Handheld meter robustness testing

[CDaniel]

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 

[dcac]

Come to think of it, when I measured the waveform 121gw puts out in caps mode, its been some time ago but I seem to remember the frequency of that waveform was much lower than it should have been according to the way 121gw setup the hy3131. It was like a 'hidden' capacitor was present slowing the waveform down. But D13 is not directly in parallel with the input jacks, but still its absence could perhaps affect the circuit more than what the calibration routine can handle.

[Dr. Frank]

Hey Joe, dear dcac, CDaniel,

Maybe you remember, that I made a strict review on the basic characteristics  of the 121GW: https://www.eevblog.com/forum/testgear/eevblog-121gw-discussion-thread/msg1687298/#msg1687298
where I quantitatively verified the specifications of DC/AC voltage and current and Ohm.
My 121GW is from the 2nd batch from Kickstarter, delivered in July 2018 with FW 1.22, now 2.02, and its hardware is still very similar to Joes latest 121GWs.

Very soon I observed and described several problems with the Ohm mode, especially, that negative Ohm values are clipped to zero, so there is a 'hidden' offset (originating from initial calibration), which affects the accuracy of low Ohm measurements, as this offset can't be zeroed.

That clipping effect of Ohm mode is obviously removed latest with FW 2.02, but still present on the capacitance mode.
That explains the big errors you all see with small pF capacitors. This is evident as a threshold effect in capacitance value, i.e. the calibrated zero value plus the actual offset will always be subtracted from the real value. Capacitors smaller than this threshold will be displayed as 0.000pF, and bigger ones will show too small a value.

If I let my 121GW sit for a while in the 10nF range, it will possibly display some pF offset with open leads, and if I then zero the reading, I always get very precise values, even with pF capacitors (e.g. 39pF).
If that obvious (zero) drift is zeroed repeatedly between subsequent readings, then the measurement of the capacitors value is consistently precise, better than the specified 2.5%.
Long measurements on this zero drift make no real sense, I think.

Btw.: I can really check the accuracy by my DE-5000 LCR meter, which is specified down to 0.3% accuracy, depending on range and test frequency.

Interestingly, the actual manual (9/2019) still specifies this mode as:




I loaded FW 1.17, and indeed, the resolution then was 3 digits only, i.e. 10pF resolution in the lowest range.

That also means formally, that +5 digits, or 50 pf variation are allowed.
So do not expect too much of accuracy, because of this misleading specification, or because of the additional digit resolution.
The specification might be updated / improved to the 4 digit resolution, which seems to be quite reasonable, if that threshold problem can be solved.

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.


Similar aspects hold for the 50.000 Ohm range, as this has even +20 digit or 20 mOhm  specified deviation.
CDaniel criticized in another thread:
'2 - in 50ohm range the last 2 digits are pretty useless . Every time you short the leads you get different reading and fluctuating'

At first, I think that you simply measured the varying contact resistance of your probe or short, whatever you have used.
This has nothing to do with the 121GW itself.

Instead, I would say in a general manner, that it is very difficult to make reasonable Ohm measurements in the 1..50 mOhm range with the 2W method only.
The 4W / Kelvin method is required to safely cancel lead and contact resistances.

Additionally, the 121GW uses about 500µA in this range only, i.e. 1 mOhm equals about 0.5µV, so proper cancellation of thermo voltages is also required, i.e. usually the OFFSET COMPENSATION function does this job.

By using solid cables, and proper zeroing (REL), e.g. by use of an appropriate short, one can cancel both, contact resistances and thermo voltage, even on the 121GW.
By this method, and in contrast to CDaniels tests, I get quite accurate and stable readings at 100 mOhm and upper settings on my decade resistance box, and the deviation or fluctuation from nominal are a few mOhms only. Frequent zeroing is also required here.

You can as well measure/detect the variation of the contact resistance of the switches.
So that low resistance resolution can be quite useful, although I would always prefer an appropriate instrument for such measurements, like my 34465A.

Frank

[CDaniel]

So basically what you are saying , that drift is acceptable bacause we can zero it and still in specifications ? Sorry but for some people is not good enough  . What meter beside this do you use and is drifting ?

I did the cap calibration both way , zero offset and then full scale gain with 10nF or just full scale gain . Same result , dead range .

We understand that this is not 4 wire resistor mesurement , but the crappy design with low current is making things much worse that could be . Indeed , the voltage across the DUT resistor is so low that is comparable with every solder joint thermo voltage . You can verify this by heating the input jack , or every solder joint in the signal path . The value is drifting very much .

[joeqsmith]

In Part 4, we compare the 121GW's bargraph, 3dB point and freq counter with other DMMs.   We then attempt to align one if the old production meters using the original procedure.   We finish up repeating  one of Dave's recent tests.   There's a few other other bits sprinkled along the way. 

