Technical and Technical IP questions for Ion

[T'ai Chi]

You're young, you probably can't help it.

'Young' is relative. I'm probably ancient compared to lifetimes of some things; like fruit flies, quality programming on tv, or the lifetime of pre-flame discussion in threads, for example.

 

[Ion]

...
He has what sounds like a good implimentation, but of course that doesn't address his fraudulent claims about FFT's vs. Goertzel.

His basic point of deception is that he'd like to snow us into thinking that one use of a Goertzel filter, giving us one line, is somehow equal to an entire FFT.
...

It is in the case that I have shown.

Because this is cleared up, I go back to what generated my claim, namely that the science came from Germany in 1958, no matter Cooley and Tukey in 1968, while U.S. does the technology for this science.

In spite of the frantic 'publish or perish' by the American universities (and the thread that triggered this one was mentioning how 'much' research is done in U.S. compared to Europe), when all the flakiness of the 'publish or persih' weeds out, then one sees solid science from Europe that stands up.

Gee, it looks like in books in Electrical Engineering this is pretty much the norm...

 

[jj]

It is in the case that I have shown.

Your "case" has shown conclusively that an FFT (that last letter T stands for "transform", not "some small part of a transform") is much, much more efficient than a full transform implemented via Goertzel.

Each of the cited references in this thread start out by saying that Goertzel is less efficient for a transform. One of them quite nicely goes on to say that it's useful when you only need a few lines out of an FFT, which is the case you use as an example.

You have, however, engaged in deception when referring to that set of lines as a 'transform'. Such a partial transform is not 1:1 and onto, does not obey Parsival's theorem, and so on, and as such simply is not a transform.

Please correct your claims, and admit that Goertzel is, as shown by all of the references, an inefficient way to calculate an FFT compared with Cooley and Tukey or even more modern methods based on twiddle factorizations.

Your claim "better than N log2 N" has been shown to be completely fallacious.

Therefore, your claim that you have "shown" any "case" involving a transform is false, and at this point it can only be a concious deception.

Therefore, your restatement of that claim is nothing but concious restatement of misinformation regarding the basic technology in your field, an atrocious behavior for anyone claiming expertise.

Because this is cleared up, I go back to what generated my claim, namely that the science came from Germany in 1958, no matter Cooley and Tukey in 1968, while U.S. does the technology for this science.

Your fraudulent claims about the FFT have already been dismissed. As the various papers refered to show, an FFT is N log2 N, and a Goertzel transform (as opposed to filterbank) is N^2 in complexity order. The evidence is clear. Your failure to acknowlege the obvious can no longer be due to either ignorance or misunderstanding, as the evidence has been both explained and cited in this thread, both by me, and by reference by a host of other authors who are acknowleged experts in the field.

In spite of the frantic 'publish or perish' by the American universities (and the thread that triggered this one was mentioning how 'much' research is done in U.S. compared to Europe), when all the flakiness of the 'publish or persih' weeds out, then one sees solid science from Europe that stands up.

You've yet to show any evidence of that set of claims. I'll be glad to add them to your "unanswered questions" thread, the one that will continue to hang around until you start to cope with some of your outrageous thefts of credit.

Gee, it looks like in books in Electrical Engineering this is pretty much the norm...

Please show me an EE book that says that you use Laplace to determine the stability of a digital FIR filter, persuant to your claim.

If you do, I will indeed agree that there is a problem with that book. If this is in fact the book you learned from, then in fact there is an atrocious book out there. Can you cite such a book, making such a claim, or are you as defenseless in your FIR claims as you are in your FFT claims?

 

[jj]

Still no answer to any of the basic issues. S**tstorm after s**tstorm, but not an answer.

 

[Ion]

Please show me an EE book that says that you use Laplace to determine the stability of a digital FIR filter, persuant to your claim.
...
Can you cite such a book, making such a claim, or are you as defenseless in your FIR claims as you are in your FFT claims?

