highpass function vs "manual" filtering

Hi,

I am using the highpass function to filter a signal (a temperature timeseries) and the result I get are resonable. I am trying to reproduce the results "maunually". Meaning I took the time sereis I did the FFT then I took the part of the frequency spectrum i am interested in and then I antitrasform the filtered spectrum. The results I get are very different (and not very reasonable from a physical point of view). So I am wondering:

what I am I doing wrong in implementing the manual fitering?

Does the highpass function follow the same steps?

Thanks in advance

Giacomo

I have attached the code and a file with the temperature time serie

Here is my code for manual filtering:

ship=readtable('shipdata_for_mat_ans');
ship=table2timetable(ship);

y = fillmissing(ship.ta,'movmean',height(ship));       % Signal (Vector Or Matrix)
t = posixtime(ship.Time);            % Time Vector 
Fs = 1/mean(diff(t));                % Sampling Frequency
Fn = Fs/2;                           % Nyquist Frequency
ym = y - mean(y);                    % Subtract Column Means From All Columns (Eliminates D-C Offset Effect)
L = numel(t);                        % Signal Lengths
FTy = fft(ym);%/L;                   % Fourier Transform (if you want it Scaled For Length uncomment the division by L)
Fv = linspace(0, 1, fix(L/2)+1)*Fn;   % Frequency Vector  
Iv = 1:numel(Fv);                      % Index Vector

sp=real(FTy(Iv,:));

sp_dc=sp(sp>=1/(24*3600));

dc=ifft(sp_dc,height(ship));

while using the highpass function i just do:

ship=readtable('shipdata_for_mat_ans');
ship=table2timetable(ship);
y = fillmissing(ship.ta,'movmean',height(ship));       % Signal (Vector Or Matrix)
dc = highpass(y,1/(24*3600),1/60,'Steepness',0.95);