dsp: p_fir: implementation error

[lfochamon]

Current p_fir implementation is incorrect. Using the inputs:

float x[10] = {1,2,3,4,5,6,7,8,9,10};
float h[8] = {1,2,3,4,5,6,7,8};
int nx = 10;
int nh = 8;

it outputs:

176.0, 156.0, 144.0, 140.0, 144.0, 156.0, 176.0, 204.0, 464.0, 444.0, 432.0,
428.0, 432.0, 444.0, 464.0, 492.0

while MATLAB gives

filter(1:8, 1,1:10) = 1.0, 4.0, 10.0, 20.0, 35.0, 56.0, 84.0, 120.0, 156.0, 192.0

Moreover, it goes beyond the buffer limit (the output should've been 10 samples).

[ebadi]

I also use Matlab to check function outputs (#148)
Not an expert in dsp, but are you sure "filter" is the right function? You may be interested to take a look at the following URL : http://se.mathworks.com/help/signal/ug/fir-filter-design.html

[lfochamon]

Hey! Well, filter is the function that implements filtering in MATLAB (both FIR and IIR). The link you sent me is for function to design filter coefficients. From what I understood (could be wrong) from the description and the current code in p_fir it should filter samples GIVEN the filter coefficients. Technically, it behaves like conv, except the output is the same length as the input. In other words, whereas conv returns nh+nx-1, filter returns nx.