self calibration vs. programmatic calibration

[Sean.C]

Hello,

I am trying to assess the value of programmatically calibrating my device vs. relying on the self-calibration accuracy.  According to the specs (I am using a pci-6225), I can expect the following:

AO

gain error:~150ppm (or 0-750uV for operation in 0-5V)

offset error:~218ppm (+/- 2180uV using 10V range [card based hard constraint])

AI

gain error:~160ppm (0-800uV for use in measuring 0-5V)

offset error:~156ppm (+/- 780uV using +/-5V range)

I am thinking about improving this by the following:

     1. connect ai to gnd and measure (store the value as an offset correction factor)

     2. set ao to 0V and measure with ai...  (store the value as an offset correction factor)

This should dramatically improve accuracy at low voltages and even have a significant impact at higher voltages (such as 4-5V).  Has anyone done this and can confirm that this works?  Is there someone from NI that can confirm that this makes sense (or conversely that it won't because I have over simplified things somehow)?

Thanks,

Sean

[ChrisR.]

What is referred to in the accuracy table as "offset" also includes nonlinearity.  So yes, you have oversimplified.  What you're suggesting doing is very similar to what NI does during self calibration, and I doubt it would improve accuracy significantly.

Chris

[John_P1]

Hi Sean,

An excellent idea, but it's one that NI has already considered.  The "Self Calibration" routine on your M-series device performs a procedure NI refers to as NI-MCAL for calibrating analog input.  Part of the MCAL process involves measuring the channel connected to an internal ground reference and is intended to reduce offset error as much as possible.

In fact, the procedure actually measures 29 points on each range of the ADC by using an on-board PWM DAC to supply a reference voltage.  It then interpolates these values and computes a 3rd order polynomial that is used by the driver to scale the raw data back into voltage.  So not only does the procedure account for offset error, but also gain error and non-linearities as well.

Analog output just uses a linear calibration during the self-cal which is achieved by reading back AO using the previously calibrated AI, similarly to how you described (the routing is done internally).

The success of this routine depends on an accurate voltage reference to be supplied to the PWM DAC.  This is why NI recomends for the board to be externally calibrated annually to help ensure that you remain within specifications.  You can utilize NI's calibration services to perform this external calibration, or if you have the equipment necessary can perform it yourself by following the published calibration procedure.  The adjustment procedure basically just involves providing a highly stable voltage reference to be used to calibrate the on-board reference.

Best Regards,

[JB]

Jon,

Thank you very much for this very interesting information. I'm used to execute the self calibration of the DAQ cards but I didn't know what this procedure does in detail.

Best regards