OWON XDM1041 the unknown multimeter...

[Kleinstein]

The thermocouple probe gives 0 V if the TC tip is at the same temperature as the cold junction. So the cold junction temperature is added, not subtracted. The type K TC constant is relatively constant, so one can assume adding the temperatures. It actually is adding the calculated voltage to take into account a possible nonlinear TC response. Here the simple approximation of a constant thermopower should be OK.

So a copper short reads the cold junction temperature used by the meter.

The cold junction sensor away from the actual terminals is a weakness in the design, but with the normal banana plugs it is anyway difficult to get an accurate measurement. There is a good reason the for accurate TC readings they use special connectors.
Measuring the temperature at the volts input terminal would be very tricky with isolation.

The readings of 26.9 - 27.7 °C  for the different probes and with 2 meters are not that bad.  With the current warm weather and some self heating from the meter this sounds reasonable. Melting ice can be used as a reasonable reference temperature. The other simple test point is boiling water.

With the standard plugs one has to accept a few degree of uncertainty.  TC probes are not made for high absolute accuracy and the 4 mm plugs and CAT rating of the meters add another complication. The initial reading can be off, as the meter may need to warm up before getting good readings. With the sensor at the main chip this would be normal.  With turbulent air flow there can be some temperature variations at the chip. The heat up curve can be a bit more complicated than just a simple smooth exponential and the temperature reading my not be very frequent.

[theHWcave]

The OWON XDM1041 temperature weirdness
I did a few measurements this morning and here are the results. Note that all this about firmware version 1.2. The graphs were taken by polling the measured value via SCPI every second into a spreadsheet.
From the cold-start and warmed-up graph you can see that the meter looks at the internal temperature sensor every 600 seconds (10 minutes). During that time (about 3 seconds) the meter does not respond to SCPI polls, i.e. 3 polls time out but at the 4th one you get 4 results (i.e. it buffered the responses internally).  Note that this “pause” on SCPI every 600 seconds only happens if you are in temperature mode. If you are in millivolts for example there is no pause.

Interestingly, the very first temperature measurement after it updated the value it holds as cold-junction temperature is always a bit high. Not sure why. This you can clearly see in the warmed-up graph.

In temperature mode, the meter displays the “measured” value at the bottom of the display (what a brilliant idea! Seriously why don’t other meters copy that).  This allows an easy check if it is “doing the right thing”.
Example: With just cables to my millivolt source plugged in, but no external voltage applied, it reads 28 deg. and -0.02 mV Using the NIST tables / polynomials for K-type, 28 deg. corresponds to 1.122 mV.  This it stores as the cold junction offset (for the next 10 minutes).

If I now apply +1 mV to the input, the meter shows 52.7 deg.C and 0.98 mV.  The meter adds the 0.98 to the stored 1.122mV and gets 2.102 mV. According to the NIST polynomial equation that would be 51.9 deg.C. Ok so its off by 0.8 deg.  Maybe they use lookup tables instead of the reasonably complex maths. 

If I apply -1 mV to the input, the meter shows 2.8 deg. C and -1.03mV. Doing the same calculation, the resulting voltage is 0.092mV and the NIST value 2.3 deg C. Again, 0.5 deg. off  but I think in principle the meter is doing the right thing.

@davebb: If your meter shows the right value at startup but not some time afterwards, it seems it may not do its 600s refresh of the cold-junction temperature.  The best way to find out is to repeat the same test I have done for the cold-start graph and see if you get that staircase (or the spikes in the warmed-up) graph. You could use the OWON software and its recording function but I am not sure how it responds to the “pause” after 10mins. Better to use a small script and simply send every second: “MEAS1?<LF> “ and store the response

[theHWcave]

@davebb: coming to think of it. An easier way to check if the cold junction temperature is properly updated is to start the meter after it was off and cold with shortened probes and wait. Over 30-60 minutes, the temperature reading should increase (by several degrees) as it heats up internally.

[Kleinstein]

The rather slow update of the cold junction temperature is a bit anoying. So one has to essentially wait at least some 30 minutes for warm up.
In addition there seem to be a few point's that are off just after a update.

I wonder if one could force an update by switching between a mV rading and a temperature reading. If the meter does not measure the temperature in mV range it may do an internal temperature reading just after changing the function. At least with program control one could this way force the internal reading more often if needed.

