Reply by Vladimir Vassilevsky●August 25, 20062006-08-25
Rick Lyons wrote:
> It took me a while to find the darned post of mine.
> My post was:
>
> From: R.Lyons@_BOGUS_ieee.org (Rick Lyons)
> Subject: Re: Multirate filter design
> Date: Tue, 11 Jul 2006 10:45:44 GMT
> I remember reading (somewhere) that it is *VERY*
> difficult to find the optimum decimation factors for
> a multi-stage (cascaded) decimation filter if the
> number of stages is greater than two.
>
> However, thanks to the pioneers of DSP, there
> is an equation for finding the optimum
> decimation factors for a two-stage (cascaded)
> decimation filter.
There it is! Thank you. It kinda solves the question about the
decimation ratios. Once we got the decimation ratios, we can consider
the resultant P(Z) for the ripples in the passband.
Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
Reply by Rick Lyons●August 25, 20062006-08-25
On Thu, 24 Aug 2006 01:04:15 GMT, Vladimir Vassilevsky
<antispam_bogus@hotmail.com> wrote:
>
>
>Rick Lyons wrote:
>
(snipped)
>
>Sorry about that. I was not reading comp.dsp for quite some time. Can
>you please copy that post to my email (the address is on the website).
>
Hi,
It took me a while to find the darned post of mine.
My post was:
From: R.Lyons@_BOGUS_ieee.org (Rick Lyons)
Subject: Re: Multirate filter design
Date: Tue, 11 Jul 2006 10:45:44 GMT
See Ya',
[-Rick-]
Reply by Bill●August 25, 20062006-08-25
Blah, Blah, Blah -----
Suhas -
Send the original data. I will compute the phase shift.
Bill
"robert bristow-johnson" <rbj@audioimagination.com> wrote in message
news:1156316293.040509.202570@b28g2000cwb.googlegroups.com...
>
> robert bristow-johnson wrote:
> > fizteh89 wrote:
> > > Cross-correlation is not the method of choice anymore.
> > >
> > > State-space embedding is...
> > >
> > > Read US Patent Application Pub. 20030088401 at
> > > http://www.uspto.gov/patft
> > > (Already allowed with all of the original and some new claims)
> >
> > hey Dmitry, i ran that number:
> >
> >
> >
> > and it says:
> >
> > Searching US Patents Collection...
> >
> > Results of Search in US Patents Collection db for:
> > PN/20030088401: 0 patents.
> >
> > No patents have matched your query
> >
> > can you give us a link that works?
>
> i've also tried the search with APN (Application Patent Number):
> APN/20030088401
>
> and still get zilch.
>
> is there an actual link you can get us?
>
> BTW, i know patents can sometimes take a couple of years to be finally
> awarded (so it isn't still a Patent Application, but the real thing).
> you've been at this for a few years. isn't it curious that you are
> still quoting us a US Patent Application Publication number (that i
> can't get to work, anyway)?
>
> r b-j
>
Reply by robert bristow-johnson●August 24, 20062006-08-24
Rick Lyons wrote:
>
> I don't know fizteh89 (who, sadly, has decided
> to keep his identity a secret),
Reply by Vladimir Vassilevsky●August 23, 20062006-08-23
Rick Lyons wrote:
>
> I don't know fizteh89 (who, sadly, has decided
> to keep his identity a secret), but he seems
> to be one of those guys who goes through life
> thinking, "I'm smart, and all of you are stupid.
> I have the answer to your question, but I'm not
> motivated to give you the answer."
Perhaps fizteh89 is just annoyed by rather trivial discussion. There are
many ways to measure the phase and which one is the best depends on many
factors. The one who started the thread does not seem to understand the
basics; my advice would be to go to the library and figure out for himself.
> Years ago I had a boss like that.
Hehe. It is much worse to have a boss who thinks "I am stupid but I make
you smart arses dance for me".
>
> As for Vladimir, gosh I don't know.
> He seems like a talented DSP engineer but he
> comes across as being in a bad mood much of the
> time. Some months ago he posted a question here
> and I spent a fair amount of time replying to his
> post giving what I thought was a detailed
> explanation of my thoughts regarding his
> post. He never replied to my post.
Sorry about that. I was not reading comp.dsp for quite some time. Can
you please copy that post to my email (the address is on the website).
