DMM produces different voltage readings when leads are swapped

[AlphZeta]

I didn't notice this before, but when I was testing the voltage from a precision voltage reference (2.500 V) the other day with my BK-2709B I got 2.500V reading when probing with the correct polarity (i.e. positive lead on positive and negative lead on negative). But when the leads were swapped I got -2.497V. I can repeat

The discrepancy is well within the maximum allowed error range but I don't think I have seen this before on other DMMs. Anyone else encountered similar behavior before?

[robrenz]

Short your leads and zero out with the relative button and then recheck both directions. It is common to have a small difference in the + and - calibration.  Even my 6.5 digit 24ppm bench meter calibration shows minute differences in the + and - values of the same voltage.

[AlphZeta]

That's the first thing I did and when the leads were shorted the reading was exact 0. I guess I was expecting if the DMM gives readings it would at least be consistent with itself (e.g. if there is a system error, it should show on both + and - side), but maybe I am wrong.

For a 6.5 DMM, I can understand why you would see such differences as it is very sensitive, but for a 3 3/4 one I haven't noticed on any other DMMs before....

[IanB]

Some of my DMMs do this too.

[robrenz]

0.003V error on 2.5V is 0.12%  your worst case meter specs would give +/-.0145V on a 2.5V reqding (.5% rdg + 2 lsd).  Be happy 

[AlphZeta]

Don't get me wrong, I love my BK-2709B.... I just expected the result to be symmetric I guess

[mikeselectricstuff]

It's not inconceivable that they add an artificial dead-band near zero to make the zero reading look spot-on and hide small offsets and also avoid things like -0 appearing. 

[saturation]

DMM have differential amplifier if not instrumentation amplifier inputs.  If the 2 input stages aren't exactly symmetrical or balanced by the resistor network, that difference likely will show up as an error.

[rsjsouza]

I was thinking about a numeric error...

If the equipment uses fixed-point arithmetic, this issue may be caused by the different number of positive and negative values when represented by integers.

The last paragraph of the section "Range and precision" of the link below explains it better than I: http://www.mathworks.com/help/toolbox/fixpoint/ug/f20705.html

[AlphZeta]

I know Dave has a 2709B as well, maybe he can share his thoughts?

[EEVblog]

I know Dave has a 2709B as well, maybe he can share his thoughts?

Mine gives 5.002 positive and 5.000 negative on my EDC voltage standard.

Dave.

[IanB]

Consider a centre-zero meter (see picture attached).

Why would a +V reading give exactly the same scale position as a -V reading? Given that they are different readings in different directions, it would be a surprise if they were the same, not a surprise that they are different. To get identical readings the whole measurement system would have to be completely and identically symmetrical.

[AlphZeta]

Mine gives 5.002 positive and 5.000 negative on my EDC voltage standard.

Thanks Dave! So looks like for the BK-2709B the positive measurement is slightly higher then the negative.

Why would a +V reading give exactly the same scale position as a -V reading?

I understand the reason why the measurement could be different, but in my opinion DMM makers should trim the result to make the measurements at least consistent with the device itself.

[mikeselectricstuff]

My old Fluke 87 Mk.1 gives about 3mV diffference between polarities in 41/2 digit mode

[M. András]

my brand new  fluke 289 gives a 0.0002volt diff at 1.7420, and -1.7418 volts when i swap the leads. didnt find better stable power source then a deeply discharged cr123a (factory cal date is jan 9 this year) wonder how accurate is that factory calibration there isnt a full range and full function cal list on the paper what i got with the meter, just few ranges and functions

[robrenz]

Cal. data on my new 8846A
Lesson learned, probably all meters vary in the LSD from plus to minus readings.
If I am splitting hairs I should use which side is more accurate for the range in use.
( I know these values are not guaranteed and could have changed since the Cal. but it sounds good  )
+100 mV     -.0005 mV
 -100 mV      -.002 mV
+1.0 V         -.000001 V
 -1.0V           .000000 V
+5.0 V         -.000010 V
 -5.0V           .000000 V
+10.0 V          .00000 V
 -10.0V          -.00001 V
+100.0 V         -.0004 V
 -100.0V          -.0001 V
+1000.0 V         -.001 V
 -1000.0V           .000 V

[ejeffrey]

I understand the reason why the measurement could be different, but in my opinion DMM makers should trim the result to make the measurements at least consistent with the device itself.

We normally imagine that devices like meters are intrinsically perfectly linear, and the only sources of error are offset and gain errors.  However, this is just not true.  This meter is simply showing a small non-linearity. Trimming it to eliminate all variation between + and - is equivalent to building a higher spec meter.

[bfritz]

I understand the reason why the measurement could be different, but in my opinion DMM makers should trim the result to make the measurements at least consistent with the device itself.

Every trim operation costs time on a tester, which is cost to the customer.  How much more are you willing to pay for this?  How much error between the two readings is acceptable?

[saturation]

The symmetry between measuring voltage with the proper then reversed polarity used to be a spec listed on DMMs through the 1960s to early 1980s.  By then, IC instrument amplifiers pretty much removed any practical difference, in robrenz listing, that different is ~ 1ppm, within his cal specs very likely for just the right polarity.

I tested the HP 3456a using the EDC MV106 as a source at 10V, 1V and 100mV, the difference is also 1 ppm or less.

[Skibane]

It's not a fault - It's a "feature"!

Measure both ways, take the average of the readings, and viola - you've got more better-er accuracy!

It's like getting a second opinion on your measurement, without needing a second meter...