Offset in filtered data when using filtfilt() Function

Hello,

when filtering my data using the filtfilt() Function, I get a offset in the filtered data.Can anyone tell me where this offset could come from and how to get rid of it?

Below is the code I use for filterdesign and actual filtering:

fg = 0.1; % Desired Cutoff frequency

fs = 1/(mean(diff(Time))); % Derive sample frequency fs from Time vector

fc = 2*fg/fs; % Calculate normalized cutoff frequency

d = designfilt("lowpassfir", 'PassbandFrequency', fc, 'StopbandFrequency',fc+0.5*fc) % Filterdesign as lowpass filter

filtered_data = filtfilt(d, unfiltered_data);

figure()

plot(Time, unfiltered_data)

hold on

plot(Time, filtered_data)

legend('unfiltered data', 'filtered data')

The figure below shows the result of some example data.

[Paul]

Hi Kletzi,

It will be easier to get help if you post the full code and data that recreastes the example in your question (or something close to it).

[Jan]

@Paul: You can reproduce the effect with dummy data:

fc = 0.4;
d = designfilt("lowpassfir", 'PassbandFrequency', fc, 'StopbandFrequency', 1.5 * fc);
x = rand(1, 200) + 20;
y = filtfilt(d, x);
plot(x, 'b'); hold('on');
plot(y, 'r');

Answer 1

[Paul]

Thanks @Jan

The filter, d, does not have unity gain at dc.

fc = 0.4;
d = designfilt("lowpassfir", 'PassbandFrequency', fc, 'StopbandFrequency', 1.5 * fc);
dcgain = sum(d.Coefficients)

dcgain = 0.9481

filtfilt attenuates the dc component of x by dcgain^2

(20 + .5)*dcgain^2  % 0.5 is the mean of rand

ans = 18.4273

x = rand(1, 200) + 20;
y = filtfilt(d, x);
plot(x, 'b'); hold('on');
plot(y, 'r');
plot(filtfilt(d,20+0.5+0*x),'g');