Re: Linear Phase via 'Phase Cloning' New Method

On Fri, 15 Feb 2008 11:32:53 -0600, "westocl" <cweston_@hotmail.com>
wrote:

>>On Fri, 15 Feb 2008 09:43:03 -0600, "westocl" <cweston_@hotmail.com>
>>wrote:
>>
>>>>westocl wrote:
>>>>
>
>>
>>So just showing up and claiming that your new algorithm is "better"
>>just raises the question "better how?".  
>
>stop saying the word better, i said 'new' and you are to qualify if its
>better. i dont work in that industry

You said it "beats a FIR", which implies better than.

If you want someone to look at your scheme and "qualify if it's
better", you need to explain more clearly what it is you're trying to
achieve and how you think you've achieve it, and what "better" means.
Otherwise it's just handwaving, IMHO.

It seems to me you're asking akin to "Tell me the widget hidden inside
this box is useful" and then not let anyone look inside the box or
even fully explain what the widget is supposed to do.

I suggest you're not going to get very pointed responses until you do.
>-
>>Negative group delay means that the output appears before the input
>>arrives. 
>
>-so are you saying that i couldnt give a whole signal a known group delay
>and delay parts of the signal more than others???  how is that not a
>relative negative group delay? 
>
>
>>I think you've gotten people's interest and curiosity, but I would
>>suggest trying to understand the context of general filter design at
>>least well enough to know how your system fits in.
>
>-no Eric your job is to understand filters.  My job is to clone phases.

I've no idea what "clone phases" means or why it would be useful to
anyone.

>I initially asked for a text file of some data from people that maybe i
>could try and process with my algorithm. I get nothing out of screwing the
>data up.  if the algoritim works, i would just give you the coeffients i
>used to tweak your data then we would have helped one anther. I learn the
>limitations of the algoritm, and you have nothng to loose.

What sort of data and what metrics should be important?  I can
generate some nice white noise for you, but you can do that, too.

If your job is to "clone phases" what properties should a test signal
have that would demonstrate such an operation, and how would one know
whether it is successful or not?

>and further more, i thought i said 'i insert a known time delay into the
>system', therefore the output will never preceed the input.

Then you'll always be dealing with a positive group delay.   

>even though you meant it for jest...Thats exactly why the algoritm didnt
>run initialy when i tried to run it.  I didnt put a delay in.  To clone a
>signals phase, it must see the signal first!

I wasn't jesting.   You made a statement that didn't make sense (IMHO
you've made several, clarification is coming slowly) and I was looking
for clarification.
Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.ericjacobsen.org

>westocl wrote:
>
>   ...
>
>> Ok, I have a question then. Is there currently a way to make Cheby, or
>> elliptical IIR filters with linear phase in the pass band?  If the
answer
>> is no, i shall explore it.
>
>It is no by definition. The phase response of a Chebychev or elliptical 
>IIR filter is part of it. There is no way to make a bicycle with three 
>wheels or one, trikes and unicycles notwithstanding.
>
>> I need a way to qualify phase cloning as being new and useful. but
>> currently i lack practical DSP knowlege and i do not want to
erroneously
>> state claims.
>
>What you claim is combining any magnitude and phase responses. I can 
>prove that impossible, but it is possible to produce a close 
>approximation. A simple way to do that without excessive delay would be 
>valuable.
>
>Jerry
>-- 
>Engineering is the art of making what you want from things you can get.
>???????????????????????????????????????????????????????????????????????
>
A simple way to do that without excessive delay would be 
>valuable.

Jerry, i appreaciate the feedback.  I shall attemp to just that.

>On Feb 20, 10:41 am, "westocl" <cwest...@hotmail.com> wrote:
>> Ok, I have a question then. Is there currently a way to make Cheby, or
>> elliptical IIR filters with linear phase in the pass band?  If the
answer
>> is no, i shall explore it.
>>
>
>Are you asking about the magnitude and phase response
>for a finite pre-determined set of frequencies?  Or
>for the response to a continuous range of frequencies?
>Are you asking about an LTI filter response within some
>given error bound and maximum delay, or something else?
>
>

Ron, in a way i mean a pre determined set of frequencies.  I shall only
target frequencies with a given sampling rate up to Fs/2, so that
simplifies things.

My ultamte goal for Phase cloning is to have a filter whose phase response
is exactly what you describe with a low pass magnitude response.  Linear
phase is just one specific application (and with my ignorance of DSP) i
thought was wat people wanted all of the time.  In general i shall attemp
to exploit that one can have any group delay respose we so desire. 

I want to eliminate the need for equalizers all togeter.