[theHWcave]

A very good point and I tested that just now. Sadly the 10 minutes cycle to refresh the cold junction temperature reference seems to run independent of what mode you are in

[davebb]

Hi I have had my meter on for some time, it is reading 32.5c with input shorted , real temp in the room is 25c
i am now going to switch the meter of to cool down and then run a test with the easy dmm software recording
but it does look as this part of the meter is duff,
Thanks Dave

[Kleinstein]

The part around the connects would warm up somewhat. So 6 degree higher than room temperature may be actually right.
Even with a seprate sensor near the ground therminal there is still the problem if the other terminal is the same temperature. From the design good TC probes would have to have low thermal conductivity to the outside, so that the assuption of a temperature relatively close to the meter internal temperature is not that bad.

[davebb]

Hi
I do not think it is correct , it reads correct when cold, and with the probe it reads correct until it jumps after about 10-20 mins , i will log when it has been off for some time,
I will post the log soon,
Thanks Dave

[davebb]

The OWON XDM1041 temperature weirdness
I did a few measurements this morning and here are the results. Note that all this about firmware version 1.2. The graphs were taken by polling the measured value via SCPI every second into a spreadsheet.
From the cold-start and warmed-up graph you can see that the meter looks at the internal temperature sensor every 600 seconds (10 minutes). During that time (about 3 seconds) the meter does not respond to SCPI polls, i.e. 3 polls time out but at the 4th one you get 4 results (i.e. it buffered the responses internally).  Note that this “pause” on SCPI every 600 seconds only happens if you are in temperature mode. If you are in millivolts for example there is no pause.

Interestingly, the very first temperature measurement after it updated the value it holds as cold-junction temperature is always a bit high. Not sure why. This you can clearly see in the warmed-up graph.

In temperature mode, the meter displays the “measured” value at the bottom of the display (what a brilliant idea! Seriously why don’t other meters copy that).  This allows an easy check if it is “doing the right thing”.
Example: With just cables to my millivolt source plugged in, but no external voltage applied, it reads 28 deg. and -0.02 mV Using the NIST tables / polynomials for K-type, 28 deg. corresponds to 1.122 mV.  This it stores as the cold junction offset (for the next 10 minutes).

If I now apply +1 mV to the input, the meter shows 52.7 deg.C and 0.98 mV.  The meter adds the 0.98 to the stored 1.122mV and gets 2.102 mV. According to the NIST polynomial equation that would be 51.9 deg.C. Ok so its off by 0.8 deg.  Maybe they use lookup tables instead of the reasonably complex maths. 

If I apply -1 mV to the input, the meter shows 2.8 deg. C and -1.03mV. Doing the same calculation, the resulting voltage is 0.092mV and the NIST value 2.3 deg C. Again, 0.5 deg. off  but I think in principle the meter is doing the right thing.

@davebb: If your meter shows the right value at startup but not some time afterwards, it seems it may not do its 600s refresh of the cold-junction temperature.  The best way to find out is to repeat the same test I have done for the cold-start graph and see if you get that staircase (or the spikes in the warmed-up) graph. You could use the OWON software and its recording function but I am not sure how it responds to the “pause” after 10mins. Better to use a small script and simply send every second: “MEAS1?<LF> “ and store the response

My be they messed mine up with  firmware version 1.3.0 that mine has ?
Dave

[davebb]

Hi i have attached the log of the meter of the internal reading front ports shorted at the meter,
Thanks Dave

[theHWcave]

@Kleinstein: The only temp. sensor is the one inside the HY3131. Its physically a fair bit away from the probe sockets and not thermally connected other than through quite a bit of PCB material and ambient air inside the meter. That said, I stuffed an external k-type sensor into one of the sockets and measured the temperature which is within 0.5 deg. the value cold junction temp shown by the meter. I guess that is kind of acceptable (given the price etc)

[theHWcave]

@davebb
what a terrible format that OWON PC software produces.... 

