Why counts appear abitrary

[unimorpheus]

With all this meter talk I thought I would pose a question I have had for some time. When it comes to DMMs I understand counts and how they relate to digits. What I don't understand is why counts are implemented in the manner in which they are. Example, a 4.5 digit meter can be a 10000, 20000, 30000, 40000 or 50000 count meter. I do understand there are specs such as 4-3/4 or 4-4/5 digits and the like to attempt to flush out the difference in counts but this convention doesn't seem to have be widely adopted. Most manufacturers seem to go for counts. What I really don't understand is if you can implement in the display space a 50000 count meter, why not just make it a 100000 count (00000-99999). You have the display space and digit count. Why the arbitrary rollover at strange places such as 4000, 6000, 20000 or 50000. If you can implement 6000 counts what prevents you from implementing 10000 counts(0000-9999). I know there is probably some very good technical explanation for this. I don't know what it is but I am sure someone here does. 

[wraper]

What the point in additional counts if they contain no useful information?

[unimorpheus]

What the point in additional counts if they contain no useful information?

When the meter changes range you lose a digit of precision. Granted the meter has to be capable of delivering this precision accurately but if the LSD is available below 6000 counts why not above 6000 counts or whatever the rollover point is.

[samgab]

I think it has to do with the front end chip that runs the DMM which is used. This is what limits the refresh rate and the number of counts, AFAIK. I'm sure I'll be corrected if this is not right. 

[IanB]

Well, suppose you had a 12 bit A/D converter to sample the measurement, then in binary this would give a reading from 0 to 4095. With appropriate scaling of the input signal you could, for example, obtain a reading from 0.000 to 4.095 V, or 4000 counts. If you now wanted to measure a 5 V signal you would need to (for example) divide the input voltage by 10, and now on the next range up you could read from 00.00 to 40.95 V.

I'm sure the reality is a bit more complicated than this, but I think it may give you an idea.

[unimorpheus]

Well, suppose you had a 12 bit A/D converter to sample the measurement, then in binary this would give a reading from 0 to 4095. With appropriate scaling of the input signal you could, for example, obtain a reading from 0.000 to 4.095 V, or 4000 counts. If you now wanted to measure a 5 V signal you would need to (for example) divide the input voltage by 10, and now on the next range up you could read from 00.00 to 40.95 V.

I'm sure the reality is a bit more complicated than this, but I think it may give you an idea.

That actually does make sense. I didn't think about the original input conversion to binary.

[kcbrown]

So why don't we have 65536 count multimeters?   

[bitseeker]

So why don't we have 65536 count multimeters?   

Those would likely be advertised as 60000 counts.

[kcbrown]

So why don't we have 65536 count multimeters?   

Those would likely be advertised as 60000 counts.

No doubt!

4000 count, 60000 count, 120000 count, and others that are a multiple of 1000 that's nearest to a power of 2 value (where the power of 10 number remains under the power of 2 number) make sense.  But what's the reason for limiting an 8k count (13 bit) ADC based meter to 6000 counts, instead of setting it up as an 8000 count meter?  Or a 16 bit ADC based meter (65536 counts) to 50000 counts?

[Wytnucls]

From a Tektronix document:

Resolution is a measure of the smallest increment that may be discerned. At first glance, it would seem that 10.000 volts measured with a 40000 count DMM would be read to a resolution of 0.001 volt. This is usually the case where the DMM's A to D converter resolution exceeds that of the display but some have less resolution than the display. In this case, the last digit could read 0, 1, 2, 3, 5 ,6, 7, 9, etc. with a linearly increasing voltage. Notice that only eight out of the ten possible values were displayed. This is an artifact due to the digital nature of the conversion. In a more extreme case, only odd or even digits are be displayed, hence the need for a resolution spec separate from the display count.

To avoid quantization artifacts on the display, most manufacturers limit the DMM count accordingly.

[wraper]

So why don't we have 65536 count multimeters?   

That would be 30000 count. Don't forget that you measure in the range from - number of counts to + number of counts. Also any decent multimeter which is calibrated digitally and not not by a ton of pots, needs excessive resolution of ADC.

[Smith]

Meters like the Fluke 87V have 4,5 digits, but normally works in 3,5 digit mode. The last digit can be read, but because of the specs the absolute deviation of the last digit is just 10 times worse than the previous digit, so it doesn't add anything to the absolute value. Instead it can tell you something about stability or increasing/decreasing measurement values.

I know some Keithley gear have these features too (ability to read one extra digit)