sbRIO 9636 - % of Full Scale locations for End points of Best Linear Fit on A/D ?

[Frustrated_NI_customer]

Have an application that requires improvemet upon the absolute accuarcy of the sbRIO 9636 A/D and plan to feed in an External precision reference with precision Bulk Metal Foil resistor ratios selected for End Points at the A/D's Best Fit Line.

Scan rate is 1kHz, Inputs are Differential and the Voltage range of interest is 0 to 5 VDC. What % of Full Scale (5V) should we set the line's ZERO and SPAN End points to obtain the best linear fit for best Absolute Accuracy?

Don't want to assume 10% and 90% if these are not the best End Point locations for dynamic compensation.

[AaronTheAE]

I have a knowledge base article here that we use to determine the absolute accuracy of our devices: http://digital.ni.com/public.nsf/allkb/8BA2242D4BCC41B286256D1D00815B90

You should be able to find the necessary information to calculate it from this specification guide: http://www.ni.com/pdf/manuals/373378d.pdf

How were you planning to determine what characteristics would specify the end points for your linear fit?

[a_clucker]

Hi,

It might help us if you post a wiring diagram and/or schematic of this to help fill in some details. Are you trying to feed this external reference in as a hardware reference line, or somehow programmatically?

Also, there is a NI Community group you might find useful: Hardware Developers Group for sbRIO and SOM. You might be able to garner responses from developers who have worked more closely with this type of concern.

Kind Regards,

[Frustrated_NI_customer]

Thanks for responding Aaron and Andrew

The plan is to dynamically compensate out DC Offset and DC Gain (Span) Errors as well as DC Thermal Drift Errors via a precision External Reference fed through neigboring channels of the AMUX going through the same Signal Conditioning and A/D Conversion chain.

I asked what Linear Transfer Equation End Points I should use over my Input Range of interest, because some Manufacturers tout impressive Non Linearity numbers by performing a Linear regression and defining a Best Fit Line that differs from the Nominal transfer equation of the Signal Condiitoning and A/D Conversion chain.

The plan is to feed two precision voltage references near my Zero and Span values ( 0 and 5VDC) and to dynamically generate a linear transfer equation to compensate the converted values coming out of the A/D.  We will not be performing costly characterization tests over a full sweep of the voltage input range or the full external ambient temperature range.  The technique planned won't compensate for INL or DNL or Channel to Channel variation through the AMUX, but it should provide adequate improvement in the absolute accuracy over temperature for our application.

I reached out to Spex in a private message and received a recommendation to use AGND as the Zero reference and a precision votage reference between 75% of Full Scale (such as a 4.096V) or as close as possible to Full Scale while avoiding clipping due to any offsets or gain errors that might be present.

Thanks again!