PXI-5105 resolution not quite as expected.

[Bandit]

I have a system with PXI5105 boards, and I have used it to sample a signal on a 30V range (+/-15V), 1megaohm input impedence, 1megasamples per second sample rate.

The PXI5105 has a 12 bit resolution, so I expect to see quantization steps of 30V / (2^12) = 7.32mV

Acquiring a steady state signal, the noise ranges over 7 discrete levels, so is spread over 6 quantization steps (see attached .jpg file)

Noise ranges over 52mV (it goes from -16 to +36mV), so measured quantization step is 52 / 6 = 8.66mV

I'm missing something obvious here. Why the discrepancy betwen calculated and measured resolution?

My only thought is that resolution isn't linear over the voltage range. I expect there is some non-linearity, but didn't expect this much.

Please fill the gap in my knowledge!

[Matt_E. · Accepted solution]

Hi Bandit,

The discrepancy you are seeing is due to the fact that the actual peak to peak input range of the board is not exactly 30V.  The overall gain of the analog front end which sits between your input signal and the ADC has initial inaccuracies and can also drift over time and temperature.  These effects can be minimized but never completely eliminated.  As a result we design the boards to have a nominal peak-to-peak input range slightly larger than the ideal ranges specified in the datasheet.  This ensures that the board will never have less than the ideal range, even as the gain drifts slightly (this way you can always be assured that if you are using the 30Vpp range and you put in a 30Vpp signal you will never saturate the ADC which results in a loss of information about the signal).  The 5105 has a precision internal calibration voltage generator that is then used to calibrate the scaled data that is returned to you.  This does result in a slight reduction in resolution (increase in the voltage/ADC code) as you have seen, but the voltage being returned to you is still accurate according to the published AC and DC accuracy specs in the datasheet.  If you need to programatically know what the calibrated resolution it is possible to get this information from the "wfm info" cluster output of the "niScope Fetch (poly).vi" when you are using the vi in a binary data fetch mode.  Sorry for the confusion and hope this helps.

Regards,

-Matt

[Bandit]

Top answer Matt! Just the info I needed. Now I know I'm not going mad.   🐵

I had wondered about the 30V range, but couldn't find anything in specs suggesting what you have described.

Not been on the forums for a while, there have been some changes. Highlighting the solution in the thread is new to me, and a great idea. Can't find the bit to rate your answer though, so I'm telling you here and now it's 5 stars    *****  

[Matt_E.]

Glad it was helpfull, please don't hesitate to post any questions in the future.  These sort of things help us find out what parts of our documentation need improvement.