Realistically, How many bits?

[kmcdevitt]

NI PXIe-4330, 8 Ch, 24-Bit, 25 kS/s Simultaneous Bridge Module.

The stated accuracy for this device is 0.1%.

To me, that is 1 part in 1000, which would be 10 bits of accuracy.

So why does NI claim 24bits, if only 10 are meaningful?

[LuisO]

The '24-bits' refers to the board's resolution, while the '0.1%' is the maximum (i.e., guaranteed) gain error, which is only part of the absolute accuracy spec (the other parts are the offset error, noise, nonlinearity, etc.)

So, while the number of bits refers to the smallest input step change you can detect, the accuracy spec (gain + offset) tell you how close your measurement will be to the 'real' physical value you connect to the inputs.

This short tutorial may be helpful:

http://zone.ni.com/devzone/cda/tut/p/id/4517

Luis

[kmcdevitt]

Luis,

The only relevant accuracy is the maximum accuracy that we would encounter during normal operation in the rated environmental envelope intended for the device.  So if it is stated that the accuracy is 0.1%, then that translates to me as 1 part in 1000, yielding 10 meaningful bits.

The PXI-4071 claims 26 bits.  It has a stated accuracy of 100uV over the -1000V to +1000V range.  Thats 1 part per 20,000,000 which translates to me as 24.25 bits.  Maybe NI should call the PXI-4071 a 24 bit device.

However, the PXIe-4330, claiming 24 bits with 10 meaningful bits, to me, seams a wee bit of a stretch.

Kevin.

[LuisO]

Kevin,

Yes, if you only care about absolute accuracy, then the most important thing is the uncertainty of 1 in 1000.  However, it is still a bit of an apples and oranges comparison to the '24-bit' label.  The quantization error due to resolution is only one of the contributors to the board's accuracy.

The number of bits specified is usually the ADC resolution, (i.e., 24 bits because the ADC in the board is 24 bits).  Even if the accuracy is 0.1%, having 24-bit resolution allows the board to have much better AC specifications than a 10 bit board would have. (Signal to noise ratio, total harmonic distortion, alias rejection,etc.).  

I think it would be misleading if they called it a "24-bit accurate" board, but there are a couple of pages of DC and AC accuracy specs. Even if the absolute accuracy is only 0.1%, knowing the board is returning 24 bit data, along with all of its implications, is beneficial (and yes, it looks good on the first page of the datasheet!! :-).

Now, I understand the comparison with PXI-4071.  However, that is a completely different animal, as the ADC resolution can be traded off for measurement speed.  Even so, as you pointed out, the number of bits (resolution) does not match up with the stated accuracy.

Luis