the ramp and the hold at high temp.   About 2 hours, compresses into a few seconds. 

I have a sinking feeling that you feel that if you took a 1lb metal block, placed it in your freezer, left it overnight, take it out the next day and held it in your hand that it would be warm because it was no longer in the freezer.   You don't seem to understand that there is a lag and it will take time for it to settle.   We need to consider the thermal mass of the meter.  The meter's case will provide some insulation. 

If you watched Part 5, where I was talking about the filters and showing some resistors that I had measured, I compared the 121 with some other meters.  Two meters in particular you should note were the UT181A and the highly modified UT61E.  These two meters were not allowed any warmup time before I began to log the data.  Compare them with meters that were allowed minutes to warm up prior to starting the test.    That UT61E would have normally wandered all over the place....

If I wanted to say look at the drift with temperature, where we would compare the internal sensor, I would step the chamber every 2 degs or so, and allow it to settle at each temperature for a half hour or so.   Maybe place a TC inside the meter near it's ambient sensor that we could monitor as well.   Then we could plot all of this data.   We would need to make sure we can record the 121GW's ambient sensor first.   

Someone had posted about how fast I ramp the temperature.   To them, -10 to 60 in an hour is a fast ramp.  It's interesting to hear different perspectives.   We would ramp -40 to 85 or -40 to 125C  in seconds.   Basically the mass of the DUT and the chamber where the limiting factor.   Look up a thermal shock chamber where we have two chambers running at two temperatures and a dumbwaiter running between them.

[newbrain]

Probably the voltage reference is shitty ... pretty obvious

Probably anything but, if it's the one on the circuit diagram.

[dcac]

It’s difficult to tell exactly as this part of your video is so time compressed, and is perhaps the overlayed chamber camera not really in sync, but:

121gw showing 0.899mV at -8.9c, control program (chamber temp) is shoving about 20c:

121gw showing 0.871mV at 20.9c, control program (chamber temp) is shoving about 43c:

121gw showing 0.997mV after 2h in 20.2c “room temperature”

so I get 0.899mV when chamber temp is shoving about 20c and then 0.997mV at 20c “room temperature”, why the big difference?

That temperature drift from -10 to 60C is very concerning, and what the hell is causing it.
...
I just don't get it.

Different parts inside the meter are changing at different rates when I ramp the box, even at this slow rate.  With this meter, it causes a fair bit of error.      I allow it to sit at the two extremes for a half hour to stabilize.  There's not a lot of room in that box and there is a decent sized fan.  At -10 we had 1.003mV, 20ish 0.997mV and 60C 0.928mV.   

But if you leave it in your car overnight in the dead of winter (much colder here than -10C) and bring it inside to use it, expect there to be a fair bit of error until it settles.  Normally we would test something like this with a shock chamber but in my home lab, you have to settle for my cardboard box.

Yes it's VERY compressed.  There's about 6 hours of data, maybe more.  It takes several hours for that peltier setup to cool down to -10.  You get to see the last half hour, the ramp and the hold at high temp.   About 2 hours, compresses into a few seconds. 

I have a sinking feeling that you feel that if you took a 1lb metal block, placed it in your freezer, left it overnight, take it out the next day and held it in your hand that it would be warm because it was no longer in the freezer.   You don't seem to understand that there is a lag and it will take time for it to settle.   We need to consider the thermal mass of the meter.  The meter's case will provide some insulation.

I realize just because the 121gw internal sensor is showing a certain temperature doesn’t mean the rest of the meters internals is at that same temp (yet). But 0.870mV at 20.2C in the thermal chamber seems very far from 0.997mV at “room temperature” 20.2C.

 
As you can see I already mentioned there's a lag.

I still think it's strange that your chamber camera could catch 121gw showing 0.871mV at 20.9C on its internal sensor.


 

[CDaniel]

The thermistor measurement inside 121GW is not that good ... I see some self heating at room temperature so at low temperature could show way higher and then heating faster than the rest of the board .

[joeqsmith]

I realize just because the 121gw internal sensor is showing a certain temperature doesn’t mean the rest of the meters internals is at that same temp (yet). But 0.870mV at 20.2C in the thermal chamber seems very far from 0.997mV at “room temperature” 20.2C.

 
As you can see I already mentioned there's a lag.

I still think it's strange that your chamber camera could catch 121gw showing 0.871mV at 20.9C on its internal sensor.

Why?  Do you feel that the meter is homogeneous in regards to temperature?  There is going to be a slight temperature gradient when looking at the top of the meter near the LCD to the bottom near the leads.   While the chamber is small and offers a fair amount of air flow, as we get closer to the Peltier's output, we are going to see a different temperature than whats going to their input.   Again, one way to remove this effect is to allow everything to stabilize.   You can't take a snapshot while things are changing and expect much more than to note that things are changing...   

