If you have a closer look at this "circuit", then you will notice, that the LM2924 cannot be the right chip here. The LM2924 contains an OPamp and a comparator. I guess they wanted to draw something like the LM324 (i.e. LM2902) which they confused to LM2924, according to the "formula" LM2902 + LM324 = LM2924?

You will also notice, that pin 4 and pin 8 are reversed, that the output pin 1 is short circuited to the supply voltage, that the X and Y outputs of upper ADXL202 are short circuited to the outputs of OPamps, and and and!

I guess the scheme shall show two differential amplifiers providing Xout = Xfilt_lower -Xfilt_upper and Yout = Yfilt_lower - Yfilt_upper. The two RC-parallel combinations should be connected across output and inverting input of corresponding OPamp, though.

But even then, they make the mistake, that Xfilt and Yfilt are PWM outputs which at least need some integrating low pass filters before they can be analogously summed.

But even if they would have done so, they were completely wrong when thinking, that Xfilt_lower - Xfilt_upper would provide any average, which they try to achieve as they write in the link.

They also write in the link, that they have enormous problems with noise, which is clear to me now: The signal they have, when subtracting the signals of two ADXL202 is just the missmatch signal, which should be zero for two identical chips...

Maybe I'm entirely wrong, but to me this cicruit makes absolutely no sense!

Kai