EEVblog 121GW Multimeter Issues

[Brumby]

Yes it is normal.

Nothing to worry about at all.

[Dr. Frank]

I attempted to repeat your test the best I could.  I am using a few Caddock TK series resistors to make up roughly 1.9M.   The two plots shown are my HP34401A (white) compared with the pre-production 121GW (red).   I have also included the CSV file off the meter (renamed to .txt to allow uploading).  The recording interval is 1 second for both meters.

Looking at the schematic, nothing pops out other than you should have a better reference.  A bit odd you don't see the problem in every range. 

Added your data set for completeness.

Hello Joe,
thanks for your tests.. the measurement of the 1.9M inside the 5450A by a 34465A, or a 3458A at NPLC 10 is as stable to the last digit, as yours with the 34401A, but w/o drift.
I could replicate that instability with other very stable 1M resistors and short, drilled cables also, but sometimes, I also get stable readings.. very strange.

So I wonder, what you mean by 'you should have a better reference', do you mean the 121GW internal reference resistors?

I doubt that these are the root cause for this instability problem.. because the Ohm mode is usually realized by a ratio measurement, so I suppose, that it's caused by the AD1.
And yes, this instability is visible part time in other ranges also, surprisingly in the 50 range, although that's driven by a high 470µA current.

Maybe you still have an old firmware running. In the latest 1-22, the counter register for the sampling frequency might be set improperly. Otherwise, 50 and 60Hz should be suppressed by 120dB @ 5Sa/sec, but 75dB only, when using a faster sampling rate... that could also be the case, as the sampling seems much too fast.
 
Maybe something inside the Ohm circuit is oscillating, or super sensitive, depending on the hardware version..

 I did not fully understand the schematics, either, and also the HY3131 datasheet is really crap., still  searching for an Application Note.

Anyhow, I really think, that this problem could be solved by software.

Frank

[joeqsmith]

I attempted to repeat your test the best I could.  I am using a few Caddock TK series resistors to make up roughly 1.9M.   The two plots shown are my HP34401A (white) compared with the pre-production 121GW (red).   I have also included the CSV file off the meter (renamed to .txt to allow uploading).  The recording interval is 1 second for both meters.

Looking at the schematic, nothing pops out other than you should have a better reference.  A bit odd you don't see the problem in every range. 

Added your data set for completeness.

Hello Joe,
thanks for your tests.. the measurement of the 1.9M inside the 5450A by a 34465A, or a 3458A at NPLC 10 is as stable to the last digit, as yours with the 34401A, but w/o drift.
I could replicate that instability with other very stable 1M resistors and short, drilled cables also, but sometimes, I also get stable readings.. very strange.

So I wonder, what you mean by 'you should have a better reference', do you mean the 121GW internal reference resistors?

I doubt that these are the root cause for this instability problem.. because the Ohm mode is usually realized by a ratio measurement, so I suppose, that it's caused by the AD1.
And yes, this instability is visible part time in other ranges also, surprisingly in the 50 range, although that's driven by a high 470µA current.

Maybe you still have an old firmware running. In the latest 1-22, the counter register for the sampling frequency might be set improperly. Otherwise, 50 and 60Hz should be suppressed by 120dB @ 5Sa/sec, but 75dB only, when using a faster sampling rate... that could also be the case, as the sampling seems much too fast.
 
Maybe something inside the Ohm circuit is oscillating, or super sensitive, depending on the hardware version..

 I did not fully understand the schematics, either, and also the HY3131 datasheet is really crap., still  searching for an Application Note.

Anyhow, I really think, that this problem could be solved by software.

Frank

I was just comparing the two versions of the schematics and noting what I saw.  ZD1 and circuit surrounding it was changed.   U14 is no longer populated.   D14 clamp was also changed.  There could be other changes.  These are just what I noticed.   

As mentioned in a previous post, the firmware is pre 1.0.  Also, I had responded to your original post previously where I tried a few quick tests at 50 and 60Hz. 

