Technical and Technical IP questions for Ion

[jj]

In http://www.randi.org/vbulletin/showthread.php?s=&threadid=35228&perpage=40&pagenumber=5 ,

Ion asserted a number of things, based on an assertion in his previous thread that americans are ugly and stupid. One of them was that, by his omission of others, that Goertzel was responsible for the modern implementation of the FFT. Why, Ion, did it take you 5 pages to admit that his work is derivitive of Cooley and Tukey's? (NB, there's nothing WRONG with that.)

Why, then, when you finally admitted this, did you cite a bunch of work from people born before the USA even existed as some kind of confused, purported evidence that the USA is a nation of IP theives who just take others' work, even though it's obviously absurd to cite such evidence.

Why did you suggest that DTMF signalling, and in particular US Touchtone (might be TM) signalling, was a European invention, based on one small aspect, that of tone detection?

Where have you addressed the issues of getting such tones across a highly nonlinear network?

Why have you not cited other uses of MF tone signalling that preceded TouchTone (TM?), and properly address the invention and origin of those methods?

Why haven't you addressed the overall system design of the Touchtone (TM?) signalling system in the USA, and instead harped seemingly endlessly on one and only one aspect of a large problem?

What is your evidence that, as you have so clearly stated, I have no idea what Fermat's theorem is?

What is your evidence, as you have implied, that you have standing to lecture me in the field of DSP? (Note, anyone can learn, but you seem to imply that in this field, you are the teacher, and I am the student.)

Please show your standing, and your evidence that *I* have no experience or understanding in this field. Patents with your name on them (subject to confirmation it's actually you), papers with your name (ditto), and products on the market may be cited here, at the very least. As that information is publically available in my case, I really feel little need to say more.

Finally, please explain the following. It's hard to reduce this to one and only one question, because of the stunning absurdity of the whole, so I will simply quote this absurdity and allow you to flounder about as you might. In the thread cited above, you've said:
-----
Now let's do the same with Finite Impulse Response digital filters that are stable based on where their poles sit after performing a Laplace (Fr.) transform on their transfer (i.e.: output divided by input) function.
-----


Ok, let's ask a few questions about this:

1) FIR filters have how many poles?

2) If they have poles, in what form are these poles expressed?

3) How is an FIR filter "unstable"?

4) In what domain are sampled filters analyzed?

5) What do Laplace transforms have to do with any of this?

6) In what way would the work of Rabiner and Gold (Digital Signal Processing) relate to this?

7) How many zeros does an FIR filter have?

8) In mathematical terms, a pole implies a recursive structure. How does this relate to the term "FIR"?


ATTENTION: This is not an attack thread. DO NOT TURN IT INTO ONE, I DO NOT WISH IT MOVED TO FLAME WARS. RESTRICT YOUR COMMENTS TO THE TECHNICAL AND IP ISSUES!

 

[Ion]

...
Ion asserted a number of things, based on an assertion in his previous thread that americans are ugly and stupid.
...

This is what I asserted in my opening post of that thread:

...
I come from Europe, where people are more educated and well-kept than the Americans.

What takes place in U.S. is islands of quality in a sea of mediocrity.

People in U.S. are still superstitious religion-wise, greedy money-wise, without social programs and unkept intellectually and in physical looks.

The islands of quality come from a quality of Americans that the Americans themselves are unaware of but that foreigners like me are aware of, and they come from the high standards of skills required in immigration to the U.S....

If you, jj are an American specialist in the intellectual D.S.P. field, then my posts remain consistent with the statement "...The islands of quality come from a quality of Americans that the Americans themselves are unaware of but that foreigners like me are aware of,..." that I made there.

It is similar to U.S.' Michael Jordan and Pete Sampras being specialists of fitness, while my posts remain consistent with that general statement.

Compare my claim "...People in U.S. are still superstitious religion-wise, greedy money-wise, without social programs and unkept intellectually and in physical looks...." with my day.

My day is of the standard that I am brought to work with scientific skills that are in shortage among the U.S. locals, I go to work in a competitive innovative field, then after work I train in swimming against the likes of who made the German national team in 1998 for the World Swimming Championships in Australia, and finally I come home more than 12 hours later to recover for the next day.

Now, let me fire up the technicalities.

