I am using an AD7712 in a load cell application and am having difficulty obtaining a high level of accuracy across the entire input range of the converter. My test method utilizes a precision full-bridge with setpoints at 0,0.5,1,1.5,2,2.5,3mV/V and across select gain settings of the amplifier (1,2,4,8,16,32,64) at a notch filter frequency of 10Hz. The sense method utilizes a kelvin connection at the bridge. At each combination of points, I average 32 readings from the AD7712 and determine the accuracy due to a change in gain and the actual measurement accuracy.
The kelvin measurement is buffered and fed to the reference inputs of the ADC with a Vexcite/2 (Vref-) and Vexcite*3/4 (Vref+). Vexcite is nominally 10V.
The reference voltage is <0.1% off of what is expected.
The Vref- input appears to be approximately 6.5mV away from the bridge zero input, which should result in a 0.26% offset error.
I currently have a gain accuracy of 0.16% and a measurement accuracy of 1.16% when I use self-calibration.
I have had a slightly improved response when I use calibration method 4 which provided a gain accuracy of 0.09% and a measurement accuracy of 0.95%.
I am running out of ideas on how to improve this measurement accuracy in my application. Does anyone have any ideas of things to look at to improve the measurement accuracy?