The parts I used for this test were not temperature controlled.  No towels were used.  I didn't let the HP warm up for very long and the room temperature changes.   Its stable enough to show your 121GW doesn't behave like the pre-production unit I have.  Maybe down the road after you get this first problem sorted out we can look at drift.   

If there was an easy way to backup the firmware on this meter using the tools I have, I would attempt to load the same version of firmware you are running.  At least split the problem in half.   I am not even sure if Dave would have the images for this old firmware available.   That may be an option as well.

[joeqsmith]

I wrote Dave about the obtaining the older image for this meter.  If he provides it, I will try to repeat the noise test with the same version of firmware. 

In the meantime, I had let the original setup stabilize and changed the value of the small resistor to bring it in a little tighter.   I used some epoxy mixed with heatsink compound and glued a section of copper to the resistor network and made a case from Pyropel wrapped in alumimum foil to keep the air off it.   Still not temperature controlled.   I let the HP warmup  for 2 hours and ran another sweep.   I had left the 121 on for about 10 minutes this time before I started to collect the data. 

The HP (red) still has a bit of drift but again, I am not using any sort of temperature control.   I would have expected the pre-production 121GW to read a lower resistance, not higher (blue).   But from what I saw, the meter was fairly sensitive to temperature which I assume is why they changed the reference. 

It looks like if I subtract off the pre-production meter's drift, we are down to maybe a count of noise.  After a half hour, it was getting pretty stable.    I have attached the raw data CSV file. 

***************  Other Data *****************

Sadly, I never purchased the interface cable for my Brymen but I thought I would show how a few other meters  compare with the pre-production 121GW.   The UNI-T UT61E had really bad temperature drift and I had compensated it (plus it has several other mods).  The UT181A was the most stable meter I had tested.   For these two meters, I did not allow any warmup time.   

The CEM DT9939 (top yellow) has a pretty bad drift problem as well so I let it warm up for 10 minutes. 

And just for the fun of it I tried the Gossen MetraHit Ultra (dark blue).   This meter has Netic in the back along with an aluminum shield in the face that I added to help tame it.   

***************  And still more data  *****************

I repeated your test at 500K.  I tried both ranges.   I didn't have a clean way to put together a 19M but I do have some low tempco 20M 2ppm parts.     

As was mentioned, the 121GW production as well as the pre-production meters are sensitive to static fields.  Looking at the 20M data,  the first spike is me walking away from the meter.   The second two spiked are me walking out and back into the office.  The last spike is walking up to the meter to turn off the logging. 

It would be interesting to see you repeat your test once the noise problem is sorted out.  I remember the pre-production meter not being all that stable with temperature.  They didn't change the reference for the fun of it. 

[firewalker]

Is this the place to report issues or is there a more official place?

[Brumby]

Here is good.

[eV1Te]

I received my meter today and I am quite pleased with most aspects of it. However the SD card logging seems broken. I have the original firmware 1.22 that came installed.

I can start the logging by holding MEM, and stop it by holding MEM again, however the saved CSV file do not contain any time stamps, so it is impossible to know when the measurements were taken.

Also looking at the display of the number of points stored during the logging, I can see that the time between each save/point is not constant, sometime it saves several points quickly, sometimes it takes longer time between points. Hence it is not possible to calculate the time for each point based solely on the "Point No."

If there is a fix for this issue that I have simply missed, please let me know.

Edit:
I noticed more things with the logging function that seems poorly optimized.
1. The function/mode that you are currently logging is printed on every row in the CSV, this is a huge waste of memory space since it could easily just be printed once at the top of the file. Switching functions during logging stops the logging, so it is not possible to have several functions in one CSV in any way.
2. I can not get it to log the values in the secondary display. For example I connect a thermocouple that is not Type K and I measure mV and have the meters internal temperature in the secondary display. I intend to process the data in post and calculate the real temperature based on the mV and the internal temperature, this is because the meter only supports type K thermocouples natively. But no matter what I do and press I can not get it to log the internal temperature.