Besides that, the good news is that your meter behaves exactly like mine. Every 10 minutes the cold junction is updated. 
18:47    25.6
18:57    jump to 29.7
19:07    jump to 31.2

So that part works as expected and the firmware did not do anything bad (or good)

Now with that established and the meter in a warmed-up state can you give:
1. the temp reading and the measure value shown on the display with probes shortened
then replace probes with K-type (the right way around ! and take care not to touch close to the tip)
2. he temp reading and the measure value shown on the display with the k-type in  and what value it should be (according to some other thermometer you have)

Thanks

[davebb]

Hi
The meter has been on for some time withe the input shorted it reads 33.4c 000.02mv, and 25.4 on another thermometer
with a k type it reads 34.6c  000.07mv, and 25.5 on another thermometer ,
When the meter is on from cold i had a reading with in 0.5deg c betwen meter with probe and thermometer
Thanks Dave

[theHWcave]

Very strange!  These values seem correct. That is, if the socket temperature (cold junction) is indeed 33.4 deg.C and the measured value with K-type is indeed 0.07mV then the readout should be 34.6 deg C. To see 25.5 in this situation, the measured value should have been -0.301 mV.  Are you sure this thermocouple is working and is a k-type? Also what are you using to plug it into the meter? Can you send a photo of the way the couple is plugged in?

Thanks

[davebb]

Hi i am using a k type with 4mm plugs from another owon OW18E handheld and the same with another one from another meter and with that meter is reads slightly lower then my room thermometer, if voltage reading is correct why does the internal temp in chip reading go so high as well after 10mins, did they use the wrong calibration settings at owon, and both read correct when the meter is cold ,
Thanks Dave
 

[davebb]

To test the internal chip sensor i used 2 good 4MM plugs and the copper core from some rg213 coax cable between the to plugs so is very short,
 The meter has been running a long time now and is reading 35.2 with the k type , and 33.4 c with the short,
the room temp is 25.5c
Thanks Dave

[Kleinstein]

The socket inside the meter should get warm. So the 33.4 C for the short are plausible. The strange thing is the low voltage measured for the thermocouple. So it the the thermocouple working, e.g. showing about the right difference for something like boiling hot water or ice water.
The numbers shown would point to a failure with the probe or maybe the 50 mV voltage range. Chances are the temperature measuremetn would use the 50 mV voltage range.  So one could do a test with a more or less stable voltage in the 20-40 mV range and measure this in both the 50 mV and 500 mV range.

[davebb]

Hi thanks for the info yes the k type is working, it does not jump temp on any other meter  i have tested with other k types and get the same problem,
i have now tested with 30.842mv from a psu and when i ended the test it was 30.927mv showing at the end of the log file When i switch over to voltage,
and that was correct with another meter,
Thanks all for your Help
Thanks Dave Browne 2E0DMB

[theHWcave]

I am using a K-type that came with the BM235. It has 2x 4mm banana plugs on one side and the K-wire pair comes directly out of the other end. For my BM869S I use a specially made adapter that has 4mm banana plugs on one side and a small 2-pin socket especially for thermocouples on the other side. My RS-component K-type  thermocouples I bought for this meter end in those plugs that fit the sockets in the adaptor.

In both cases the cold-junction (i.e. the non-sensor end of the K-type wires) actually happens outside the meter.  Depending how “hot” the XDM1041 terminals are and how well the banana plugs conduct that heat, the cold-junction may be quite a bit lower in temperature than what the XDM1041 thinks it is. In handheld meters this seems to be less of a problem because they are seldom on for long periods and generally more power optimised. This means the inside temperature of a handheld is closer to ambient.

This could well be the problem of your meter which seems to run hotter than mine. I have never seen more than 28 deg after hours of operation but you seem to run at 31 or so.  It shows correct for the first couple of 10-minute cycles until it adjusts the cold-junction to 31 deg which is way higher than the real ambient temp.

The first question is why does it get hot inside? The power supply is always on, regardless of the power switch so that is not it. The HY3131 with the actual temp. sensor consumes max 0.5W and the processor even less. In the end I had to open the meter and examine it with a crude self-made thermal imaging camera (MLX90640 sensor & Raspberry Pi). The only thing that gets quite hot is a 1 ohm SMD resistor next to the LCD connector. When I turned the backlight to the lowest setting, the heat disappeared.

So here is a test you can do. Run the meter at the lowest backlight intensity and check if you getting better accuracy. I have done this now for 30 mins and the temperature has dropped by 2 deg.