The thermistor measurement inside 121GW is not that good ... I see some self heating at room temperature so at low temperature could show way higher and then heating faster than the rest of the board .

At -10C the 121 was showing a much lower temperature, not higher. 

[Fungus]

It even seems like the CAT rating could be bogus.

Fungus, just keep in mind the CAT requires the operator survivability, not the meter.

True, but a loud bang and a jolt in your hand can easily make you fall off a ladder with surprise (or take a step back, put your foot in a bucket, fall over and grab a busbar on your way down...)

If this meter has sparks jumping around inside and tracks being vaporized at 2000V then it's not going to be pleasant to hold in your hand when 6000V@ 3000A hits it (ie. CAT III 600V). I don't know about you but I want a lack of surprises in my tools when I'm around dangerous things.

[joeqsmith]

I don't see a way to get the 121GW to log the secondary display to the SD card but looking at the CSV file, it seems like they have some provisions to support it.  I can however log the internal temperature with BLE.   I already have a way to program temperature profiles into my meat packing box.  These two programs would need to be combined to keep everything synchronous.   

If the goal is to replicated what Dave shows, but using incremental steps in the temperature along with the reported internal temperature,  I think we should use his same temperature range and voltage level.  I see no reason why it would behave any different but we know there are some differences between the meters that were sent to me and what he had tested so maybe.


Getting back to the self heating....  It's not uncommon to let TE warmup prior to use and with some of these meters, like the 121, it does seem to move around a fair bit.   That's once of the nice things about meters like the 181A.  It seems you turn them on and they are ready to go.    Of course, I could say that about the free HF meters too.       Plot shown is meter turned on cold, and allowed to run for a half hour.  About a 2 degree C rise.   

[rsjsouza]

It even seems like the CAT rating could be bogus.

Fungus, just keep in mind the CAT requires the operator survivability, not the meter.

True, but a loud bang and a jolt in your hand can easily make you fall off a ladder with surprise (or take a step back, put your foot in a bucket, fall over and grab a busbar on your way down...)

Oh, certainly so. I was just reinstating something that is an enabler of a very wide wiggle room between something that passes the independent testing and something that is seriously damaged.

If this meter has sparks jumping around inside and tracks being vaporized at 2000V then it's not going to be pleasant to hold in your hand when 6000V@ 3000A hits it (ie. CAT III 600V). I don't know about you but I want a lack of surprises in my tools when I'm around dangerous things.

But, but, but... Everyone LOVES surprises!

[dcac]

I realize just because the 121gw internal sensor is showing a certain temperature doesn’t mean the rest of the meters internals is at that same temp (yet). But 0.870mV at 20.2C in the thermal chamber seems very far from 0.997mV at “room temperature” 20.2C.

 
As you can see I already mentioned there's a lag.

I still think it's strange that your chamber camera could catch 121gw showing 0.871mV at 20.9C on its internal sensor.

Why? Do you feel that the meter is homogeneous in regards to temperature?  There is going to be a slight temperature gradient when looking at the top of the meter near the LCD to the bottom near the leads.   While the chamber is small and offers a fair amount of air flow, as we get closer to the Peltier's output, we are going to see a different temperature than whats going to their input.   Again, one way to remove this effect is to allow everything to stabilize.   You can't take a snapshot while things are changing and expect much more than to note that things are changing...   

I should perhaps have said: it take some time until the rest of the meter internals been affected by the temperatur the sensor is showing. So no of course there are temperature differences all over the PCB. And this probably change even if you just change the orientation of the meter. Turn it i.e. upside down and less heat will affect the NTC.

But still with regard to ambient temperature changes, the differences in temperatures inside the meter will change equally (or nearly equally), if, as you said, you just wait long enough for it to happen.

I can't really comment more than this as I haven't seen the thermal cycling in more real time. It would be interesting to know why this huge drift is happening though.
 

[joeqsmith]

I doubt the meters reported internal temperature has anything to do with the voltage drift.   You could possibly try and monitor a few spots inside the meter while sweeping the temperature to see what changes.  Seems like a lot of work for little gains, but I've spent time on far worse things.

Looks like it takes about a half hour to warm up.  Even after a 10min warmup I was seeing a bit of drift in the resistance measurements.  Again, pretty typical.  The odd ball is really just how stable that UNI-T UT181A is.   

[CDaniel]

Cooling spray and soldering iron , the easy way to detect such problems 

[floobydust]

The thermistor measurement inside 121GW is not that good ...

I think the 121GW CJC circuit is powered (thermistor) from one Vreg (VDD U1) yet the MCU's A/D is off another (VDDP U2). So that drift would be a delta between two vanilla voltage regs. The precision vref is not used. The MCU's ADC input impedance is also 50k so that contributes. The MCU also has a calibrated on-ship temp sensor.