>
> In any case, maybe Vlad will post a reply here and
> let us know what he thinks.
>
I think that there are many more interesting topics then the scholastic
Nyiqust theorem. For example, let's take a linear phase lowpass filter
for antialiasing. What is the minimum group delay of that filter, if the
3dB bandwidth is X percent of Nyiquist, and the stopband attenuation is
Y dB? AFAIR Rabiner has some half-empirical formulas for that.
Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
Reply by Rick Lyons●August 23, 20062006-08-23
On 22 Aug 2006 21:01:14 -0700, "Rune Allnor" <allnor@tele.ntnu.no>
wrote:
>
>Vladimir Vassilevsky wrote:
>> fizteh89 wrote:
>>
>>
>> >
>> > You can safely disregard all comments by comp.dsp "experts": some of
>> > them are clueless, others are jealous :)
>> >
>>
>> You nailed them! 33 wizards solwing a great problem of 2 x 2 = 4
>> Thank you for the excellent comment.
>>
>> VLV
>
>The comp.dsp regulars aren't more "clueless" than that both of you
>find it worth your while to hang around, apparently...
>
>Rune
Hi Rune,
I agree with you. Anyone who says that
the guys here on comp.dsp are "clueless"
certainly doesn't recognize the knowledge
and advice of the talented & experienced
professionals in the art of DSP here.
The Information -to- Noise ratio on
comp.dsp is very high.
I don't know fizteh89 (who, sadly, has decided
to keep his identity a secret), but he seems
to be one of those guys who goes through life
thinking, "I'm smart, and all of you are stupid.
I have the answer to your question, but I'm not
motivated to give you the answer."
Years ago I had a boss like that.
As for Vladimir, gosh I don't know.
He seems like a talented DSP engineer but he
comes across as being in a bad mood much of the
time. Some months ago he posted a question here
and I spent a fair amount of time replying to his
post giving what I thought was a detailed
explanation of my thoughts regarding his
post. He never replied to my post. I had the
distinct feeling that my post was a waste of
time on my part. I wonder if Vlad
saw my post all. Maybe he didn't.
In any case, maybe Vlad will post a reply here and
let us know what he thinks.
See Ya' Rune,
[-Rick-]
Reply by fizteh89●August 23, 20062006-08-23
robert bristow-johnson wrote:
> it's not radically different. i already shown in that 2002 post (that
> i looked up and linked) that is is a variation of AMDF where you
> compute an AMDF with only 3 terms in the summation, then you subtract
> it from some "r", (a decreasing function that turns minimums into
> peaks) then you run that through the (Heaviside) unit step fuction
> (which for some signals has a desirable property of de-emphasizing
> false peaks, but since it throws away information, can make the
> algorithm susceptable to being fooled by waveforms that have minute
> differences that is lost by this unit step function), then you add it
> up to get a histogram.
>
> it is a twist on AMDF. ASDF is another twist on AMDF.
> Auto-correlation is an inverted ASDF.
>
> it's the *same* basic paradigm (not a new paradigm as you have
> repeatedly claimed) where you take your waveform and compare it somehow
> to time-delayed copies of the same waveform with adjustable time-delay
> ("lag"). at lags where the comparison shows great similarity, you have
> a good candidate for a period of the quasi-periodic waveform. all
> sorts of
>
You showed nothing... In fact, you couldn't even read the patent
application, much less understand cited references from a different
field of study.
Yeah, ASDF is just a twist on AMDF, and correlation is just a twist on
ASDF, and spectrum is just a twist on correlation and cepstrum is just
a twist on spectrum and we are all just green monkeys fooling around...
Reply by robert bristow-johnson●August 23, 20062006-08-23
fizteh89 wrote:
> >>Bill wrote:
> >> Use the Matlab "xcorr" function. The cross-correlation is maximal when
> >> the two signals are shifted with respect to each other by some amount. The
> >> amount of shift that produces the maximum is the amount by which one signal
> >> lags, or leads, the other. Be careful - periodic signals may have periodic
> >> peaks in the cross-correlation. Check the Matlab documentation for the
> >> details.
> >robert bristow-johnson wrote:
> > this, and Symon's suggestion.
