Hi,
the compensation allows to handle device to device variations, influence of external circuitry and temperature effects, whilst still achieving precise results.
The compensation is a mathematical correction of the measured values, in this case a linear correction using a multiplication factor (gain compensation) and by adding a factor (offset compensation). The compensation is channel specific, that why you see different entries in above table.
Lets take the VSENSE2 as example (and explain the principle, not the detailed implementation):
going through the picture from left to right...
- the input voltage vin is scaled by an internal voltage divider vin * 1/28 (desired sigma delta converter input voltage is 0.0 - 1.0V)
- the scaled vin is sampled by a sigma delta converter, which resolution is sdres = Vref / (2^16-1) = 1.25V / 65535 ~ 19uV
- the compensation is using a gain and a offset value out = in * gain + offset (nominal gain = 1.0679, offset = 0)
This results into the following formula:
ACQ_VOLT = ( vin * 1/28 * (2^16-1) / Vref ) * gain + offset
with nominal values:
ACQ_VOLT = vin * 1/28 * 65535/1.25V * 1.0679 + 0 V
or a resulting channel resolution:
vsense2res = 28 * 1.25V / 65535 / 1.0679 = 0.5 mV / lsb
Having the flexibility to modify gain and offset values allows to maintain the 0.5mV/lsb resolution with errors in the system.
Example: If the scaling is 2 % to low 1/(0.98* 28) a 2% high gain compensation factor 1.0679 * 1.02 would maintain the 0.5mV/lsb resolution.
As the compensation is done in the digital domain the representation of the compensation code has to be known. The table above provides this information. The nominal gain of 1.0679 is represented with a gain code of 0x200. The resolution of the gain code is LSB = 0.000488.
The gain factor for e.g. 0x210 can be calculated:
gain = 1.0679 + (gaincode - default code) * LSB = 1.0679 + 16 * 0.000488 = 1.075708
Rgds
Wolfgang