(Bear in mind though that my education is not perfect, that you will see scientific mistakes here, and that what I tell people around me, stands

-i.e.: "If I were to redo my life, I would do the same things, but better."-)

...
One of them was that, by his omission of others, that Goertzel was responsible for the modern implementation of the FFT. Why, Ion, did it take you 5 pages to admit that his work is derivitive of Cooley and Tukey's? (NB, there's nothing WRONG with that.)

Why, then, when you finally admitted this, did you cite a bunch of work from people born before the USA even existed as some kind of confused, purported evidence that the USA is a nation of IP theives who just take others' work, even though it's obviously absurd to cite such evidence.
...

This is incorrect:

Goertzel in 1958 has the fastest Discrete Fourier Transform so far, traced back to -roughly in milestones- Danielson and Lanczos in 1942, Runge in 1905, Gauss in 1805, and the creator, Fourier.

Cooley and Tukey published their algorithm in 1965, after Goertzel's work in 1958.

...
Why did you suggest that DTMF signalling, and in particular US Touchtone (might be TM) signalling, was a European invention, based on one small aspect, that of tone detection?
...
Why haven't you addressed the overall system design of the Touchtone (TM?) signalling system in the USA, and instead harped seemingly endlessly on one and only one aspect of a large problem?
...

It's one precise example, among many.

If you know more, do tell.

...
What is your evidence that, as you have so clearly stated, I have no idea what Fermat's theorem is?
...
What is your evidence, as you have implied, that you have standing to lecture me in the field of DSP? (Note, anyone can learn, but you seem to imply that in this field, you are the teacher, and I am the student.)

Please show your standing, and your evidence that *I* have no experience or understanding in this field.
...

The only introduction of yourself comes from:

...
And those of us, born from stock long in the USA, whose families come from some of the poorest, least-educated parts of the Appalachans, are to think what of you and your apparent bigotry?

in a tangent thread where we posted preceding my thread by one day.

...
Ok, let's ask a few questions about this:

1) FIR filters have how many poles?

2) If they have poles, in what form are these poles expressed?

3) How is an FIR filter "unstable"?

4) In what domain are sampled filters analyzed?

5) What do Laplace transforms have to do with any of this?

6) In what way would the work of Rabiner and Gold (Digital Signal Processing) relate to this?

7) How many zeros does an FIR filter have?

8) In mathematical terms, a pole implies a recursive structure. How does this relate to the term "FIR"?
...

1), 2), 3), 4), 5), 7), and 8) are addressed in one example of filter that I have worked on
....................................................................................................

It is taught that Finite Impulse Response (F.I.R.) filters can not be unstable.

In designing an F.I.R. filter with interpolation points uniformely spaced between 0 and pi, the Discrete Fourier Transform can be used.

An example of Discrete Fourier Transform (D.F.T.) is the algorithm below.

A N points D.F.T. of frequency k is:

(I don't have mathematic symbols on this computer, so I use plain language which makes it harder to follow though)

X[k] = (Summation from n = 0 to N-1 of) x.exp(-j.k.pi/N)

= (Summation from r = -infinity to +infinity of) x(r).exp(-j.2.pi.(-k.(n-r))/N).u[n-r] for n=N

(where u(n-r) is the unit step)

= denoted as kth. tap of y[n] for n=N

It is a discrete convolution of the finite-duration sequence x[n] with the sequence kth. tap of y[n].

A system with impulse response of exp(j.2.pi.k.n/N).u is depicted -in the frequency domain- by:

kth. tap of Y(z) = X(z) + (x(z)exp(-j.2.pi/N).(-k))/z

Applying the inverse Z transform, gives in time domain:

kth. tap of y = x + x(n-1).(Nth. tap of exp(j.2.pi.k/N))

thus the transfer function of the system:

kth. tap of H(z) = ((1-exp(j.2.pi.2.k/N))/z.z)/((1-exp(j.2.pi.k/N))/z)

= ((1-exp(-j.2.pi.k/N))/z)/(1-2cos(2.pi.k/N)/z + 1/z.z)

The numerator in this formula for the transfer function has the zeros and the denominator has the poles.

So, this is a D.F.T. used in designing stable F.I.R.s, however it has zeros and poles like if it can be unstable.

