Can't get a STABLE highpass filter with a 0.01 Hz cutoff frequency

I have some strain data which has a very low frequency drift. The drift is for the most part linear. I need to be able to eliminate the drift. Using detrend is not a good solution for me as my real data (real strain) can have massive variations. These variations tend to be low frequency in general.

I would like to be able to use a highpass filter with a cutoff of around 0.01 Hz. Sampling frequency is 4000 Hz. I have used all possible combinations I could think of in the FDA tool, but none of them have turned out to be stable. I imagine I'm not the first one to need to filter out very low frequency drift while keeping real low frequency data (0.5 Hz and higher). Any advice?

Thanks a lot!

Answer 1

[Wayne King]

I can get a stable highpass design with the following

 d = fdesign.highpass('Fst,Fp,Ast,Ap',0.1,0.2,40,1,4000);
 Hd = design(d,'butter');

Then you just filter with

output = filter(Hd,input);

as just one example, but are you sure you need a highpass filter here? How about differencing the time series?

Try the first and second differences

output = diff(input);

or

   output = diff(input,2);