Although I think mine is generally ok as long as I use the BM235 probe which probably gets warmed quickly through the sockets, I am seriously thinking of opening some of the “vents” at the back with a drill. They seem to be more style than function.

[davebb]

Hi
ok i will try lowering the back light and see if that helps
but i did run the meter open and it did not make much difference,
Thanks for your help
Dave

[davebb]

Hi yes you are correct it has changed the temp by about 2 dg/c by turning the screen brightens to 0, and i have moved the meter in to free space only reading about 2.5dg/c high now,
Please let me know if you find anything more,
Thanks for your Help
Dave

[Misk]

Sorry for my english. I buy OWON XDM1041 and I have a problem with it. Problem in this video https://www.youtube.com/watch?v=rcfSdUYxkVM. The measuring of megaoms resistors is bad and unstable. I think the HY3131 is bad. Can anyone measuring the 10MOm resistor on a analog board of XDM1041 near the HY3131 (conected to 5 pin from end). I measuring 8 MOm and the risistance change, when I repositing my body and other objects. When I touch the board it change to 6MOm. The resistor 10Mom is good. I test them separetly. I need to know, how the good HY3131 work.
The return of the device is expensive for me. Other function work satisfactorily for me.

[Kleinstein]

The resistor reading seems to change with the meter taken of the ground. This could indicate something like a problem with EMI or maybe too much mains hum.

[davebb]

Hi Have you tried a PT100 to see if that works ok, does the meter still use the same temp in the HY3131 for 10min cal
Thanks Dave

[theHWcave]

The meter was running the whole day today using the BM235 K-type probe with its sensor next to a PT100 which was connected to the 34401A. I recorded the values from both  every 30s over almost 10 hours. The XDM1041 was consistently about 1.8 deg. too low but followed the temperature variations exactly. This was using the lowest backlight.  When I plugged the same K-type into the BM869S (without changing anything else) it showed about 1 deg lower that the 34401A's PT100 reading. So I would say 1 deg error due to the K-type probe itself (totally expected) and 1 deg due to the XDM1041 thinking the ambient temp. is higher than it is. Although backlight is low, some things still produce a little heat and it builds up because there is very little airflow.  Next step (weekend probably) is to widen the cooling vents in the DDM1041 and see if I can get the internal temperature lower.

The PT100 is another story. I have a proper 3-wire one from RS-Components so reasonable quality.  Not sure if you are familiar with 3-wire PT100 but basically you need to eliminate the resistance of the wires to the sensor head. PT100 is using resistance and 100 Ohm = 0 deg. C. The change in resistance is very little from 100 to 110 Ohm is the equivalent of 0 deg C to over 25.5 deg C so you can see that milliohms count here. There are special circuits for 3-wire probes but I am using the fact that 2 wires are connected at the probe end. So if you use these and measure the resistance you get a value of 2x the resistance of the length of the probe cable. I use that to null the meter and then replace one of the wires  at the meter with the 3rd one (that goes through the probe).  And this is where the XDM1041 fails. You can null the PT100 probe in resistance mode, no problem. But when you go to Temperature mode and select PT100 it ignores the REL value for resistance and shows the value that includes the wire resistance.  REL in Temperature mode nulls the temperature value not the resistance.  Using PT100 on a XDM1041 iin temperature mode is useless unless you use an extremely short or very thick cable to the sensor head. I mean that is doable but why?  The alternative is to use the PT100 in resistance mode (with proper null of cable resistance) and convert the resistance to temperature by hand (or script when reading from SCPI). This  is by the way exactly what I have done for the 34401A because the 34401A has no PT100 mode so I am simply reading the resistance and convert that using the PT100 formula as long as you go only positive (> 0 deg C) its not too complicated.

# convert ohm to deg.C using PT100 probe
def PT100(ohm):
   TC = 0.00385
   A  = 3.9083E-03
   B  = 5.775E-07
   R0 = 100.0
   return (-A + math.sqrt(A*A - 4*B*(1-ohm/R0)))/(2*B)

Does it do the 10 minute - cold-junction update in PT100 mode? I have not tested it but I am almost sure it doesn't. I mean it may still do it in the background but since PT100 does not need the ambient temperature it should not show. But I will test it tomorrow and report