Forums

EJ: Remez / PM for nonlinear phase

Started by Fred Marshall January 7, 2004
Eric,

I just now focused on your suggested process:

"generate the usual Remez
input grid, except now create real and imaginary components.  Send the
real portion into the Remez routine configured for normal filter
design, and send the imaginary component into the Remez routine with
the Hilbert bit set (or however your nearest Remez function needs to
be set to do this).  Take the resulting coefficient sets and add them
together.  Presto!  You have an asymmetric filter which (presumably)
meets the criteria that you gave the Remez function, including
stopband attenuation."

This is intriguing.  I could work some more on figuring out how and why and
under what conditions this might work.  But, I figured I might as well ask
for further illumination that you might give.

Fred


Fred Marshall wrote:

> Eric, > > I just now focused on your suggested process: > > "generate the usual Remez > input grid, except now create real and imaginary components. Send the > real portion into the Remez routine configured for normal filter > design, and send the imaginary component into the Remez routine with > the Hilbert bit set (or however your nearest Remez function needs to > be set to do this). Take the resulting coefficient sets and add them > together. Presto! You have an asymmetric filter which (presumably) > meets the criteria that you gave the Remez function, including > stopband attenuation." > > This is intriguing. I could work some more on figuring out how and why and > under what conditions this might work. But, I figured I might as well ask > for further illumination that you might give. > > Fred
http://dspguru.com/comp.dsp/tricks/dsn/nlp_fir.htm Enjoy! Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Fred Marshall wrote:

> Eric, > > I just now focused on your suggested process: > > "generate the usual Remez > input grid, except now create real and imaginary components. Send the > real portion into the Remez routine configured for normal filter > design, and send the imaginary component into the Remez routine with > the Hilbert bit set (or however your nearest Remez function needs to > be set to do this). Take the resulting coefficient sets and add them > together. Presto! You have an asymmetric filter which (presumably) > meets the criteria that you gave the Remez function, including > stopband attenuation." > > This is intriguing. I could work some more on figuring out how and why and > under what conditions this might work. But, I figured I might as well ask > for further illumination that you might give. > > Fred
http://dspguru.com/comp.dsp/tricks/dsn/nlp_fir.htm Enjoy! Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Fred Marshall wrote:

> Eric, > > I just now focused on your suggested process: > > "generate the usual Remez > input grid, except now create real and imaginary components. Send the > real portion into the Remez routine configured for normal filter > design, and send the imaginary component into the Remez routine with > the Hilbert bit set (or however your nearest Remez function needs to > be set to do this). Take the resulting coefficient sets and add them > together. Presto! You have an asymmetric filter which (presumably) > meets the criteria that you gave the Remez function, including > stopband attenuation." > > This is intriguing. I could work some more on figuring out how and why and > under what conditions this might work. But, I figured I might as well ask > for further illumination that you might give. > > Fred
http://dspguru.com/comp.dsp/tricks/dsn/nlp_fir.htm Enjoy! Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Hi!

> > This is intriguing. I could work some more on figuring out how and why
and
> > under what conditions this might work. But, I figured I might as well
ask
> > for further illumination that you might give.
Ok. So the phase is not linear. But can its response also be specified? Is there any practical value to this approach? Thanks! Atmapuri
Atmapuri wrote:

> Hi! > > >>>This is intriguing. I could work some more on figuring out how and why > > and > >>>under what conditions this might work. But, I figured I might as well > > ask > >>>for further illumination that you might give. > > > Ok. So the phase is not linear. But can its response also be specified? > Is there any practical value to this approach? > > Thanks! > Atmapuri
Filters made this way will have "prompt" outputs; the impulse response is asymmetric and primarily early. That makes such filters more suitable for servo and other feedback loops. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������