[Dr. Frank]

So basically what you are saying , that drift is acceptable bacause we can zero it and still in specifications ? Sorry but for some people is not good enough  . What meter beside this do you use and is drifting ?

Please, do not imply things, which I have never said.

Whenever you make measurements with high resolution and for high precision, it's always good metrological practice to first zero the meter, or to cancel offsets by other methods. That is even mentioned in the manual of the 121GW for the 50 Ohm range, and for 50 mV_dc, but also in manuals from more sophisticated instruments.
 
I have provided an explanation and effective methods which will help users to achieve precise measurements with the 121GW, by mitigating its physical or circuit limitations.

I see no use case for making drift measurements, especially if you simply want to determine the value of a capacitor, or of a resistance. I also can not reproduce such an excessive drift, that would prohibit precise measurements.

We understand that this is not 4 wire resistor mesurement , but the crappy design with low current is making things much worse that could be . Indeed , the voltage across the DUT resistor is so low that is comparable with every solder joint thermo voltage . You can verify this by heating the input jack , or every solder joint in the signal path . The value is drifting very much .

Again, I do not share your opinion about the design.
It is like it is, and regardless of its shortcomings, it's anyhow possible to make useful and quite precise low level measurements, especially in relation to the form factor and price area of this DMM.

It's as well a good metrological practice to avoid or cancel thermo voltages in your measurement setup, so I do not understand, what you want to contribute to this aspect.


Maybe you should simply sell your 121GW, if you are personally dissatisfied.
Your permanent complaints, w/o any constructive hints / ideas really do not help anybody here in this forum.

Frank

[joeqsmith]

I have included my attempt to reproduce Dave's speed test in Part 4.   Dave does claim the "...I checked, the Keysight "drifts" in a similar way,.... "
https://youtu.be/-eUkSufMK5A?t=460
I overlaid Dave's two meters to better show how the 121 compares with the Keysight meter.    I also included more information about my test.   

I also show the effect of aligning the capacitance function of the old production hardware using the procedure Dave supplied with the prototype.
https://youtu.be/-eUkSufMK5A?t=905 

Looking in the latest manual for the 121GW, page 53, VA connects is not going to yield good results.  I covered it over 2 years ago.  You would think someone would have taken the time to correct it by now. 

The biggest problem I saw so far with the old production hardware, with 1.57 or 2.02 firmware installed was that it failed to give any indication to the user that  high voltages could be present in all conditions.  IMO, this is really bad.    A fellow member had pointed out that I had seen this with the prototype as well.   I guess a few of you remember important things like this.    This should NEVER happen.

The attached is from EN 61010-2-033:2012.   See note b).    Seeing stuff like this and the Gossen with its latching relays getting through chips away at my trust in them. 

[rsjsouza]

The biggest problem I saw so far with the old production hardware, with 1.57 or 2.02 firmware installed was that it failed to give any indication to the user that  high voltages could be present in all conditions.  IMO, this is really bad.    A fellow member had pointed out that I had seen this with the prototype as well.   I guess a few of you remember important things like this.    This should NEVER happen.

The attached is from EN 61010-2-033:2012.   See note b).    Seeing stuff like this and the Gossen with its latching relays getting through chips away at my trust in them.

(Sorry, I am having trouble to catch up with all things...)
Do you mean that 121GW fails to present a OL on the display? Or only on certain ranges?

[joeqsmith]

The biggest problem I saw so far with the old production hardware, with 1.57 or 2.02 firmware installed was that it failed to give any indication to the user that  high voltages could be present in all conditions.  IMO, this is really bad.    A fellow member had pointed out that I had seen this with the prototype as well.   I guess a few of you remember important things like this.    This should NEVER happen.

The attached is from EN 61010-2-033:2012.   See note b).    Seeing stuff like this and the Gossen with its latching relays getting through chips away at my trust in them.

(Sorry, I am having trouble to catch up with all things...)
Do you mean that 121GW fails to present a OL on the display? Or only on certain ranges?

[Monkeh]

Not in any defense of any issues with meters, and this has probably been pointed out before..

These sorts of issues (the Gossen issue being a grand example) are exactly the reason safe isolation procedure involves verifying the meter. Prove meter, test circuit, prove meter, rinse and repeat.

[rsjsouza]

So, it is not really a problem with indicating an overvoltage condition but instead a problem in effectively finding the correct range to display. Given the read voltage on production #2 meter is so off, I imagine the circuitry/firmware completely failed to read the inputs. 

Given the cost of every certification run is quite high, I have my doubts if every firmware revision goes through that process...

[joeqsmith]

So, it is not really a problem with indicating an overvoltage condition but instead a problem in effectively finding the correct range to display. Given the read voltage on production #2 meter is so off, I imagine the circuitry/firmware completely failed to read the inputs. 

