6.5 digit multimeter precision ?

[AndyTT]

Hi All,

I have a DMM (Keysight 34465A) which has 6 and 1/2 digits of precision. Why when used with labview can we increase the number of digits to more than 6.5?

In the example shown in the screenshot I have changed the number of digits to 18! The meter is limits to 6.5 so where do they come from?

Thanks

Andy

[JÞB]

The magic of IEEE 754

Most of the set of "Real Numbers" cannot be exactly represented in a 64 Bit floating point format. There are errors in showing them out to that many digits.  Your meter did not get more accurate.

[deceased]

I think in multimeter land nowadays, they calculate the ENOB value (effective number of bits), then use this to calculate the "digits of precision".

For example, if the ENOB is 22 bit they calculate Log(2^22) to give 6.6 "digits".

[AndyTT]

I wasn't intending to use 18 points, it was only an experiment to see how far it would go. I quite expected it to be filled with all zeros after the 6th digit. I find these errors quite misleading, especially if you are looking to average results.

In my case I was looking to get one extra digit, (7 digits) by averaging n samples. Without averaging anything I can change the number of digits to 7 and get a value but it appears its not the correct value.

[Henrik_Volkers]

You can mean values to get some 'extra' digits... , however you migth gain some extra resolution, but rarely accuracy.

The possible accuracy is at least determined by the stability of the internal used reference. (LM399, LTZ..., ... ).

You will need a better DC reference to check, or a 8.5 digit voltmeter.. say hello to voltnuttery 😄

(IF the internal hardware is better than 6.5 digits, be shure it would be claimed...)

[billko]

Sure, it's misleading, but it's also not a LabVIEW issue.  For more info, check this link out.

[billko]

---

@Henrik_Volkers wrote:

You can mean values to get some 'extra' digits... , however you migth gain some extra resolution, but rarely accuracy.

The possible accuracy is at least determined by the stability of the internal used reference. (LM399, LTZ..., ... ).

You will need a better DC reference to check, or a 8.5 digit voltmeter.. say hello to voltnuttery 😄

(IF the internal hardware is better than 6.5 digits, be shure it would be claimed...)

---

I don't know if you can even say, "rarely"  I don't think it's even possible at all.  You can't have an answer that is of greater accuracy than the least accurate of the measurements when processing data.  (In other words, rounding the answer is in order.)

[AndyTT]

Hi Bilko,

Could you fix the link the referenced, it just opens this page again for me.

Thanks

Andy

[crossrulz]

---

@billko wrote:

---

@Henrik_Volkers wrote:

You can mean values to get some 'extra' digits... , however you migth gain some extra resolution, but rarely accuracy.

The possible accuracy is at least determined by the stability of the internal used reference. (LM399, LTZ..., ... ).

You will need a better DC reference to check, or a 8.5 digit voltmeter.. say hello to voltnuttery 😄

(IF the internal hardware is better than 6.5 digits, be shure it would be claimed...)

---

I don't know if you can even say, "rarely"  I don't think it's even possible at all.  You can't have an answer that is of greater accuracy than the least accurate of the measurements when processing data.  (In other words, rounding the answer is in order.)

---

Thinking back to my grad school days here...

You are not gaining accuracy, but are gaining confidence in a measurement.  On the surface, it sounds like the same thing, but it is not.

[billko]

---

@AndyTT wrote:

Hi Bilko,

Could you fix the link the referenced, it just opens this page again for me.

Thanks

Andy

---

Hint: It doesn't need fixing.