Re: question about integration of spectrum

hi all,
i want to get a integral spectrum, i have two choices:

1:do integration to time domain then do fft to get the spectrum
2:do integration to fft spectrum directly

Integration in time domain, which is implemented in my project, has a few
non-evident (but resolved) issues: DC reject before and after integration,
overflow and quantization, etc.

i do not know which is better, can anyone give me some advices.

and if we do integration to a spectrum ( multiply 1/(j*w) ), how can we
deal with 0 Hz?

The integrator has infinity gain at DC. Thus, DC must be filtered out
before the integration. Moreover, DC must be filtered out also after the
integration.


.

Loading