Tews Technologies TMPE627, User Manual

Page: 21 / 34

TMPE627 User Manual Issue 1.0.2 Page 21 of 34
ADC & DAC Correction 4.12
There are two errors that affect the DC accuracy of the ADCs and DACs.
Offset Error:
The Offset Error is the deviation from an ideal zero. For ADCs it is the data value when converting with the
input connected to its own ground in single-ended mode, or with shorted inputs in differential mode. For
DACs it is the difference between the ideal and actual DAC output with zero code as digital input. This error
can be corrected by subtracting the known error from the datum.
Gain Error:
The Gain Error is the deviation in slope of the ideal ADC or DAC transfer characteristic. This error can be
corrected by multiplying the datum with the correction factor.
The TMPE627 provides offset and gain correction values for each channel and voltage range. The correction
values are based on a linear regression.
The correction values are obtained during factory calibration and are stored in an on-board EEPROM as 2-
complement 2-byte-wide values in the range from -32768 to +32767. To achieve a higher accuracy, they are
scaled to ¼ LSB.
4.12.1 Off-Module Correction
The correction can be done off-module in software, using the following correction formula:
4
1
4
1
2
1
Re
⋅ −






⋅ − ⋅ =
corr
solution
corr
Offset
Gain
Value Data
Figure 4-3: Correction Formula
Data is the corrected result (for AD channels this is the ADC input value, for DA channels this is the value
that has to be written to the DAC to achieve the desired output value).
Value is the uncorrected datum (for AD channels this is the ADC reading, for DA channels this is the desired
output value).
Gain
corr
and Offset
corr
are the correction values stored in the Calibration Value ROM.
Resolution is the data converter resolution in bit (for the TMPE627 it is 16 for both AD and DA)
Floating point arithmetic or scaled integer arithmetic must be used to avoid rounding errors in
computing above formula.
Due to inherent device deviation, the extremes of the full scale range may not be fully reachable,
even after calibration.