Hey not-so-smart, one designs an F.I.R. from I.I.R.s susceptible to instability.

 

[jj]

Hey not-so-smart, one designs an F.I.R. from I.I.R.s susceptible to instability.

Well, that's false, too, but that's not what you claimed originally. What you originally claimed was that *Laplace* transforms are how one determines the *stability* of the *poles* of the *digital FIR filter*

Fact: FIR filters have no poles.
Fact: They can not be unstable.
Fact: They are in the 'z' domain, not in the Laplace domain.
Fact: Laplace has not a thing to do with it.

Show me a book that states your claim that digital FIR filters are unstable, that they have poles, and that you use Laplace transforms to figure out the stability.

THAT was your claim. Address it, or admit that your original claim was utterly false.

And one does not necessarily design FIR's from IIR's, either. That's just another mistake. One can, but in fact one rarely does, for lots of reasons, including those not so obvious.

One need look as far as Remez to see an example of a filter design method, a well-known, commonly used method, that never involves FIR filters.

Furthermore, one can design from a Sync function, an infinite, all-zero STABLE function with finite energy, any one of a number of classes of filters, via windowing (Kaiser, Hann, Hamming, Chebychev ...), without ever having poles in the equation, even for a filter starting out with infinite length, but that is unrealizable since it requires infinite delay.

Please stop spreading misinformation about your claimed subject of expertise, it is a disservice to your professional community.

 

[Ion]

...
Please stop spreading misinformation about your claimed subject of expertise, it is a disservice to your professional community.

Please stop spreading misinformation about your claimed subject of expertise since I showed Goertzel to you, it is a disservice to your professional community.

 

[jj]

Please stop spreading misinformation about your claimed subject of expertise since I showed Goertzel to you, it is a disservice to your professional community.

Wow, you can repeat what the other guy says, now, can't you?

You haven't "shown" me Goertzel, and you've "shown" that your own claims about Goertzel vs. the FFT are outright fraudulent.

I've helped you demonstrate the fraudulent nature of your claims.

You should thank me. Perhaps in the future you won't embarrass yourself to this extent.

Now, please admit fully that:

1) Goerzel, implementing a transform (as opposed to a selective filter bank) is much less efficient, computationwise, than an FFT, as it is of order N^2, instead of N log2N. Goertzel, used as a transform as opposed to a highly decimated (in frequency) filterbank, is quite inefficient, and further demostrates rounding and overload issues in real implimentations requiring numerical accuracy.

2) FIR filters have no poles, despite your claim.

3) FIR filters can not be unstable, despite your claim.

4) Laplace analysis has no reading on digital FIR filters, despite your claim.

5) The way that you would determine the stability of something in the digital (sampled) domain, even if it had poles, would not be Laplace,despite your claim.

6) FIR filters are not necessarily designed from IIR filters, despite your claim.

On those 6 issues, you are trivially shown to be utterly wrong, and your actions in repeating these inaccurate assertions is a very serious disservice to the professional community.

Cease and desist.

 

[WildCat]

Please stop spreading misinformation about your claimed subject of expertise since I showed Goertzel to you, it is a disservice to your professional community.

Wow, that French education is really paying off, lon. Your debating skills are now at the level of a US 1st-grader.

Are you ever going to admit that Cooley and Tukey's work went miles beyond Goertzel's?

And that you came to work in the US because the economy here is superior to France's devolving socialist paradise? Oh well, you really don't have to. Actions speak louder than words.

 

[jj]

Wow, that French education is really paying off, lon. Your debating skills are now at the level of a US 1st-grader.

As would his emotional reactions, I dare say.

Are you ever going to admit that Cooley and Tukey's work went miles beyond Goertzel's?

Naah, then he'll have to admit that a fat old white boy from the hills whupped his butt in a debate. He'll die of old age trying to live down this one.

