Really? Pulse Width Modulation? The process? Not a specific Pulse Width Modulated signal!
If yes, then show it! I don't have none of your thesis. And then do the rest of the analysis, instead of expecting me to do! (I didn't ask you to discuss about it, you insisted to talk. Then now finish what you've started!)
Analog? Do you think this tells anything?!? Uniformly and naturally sampled PWM are both can be analog!
And if "no", (so naturally sampled PWM), then what kind of sampling theory do you want to apply? There must be something for it, but have you learned about it? Because I've learned only about periodical (=uniform) sampling! (And red something about random sampling, but it is hardly usable for such a huge dinamic range.)
"The transposing"? Transposing of what, and with what properties? Simply the existance of transposion is crying too few to make any high precision analysis based on that!
Thanks, this is more understandable! Now somehow I ought have to transform the other sentence too, and try to understand the connection with others, but since I disagree with you, it's almost impossible for me. There are sooo much contradictions with the facts I know, and there are sooo much unanswered questions!
I uselessly think about it, if you don't refer things with my terminology. You should define modules and signals, and then refer to them! Don't expect me to think the same as you by default! Peoples are different! Told toughts are needed to make agreement!
Pafi!
You are translating your questions to make new ones to be unanswered. Search for the meaning behind my answers and not just for the critics. Also read the articles I wrote about if you would, and it may will be brighter. What I wrote to be facts are facts, and what I assume are assumptions. If you can't think fast in frequency plane, the fastest solution is to take a paper and draw the spectras with the modulation and sampling effects.
What properties are you inquired about?
What could you call high-precision? The transponding can be described analitically with convolution. Isn't it enough?
Do you know anybody who make signal hold before an analog PWM in audio amplifier?
The connection with the other facts is the controller is also fed with the carrier components. And I assume the sidebands are being transponded to the baseband because of the triangle comparation caused sampling. Thus the sidebands will be forced to the output in the baseband, by the controller, although they weren't in the input signal. This is a fact with sampled feedback. But there is also sampling in analog PWM with continuous feedback, therefore I assume the fact exists in analog PWM too.
The following paper is also dealing with the same topic. The outcome is a special loop-filter topology that reduces these effects. Unfortunaltely it can't be downloaded for free.
http://www.aes.org/e-lib/browse.cfm?elib=13497
There is not even anything on that subject on the TI homepage though the two writers are TI employees.
Regards
Charles
http://www.aes.org/e-lib/browse.cfm?elib=13497
There is not even anything on that subject on the TI homepage though the two writers are TI employees.
Regards
Charles
Yes. I also haven't found any free published article about it. I saw impressing THD values of Zetex chipset, but there was no documentation about anything. Then I asked them to send the HDL description of the modulator for me, but it was never sent.
I'm curious about what the TI guys called to quadrature sampling.
I'm curious about what the TI guys called to quadrature sampling.
Gyula!
I don't understand this sentence. Please write in hungarian! ("úgy fordítod le a kérdéseidet, hogy újakat hozzanak létre, amiket nem kell megválaszolni", vagy "úgy fordítod le (értelmezed) a kérdéseidet, mint amik újak és megválaszolatlanok", vagy "úgy fordítod le a kérdéseidet, hogy"? Vagy csak simán elírtad, és az utoló szó "answered" akart lenni? Szóval hogy is szólna ez magyarul? És melyik kérdésekre vonatkozik?
Edit: after 5 minutes more of thinking I think I've finally managed to translate it formally, but it doesn't make any sense for me: "úgy fordítod le a kérdéseidet, hogy az újakat megválaszolatlanná tegyék" I don't know what it refers to. I don't know how could have I done such a thing. Please clearify!)
I always try to, but I couldn't find such a meaning, wich doesn't involve some other questions/contradictions, this because I can't accept any of these possible meanings.
I would, if I knew where they are! That's why I asked you to show it.
Why I had to do it? You wanted to convince me the whole thing is easily possible! Then do it, not make me do!
Is the transposed signal the perfect copy (or mirror) of original, or is it just something similar?
Periodical sampling makes perfect copyes, but natural sampling absolutely not!
Yes, it would be enough, if you showed me it for natural sampling! Convolution of what? Sampling time-points are function of the continuous signal! Can you describe its spectra analitically? How?
The question was not about building, but about theory. Actually I planned to make, and tryed to analyze an analog, periodically sampled PWM amp (exactly because of the before mentioned perfect transposion), but finally I found it to be inferior compared to naturally sampled one.
But this is unimportant, because there was a strong reason to believe you were talking about periodically PWM: you said "you should apply the sampling theory". Sampling theory can be applied only to periodical sampling. (At least, I don't know other, universally usable sampling theory.)
P.S.:
I'm sorry to writen soooo many times the same thing (shannon theorem applies only to periodical sampling), but I wanted it to be acknowledged finally, since this is the basic of our argument!
