Since the beginning of the project, a very simple circuit has been used to couple the SM24 sensor. It consisted of three 1 kΩ resistors: the one in middle serves as the termination shunt for the SM24, as indicated in the datasheet, and the other two resistors are used to limit the maximum current and protect the ADC.
Using that solution, there are two things, which are bothering:
1. Because AINCOM was coupled with REFOUT (2.5 V) by software, using the level shift VBIAS function, the input signal had no zero mean. It is possible to compensate that by subtracting the mean of the input signal from the current signal value, but it was still not very elegant, besides some other disadvantages with spectral analysis.
2. The input signal had a very high 50 Hz noise.
Therefore, a better but still simple solution was used. Hereto, the middle 1 kΩ termination shunt was replaced by two 470 Ω resistors in series, a capacitor added parallel to the outlet of the SM24 in order to suppress high frequency noise and the REFOUT (2.5 V) connected with the middle of two 470 Ω resistors (see circuit figure).
Now the input signal is symmetric, has a zero mean and the 50 Hz noise is greatly reduced, which leads to much better results.
In order to further reduce the noise, the build in low noise amplifier (PGA) was activated with a gain of 32 and the sample rate was set to 100 Hz.
Using a 100 Hz sample rate instead of 400 Hz increases the noise free bits from 20 to 21, see data sheet.
Using the PGA with a gain of 32 increases the input voltage by a factor of 32, therefore reduces the noise, but decreases the absolute measuring range by 5 bits. Since 16 bits noise free bits remain for measurement, the reduction of the noise is worth the reduction of the resolution.