What a pitchfork festival. Ouch 
The main DCV temperature drift would be hard to track down. Even MLCC's going low value at low temperatures, can make noise increase. Have to take the meter apart with test wires out of the temp chamber and check the NJM Vregs and Vref etc. for drift

[dcac]

Looks like it takes about a half hour to warm up.  Even after a 10min warmup I was seeing a bit of drift in the resistance measurements.  Again, pretty typical.  The odd ball is really just how stable that UNI-T UT181A is.   

Yeah this is my impression too, the relatively low power consumption of the 121 is working against it here, should perhaps install a resistor in it just to burn some heat and bring it faster to (more) stable conditions. It could of course also be compensated for in software, if it has a repeatable enough behavior, I wonder if this is done in other MCU driven DMM's.

[CDaniel]

Resistance could have other reasons to drift , the circuit is more complex  , voltage is important for finding the root cause .

[joeqsmith]

White: Chamber's temperature
Red: 121's reported ambient temperature
Green: 121's reported input voltage

I'm using a 1.0 volt reference rather than 4.5.  Data was logged over BLE.  The sample rate is roughly 2Hz (each 5deg step is roughly a half hour to give some frame of reference).

It's pretty much as stable as what Dave had posted using the 5V range.   Standard deviation is roughly 250uV.    Of course this is the Volt scale, not the mV where I test them.   

Notice how close the ambient tracks compared with what I previously show.   There is one major difference.  This is Production 2 which was tested to failure,  back from the dead, sort of.  Production 1 seemed to have a problem with the selector switch when I first looked at it.  I may have a look at it.  The build quality of Production 2 wasn't all the great.  Maybe there is something else wrong with it.

[CDaniel]

So it is good for volt scale , millivolt uses that transil diode D13 . If the leakage current is significant and variable with temperature a small variable voltage divider to ground is formed with the input thermistor + resistor 1K2+1K . Of course just a theory .

[joeqsmith]

So it is good for volt scale , millivolt uses that transil diode D13 . If the leakage current is significant and variable with temperature a small variable voltage divider to ground is formed with the input thermistor + resistor 1K2+1K . Of course just a theory .

I know Dave and a few others were not big on me running tests down in the mV range but I want to see how they look with low signals.  Normally, I would just short the inputs.   

It could very well be the TVS but there is a trend like we saw in the mV range.  Like you said, freeze spray would hunt that one down.    It's certainly much better than the prototype was.   

[dcac]

I can't recalibrate ... for some reason using a 10nF like in instructions result in 0.000nF readings in this range for any cap , so the only option is to reload the saved calibration . Maybe it's a bug , I tried many times even with different values , I don't think I did something wrong 

Also pay attention if you try to re-calibrate the capacitance ranges.
There is no entry for zero value calibration in the small table inside the chapter ZERO OFFSET CALIBRATION.
In the big colored table, though, there is a scrambled entry, 'R1 : ', which might indicate that the first calibration point with open leads sets the offset, and the 2nd one with the nominal reference capacitor sets the gain.
That might explain CDaniels problem, that 0.000 is displayed for a 10nF capacitor, after calibration.

Frank

I'm not sure if the capacitance calibration procedure ever been properly documented, I can't find it anyway, but here is what worked for me:

First make sure you saved current calibration to the SD card.

Then you need a 1nF, 10nF and a 100nF reference cap.

Enter calibration mode, select lowest cap range and connect a 1nF cap, meter will display a value probably somewhat less than 1nF, press Setup to perform offset cal, after count down meter will still display an offset value. Now change the 1nF to a 10nF cap, press Mem to perform gain cal, after count down meter will store both offset and gain calibration and should now display 10nF.

Select next cap range and repeat the procedure but using 10nF to start with and then change it to 100nF.     

The rest of the cap ranges only seem to have gain calibration and if needed use the values from the table in the manual.

Also this procedure is probably only valid for FW 1.57 and up.

[CDaniel]

Yes , is working that way , have you discovered by chance ?  Weird anyways ... the offset calibration is for zeroing the meter with nothing connected for stray capacitance or whatever
Now I have the meter without D13 and recalibrated for caps ... so far is not drifting .

[dcac]

Yes , is working that way , have you discovered by chance ?  Weird anyways ... the offset calibration is for zeroing the meter with nothing connected for stray capacitance or whatever
Now I have the meter without D13 and recalibrated for caps ... so far is not drifting .

Sorry I can’t reveal my sources, you know, I could tell you but then I’ll have to.....

I only re-calibrated my meter to bring the huge negative offset closer to or slightly above zero.

And I would never use it without D13 installed. D13 is essentially the only real protection it has in i.e. mV/Hz/Ohms modes. Even Joe’s grill starter would likely fry your hy3131 in an instant.

Great the cal procedure worked though.

[CDaniel]

I knew it wasn't by chance ... 
That input has in series 3x300Kohm that would drastically limit the current , not 10M like the main input but still ... and on the pcb is the place where the 2 transistors clamp could be soldered .
Anyway better to burn it , than to see that stupid drift