[Mark]

http://ccrma.stanford.edu/~jos/filte...lly_Ideal.html

Mark

[William Sommerwerck]

"Mark" wrote in message
...

http://ccrma.stanford.edu/~jos/filte...lly_Ideal.html

The "correct" answer is that it depends on what you're using the filtering
for. Most -- but not all -- deviations from flat amplitude response are
minimum-phase, and require minimum-phase correction.

[Mark]

On Feb 13, 2:04*pm, Mark wrote:
http://ccrma.stanford.edu/~jos/filte...lly_Ideal.html

Mark

this supports my contention that brick wall (steep) filtering always
has a bad transient response (ringing). The only way to make a filter
with better transient response (less ringing) is to make the
transition band more gradual. Changing the phase response can move
the ringing around, (before or after the main response) but does not
eliminate it.

Mark

[Scott Dorsey]

In article ,
Mark wrote:
On Feb 13, 2:04=A0pm, Mark wrote:
http://ccrma.stanford.edu/~jos/filte...lly_Ideal.html

this supports my contention that brick wall (steep) filtering always
has a bad transient response (ringing). The only way to make a filter
with better transient response (less ringing) is to make the
transition band more gradual. Changing the phase response can move
the ringing around, (before or after the main response) but does not
eliminate it.

In the analogue domain, this is the case. In the digital domain where
we can build FIR filters, it's not the case at all. Consequently, we
use oversampling on digital converters to move as much of the anti-aliasing
and reconstruction filter into the digital domain.

I strongly recommend the discussion of filtering in _Signals and Systems_
which is a good introduction to the math.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[William Sommerwerck]

"Mark" wrote in message
...
On Feb 13, 2:04 pm, Mark wrote:
http://ccrma.stanford.edu/~jos/filte...lly_Ideal.html

Mark

this supports my contention that brick wall (steep) filtering always
has a bad transient response (ringing). The only way to make a filter
with better transient response (less ringing) is to make the
transition band more gradual. Changing the phase response can move
the ringing around, (before or after the main response) but does not
eliminate it.

This is not completely true.

[Mark]

On Feb 14, 1:53*pm, (Scott Dorsey) wrote:
In article ,

Mark wrote:
On Feb 13, 2:04=A0pm, Mark wrote:
http://ccrma.stanford.edu/~jos/filte...lly_Ideal.html

this supports my contention that brick wall (steep) filtering always
has a bad transient response (ringing). *The only way to make a filter
with better transient response (less ringing) is to make the
transition band more gradual. *Changing the phase response can move
the ringing around, (before or after the main response) *but does not
eliminate it.

In the analogue domain, this is the case. *In the digital domain where
we can build FIR filters, it's not the case at all. *Consequently, we
use oversampling on digital converters to move as much of the anti-aliasing
and reconstruction filter into the digital domain.

I strongly recommend the discussion of filtering in _Signals and Systems_
which is a good introduction to the math.
--scott
--
"C'est un Nagra. *C'est suisse, et tres, tres precis."

I'm sorry to disagree. An FIR filter with a linear phase will still
have ringing if the frequency resonse is steep. This is shown in the
example in the Stanford site. The min phase filter (typical of an
anlog filter) has post ringing. A linear phase filter (typical of a
digital FIR filter) has both pre and post ringing. The duration is
less but the point of the artcile is that pre ringing is more
objectionalbe compard to post ringing. In any case the linear phase
filter still does have ringing.

The bottom part of Fig 11.2 is a linear phase filter like a typical
FIR filter.


Mark