[MadModder]

Both date and time is on the first row in the file, then it says what interval is used. So it is not impossible, just somehwat cumbersome. Add another column, copy the time from the top row, then next row is the one above+(interval in seconds)*(1/86400). If using excel or the like. Otherwise I suppose a script could be used...
But it would be very easy codewise to implement the timestamp into the log file, so we don't have to mess around like that.

[Dr. Frank]

..  I updated the pre-production meter to 1.02 then saved the calibration data.  I then proceeded up upgrade the meter to 1.22.   

I remember reading something early on about the firmware may have required the meters to be realigned.   This appears to be the case but it should be good enough to see if running the latest firmware will cause the noise problem.   

...

*********** Update *************

...

One last note,  it does appear that at least with a 0.5 ohm resistor, with the 1.22 firmware installed, the readings are VERY unstable.  The meter will display 0.5 and noise for the other two places.  After downgrading, the meter again changes a couple of counts at most.   So it does appear that the noise people are seeing could very well be due to the firmware.   The other thing I noticed is the display update rate for the low values is much faster with the 1.22 firmware.

Joe, thank you very much, once again.

So you could at least confirm that the 50 Ohm range is noisy / unstable with FW 1-22, and that the sample rate is too fast.

I did it the other way round, downgraded to 1-05, but the readings in 5M range were as unstable as with 1-22.
I did not check the 50 Ohm range, though.

Then on my bench in the basement, I arranged the 121GW, the cables and a small resistor box with a 1M Ohm PWW, T.C. 1 resistor inside, so that the reading was stable to +/- 2 counts at most. Moving the whole setting 20 cm to the left or to the right, or changing the angle of the cables from perpendicular to horizontal on the bench, created again that heavy noise disturbance.

In that 'quiet' configuration, when I approach the cables with a mains transformer (inside a table lamp @ 50Hz, no SMPSU), the reading again will go wild.
That's the final proof for me, that 50Hz pickup really creates these misreadings, i.e. also the gain shift in the 50M range.

I will try to make a setup to quantitatively measure the mains suppression at 50Hz and at 60Hz, probably by a coil around the measuring cable, and inducing controlled levels of 50 and 60Hz signals from an oscillator.

Frank

[AccountRemovedPerUsersRequest]

121GW just arrived to Scandinavia. One detail paid my attention: I barely can read red labels. Too dark tone compared to the background.

Axel.

[eV1Te]

Both date and time is on the first row in the file, then it says what interval is used. So it is not impossible, just somehwat cumbersome. Add another column, copy the time from the top row, then next row is the one above+(interval in seconds)*(1/86400). If using excel or the like. Otherwise I suppose a script could be used...
But it would be very easy codewise to implement the timestamp into the log file, so we don't have to mess around like that.

The default setting is to sample and save points as fast as the ADC allows, hence the interval is not fixed but depends on the mode, range, etc. It is therefore impossible to calculate the time for each point using the default setting.

However it is true that if you change the settings and limit yourself to a slower sampling speed, i.e. not using the full potential of the meter, it does seem like the timing is consistent so that can assume a fixed number of seconds between each point. I tried it over 4 minutes using a stopwatch and I could not detect any drift or dropped points by eye at least. Hence I guess this is the recommended method for now.

In order to allow for faster sampling I guess you need the actual time in a column (seconds since start is probably easiest), as it would make analysis much easier without having to process the time manually every time. Also as a separate improvement I would remove the "mode" from being printed on every row, either print it once in the header or only print it when it changes (if you intend to include mode switching during logging in the future). This would save approx 10-30 % in space depending on the mode used.

I also noticed that the meter does not autorange during logging (it goes into manual ranging when starting the logging process). This is definitely a feature that I would like to see in future updates.