Stable in the Fourier domain means poles within the unit circle, and in the Laplace domain means on the left half plane of the real axis.

The total computation of this D.F.T. is:

2.(N+2) real multiplications and 4.(N+1) real additions, which is less than N log N.

...
Patents with your name on them (subject to confirmation it's actually you), papers with your name (ditto), and products on the market may be cited here, at the very least. As that information is publically available in my case, I really feel little need to say more.
...

Examples of filters I have worked on
.................................................................

In a Low-Delay Vector Excitation Coding algorithm for speech compression, I did work on

.) an Infinite Response Filter (I.I.R.) all-pole lattice filter of order 20 for a 'Short-Term Predictor', stable when adapting its reflection coefficients,

and

.) a pole-zero weighting filter used to shape the frequency distribution of the quantization noise, by applying input speech to a tenth order all-zero lattice filter with coefficients adapted by the Least Mean Square method and used by a transversal filter that gives the transfer function of the weighting filter.

 

[Ion]

To emphasize this:

...
8) In mathematical terms, a pole implies a recursive structure. How does this relate to the term "FIR"?
...

this formula of a D.F.T. for an F.I.R.:

....
thus the transfer function of the system:

kth. tap of H(z) = ((1-exp(j.2.pi.2.k/N))/z.z)/((1-exp(j.2.pi.k/N))/z)

= ((1-exp(-j.2.pi.k/N))/z)/(1-2cos(2.pi.k/N)/z + 1/z.z)

The numerator in this formula for the transfer function has the zeros and the denominator has the poles.

So, this is a D.F.T. used in designing stable F.I.R.s, however it has zeros and poles like if it can be unstable.
...

has a recursion with a forward loop and a backward loop executed every many samples.

 

[jj]

(regarding Cooley and Tukey)

This is incorrect:

Goertzel in 1958 has the fastest Discrete Fourier Transform so far, traced back to -roughly in milestones- Danielson and Lanczos in 1942, Runge in 1905, Gauss in 1805, and the creator, Fourier.

Cooley and Tukey published their algorithm in 1965, after Goertzel's work in 1958.


You're leaving out an important detail, quite aside from your citation of people born before the USA existed. Please, now, since you're asserting expertise, you tell us the important fact you left out.

Explain why it makes your example, well, I'll be kind and say "confused", for extra credit.

What's more, please explain why it took you so long to even mention Cooley and Tukey, won't you?

(NB, for the reader, I will explain this issue via PM, if you like, if you're not Ion, who I believe is well aware of what s/he is saying.)

If you know more, do tell.

I already have. You didn't notice.

The only introduction of yourself comes from:

You, sir, are entirely negligent. The answer to "my introduction" or whatever you want to call it is directly and fully within your reach every time you read an article. It's there all the time, changes about once a year, and only to update information. It's entirely within your reach, and only your overweening arrogance (or fear at finding out what kind of blender you've jumped into) would appear to prevent you from pushing on the 'www' button at the bottom of your screen.
.
It is taught that Finite Impulse Response (F.I.R.) filters can not be unstable.

No, it is not "taught". It is demonstrated mathematically. That is, unless you're going to reuse the term FIR to mean something else. FIR means all of the vectors involved are FINITE. Not cancelled by some other infinite vector, etc, but FINITE.

In designing an F.I.R. filter with interpolation points uniformely spaced between 0 and pi, the Discrete Fourier Transform can be used.

Filter DESIGN? We're not talking about FILTER DESIGN, we're talking about your assertion that FIR FILTERS CAN BE UNSTABLE, and that LAPLACE is how you examine their stabilty!

Regarding design:
Are you talking about Remez here? Yule-Walker? Are you doing LMS? Are you about to claim that an FIR filter inside a loop (which means it's an IIR filter, really, not an FIR) can be unstable? Would you mind being a bit more specific.

I share your hatred of doing mathematical expressions in text. It doesn't work.

The numerator in this formula for the transfer function has the zeros and the denominator has the poles.


Yep, sure does have poles in the denominator. Is there a reason you're abusing the terminology and calling this an FIR filter? If this is a term inside a filter design, that's outside of the scope of your original claim. I'm not going to bother to track your math, it's not worth it. Say what you mean, instead.

