This is odd! Would that be the case with an analogue meter?

Not perhaps so obviously, but yes, when you are crammed in near the zero of the scale. If the moving coil meter has say a 100 degree maximum deflection (makes the sums easy but is about right for most real meter movements) ,  then three significant figures  would be an angular change of  0.1 degree. Now 1 degree is 1mm graduations on a 50mm long needle, so unless you have a massive instrument, then  I suggest that while 2 significant figures (1%) is quite easy, with mirrors to avoid parallax and so on, three is probably impossible.  So we can see a change of 1%,  if the thing being measured is near full scale, or for the hard case the spec must meet,  if it is just after a range change, then it depends on the range steps. Good analogue meters had a 1,3,10,30,100 scaling, so at worst the thing you are measuring might be just over 100% of the range below, and so about 30-35% of full scale so that 1 degree mark is more like 3%, and then we come to the equivalent of the absolute offset of 5 counts - how accurately can you 'zero' the analogue meter  probably a degree or so again?


Analogue to Digital conversion (ADC)s have the same problem and as far as I know all common digital meters either use dual slope integrations or successive approximation (  A long read here for what that means ) Both types suffer from non-linearity, step size errors, and DC  offsets to some extent, so you have some equivalent abberations, before you even look at how well it has been calibrated and against what, and the uncertainty of the actual circuit being tested - as described above.

They do pretty well all things considered.

This is odd! Would that be the case with an analogue meter?

Not perhaps so obviously, but yes, when you are crammed in near the zero of the scale. If the moving coil meter has say a 100 degree maximum deflection (makes the sums easy but is about right for most real meter movements) ,  then three significant figures  would be an angular change of  0.1 degree. Now 1 degree is 1mm graduations on a 50mm long needle, so unless you have a massive instrument, then  I suggest that while 2 significant figures (1%) is quite easy, with mirrors to avoid parallax and so on, three is probably impossible.  So we can see a change of 1%,  if the thing being measured is near full scale, or for the hard case the spec must meet,  if it is just after a range change, then it depends on the range steps. Good analogue meters had a 1,3,10,30,100 scaling, so at worst the thing you are measuring might be just over 100% of the range below, and so about 30-35% of full scale so that 1 degree mark is more like 3%, and then we come to the equivalent of the absolute offset of 5 counts - how accurately can you 'zero' the analogue meter  probably a degree or so again?


Analogue to Digital conversion (ADC)s have the same problem and as far as I know all common digital meters either use dual slope integrations or successive approximation (  A long read here for what that means ) Both types suffer from non-linearity, step size errors, and DC  offsets to some extent, so you have some equivalent abberations, before you even look at how well it has been calibrated and against what, and the uncertainty of the actual circuit being tested - as described above.

They do pretty well all things considered.