Multimeter comparison, within spec?

[MagooT]

Hi!
I was measuring some voltages with two meters and thought the readings looked a bit off, so i did some more measurements. I have these three meters:
Extech 430 (+-0.5% +2 digits)
Fluke 87V (+- 0.05% +1 digit)
Gossen Metrawatt Metrahit X-tra (+-0.05 +3 digits)

I did some measurements, hooked all up to the same power supply.
Measurement 1:
Extech 0.996V
Fluke 1.0005V
Metrahit 1.0014V

Measurement 2:
Extech 2.989V
Fluke 3.001V
Metrahit 3.010V

I was mainly thinking about the difference between the Fluke and Gossen meters,
is it possible from this information to calculate if one of them is out of spec?

Thank you for any help.

[retiredcaps]

In high resolution mode, the Fluke 87V accuracy specifications for DCV is +/-0.05% + 10 digits.

[Lightages]

Well you gave some measurements but what was the actual voltage being fed to the meters? If you don't know the voltage to an accuracy better than the best accuracy of the meters then you have no way of knowing which is in spec or out.

[MagooT]

Thanks for your replies.
Unfortunately i don't know the actual voltage, i don't have an accurate voltage source.
I just thought for example the 9mV difference between the gossen and the fluke on the second measurement looked like a lot.

Is there any affordable voltage standard that could be used to check this?

[Wytnucls]

As retirecaps pointed out, the Fluke 87's accuracy is not that great at 0.05%+10, so it is still within specs, assuming the Gossen is closest to true voltage, like it should be.

[Neilm]

Were all the meters connected at the same time? If they were not then any difference in input impedance could change the actual voltage on the terminals hence the different readings.

[MagooT]

Wytnucls> Is the +10 digits also for the 6V range, without high res mode? From the datasheet i understood it should be +1 digit, but i could be missing something

Neilm> Yes, all meters were connected simultaneously and i left them on for about 5-10 minutes to make sure the readings were stable, which they were.

[Wytnucls]

+10 for the 20,000 count high res mode, +1 for the normal 6,000 count, regardless of voltage range.
(except for mV range where accuracy is worse at 0.1%+1)