And that you came to work in the US because the economy here is superior to France's devolving socialist paradise? Oh well, you really don't have to. Actions speak louder than words.

Well, he came from someplace in worse shape than France, if his profile can be believed. It's a shame he can't learn from his elders, I'm not sure where he got that handicap.

 

[Ion]

Take this quiz again, jj:

...
Take this quiz, jj:

1.) F.F.T algorithms are based on the principle of decomposing the computation of the Discrete Fourier Transform of a sequence of length N into successively smaller Discrete Fourier Transforms.

Do you agree with this statement, yes or no?

2.) I posted this:

This means that the D.F.T result, X[k], is equivalent to the kth. tap of y[n] when n = N.

It implies that the algorithm generates one output result only after N input samples have ocurred:

since any other value of the kth. tap y[n], in which n is different than N, does not contribute to the end result of the D.F.T., X[k], there is no need to compute kth. tap of y[n] until n = N.

Do you agree, yes or no?

3.) If you answered yes to 2.), then it is a smaller D.F.T..

Do you agree with this statement, yes or no?

4.) If you answered yes to 1.) and 3.), then the statements

"...F.F.T algorithms are based on the principle of decomposing the computation of the Discrete Fourier Transform of a sequence of length N into successively smaller Discrete Fourier Transforms..."

and

"...it is a smaller D.F.T."

are valid.

Q.E.D..

Then, when you pass scientifically that one, consider this:

It is in the case that I have shown.

Because this is cleared up, I go back to what generated my claim, namely that the science came from Germany in 1958, no matter Cooley and Tukey in 1968, while U.S. does the technology for this science.

In spite of the frantic 'publish or perish' by the American universities (and the thread that triggered this one was mentioning how 'much' research is done in U.S. compared to Europe), when all the flakiness of the 'publish or persih' weeds out, then one sees solid science from Europe that stands up.

Gee, it looks like in books in Electrical Engineering this is pretty much the norm...

So far, the score is between me and jj is 2-0 in my favor.

 

[Ion]

Wow, that French education is really paying off, lon. Your debating skills are now at the level of a US 1st-grader.
...
Actions speak louder than words.

I agree that actions speak louder than words.

That's why they brought me in U.S. to do the work in D.S.P..

Gee, it even looks that the 56k modem on computers -maybe on yours' too- has my D.T.M.F. Detection design and code implementation in it.

Because you cannot do it, you cannot get paid professionally for the D.T.M.F. Detection work like me and my actions speak louder than words.

 

[Ion]

Well, he came from someplace in worse shape than France, if his profile can be believed.
...

You forgot that "...the least educated parts of Apalachians..." are in worse shape than my background?

Look at you, retired and overwhelmed in competence.

In contrast, they imported me to do D.S.P..

 

[jj]

So far, the score is between me and jj is 2-0 in my favor.

The root of "science" is knowledge. When you demonstrate some, get back to me.

Frankly, it appears that you think that you can simply lie your way out of this one. Unfortunately for you, there are a host of web references convenient to people that state clearly that Goertzel is N^2 and Cooley-Tukey is N log2N.

There are also any number of books, papers, etc, that say the same thing.

Against that, we have your assertion that a transform can consist of only 1 line of N, that a transform can be a transform if it's not 1:1 and onto, no mention of completeness in your world, and that Goertzel is order N, instead of order N^2.

You're not going to succeed in your attempt to pass off your fraudulent claim, because even barely interested people can find out the facts of the case from something independent either of us.

Are you an IEEE member, by any chance?

 

[Ion]

This:

The root of "science" is knowledge. When you demonstrate some, get back to me.
...

applies to you jj.

Look in a mirror, and shape up.

 

[jj]

You forgot that "...the least educated parts of Apalachians..." are in worse shape than my background?

Really? We didn't have a megomaniacal dictator who was telling everyone to do utterly woowoo things. That's not your fault, either, I was trying to be sympathetic. You aren't to blame for your background, even though you quite snottily blame others for their background. You are the epitomy of a bigot in that regard. It's quite clear that you just can't imagine that you lost so profoundly to a fat old white boy from the hills.

