I took over a legacy design, successfully implementing two LWDFs - 1) lopass cutoff at 350Hz and 2) bandpass from 8KHz to 13KHz.
The low pass is a problem. Here's why.
During low pass operation, a Sallen-Key LPF with cutoff of 2KHz is implemented using OA1.
When OA1 is enabled, there is an asymptotic capacitor discharge mixed with the audio.
The LWDF expects all data points (64 of them) to be centered around mid-rail (e.g. the data must be + and -), otherwise the output of the LWDF contains DC bias (of course).
What I'm thinking is to pre-process my 64 data points (taken at 500uSec intervals with the ADC10) to eliminate this asymptote...and it seems that another LWDF hi-pass (above 50Hz) would be a neat way to do it?
Problem is, I'm not good enough at LWDF techniques to understand slaa331 well enough to allow me to implement it.
Comments or any ideas or help is most welcome.
Thanks,
Frank
P.S. the original data acquisition is 64 single conversions at 500uSec intervals (TIMERB driven) - for a total of 32mSec worth of data. I want to measure how much 50-450 Hz energy is on the audio line during those 32 msec. That's the ultimate goal, and LWDFs need not be used - but being a legacy design, I am not allowed to modify the hardware (e.g. get rid of the Sallen-Key filter).