For that matter, we're still talking about filter design, I see, so are we talking Remez here, Kaiser, etc. Would you mind being just a BIT specific instead of blowing all this unnecessary smoke?



So, this is a D.F.T. used in designing stable F.I.R.s, however it has zeros and poles like if it can be unstable.

And that's a completely DISHONEST statement, if it's being used to justify your claim that FIR filters are unstable.

Now, you've run screaming from "unstable FIR's" to unstable FIR design.

That's a different thing entirely. Filter design algorithms certainly can be unstable, but that isn't what you claimed.

Stable in the Fourier domain means poles within the unit circle, and in the Laplace domain means on the left half plane of the real axis.

Laplace is entirely irrelevant, as is Fourier, when we're talking about 'z' transforms.. As you've already stipulated, we're in the 'z' domain, ergo your citation of Laplace was utterly irrelevant.

Laplace has nothing to do with it unless you're going to try to apply a mapping. Which will that be, Impulse Invariant? Matchz? Bilinear?

As there are many mappings of one to the other, you simply can't say that Laplace has ANY meaning at all. The germane question is that of the radius of the pole (in the IIR filter), is it >1? If it is, it's unstable. If it's ==1 it's conditionally stable. No Laplace need apply. There's not an 's' in this whole formulation.

Your citation of Laplace was entirely as bogus as your claim that there are unstable FIR filters.

The total computation of this D.F.T. is:

2.(N+2) real multiplications and 4.(N+1) real additions, which is less than N log N.

If you can't baffle them with BS, try to baffle them with completely irrelevant data, eh? Yes, I know how many multiply/adds it takes to do a good FFT. I've known that for a while, too. Tell me, lad, what year were you born? It might JUST be true that I knew that before you were even born.

NB to others here: He keeps pointing out how the FFT is more efficient. Nobody's arguing that. He's cited the N log N nature of the FFT about 14 times like it's some mantra that he can keep chanting.

Now, tell me, Ion, I have 256 points of data. I want to take those 256 points of data, and calculate the complex 129 line spectrum (well, 127 complex, 2 real, to be specific). How many operations will this take with Goertzel? How many with a rolling-radix 2/4/8 FFT, see Proakis' book for an example of that.

I want the whole spectrum, mind you, not one or two lines.
(cough)

Examples of filters I have worked on
.................................................................

In a Low-Delay Vector Excitation Coding algorithm for speech compression, I did work on

.) an Infinite Response Filter (I.I.R.) all-pole lattice filter of order 20 for a 'Short-Term Predictor', stable when adapting its reflection coefficients,


You were blowing all sorts of smoke about FIR filters being unstable (they aren't, although they can be ill-conditioned, something else altogether, they can't be unstable), then you tried to move to FIR design instead of FIR filters, having realized you blew the first one entirely. THEN you tried to justify mentioning Laplace, when Laplace has nothing to do with it. Now we're at data-dependent IIR filters! Goodness!

Not only that, what you're referring to goes back to LPC10 days (actually farther). Would you like to credit the people who came up with direct and lattice form predictors? How about it? Who were they?

You said FIR filters can be unstable. What are we talking about now?

A data-dependent IIR filter!!!! Not just an IIR filter, but a data-dependent IIR Filter.

Talk about changing the subject. You're not changing the horse in midstream, you're leaping from one mule to another in class 6 white water here. Make sure you hit the saddle, you don't want to land in the water here!

Yes, they can certainly be unstable, FIR filters can. Did you mean that in the first place? If so, why didn't you just admit that your first statement was technically wrong, instead, and say you meant IIR instead?


and

.) a pole-zero weighting filter used to shape the frequency distribution of the quantization noise, by applying input speech to a tenth order all-zero lattice filter with coefficients adapted by the Least Mean Square method and used by a transversal filter that gives the transfer function of the weighting filter.

Wow! A noise-shaping filter!

Um, didn't I ask you about one of thise about 20 articles back?

(yes, I did)

Look. I'm tired of your evasion.

You simply made a mistake when you said "unstable FIR". You further made a mis-statement when you suggested that Laplace was a germane way to test for stability of digital filters.

Now, you've realized that, and tried to slip over from filters to filter design, and then from filter design to IIR filters, and then to data dependent IIR filters, while APPEARING to defend the indefensible.

