Re: [Openeeg-list] filter dealy

[Jarek F. <jf...@ya...>]

Devidas,

Similar question was answered by Peter Schmalkoke on 23 Apr 2005.

I copy it below, it should be helpful.

Regards
Jarek

"Per definition a pure
sine wave does not have a start or end. Thus, when regarding a single frequency, there
can only be a statement about a constant phase shift, which depends on the signal
frequency, the cut-off frequency, the filter order and the Q factor. But
neither the phase shift itself nor any of its derivation over frequency is a constant
over the frequency range of interest.

I think that most probably the question was meant to refer to the delay time that occurs
with an alteration of amplitude. Such an event can not be described
as consisting of one pure frequency, but must rather be regarded as comprising
a frequency band. When we want one simple number for the delay time, then there must be
some kind of definition of how the onset time is determined and how to integrate the
response to a frequency band.

With an low pass filter a simple measure for the response delay could be the delay time
until a DC step occurs at the output with 1/2 of the final amplitude.
An estimation for an 8th-order lowpass filter like the MAX7480 could be from answering
the question how long it takes for a 1st-order lowpass to reach the
8th root of 1/2 at the output.

A more accurate answer would comprise a plot of delay time versus input frequency,
whereby the input signals should be sharply switched sine waves,
which are then bandpass filtered at a constant bandwidth, and the definition of onset
time could be the moment when the amplitude reaches 1/2 of the final
amplitude. An over-all delay time might then be computed from averaging over the
frequency range of interest with the amplitude response serving as weighing factors.
Havn't done that, though."