I don't understand this sentence. Please write in hungarian! ("úgy fordítod le a kérdéseidet, hogy újakat hozzanak létre, amiket nem kell megválaszolni", vagy "úgy fordítod le (értelmezed) a kérdéseidet, mint amik újak és megválaszolatlanok", vagy "úgy fordítod le a kérdéseidet, hogy"? Vagy csak simán elírtad, és az utoló szó "answered" akart lenni? Szóval hogy is szólna ez magyarul? És melyik kérdésekre vonatkozik?
Edit: after 5 minutes more of thinking I think I've finally managed to translate it formally, but it doesn't make any sense for me: "úgy fordítod le a kérdéseidet, hogy az újakat megválaszolatlanná tegyék" I don't know what it refers to. I don't know how could have I done such a thing. Please clearify!)
I always try to, but I couldn't find such a meaning, wich doesn't involve some other questions/contradictions, this because I can't accept any of these possible meanings.
I would, if I knew where they are! That's why I asked you to show it.
Why I had to do it? You wanted to convince me the whole thing is easily possible! Then do it, not make me do!
Is the transposed signal the perfect copy (or mirror) of original, or is it just something similar?
Periodical sampling makes perfect copyes, but natural sampling absolutely not!
Yes, it would be enough, if you showed me it for natural sampling! Convolution of what? Sampling time-points are function of the continuous signal! Can you describe its spectra analitically? How?
The question was not about building, but about theory. Actually I planned to make, and tryed to analyze an analog, periodically sampled PWM amp (exactly because of the before mentioned perfect transposion), but finally I found it to be inferior compared to naturally sampled one.
But this is unimportant, because there was a strong reason to believe you were talking about periodically PWM: you said "you should apply the sampling theory". Sampling theory can be applied only to periodical sampling. (At least, I don't know other, universally usable sampling theory.)
P.S.:
I'm sorry to writen soooo many times the same thing (shannon theorem applies only to periodical sampling), but I wanted it to be acknowledged finally, since this is the basic of our argument!
Strange that they call the triangle comparation sampling in the abstract, and then wrote quadrature sampling. Interesting. If that nomination is for sampling the feedback, among the use of another A/D, the quadrature sampling also introduce significant distortion. Because the samples will not be captured at the same time. So the spectras will be quadratured but not belong to each other in the time-plane. If these signals are considered as real and imaginary part of a complex signal, the distortion will be frequency-dependent, and rising with the frequency. The solution can be interpolating one of them to capture the true I-Q pairs. This means four-time interpolation one of them and then four times of decimation with selecting the pairs belong together. The proper interpolation needs smooth amplitude characteristics and linear phase, what can be done with high-order filters and magnitudes faster signal processing compared to the feedback sampling frequency. Otherwise the smooth characteristics can't be reached, what causes distortion in the complex envelope. So I'm curious about what could they refer to.
Just a quess without knowing anything about that article:
Can't be quadrature sampling simply that they use two (or rather 4) shifted (by 90 degrees) triangle for comparation?
http://users.hszk.bme.hu/~sp215/elektro/eszosztas/PWMquad schematic.gif
Others call it 5 level, or ClassBD modulation.
Can't be quadrature sampling simply that they use two (or rather 4) shifted (by 90 degrees) triangle for comparation?
http://users.hszk.bme.hu/~sp215/elektro/eszosztas/PWMquad schematic.gif
Others call it 5 level, or ClassBD modulation.
I don't know. I think using quadrature sampling is possible for the carrier and its harmonics, but the control of the baseband has to be made with scalar values, because the quadrature sampling works only for carrier-based signals. It's based on the transponding of carrier and its harmonics.
But they called the triangle comparation sampling, so phase shifted carrier pwm is also a possibility. Maybe some sideband can be cancelled by the phase shift of PSCPWM.
But they called the triangle comparation sampling, so phase shifted carrier pwm is also a possibility. Maybe some sideband can be cancelled by the phase shift of PSCPWM.
What I was waiting for an answer to is the other topic.
I don't know much about sideband cancellation, but I think it's more important (I mean: more obvious result) that carrier freq (f0), and f0*(2n+1), and 2*f0(2n-1) harmonics are supressed this way completely (theoretically), so effectively only 4*f0 (and it's harmonics) remains as "carrier". (And its sidebands of course, but unfortunately they have much wider spectra than a simple PWM signal with 4*f0 carrier.)
I don't know much about sideband cancellation, but I think it's more important (I mean: more obvious result) that carrier freq (f0), and f0*(2n+1), and 2*f0(2n-1) harmonics are supressed this way completely (theoretically), so effectively only 4*f0 (and it's harmonics) remains as "carrier". (And its sidebands of course, but unfortunately they have much wider spectra than a simple PWM signal with 4*f0 carrier.)
I was start to read the article, but I had stopped it at about page 111. Currently I haven't got enough time to read the whole text. If you don't find, read this:
http://www.icepower.bang-olufsen.com/files/ph.d.thesis/Chapter_2.pdf
A quote from page 10 can guide you:
"The basic theory for the development of double Fourier series expressions to represent
double periodic waveforms is described in detail Appendix B.1. DFS expressions for both
the differential and common mode for the four schemes are derived in Appendix B.2."
