AnsweredAssumed Answered

Phase distortion in low pass butterworth filter design

Question asked by Bro-brown on May 21, 2014
Latest reply on May 21, 2014 by Matt.Duff

Portable  3 kHz +/-25g accelerometer DAQ

 

Can you assist  a  4 pole butterworth filter  get around for the phase shift distortion  it creates as a filter amplifier stage in this low power DAQ

We needs the mems accelerometer waveform to remain as an intact time invariant  amplified signal for all frequencies in its signal waveform. so the specific frequency peaks for selected frequencies in that bandwidth of interest remain  synchronised to the same linear time scale as the mems analogue signal. This logging approach allows us to research  the needles in the haystack that combine to create  the significant portion of the peak acceleration.Most portable data loggers  I that band and g range shuffle  specific freq peaks around so much relative to each other  before and during  ADC that  the digital waveform is chalk compared to cheese when the mems output is overlaid on the reconstructed digitised waveform

 

These quick fixes appear infeasible

(A) The quick fix of slicing the data after the buffer stage and scrapping  the analogue butterworth filter. It has a very serious  downside

Including a  10  fold increase in ADC SPS as data space upstream and a LOSS OF  1 MORE LSB due to decreased  signal to noise ratio. Then software overheads are needed as  a  Fourier analysis to disassemble the freq above the bandpass   Before an an integrator is used to rebuild the original scaled  mems waveform back into a amplitude (voltage) vs time signal

(B) The other fix of broadening the low pass filter bandwidth so the % of  wanted freq to unwanted that is  passed creates a near linear shift in the band of interest does worse than fix (A) It   escalates our  SPS  40 fold and leaves even more digital filtering overheads  whilst making us  shed even more LSB's  as errors

Question to those who sympathise with the problem

If  dynamic analogue phase compensation is the approach  we need to take can you suggest ways to achieve it or further reading we can do to map the phase shift reversed  to each frequency the low pass analogue signal band

We must keep at least 70 dB/decade rolloff in the filter stage to attenuate the unwanted frequencies fed to the ADC to  stay with the 1LSB loss.

Bro Brown

Outcomes