Jon said:
Joseph,
A FIR filter can be made to have linear phase. Any FIR filter whose
coefficents are symmetrical with respect to the center coefficent wil
have linear phase. For example, let A(c) = the coefficient of the
middle sample. Then if A(c-1) = A(c+1), A(c-2) = A(c+2), etc the
filter will have linear phase. All FIRs do not necessarily have
linear phase. No analog system can be made to have linear phase.
Regards,
Kral
There seem to be different meanings of 'linear phase' if used by digital and
analog guys respectivly.
In digital is meant a filter without any phase shift (exept a certain
unavoidable constant delay).
An analog guy will understand 'linear phase', that the phase is increasing
linearly with frequency. Unadvertedly this will lead to the same as above, a
constant delay for all frequencies.
The thing is, the digital guy forgets about the delay and is proud that his
filter has no phase change at all. And the analog guy forgets about the
filter and concentrates on the delay...
The analog loves OTOH a *minimum* phase filter. This type is related to the
OP-question, because in the analog world every filter needs a phase change
to perform its function. For a filter of 1st order this will be +90° for a
HP or -90° for a LP at the -3dB points and +/-180° at infinite/zero. A 2nd
order filter has double of these values. Most natural processes (pendulum,
mass/spring etc.) behave as minimum phase systems.
A minimum phase filter has the shortest possible delay. Sometimes a very
important attribute.
We can calculate the phase from the amplitude in a frequency plot for a
minimum phase system, this might be helpful in certain processing
situations.