Edit:
More issues I noticed today:

• The claimed "update rate" is 5 samples/s nominal, however my display only updates 1-2 times/s, hence the meter feels slow and unresponsive, is this an error in the manual or is the meter supposed to be faster?
• When manually selecting the 10 A current range, the display tells you that you are in the 1000 A range. Perhaps there is no 10 indicator in the LCD and this is intended, if that is the case please update the manual.
• There is a 1000 V range, however the manual specifies maximum 600 V for the input protection. Perhaps this is the same issue as above? Please update the manual.
• There are no range indicators in the display for Frequency, Period, Capacitance. Perhaps indicators is lacking in the LCD? In that case I suggest to always show leading zeros in these modes when in a manual range. For example the 100 nF range, write 07.6 nF, instead of just 7.6 nF in order to indicate the amount of headroom available.

[Brumby]

One detail paid my attention: I barely can read red labels. Too dark tone compared to the background.

This is the reason I changed the colour of the red markings on my recent crap meter refacing project to something more orange.  Pure red on grey or black does not have the greatest contrast.

[MadModder]

The default setting is to sample and save points as fast as the ADC allows, hence the interval is not fixed but depends on the mode, range, etc. It is therefore impossible to calculate the time for each point using the default setting.

However it is true that if you change the settings and limit yourself to a slower sampling speed, i.e. not using the full potential of the meter, it does seem like the timing is consistent so that can assume a fixed number of seconds between each point. I tried it over 4 minutes using a stopwatch and I could not detect any drift or dropped points by eye at least. Hence I guess this is the recommended method for now.

In order to allow for faster sampling I guess you need the actual time in a column (seconds since start is probably easiest), as it would make analysis much easier without having to process the time manually every time. Also as a separate improvement I would remove the "mode" from being printed on every row, either print it once in the header or only print it when it changes (if you intend to include mode switching during logging in the future). This would save approx 10-30 % in space depending on the mode used.

I also noticed that the meter does not autorange during logging (it goes into manual ranging when starting the logging process). This is definitely a feature that I would like to see in future updates.

Edit:
More issues I noticed today:

• The claimed "update rate" is 5 samples/s nominal, however my display only updates 1-2 times/s, hence the meter feels slow and unresponsive, is this an error in the manual or is the meter supposed to be faster?
• When manually selecting the 10 A current range, the display tells you that you are in the 1000 A range. Perhaps there is no 10 indicator in the LCD and this is intended, if that is the case please update the manual.
• There is a 1000 V range, however the manual specifies maximum 600 V for the input protection. Perhaps this is the same issue as above? Please update the manual.
• There are no range indicators in the display for Frequency, Period, Capacitance. Perhaps indicators is lacking in the LCD? In that case I suggest to always show leading zeros in these modes when in a manual range. For example the 100 nF range, write 07.6 nF, instead of just 7.6 nF in order to indicate the amount of headroom available.

Yes, you're absolutely right about the "free flow" logging. Didn't think about that. That is in this format impossible to use for example in a diagram with a time axis. Or to just reference if one wants to point at a specific time when searching for irregularities. In some cases I would even go so far to say useless.

I can not find any numbered range indicators on the display other than 500 and 1000. (max contrast and backlight helps). 500 is three segments. But the 1000 I believe is only one big segment? 
  for leading zeroes instead.

And there's a lot in the manual saying things are set in a way as default, when it's not. For example it says logging delay is 1 sample/s, when it is 0. It describes things as if the beep is on by default. It is off.

[Dr. Frank]

I've tested the 50/60Hz mains suppression of the Ohm ranges.
At first, there's a log of this 'quiet' arrangement, about +/- 5 counts noise, although the direct reading is more like +/- 1 count.

Then I coupled a noise signal via a ferrite core into one of the Ohms cables.

That creates now +/- 100 counts of noise readings, but there is no difference between 50 or 60Hz.
I alternated between 50 and 60Hz each 50 samples.

So I conclude, that for Ohm mode, there is no proper mains suppression programmed into the HY3131.