Then substitute the triangular functions with complex exponential forms in the IFS, reduce the sum to integration of the coefficients multiplied by frequency delayed dirac-impulses and you get the spectra.
http://www.icepower.bang-olufsen.com/files/ph.d.thesis/Chapter_2.pdf
A quote from page 10 can guide you:
"The basic theory for the development of double Fourier series expressions to represent
double periodic waveforms is described in detail Appendix B.1. DFS expressions for both
the differential and common mode for the four schemes are derived in Appendix B.2."
Then substitute the triangular functions with complex exponential forms in the IFS, reduce the sum to integration of the coefficients multiplied by frequency delayed dirac-impulses and you get the spectra.
Hi!
You think about the triangle comparation can be done by twice speed sawtooth. This also can be done with the same speed sawtooth compared with symmetric input. But the difference of the UPWM and NPWM is not this. Simply, the UPWM compares to an input what is sampled periodically, with equal distances. So a hold effect will happen until the crossing points, which causes distortion. The input signal of NPWM is analog and the crossing points select the samples, which causes a periodical, but non-equal time distance sampled signal.
The UPWM is easily useable in digital PWMs. The NPWM is obvious for analog input signal. But they are useable in those kind of systems vice versa. You can use NPWM in a discrete-time system if you interpolate the input samples, and the UPWM is useable in analog systems with a signal sampler and holder in the PWM's input.
You think about the triangle comparation can be done by twice speed sawtooth. This also can be done with the same speed sawtooth compared with symmetric input. But the difference of the UPWM and NPWM is not this. Simply, the UPWM compares to an input what is sampled periodically, with equal distances. So a hold effect will happen until the crossing points, which causes distortion. The input signal of NPWM is analog and the crossing points select the samples, which causes a periodical, but non-equal time distance sampled signal.
The UPWM is easily useable in digital PWMs. The NPWM is obvious for analog input signal. But they are useable in those kind of systems vice versa. You can use NPWM in a discrete-time system if you interpolate the input samples, and the UPWM is useable in analog systems with a signal sampler and holder in the PWM's input.
Gyula! (to post #34)
Your attitude is absolutely not fair!
It was not me, who told this thing is easily possible, but it was you! So you have to do this s*t, not me!
Don't send me to a tour around the world to find your statements! This is ridiculous! This is not my job! If you have answers, tell them! If you don't have, then just tell the trueth!
Your attitude is absolutely not fair!
It was not me, who told this thing is easily possible, but it was you! So you have to do this s*t, not me!
Don't send me to a tour around the world to find your statements! This is ridiculous! This is not my job! If you have answers, tell them! If you don't have, then just tell the trueth!
Pafi!
Maybe you haven't found the appendix. I suggest you another link which contains a precise description about the derivation of DFS with example:
http://www.icepower.bang-olufsen.com/files/convention/4446.pdf
I beg your pardon but I've got thousands of another work to do. If you have understood the process, you should consider my explanations from post #14.
Maybe you haven't found the appendix. I suggest you another link which contains a precise description about the derivation of DFS with example:
http://www.icepower.bang-olufsen.com/files/convention/4446.pdf
I beg your pardon but I've got thousands of another work to do. If you have understood the process, you should consider my explanations from post #14.
Gyula!
As you may noticed, if you really red these articles, these equations describes only the PWM of one sinusoidal audio signal, not full-band PWM process, so they are incapable of making calculations in fed back amplifiers, since there is an infinitely complex spectra on the input of comparator! Despite of the extremely simple input they are already doubly infinite series!
So stop bullshitting me!
As you may noticed, if you really red these articles, these equations describes only the PWM of one sinusoidal audio signal, not full-band PWM process, so they are incapable of making calculations in fed back amplifiers, since there is an infinitely complex spectra on the input of comparator! Despite of the extremely simple input they are already doubly infinite series!
So stop bullshitting me!
I have this UPWM bitstream, I used | to divide the periods.
00111111|00001111|00000011|
Period by period I recreate the bitstream doubling the number of samples, ordering the bits forward then backward i.e 00111111 -> 00111111 + 11111100.
00111111|11111100|00001111|11110000|00000011|11000000|
Now I play the bits at twice the clockspeed (00=0, 11=1): -
01111110|00111100|00011000
This is a close approximation to NPWM, n'est ce pas? The higher the sample rate the closer the approximation.
w
00111111|00001111|00000011|
Period by period I recreate the bitstream doubling the number of samples, ordering the bits forward then backward i.e 00111111 -> 00111111 + 11111100.
00111111|11111100|00001111|11110000|00000011|11000000|
Now I play the bits at twice the clockspeed (00=0, 11=1): -
01111110|00111100|00011000
This is a close approximation to NPWM, n'est ce pas? The higher the sample rate the closer the approximation.
w
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.