Why don't you just give it up. Admit you're beat entirely on Goertzel, on FIR filters, on IIR filters, and on your entire premise that it's *only* immigrants who make the USA fly.

 

[jj]

To emphasize this:

this formula of a D.F.T. for an F.I.R.:

has a recursion with a forward loop and a backward loop executed every many samples.

No, that's not a "DFT for an FIR" you're simply using an IIR to model the basis vector of a DFT.

And in any case, that basis vector is not finite, so it's not FIR in any case. You're simply truncating it. So stop playing with words, you know what you're really doing.

 

[clk]

No offense to you two, but I find this thread highly amusing, because only on these forums will you find people passionately arguing about signal processing history and methods.

 

[jj]

No offense to you two, but I find this thread highly amusing, because only on these forums will you find people passionately arguing about signal processing history and methods.

What we're actually arguing about is his complete misattribution of credit, algorithm complexity, algorithm capability, etc, all of which read back to his assertion that (paraphrased) americans are stupid and ugly.

What we've shown is that he's evasive, misleading, confused, and makes mistakes.

 

[DanishDynamite]

jj:

...his assertion that (paraphrased) americans are stupid and ugly.

I'm not sure I understand. Is there some reason to suspect that this isn't true?

(Sorry, couldn't resist)

 

[jj]

jj:I'm not sure I understand. Is there some reason to suspect that this isn't true?

(Sorry, couldn't resist)

Heeesh. That Hertz. I guess what Kant around Gauss around. You'll Popper right up, say Watt, and put deCartes before deHorse? Well, Max, I just B curl D.

 

[DanishDynamite]

jj:

Heeesh. That Hertz. I guess what Kant around Gauss around. You'll Popper right up, say Watt, and put deCartes before deHorse? Well, Max, I just B curl D.

Worst. Pun. Ever.

 

[Brown]

Volt's all the ruckus? Coulomb down! Fourier information, there's no need for Joule to be angry.

 

[jj]

Volt's all the ruckus? Coulomb down! Fourier information, there's no need for Joule to be angry.

Whoa, you get a charge out of that? Must be way amped up, to try to elecTron. Wire we doing this, anyhow, to gauge somebody's resistance?

 

[jj]

jj:Worst. Pun. Ever.

But, but you already said that at least once about some OTHER pun I made.

Are you going AC on us now, is this a new phase, or just a power factor for you?

 

[DanishDynamite]

jj:

But, but you already said that at least once about some OTHER pun I made.

I don't see a contradiction. They just keep getting worse.

 

[jj]

jj:I don't see a contradiction. They just keep getting worse.

Man, that's a lot of pressure. I have to keep getting worse!?!?!?

 

[jj]

Oh, yes, I should note, the 1-line transform Ion is pointing out is in fact earlier than Cooley-Tukey, but has nothing to do with an FFT. It's a one-line transform. (meaning it only calculates one line per pole pair, requiring order N operations per line).

He said he was talking about FFT's, and there is some later work regarding multiple factorizations I thought he was referring to. It's quite useful, too, of course, but doesn't predate Cooley and Tukey.

Oh well, it's hard to pin him down, but we're getting there. His FFT isn't, and he hasn't replied since he logged in this afternoon.

We shall see, then, what answers come of this.

 

[T'ai Chi]

Tukey.

Yes! I recognized something in this thread!!!

 

[jj]

Yes! I recognized something in this thread!!! [/B]

Still waiting for Ion.

After his/her single attempt at a blizzard, s/he seems to have gone AWOL.

So, WhoChi, I also thought I said 'Shannon', maybe you missed that one.

 

[BTox]

No offense to you two, but I find this thread highly amusing, because only on these forums will you find people passionately arguing about signal processing history and methods.

This whole thread is freakin' greek to me.

 

[jj]

This whole thread is freakin' greek to me.

Actually, it's more French and German.

The fact that you don't know much about it, though, is just what Ion is counting on. He's made a bunch of very streched arguments to support his contention that americans are ugly and stupid, I took him to task, and he's trying to confuse the issue with a great big cloud of foul wind.

You know, the kind where the basis is factual, or mostly so, but his interpretation is farsical... It's a lot like some other folks we see here.