Received: by 10.68.230.98 with SMTP id sx2mr9930758pbc.1.1335730667384;
        Sun, 29 Apr 2012 13:17:47 -0700 (PDT)
Path: r9ni113200pbh.0!nntp.google.com!news2.google.com!volia.net!news2.volia.net!feed-A.news.volia.net!fu-berlin.de!uni-berlin.de!individual.net!not-for-mail
From: Greg Berchin <gjberc...@chatter.net.invalid>
Newsgroups: comp.dsp
Subject: Re: Reconstruction of min-phase system from non-uniform phase samples
Date: Sun, 29 Apr 2012 15:17:02 -0500
Organization: Greg
Lines: 35
Message-ID: <bu7rp7hslitc4m06jd77jldagpm8e2pqqq@4ax.com>
References: <48abe980-519c-4cb4-97e8-4266b8a8be8a@r13g2000vbg.googlegroups.com> <874ns2l7aj.fsf@randy.site>
Reply-To: gjberc...@chatter.net.invalid
Mime-Version: 1.0
X-Trace: individual.net ERguwG5mh4Dw6bMuja3F+QDsaDIxjb8Cs8KQVRUyEb18z56xyr
Cancel-Lock: sha1:Tw0u300KWXNjT7lsnTno01O1GJc=
X-Newsreader: Forte Agent 4.2/32.1118
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

On Sun, 29 Apr 2012 14:30:44 -0400, Randy Yates <ya...@digitalsignallabs.com>
wrote:

>Robert Adams <robert.ad...@analog.com> writes:
>
>> I am trying to reconstruct the magnitude/phase of an unknown
>> continuous-time (S-domain) minimum-phase system based on samples of
>> the phase. The problem is that the frequencies at which the phase is
>> known are random. Any ideas on how to approach this?
>
>This sounds like a good problem for the FDLS algorithm (Frequency Domain
>Least Squares) which our very own (here on comp.dsp) Greg Berchin has
>been the champion of for many years. 

Randy, you beat me to the answer. FDLS should be able to handle this once the
magnitude is extracted from the phase, using the logarithmic relationship that
RB-J mentioned in another post. The only problem that I see is that FDLS
generates a z-domain model, while Robert Adams has an s-domain system. Thus,
FDLS can, at best, only approximate the response.

>This seems to be a Matlab function implementing FDLS: 
>
>  http://www2.units.it/ramponi/teaching/DSP/materiale/ES_5_2.m

That version WAS NOT WRITTEN BY ME. (More correctly, it was MODIFIED by someone
else from a version that WAS written by me.) The original generic MATLAB code,
and/or a C-language equivalent, is available directly from me. (Change "chatter"
to "charter" and eliminate ".invalid".)

>I believe Greg also had an article in IEEE Signal Processing Magazine on
>this (in Rick Lyons' "DSP Tips and Tricks" column) a few years back.

That's available from me, as well.

Greg Berchin