Manual
71M6541 Demo Board REV 3.0 User’s Manual 
45    Rev 4.0 
2.3.7  COMPENSATING FOR NON-LINEARITIES 
Nonlinearity is most noticeable at low currents, as shown in Figure 2-9, and can result from input noise and 
truncation. Nonlinearities can be eliminated using the QUANT variable. 
Figure 2-9: Non-Linearity Caused by Quantification Noise 
The error can be seen as the presence of a virtual constant noise current. While 10mA hardly contribute any er-
ror at currents of 10A and above, the noise becomes dominant at small currents. 
The value to be used for QUANT can be determined by the following formula: 
LSBIMAXVMAX
IV
error
QUANT
⋅⋅
⋅
−=
100
Where error = observed error at a given voltage (V) and current (I), 
VMAX = voltage scaling factor, as described in section 1.8.3, 
IMAX = current scaling factor, as described in section 1.8.3, 
LSB = QUANT LSB value = 7.4162*10
-10
W 
Note that different values for the LSB of QUANT apply, depending on which type of code is used. The LSB val-
ues are listed in the IC data sheet for standard CE codes. 
Example: Assuming an observed error for a meter with local sensors as in Figure 2-9, we determine the error at 
1A to be +0.5%. If VMAX is 600V and IMAX = 208A, and if the measurement was taken at 240V, we determine 
QUANT as follows: 
QUANT LSB = 1.04173*10
-9
 VMAX IMAX = 1.3*10
-4
9230
_
1240
100
5.0
−=
⋅
−=
LSBQUANT
QUANT
QUANT is to be written to the CE location given by the IC data sheet. It does not matter which current value is 
chosen as long as the corresponding error value is significant (1% error at 1.0 A used in the above equation will 
produce the same result for 
QUANT). 
Input noise and truncation can cause similar errors in the VAR calculation that can be eliminated using the 
QUANT_VAR variable. QUANT_VAR is determined using the same formula as QUANT. 
The internal power supply generates a ripple on the supply and ground nets that is 90° phase shifted with re-
spect to the AC supply voltage. This affects the accuracy of the VARh measurements. If optimization of the 
VARh accuracy is required, this can be done by writing a value into the 
QUANT_VAR register of the CE 
0
2
4
6
8
10
12
0.1 1 10 100
I [A]
error [%]
error










