DMM DC readings inaccurate when leads are reversed

[nukie]

Hey guys,
I am testing a recently acquired 20000 counts DMM based on the Maxim MAX134. I found out that when I reversed the leads the negative reading does not match the actual positive reading. It's off by a little in the mV so its negligible.
 
My question is this normal? My fluke and HPs don't do this...

EDIT: I jumped the gun quite a bit, my bench meters reads identical results both ways but not portable meters.

[Kiriakos-GR]

What is the voltage reference source that you use , for your comparison ?

[nukie]

Right I did more testing, funny I'm only doing these test after owning the bench meters for so many years just because of a newly acquired DMM. I won't reveal the DUT yet until I learn how to calibrate it.

As I edited/mention in my original post, my old and trusty Fluke 8840A, HP 3468A and HP 34401A displays identical DCV results the only difference is the negative symbol when I reverse the leads. I used Fluke leads and a second pair of Chauvin Arnoux leads for confirmation.

For my Fluke 187, 189, Uni-T UT71D all exhibits some sort of difference in the mV range. The voltage source is a Geller SVR with a custom trimmed 5.0000V Analog Devices AD586KQ voltage reference. I don't think precision matters much but the noise does.

These are the results
Fluke 187 - 4.9997 (-4.9995)
Fluke 189 - 5.0000 (-4.9998)
Uni-T UT71D - 4.997 (-5.000)
unknown DMM - 5.000 (-4.983)

The Fluke 189 was recently calibrated.

I couldn't try a battery because the voltage drops as I try to take a reading. I tried a HP E3610A power supply at 0.9xxV. The Flukes and Unknown DMM passed with flying colors, identical results in the positive and negative range. So this means the voltage source is the problem. The Uni-T was reading 0.902(-0.906), the Uni-T never impress me much, it's has very poor repeatability in the low mV range. Just so you know, I was trying to measure current with sense resistor so I need accurate mV readings.

[Kiriakos-GR]

I don't think precision matters much but the noise does.

Well you had motivated me enough, so to test my Fluke 28II against the old 8050A ,
I  measured  10V , and from my bench power supply , and from my reference ( battery powered ) source.

There was no change about  noise , like comparing an clean DC source  against one PSU ,
that it does have even a minimum level of noise.

Even so, in both cases, I got the opposite numbers from yours.
At minus 10V ,  I got in both the fifth digit, to be plus one digit .. than when both measuring an positive value.
  
I am 42 years old , and every single day I discover just another crazy thing..  

[Mechatrommer]

I am 42 years old , and every single day I discover just another crazy thing.. 

you only have found 10% of it.

[vk6zgo]

Wait a minute----

From your own figures:-

The difference in reading between polarities is for the 189 --- 0.004%
                                                                              187----0.004%
                                                                           UNI-T----0.06%
                                       &, worst case,for the Unknown-----0.34%

For any range,that is more than adequate.

Like any meter,the error will be proportional to the range you are on,so if you are measuring in millivolts,you are obviously using a different (appropriate) range,& should experience errors of the same percentage as above,not the same absolute values.

VK6ZGO

[Zero999]

The reason why I think this happens is because the ADC inside the meter isn't perfectly linear so the error differs at different parts of the scale. Another reason may be the input has a precision rectifier before the ADC and the rectifier gain characteristic isn't perfectly symmetrical.

[vk6zgo]

The reason why I think this happens is because the ADC inside the meter isn't perfectly linear so the error differs at different parts of the scale. Another reason may be the input has a precision rectifier before the ADC and the rectifier gain characteristic isn't perfectly symmetrical.

That's what I always assumed to be the case.It certainly sounds like the most logical answer.

VK6ZGO

[saturation]

In ye olden days, there used to be a spec for meters that showed the difference between neg and pos readings, of the same voltage.  However, today, the difference is often very small in good DMMs, ~ 0. 

When you do such checks, the overall accuracy of the DMM shouldn't change.  For example, for the Fluke 189, even if you don't add add the LSD float to the VDC accuracy, it shouldn't cause the accuracy to change beyond 0.025%.  Assuming the first reading is accurate to all digits, the Fluke's appears way within spec.

The unknown DUT however, seems off, but it depends on its rated accuracy for that range.