Given the cost of every certification run is quite high, I have my doubts if every firmware revision goes through that process...

Guessing you skipped a bit as I clearly show the problem is there even with 1.0 firmware installed.    This problem dates back before the prototype 121GW I had looked at.   I didn't make a big deal about it back then as it was, after all, a prototype.   I should mention that someone recently wrote who had remembered seeing that early video.   

Showing a low voltage in the presence of hazardous levels is exactly one of the examples used to describe an ambiguous indication.   Again, I am not a safety expert and there may be a reason conditions like these get a pass.  Personally, I think it dilutes the cert.   

I would expect companies that take these safety standards serious re-certify their product as changes made.   

[Kosmic]

Apparently the last DMM made by Solartron (SI 7063) had a similar problem. In rare case the decimal point was not placed at on the right spot. So, you could read 10V in presence of 1000V. Since the bug was in the ASIC, they endedup scrapping the whole product and destroying all produced SI 7063.

https://www.eevblog.com/forum/testgear/my-_new_-old-schlumberger-si-7063-dmm/msg2591991/#msg2591991

[rsjsouza]

So, it is not really a problem with indicating an overvoltage condition but instead a problem in effectively finding the correct range to display. Given the read voltage on production #2 meter is so off, I imagine the circuitry/firmware completely failed to read the inputs. 

Given the cost of every certification run is quite high, I have my doubts if every firmware revision goes through that process...

Guessing you skipped a bit as I clearly show the problem is there even with 1.0 firmware installed.    This problem dates back before the prototype 121GW I had looked at.   I didn't make a big deal about it back then as it was, after all, a prototype.   I should mention that someone recently wrote who had remembered seeing that early video.   

I watched the entire video, but I just mentioned #2 as it was the first that showed teh issue.

Showing a low voltage in the presence of hazardous levels is exactly one of the examples used to describe an ambiguous indication.   Again, I am not a safety expert and there may be a reason conditions like these get a pass.  Personally, I think it dilutes the cert.   

My observation is simply my guess as to why there is no OL/hazardous indication: not a deliberate omission but I believe the meter is simply unaware of a high voltage on its inputs due to the autorange bug. In the end, the effect is the same regardless of the goof up.

I would expect companies that take these safety standards serious re-certify their product as changes made.   

Completely agree, although I suspect the rate of firmware releases does not allow for such intense testing - it costs real $$$ real fast. (although I didn't follow the FW evolution that close)

[dcac]

Hey Joe, dear dcac, CDaniel,

That clipping effect of Ohm mode is obviously removed latest with FW 2.02, but still present on the capacitance mode.
That explains the big errors you all see with small pF capacitors. This is evident as a threshold effect in capacitance value, i.e. the calibrated zero value plus the actual offset will always be subtracted from the real value. Capacitors smaller than this threshold will be displayed as 0.000pF, and bigger ones will show too small a value.

If I let my 121GW sit for a while in the 10nF range, it will possibly display some pF offset with open leads, and if I then zero the reading, I always get very precise values, even with pF capacitors (e.g. 39pF).
If that obvious (zero) drift is zeroed repeatedly between subsequent readings, then the measurement of the capacitors value is consistently precise, better than the specified 2.5%.
Long measurements on this zero drift make no real sense, I think.

Btw.: I can really check the accuracy by my DE-5000 LCR meter, which is specified down to 0.3% accuracy, depending on range and test frequency.

Interestingly, the actual manual (9/2019) still specifies this mode as:




I loaded FW 1.17, and indeed, the resolution then was 3 digits only, i.e. 10pF resolution in the lowest range.

That also means formally, that +5 digits, or 50 pf variation are allowed.
So do not expect too much of accuracy, because of this misleading specification, or because of the additional digit resolution.
The specification might be updated / improved to the 4 digit resolution, which seems to be quite reasonable, if that threshold problem can be solved.

Frank

Hi Dr. Frank

If I connect a 330pF mica capacitor to my 121gw to lift offset above the zero clipping, it will initially read 178pF and within minutes increase to about 195pF. This is without touching or holding the meter so the increase is solely from internal temperature rise in the 121gw. If I let the meter sit for more than 20 minutes or so it will very slowly start to approach to correct value of 330pF. I’m fairly confident I can contribute this behavior to capacitance drift in the D13 TVS diode, but wouldn’t be surprised if there are 121gw’s were D13 is not at all as bad as this, I got an unlucky D13 I guess.

The 121gw user manual doesn't seem to mention any warmup period, not even for the calibration procedure, from when to expect the meter to be accurate/stable. But from using many different handhold DMM’s I’ve come to expect them to be fairly accurate after 3-5 minutes, usually they are accurate much earlier than this.

Edit: Not so sure it's D13 after all, see post below.