You're not the only one, of course, it has happened before, once or twice, remember that guy travelling in Arkansas? As the farmer said, "I ain't lost".

Look at you, retired

Don't read too well, do you, ignorant child? I'm retired from the FIRST technical job I had, yes. That's not the whole story. What's wrong, you are afraid to tell the whole truth here?

and overwhelmed in competence.

That seems odd, coming from someone whose bizzare assertions are a near-perfect fit for off-the-wall science. Let's see:

You insist you're right and everyone else is wrong. (That's what you are doing, because everyone including me points out that Goertzel is order n^2, and you persist in claiming it's order n. Everyone is wrong but you in your world.)

You rail against the system, and say that it's "publish or perish". None the less you don't seem to know what was published. It took you HOW long to get around to even addressing Cooley and Tukey? It took you HOW long to notice the difference between an FFT (order N log N) and a DFT (order N^2 like Goertzel).

You are convinced of your own brilliance, but haven't shown anything beyond a bit of "publish or perish" yourself, and even those "citations" were untracible and unverifiable.

You pick fights over typos and spelling because you can't find anything else to attack.

You relentlessly insist on your superiority against overwhelming evidence.

In contrast, they imported me to do D.S.P..

Define "they", if you don't mind.

Who is "they"?

Really!

Enquiring minds want to know.

Up until now you've chosen to hide behind your pseudonym.

Even when "citing" a paper you say you've written, you conveniently leave out the authors, making your claim untestable. Citations, please, authors, dates, journals, ...

No more "I did a noise-shaping filter" and "I did a stability analysis of a lattice filter". Oh, that last one, really now.

By the way, are you ready to buy me that beer at ICASSP? You won't have any trouble finding me there.

 

[jj]

Repeat: Technical Questions for Ion, revisisted.

Ion, until you answer all of the following questions, any technical claim you might make must be questioned because of the obvious mistakes in this part of your "work" here:

1) Why do you assert that Goertzel is order when every authority one can cite says quite clearly that it's order (n*n) or
(n^2)?

2) Why do you assert that a DTMF detector uses a transform, when in fact it uses a very undersampled filterbank?

3) When you asserted that FIR filters were unstable and had poles, what did you really mean? I ask the question this way because the literature on the subject, some of which I've written, unequivocably says that FIR filters have no poles, and can not be unstable.

4) When you asserted that Laplace had something to do with the stability of the poles in the digital FIR filter, why did you fail to mention that Laplace reads on time-continuous systems, and digital systems are not time-continuous systems. Digital filtering is conventionally written in the 'z' domain, and the 's' domain does not apply.

5) Why did you mention half-planes in Laplace as a mechanism for determining the stability of a z domain filter, when one is in the Laplace (s) domain, and the other in the 'z' domain?

6) Why did you leave out any mention of the unit circle until after I prompted that?

7) Why have you failed to respond to any discussion of the properties of anything that is rightly called a transform? You haven't responded to the term "frame", nor "orthonormal", nor "1:1 and onto"...

8) What are the most important properties of Laplace, Z, and Fourier transforms in terms of analysis. You've invoked them all, you should trivially be able to specify domain, germane properties, and methods of implimentation, so do so.


Until you do all of these (and more) successfully, you are in the situation where you assert "I'm right, everyone else is wrong". It's something that people in CSICOP and JREF often see, and it's not a very pleasant thing.

Why do you insist on that appearance?

 

[Tez]

Well, I dont follow the rest of the discussion.

I can vouch for the fact that the DFT is O(N^2), the FFT is O(N log N) and Peter Shor's (quantum) algorithm for the DFT is O((log N)^k), where I believe k=3 for the best version so far...

 

[Ion]

Re: Repeat: Technical Questions for Ion, revisisted.

