ADS1130: ADS1130 is giving me half the expected result

Hi Krishan,

You have miscalculated full-scale range of the ADS1130 which is +/- (1/2) Vref/64. When connecting pin 4 to DVDD you are changing the operational parameters of the ADS1130 which is outside the device specifications and tested mode of operation.  If you want higher gain and performance you should use the ADS1230.

Best regards,

Bob B

Dear Krishan,

as Bob mentioned, the confusion is most likely coming from your calculations.
The ADS1130 offers a PGA Gain=64 and the input voltage range of the ADS1130 is specified as VIN = +/-0.5*VREF/64.
The LSB size of the ADC is 2*(0.5*VREF/64) / 2^18 = (5V/64) / 2^18 = 298nV.
That means at full capacity you will read a code of (5V*1.955mV/V) / 298nV = 32802 which is almost exactly what you are seeing.
That means the results your are getting are absolutely correct.

Your competitor seems to be operating the ADS1130 in an officially unsupported and untested mode.

Regards,

Hi Krishan,

Regarding 1, although the ADS1130 is similar to the ADS1230 certain features are not available for the ADS1130, like the internal offset calibration.  For your application this is not required as there is also sensor offset error of the bridge that must also be removed.  The best thing for you to do is a 2 point calibration where the no load condition of the bridge is 0 weight.  The returned code with 0 weight is subtracted from any further result.  The second point for calibration should be with a known calibrated weight applied that is near the value of the full-scale capacity of the load cell.  The slope of the no load and known load gives you the grams/code of the measurement.

Regarding 2, you should be able to achieve 2 gram resolution for 40kg load cell at 10sps.  Your best case resolution will be with shorted inputs.  If you have layout or supply noise issues this will show up as a noisy result. You should be able to see no more than 1 code (bit) of noise in the shorted input case.  When attaching the sensor to your PCB, there may be additional noise sources such as power line cycle noise, or EMI/RFI.  This external noise will directly impact your measurement and will need to be prevented from entering your ADC input.

For 3, you generally have a good starting point, but due to any external noise sources that may enter your system you may need to add additional filtering components.  What components you add will depend on the sources of noise you encounter.  Also, as the measurement is ratiometric where noise/drift of the excitation source should cancel in the measurement, there is an assumption that the noise and filtering is the same for both the reference input and the analog input.  This is often difficult to achieve without using a 6-wire load cell where the reference input is a Kelvin connection from the sensor.  Also note that the reference RC filtering differs in your circuit with respect to the analog input filtering (primarily the differential caps).

Best regards,

Bob B

Hi Krishan,

I'm not surprised that you are seeing a linearity error when you are calculating the gain slope with only 5% of the full-scale capacity of the load cell.  This small of a voltage output for the load cell will not account for any non-linearity of the load cell itself, nor will it account for any gain error of the ADC.  Also remember that even with full-capacity applied to the load cell it is only about 1/8 of the full-scale range of the ADC.  So 2kg is too small to determine the gain slope.

Another thing is you cannot really tolerate any additional system noise.  Averaging may not help.  Without averaging you should see return values that fluctuate by no more than one code.  You are seeing this with averaging, so I suspect that your system is noisy.

Best regards,

Bob B