I have read about delta-sigma converters, and these do not have a natural mains suppression like dual / multiple slope converters.
Due to the architecture, they need a digital filter, which has to be programmed to the 48...62 Hz frequency range to notch these by about 120dB, as given by the datasheet.

Obviously, that fails at Ohm mode, as DCV suppresses 50..60Hz AC signals down to the last digit.

I have no problem at all with the MEM function, all logs were done correctly, no failure happened up to now.

But I have measured the logging interval at ln 0 to be 4 Sa/sec, or 250ms, but not 5 Sa/sec, as implied in the manual.
That's important to know, if one makes timed logging of data.

Frank

[Woodsy]

Arrived this morning. Flabbergasted. Cant use it, NO PROBES and no zippy-up case thing...

Package contained:
No bullshit packaging box.
One piece of cardboard with a serial number.
Bubble wrap sleeve containing - 121GW MM with blue case - Batteries already installed.
Bubble wrap sleeve containing - One thermocouple.

Newsletter said: "The 121GW is finally in stock for all those who have constantly been asking.
AND they are on special AND come with a free carry case for this first offer!"

Website says "NOTE: The meter may come with either Brymen (gold) probes or UEI (nickel) probes depending upon availability at the time. Both are high quality silicone rubber probes."

Oh, and the phone number on the "Certificate of Conformity".. The number you have dialed is not in service.

WTF?!

EDIT: For complete transparency there was also a sachet of silica gel and a piece of protective plastic on the LCD.

EDIT2: Package appeared to be sealed when I opened it.

EDIT3:

[EEVblog]

Arrived this morning. Flabbergasted. Cant use it, NO PROBES and no zippy-up case thing...

Sorry about that, I screwed up.
I used meters from a box that I thought were fully packed with the cases and probes 
This was only Oz orders.
Everyone will be shipped probes and a case.

[Zucca]

What other DMM use the HY3131?

Uni-T UT171A... then?

[firewalker]

Can someone clarify the DC+AC mode? Should the meter show on the secondary display the AC or  DC component? Are the segments in red circle for that purpose?\




Alexander.

[HKJ]

Can someone clarify the DC+AC mode? Should the meter show on the secondary display the AC or  DC component? Are the segments in red circle for that purpose?\

AC+DC mode means the meter measured the AC and the DC part of the voltage and then sums them together before showing them.
The meter cannot show both AC and DC at the same time.

[EEVblog]

What other DMM use the HY3131?

Keysight U1282A

[dcac]

What other DMM use the HY3131?

Keysight U1282A

Has anyone noticed similar stability issues with resistance measurements in U1282A?.

It seems odd if only Keysight could get the hy3131 working reasonably well - and UEI can’t.

[firewalker]

What the AC DC sgments next to the secondary display are for? Many manuals explain all the segments of the screen. It would be nice touch.

Alexander.

[HKJ]

AC+DC mode means the meter measured the AC and the DC part of the voltage and then sums them together before showing them.

That's not true.   To avoid having Dave remove this post as a non-issue, the manual should be updated to show it as a feature and the proper equation should be added as well.

That I leave out the actual formula to use for summing do not make it wrong.

What the AC DC sgments next to the secondary display are for? Many manuals explain all the segments of the screen. It would be nice touch.

Hopefully for a software update sometime in the future (Many meters have displays with unused segments).

[AccountRemovedPerUsersRequest]

Brumby: Yep. That color looks better.
I can't read red labels on my workbench at all. Agilent orange is fine to read tho.
I would not expect such an issue because they're supposed to be number of protos and testers thru the dev process.

A.

[Dr. Frank]

Hello seppy,

here's a last erratum:

Page 70:

5. “IAP” will appear on the screen, then press <SETUP> to begin the firm-ware upgrade.

=> remove whole paragraph '6.' and renumber '7.' to 6.

explanation: After switching on with HOLD and MEM pressed, the 121GW will directly start in the IAP / upgrade mode, so only one press of SETUP is required to start the update process.

Frank