>
> robert bristow-johnson wrote:
> > i think, Dmitry, this is evidence of your cluelessness. nobody else is
> > talking about looking for peaks in a cross-correlation. we're talking
> > about cross-correlation at a lag of zero. this is not a pitch
> > detection problem.
> >
>
> Maximum and peak are synonyms in my vocabulary. Aren't they, Robert ?
it might have something to do with pitch detection. but nothing to do
with measuring phase-shift unless you're gonna plot those two sinusoids
and measure the time between peaks of the waveform and divide by the
period (and multiply that result by 360 degrees). but that method,
commonly done with a scope, is not robust.
> As far as posting my own solution usign state-space embedding instead
> of cross-correlation between two signals... Well, maybe I'll do it, if
> I can find enough motivation.
then you're blowing smoke.
> I am not talking about material motivation here, just a general feeling
> I have about this whole thing: a novice and complete outsider in some
> field of science proposing something new and radically different...
it's not radically different. i already shown in that 2002 post (that
i looked up and linked) that is is a variation of AMDF where you
compute an AMDF with only 3 terms in the summation, then you subtract
it from some "r", (a decreasing function that turns minimums into
peaks) then you run that through the (Heaviside) unit step fuction
(which for some signals has a desirable property of de-emphasizing
false peaks, but since it throws away information, can make the
algorithm susceptable to being fooled by waveforms that have minute
differences that is lost by this unit step function), then you add it
up to get a histogram.
it is a twist on AMDF. ASDF is another twist on AMDF.
Auto-correlation is an inverted ASDF.
it's the *same* basic paradigm (not a new paradigm as you have
repeatedly claimed) where you take your waveform and compare it somehow
to time-delayed copies of the same waveform with adjustable time-delay
("lag"). at lags where the comparison shows great similarity, you have
a good candidate for a period of the quasi-periodic waveform. all
sorts of
> Just try it once for yourself to see how it feels...
i know how it feels to have work of mine ignored. in 1996 i took the
loose ends of my Ph.D. research (some of which i had been thinking
about since 1979), snipped off the loose ends that were a dead end (but
was necessary to solve for my dissertation to be accepted) and tied
together the loose ends that made some sense and published it as an AES
preprint and since have revised it. it's the "Wavetable 101" paper.
it's been pretty much ignored and i don't have the where-with-all to
build a company to build a synth based on the ideas. only some old C
code demoing it (and the demos were pretty successful in showing that
the math works). on the other hand, another simple text document that
i put together on a whim, just because no one else had (other than
Zolzer, but my equations are simpler and more consistent than his), and
that became the most cited document i had ever written. very weird:
whim gets all this attention, would-be dissertation material gets
ignored.
but there's still a difference between how you and i represent our
work.
r b-j
Reply by fizteh89●August 23, 20062006-08-23
r b-j wrote:
> typical inpenetrable patent drivel (or patent-speak), i've been there
> before.
Hm... My impression was that all those people at ICASSP and elsewhere,
including PTO examiner, didn't have much trouble understanding the main
concept behind the method... not to mention the source code...
Hm.... Maybe two of you can band together and manufacture some great
"prior art" ? He knows how to do it....
Reply by fizteh89●August 23, 20062006-08-23
>>Bill wrote:
>> Use the Matlab "xcorr" function. The cross-correlation is maximal when
>> the two signals are shifted with respect to each other by some amount. The
>> amount of shift that produces the maximum is the amount by which one signal
>> lags, or leads, the other. Be careful - periodic signals may have periodic
>> peaks in the cross-correlation. Check the Matlab documentation for the
>> details.
>robert bristow-johnson wrote:
> this, and Symon's suggestion.
robert bristow-johnson wrote:
> i think, Dmitry, this is evidence of your cluelessness. nobody else is
> talking about looking for peaks in a cross-correlation. we're talking
> about cross-correlation at a lag of zero. this is not a pitch
> detection problem.
>
Maximum and peak are synonyms in my vocabulary. Aren't they, Robert ?
As far as posting my own solution usign state-space embedding instead
of cross-correlation between two signals... Well, maybe I'll do it, if
I can find enough motivation.
I am not talking about material motivation here, just a general feeling
I have about this whole thing: a novice and complete outsider in some
field of science proposing something new and radically different...
Just try it once for yourself to see how it feels...
Anyway, lack of motivation is what I am facing these days...