# Accuracy error of ammeter from specification : is it really just a constant offset?

On ammeter specifications it is often written accuracy specification $$(....\% rdg+ ... digits)$$

While the ammeter can have a certain last digit on the screen, which determine the resolution if the instrument

My question is : is the error expressed in the specification really just an offset (i.e. a constant shift on the current) for all the measurements made with the ammeter?

In other words is it just a systematic error?

Or is it actually something more than just an offset?

I mean, if it was just a constant offset, then its value could be provided to correct all measurements and there would not be need of this error at all!

• The actual error is below that number, but where is kinda random, there is much more involved like temperature etc. – PlasmaHH May 31 '17 at 11:08

## 2 Answers

You can not assume anything more about the error than what the datasheet tells you. So no, it is incorrect to assume that all the error is a offset, and that this offset is the same over the whole measurement range. You don't get to make up your own specs.

No, the quoted error is not just a knowable offset.

If you are thinking of an old Analog Meter, the needle appears infinitely accurate, with just an offset and a gain error. And if the meter was just bigger, you could see better, and you could see if the answer was really 10.01, or 10.015.

Digital meters aren't like that. They've got a 10bit (or 8bit or 12bit or whatever) ADC inside. If the LSB was really exactly on target, you wouldn't have a 10 bit ADC: you'd have an 11 bit ADC. If the ADC was so good it could report extra bits, it would report extra bits.

On a real 10 bit ADC used in a meter, the only things you know about the LSB is that it's bigger than smaller numbers, smaller than bigger numbers, and doesn't have any double-steps ("no missing bits").

A digital meter reports all of the number it gets from the ADC. It doesn't know any more than that. If it reads 10.1, you know that it's more than 10.0, and less than 10.2, and that there will be points that reads 10.0,10.1,10.2 ("no missing bits") That's all the ADC can tell you.

An expensive meter will also tell you that the ADC reading has no more than 1 LSB extra error at any point: the calibration will be flat across the whole range.

A cheap (or high bit number) ADC can't even tell you that: you have the uncertainty of the width of the ADC bit, plus the line may curve out a bit, so parts of the curve are low and parts are high. This will give you two-step errors in the display value.

ADC with larger error bands are available (particularly high bit number ADCs), but aren't used in most meters. High bit number ADCs may have missing bits, or even out of order bits, but that isn't generally usefull for a cheap meter.

And the display value is scaled, so there is always a % scaling error as well as the ADC error.