Hi Adrian, The ADC and DMM may each have their own offsets and gain errors which combined add up to 0.75 mV difference. I don't see any issues there. 1) The offset calibration measurement is good. However, I think the gain calibration may be causing some issues... With an approximate 5V supply acting as the reference source and a gain of 0.5V, the ADS125H02 can tolerate a maximum +/- 8V input, due to the output swing limitations on the Stage-two PGA. If it weren't for this output swing limitation, you could apply a 10V full-scale signal and perform gain calibration, but because of this limitation you'll may have to calibrate with lower voltage and then compensate for the fact that the voltage is not at 100% full-scale... NOTE: Gain calibration is only beneficial if you can supply a precise full-scale signal. The signal must be more precise than the ADC's calibrated gain error, otherwise calibration will make the gain error worse! The ADS125H02 has a typical uncalibrated gain error of +/- 0.1% (and max +/- 0.7%). NOTE: The 5V LDO on the ADS125H02EVM has an output accuracy in the range of +/-2.5%; therefore, it should not be used as a calibration source. If you do decide to calibrate the 10V ADC range (8V PGA range), then you'll have to find a precise voltage source somewhere in the range of 5-8V. You can use the GANCAL command; however, note that the ADC will calculate the gain scaling factor under the assumption that you've applied a 100% full-scale signal, and will result in a scaling factor larger than "1". Since the input signal is less than full-scale (for a 5V input this is only 50% full-scale and the computed scaling factor will be approximately "2"), you will need to multiple the FSCAL[2:0] scaling factor by 50% (for a 5V input, or 80% for an 8V signal) to correct for this. 2) What values were programmed into the OFCAL[2:0] and FSCAL[2:0] registers after you performed calibration, and how much offset did you measure with short-ed input after calibration? 3) I would recommend reading back the OFCAL[2:0] and FSCAL[2:0] registers to ensure that reasonable values were set. OFCAL should be a small number (note: this is a signed number, so small negative numbers may appear to have a large unsigned decimal value), and FSCAL should not be too far off (i.e. more than +/- 1%) from 400000h (1 V/V).
↧