[Dr. Frank]

Hi dcac,
Thank you for your observations.

On my 121GW, I see a similar behavior (as briefly indicated), but not that extreme.
Initially, it also reads exactly 0.000 nF, and either after a while, or after using other modes, it will come up with some non zero reading of several 10pF.

Only then, it's possible to use REL, and get quite precise measurements.
Mine does not drift that much and not so long.

Probably it's not the leakage of the TVS, but its capacitance.
If I interpret it correctly, the SM6T22CA has around 1nF itself @ 1V, (see diagram), which will be in parallel to the DUT, and this high capacity offset will probably be calibrated into the initial zero cal point.
That 1nF capacitance will slightly vary over temperature and voltage, but that transforms into a big variance on pF scale, which explains the observed effects.

If you replace that TVS by something else, a new zero calibration is required.

Therefore I would greatly appreciate, if Joe could publish the correct calibration procedure for zero, which would reset the threshold.

Frank

[CDaniel]

I wrote how much the reading is changing , from 510pF to 200pF , but the transil diode has only 50pF measured with a LCR bridge at 100Hz , 0V bias . It's not straightforward , probably has some leakeage resistance and the capacitance and leakage resistance is dependent on the voltage , frequency applied and of course temperature .
The maximum drift i saw in my meter was about 80pF , typically is lower , but extremely annoying to know that you have a meter that drifts . This a matter of principle , not that I don't have the means to measure caps .
When the calibration procedure for caps will become available I will investigate more .

[dcac]

Well I took the back cover of my 121gw and repeated the test with the 330pF mica and the behavior started out just as before. Then I put my finger directly on D13 probably elevating its temperature several degrees expecting the reading to quickly increase, but no change at all was really noticeable, all that happened was the reading jumping around from interference of my finger touching the circuit, but as soon as I removed it the value stabilized again. But still slowly increasing just like before. I had already checked the mica cap with another meter and it was perfectly stable. Now I really don’t know. I tried breathing some hot air over the PCB and that did seem to have an effect but from what component I could not tell.

[CDaniel]

Remove it and confirm or not that the drift is gone .
I didn't said temperature is the main cause for the drift , the capacitance is changing with temperature but I touch it with the soldering iron .
The capacitance is changing with the voltage too , could be a small DC voltage across it slowly creeping up or down and when you cycle trough functions is reseting to the initial value ...
Just don't try to recalibrate wiithout saving first the calibration data 

[joeqsmith]

Have those of you wanting the old procedure requested it from Dave?    I haven't seen anything from him after I brought it up.   

I have started to work on part 5 (firmware).    My plan is repeat the same tests I did with the prototype with a few changes.

1.9Mohm resistor stability/noise for various versions of firmware.    This 1.9M was something I had setup to repeat one of Frank's tests.   After seeing Dave's mechanical switch test,  I also plan to use the 40 ohm.  I doubt I will check both meters as the hardware appears to be the same between them.    Depending how things go, if it looks like the old production hardware behaves the same as the prototype, I may not run all the versions of firmware.   

If there is anything else you would like to see, feel free to ask.   

******
It appears 1.05 is when they introduced the extra digit for the 50 ohm range. 

[dcac]

******
It appears 1.05 is when they introduced the extra digit for the 50 ohm range.

Not exactly sure what you mean but it's been my understanding all FW released to the public, or since 1.02 anyway, have had the same resolution in all ohms ranges and 50 ohms range reading 50.000 ohms.

[Muttley Snickers]

If there is anything else you would like to see, feel free to ask.   

At the start of your last video where you had all the meters lined up for comparison I thought perhaps you may have been able conduct a similar test but set all the meters for Min/Max so as to gauge which meters can not only display a reading but capture it as well.

[joeqsmith]

******
It appears 1.05 is when they introduced the extra digit for the 50 ohm range.

Not exactly sure what you mean but it's been my understanding all FW released to the public, or since 1.02 anyway, have had the same resolution in all ohms ranges and 50 ohms range reading 50.000 ohms.

I went from 1.00 to 1.05.   I will add 1.02 to the mix and see if it also has the higher res.
***
Yes, 1.02 does indeed support the higher resolution as well.

[joeqsmith]

If there is anything else you would like to see, feel free to ask.   

At the start of your last video where you had all the meters lined up for comparison I thought perhaps you may have been able conduct a similar test but set all the meters for Min/Max so as to gauge which meters can not only display a reading but capture it as well.

The one where I talked about the 61010 standard and applied the DC biased 60Hz AC waveform?  If so, I am not sure that the meters can all capture the min/max in the modes I was using.   Let me know and I will check. 

[Muttley Snickers]

Yes that was the test, I frequently use this function and wondered what the result would be from a meter which was unable to settle or stabilise for the capture to occur.