I'm looking for better insight on use of all-pass filters for changing phase without affecting amplitude. Are there any recommended books? Or are there fields where this is used often? Also, what are the usual side effects? I guess I could expect some pre-ring and the same types of artifacts inherent to regular filters.
All-pass filters for phase correction
Started by ●March 6, 2015
Reply by ●March 6, 20152015-03-06
There's a classic tutorial paper by Sanjit Mitra and Regalia, you can search and find it easily. If you have access to a good optimizer it's fun to use it to match arbitrary phase curves. Yes, you will get many different types of impulse responses, and if you use them to phase compensate an IIR you will get pre-ringing just like a normal linear phase fir. If you are trying to match some arbitrary phase response over a very wide frequency range that approaches Fs/2 or DC then you will need a very high order and it will ring for a long time. One trick you can play is to phase-correct a low pass iir only over the passband, which can give you shorter group delay than a normal linear -phase fir, although I think it's also possible to trick Parks-Mclellan to give you the same result (not 100% sure about this). Bob
Reply by ●March 6, 20152015-03-06
On Friday, March 6, 2015 at 6:25:47 AM UTC-5, radam...@gmail.com wrote:> There's a classic tutorial paper by Sanjit Mitra and Regalia, you can search and find it easily. If you have access to a good optimizer it's fun to use it to match arbitrary phase curves. Yes, you will get many different types of impulse responses, and if you use them to phase compensate an IIR you will get pre-ringing just like a normal linear phase fir. If you are trying to match some arbitrary phase response over a very wide frequency range that approaches Fs/2 or DC then you will need a very high order and it will ring for a long time. One trick you can play is to phase-correct a low pass iir only over the passband, which can give you shorter group delay than a normal linear -phase fir, although I think it's also possible to trick Parks-Mclellan to give you the same result (not 100% sure about this). > > BobI think Vaidyanathan had a paper in the IEEE about the use of All-Pass filter as a building block of filter design. You may also want to check out Fred Harris' book "Multirate Signal Processing for Communication Systems" it had some sections dealing with the use of All-Pass sections in filter designs. Cheers, David
Reply by ●March 6, 20152015-03-06
On 3/6/15 6:25 AM, radams2000@gmail.com wrote:> One trick you can play is to phase-correct a low pass iir only over the passband, which can give you shorter group delay than a normal linear -phase fir, although I think it's also possible to trick Parks-Mclellan to give you the same result (not 100% sure about this).Eric Jacobsen wrote about a trick for P-McC to design an arbitrary magnitude and phase FIR. http://www.dspguru.com/dsp/tricks/using-parks-mcclellan-to-design-non-linear-phase-fir-filter dunno about using P-McC for IIR. i think, instead, you need to use Prony http://www.mathworks.com/help/signal/ref/prony.html or Greg Berchin's FLSD https://groups.google.com/forum/#!msg/comp.dsp/UBsxiYrnT0c/eFH7d45-MCwJ . i think i got the acronym down now. -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
Reply by ●March 6, 20152015-03-06
On Fri, 06 Mar 2015 09:04:12 -0500, robert bristow-johnson <rbj@audioimagination.com> wrote:>Greg >Berchin's FLSD >https://groups.google.com/forum/#!msg/comp.dsp/UBsxiYrnT0c/eFH7d45-MCwJ >. i think i got the acronym down now.FDLS -- Frequency Domain Least Squares I'm a little reluctant to even consider what "FLSD" might stand for! Anyone who wants a copy of the Matlab or C example code, just send me a private email request. Change "chatter" to "charter" and remove ".invalid". Greg
Reply by ●March 6, 20152015-03-06
On 3/6/15 9:11 AM, Greg Berchin wrote:> On Fri, 06 Mar 2015 09:04:12 -0500, robert bristow-johnson > <rbj@audioimagination.com> wrote: > >> Greg >> Berchin's FLSD >> https://groups.google.com/forum/#!msg/comp.dsp/UBsxiYrnT0c/eFH7d45-MCwJ >> . i think i got the acronym down now. > > FDLS -- Frequency Domain Least Squares > > I'm a little reluctant to even consider what "FLSD" might stand for! > > Anyone who wants a copy of the Matlab or C example code, just send me a > private email request. Change "chatter" to "charter" and remove > ".invalid". >i'm sorry Greg; i get it wrong each and every time. sheesh. -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
Reply by ●March 6, 20152015-03-06
On Fri, 06 Mar 2015 09:28:59 -0500, robert bristow-johnson <rbj@audioimagination.com> wrote:>i'm sorry Greg; i get it wrong each and every time.No problem, Robert. For me, it's an important thing. For everyone else, it's just a thing. Greg
Reply by ●March 6, 20152015-03-06
On 3/6/15 10:56 AM, Greg Berchin wrote:> On Fri, 06 Mar 2015 09:28:59 -0500, robert bristow-johnson > <rbj@audioimagination.com> wrote: > >> i'm sorry Greg; i get it wrong each and every time. > > No problem, Robert. For me, it's an important thing. For everyone else, > it's just a thing.well, it's kinda like mis-recalling someone's name, time and time again. there's a guy i see every year at the AES. his name is David Rick. nice guy. but sooo many years i repeatedly mis-remembered his (first) name as "Rick". like "hi Rick!" like "duh". -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
Reply by ●March 7, 20152015-03-07
On Fri, 06 Mar 2015 09:56:01 -0600, Greg Berchin <gjberchin@chatter.net.invalid> wrote:>On Fri, 06 Mar 2015 09:28:59 -0500, robert bristow-johnson ><rbj@audioimagination.com> wrote: > >>i'm sorry Greg; i get it wrong each and every time. > >No problem, Robert. For me, it's an important thing. For everyone else, >it's just a thing. > >GregDid the FDLS method ever get published in SP magazine like the thread was mentioning? Mark
Reply by ●March 8, 20152015-03-08
On 3/7/15 8:00 PM, Mac Decman wrote:> On Fri, 06 Mar 2015 09:56:01 -0600, Greg Berchin > <gjberchin@chatter.net.invalid> wrote: > >> On Fri, 06 Mar 2015 09:28:59 -0500, robert bristow-johnson >> <rbj@audioimagination.com> wrote: >> >>> i'm sorry Greg; i get it wrong each and every time. >> >> No problem, Robert. For me, it's an important thing. For everyone else, >> it's just a thing. >> >> Greg > > Did the FDLS method ever get published in SP magazine like the thread > was mentioning? >i think it got published in this: http://www.wiley.com/WileyCDA/WileyTitle/productCd-1118316932.html if not also in the SP mag. -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