Ion, until you answer all of the following questions, any technical claim you might make must be questioned because of the obvious mistakes in this part of your "work" here:
...

jj,

until you manage to pass the quiz below:

...
Take this quiz, jj:

1.) F.F.T algorithms are based on the principle of decomposing the computation of the Discrete Fourier Transform of a sequence of length N into successively smaller Discrete Fourier Transforms.

Do you agree with this statement, yes or no?

2.) I posted this:

This means that the D.F.T result, X[k], is equivalent to the kth. tap of y[n] when n = N.

It implies that the algorithm generates one output result only after N input samples have ocurred:

since any other value of the kth. tap y[n], in which n is different than N, does not contribute to the end result of the D.F.T., X[k], there is no need to compute kth. tap of y[n] until n = N.

Do you agree, yes or no?

3.) If you answered yes to 2.), then it is a smaller D.F.T..

Do you agree with this statement, yes or no?

4.) If you answered yes to 1.) and 3.), then the statements

"...F.F.T algorithms are based on the principle of decomposing the computation of the Discrete Fourier Transform of a sequence of length N into successively smaller Discrete Fourier Transforms..."

and

"...it is a smaller D.F.T."

are valid.

Q.E.D..

don't claim that you have any professional expertise in D.S.P..

So far, you manage only 'F's, never mind being imported to work for a living in D.S.P. like me.

 

[jj]

Re: Re: Repeat: Technical Questions for Ion, revisisted.

jj,So far, you manage only 'F's, never mind being imported to work for a living in D.S.P. like me.

Your failure to reply to the questions of substance is noted.

Your ability to state incoherent questions is also noted. Your questions would mostly seem to be "preaching to the choir" but of course they address none of the fraudulent and false claims that you have made, from your claim that "Goertzel is better than FFT" (FFT means Fast Fourier Transform, not "part of a fast fourier transform) (which you can easily justify if you don't actually need a whole transform, only a few lines, but which is then a filter bank, not a transform) to your claim that there are "unstable digital FIR filters".

You again attempt to represent a heavily (frequency) subsampled DFT as a full DFT. As we both know that such a thing is not in any fashion a transform, but rather a filter bank.

As we know, such a filterbank is not 1:1 and onto. (I except OBT's from that statement, of course, but that is not what you're discussing, you're discussing something much more heavily subsampled in frequency, with non-uniform subsampling, and no overlap.)

The decomposition of the FFT into smaller FFT's (not DFT's, FFT's, since you started out talking about efficiency) is not an issue, it is an obvious fact. It is obvious from the most very basic work of Cooley and Tukey, among others, that you repeatedly deny innovation to.

The fact that when you get to 2-point FFT's, Goertzel and everything else converge to sum and difference is a well-known, but not particularly germane, result for the simple reason that the order-two case is a known degenerate case. Of course, what you have is not an 'n input, n output' device in your DTMF detector, you have an N input, much-less-than-n output.

All DFT's, no matter how sampled and how implemented, have the same number of input and output points. (Yes, you can argue about complex vs. real transforms, but unless you want to simply obscure facts, that exactly proves my point, in context.)

If you're trying to argue that Goertzel's algorithm is PART of what Cooley and Tukey used when they invented the FFT, that's not in dispute. What's the fuss about.

That does not change the facts.

Doing a transform via Goertzel is order 'n^2'.
Doing a transform via Cooley/Tukey is order 'n log2 n'.

And that's the facts.

You have claimed, repeatedly, that you have shown otherwise, but in fact you have yet to show how to calculate more than one line at a time, order 'n'. Obviously, if you then calculate 'n' lines at the same time, you wind up with order 'n*n', just like all of the references cited show.

So, once again, your assertion is shown to be wrong. Your claims are wrong, your theft of credit is unethical, your repeated misinformation is unprofessional, and your willingness to insist that "everyone else is wrong" looks quite profoundly unbalanced.