Re: [music-dsp] Analog versus digital systems (Ezra Buchla)
Sure, and I didn't call you dumb. I'm not dumb for being left out the Russian space program, or wanting to write a book without a alpha-degree or what else.. It's a hard area, which I why I've tried to make a few noticeable remarks, to shake the particular subject up a bit, so to speak. If you're serious, I don't mind summing a few things up with practical application, no problem. Might even be fun. But not at this moment. Maybe tomorrow. One main thing for now, there's an interesting matter about the Switched Filter, primarily at the moment to broaden the attention, and to make the payload of this message enter in a more general digital and signal processing domain for musical (and other) applications: - Most digital systems as we now know them are Time Quantized (usually equally spaced times steps with impulse-based samples (for the sake of the Reconstruction theorem)), and Amplitude Quantized ("vertical" steps, like 16 bits for instance), for the sake of being able to use digital computers. - As a interesting comparison (for the effect) there are also Time-Quantized, Amplitude-Continuous processing chips since at least the early (Time Division Multiplexing) digital phones, of which my switched filter is an example. Clear difference: no vertical resolution to create quantization noise, and it is possible to nowadays buy chips that have digitally programmable *analog* signal path chips that may benefit from such tricks, possibly applicable in modern "digital: amplifiers. - Of course there are, at least since music in the 60s I think, machines which do amplitude quantization only, like a "grungelize" sound, applying a staircase distortion to the signal, but *without* sampling the signal in the time dimension. - For modern attempts as clean analog signal paths *with* time-varying filters one can make use of digitally programmable amplifiers, like the Crystal/Burr-Brown half-dB step, fast digital control amplifier chips (a project of mine here, at about 1/3th of the page, only if you're interested: http://www.theover.org/Audio/index2.html ), using these chips instead of "Operational Trans-conductance Amplifier"s, digitally controlled analog filters may be created. The switching technique could add more resolution to this solution, and requires fast digital control and an analysis of the effects of applying digital dithering of such kind. - Of course the original of many software implementations of digitally controlled State Varying Filters (besides the fully analog Moogs and others), from early computer-controlled analog signal path music synthesizers on, has all kinds of tuning built in the control path of the filters (and other components like oscillators and VCAs). So a part of the question I asked myself was: how is it that those analog filters sound good (or bad if that's the will of the sound programmer), given that a part of the control signal path comes from digital control. For instance: what are the properties of the DA convertor that changes a know-turn on the computer side of the instrument, into an electrical signal to drive the "cut-off" control line of the filter circuit. Gr. T.V. -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
Re: [music-dsp] Analog versus digital systems (Ezra Buchla)
or, oh! would you suggest that we stick to an analog design like the switched capacitors? but there are still lots of reasons to try and keep everything on DSP in terms of part count in a device. again, i think of this space as very informal discussion, s please no hard feelings, ezra b On Sun, Nov 10, 2013 at 4:06 PM, Ezra Buchla wrote: > i'm sorry mr. vereslt, i do apologize for insulting you. to be honest, > i made a mistake with the e-mail and thought this was still under that > other thread. and so i tried to read your code, [ which if course is > very interesting and informative to me (since i am not nearly at your > level as an engineer), it shows some fundamental aspects of > high-resolution sample reconstruction rendered in a highly-efficient > way ] - but it is not so directly relevant to coefficient changing > within the filter, is it? > > if it is, can you put the relationship a little more plainly so that > dummies like me could understand? > > for example, would you bandlimit an SVF coefficient change signal > using sinc or something? perhaps by offline-processing it and using > buffer playback? > > is this stuff only relevant at 64 bits (high vertical resolution)? > what if we obtain much greater processing efficiency at 32? > > switched capacitors is interesting to me too because i have been > thinking a lot about tunable sampling rates. this schematic itself is > old news though, is it the precise behavior of these integrators while > changing the clock rate that you wand to point out? > > you see, i really respect your intelligence and erudition on this > stuff, so i woud love to hear more explanation, just without being so > angry. i have of course read the code on your website before now, but > still need to know more specifically what part you refer to. > > i mean to try and starve troll "in all of us," i am glad that your > pursue correctness aggressively, it is a benefit for all. and so > telling you that i'm from a certain area of the world that you make > fun of, and so is dave smith for that matter, and that it is easier to > pay attention when the tone is more civil. thank you for reminding me > of it as well. > > sincerely, and thanks, > ezra b > > On Sun, Nov 10, 2013 at 3:31 PM, Theo Verelst wrote: >> ... >>>a) of course we all know about the sampling theorem and sync >>>interpolation. truly! >>> >>>b) though i enjoy reading ASM synth code, i don't see anything here >>>that is interesting. ... >> >> I'm sorry to say, but while of course I don't feel all too much of it, and >> of course that isn't a reason to use my free speech necessarily for placing >> a correction, but that little snippet is quite insulting, given the story >> thus far. So outside of rethoric, that is a technical/scientific insult of >> the first order that you're trying to force my direction. >> >> I don't really take the insult, and am glad there's serious discussion, and >> people feeling inspired to share maxima code, etc. and apparently not >> overwhelmed or something, so they post about what interests them, so on the >> average, I am glad about the results. >> >> I don't feel like scientifically defending the quotes I few simple quotes I >> posted. >> >> As a serious remark about the content of many of the musical and signal >> processing subjects: it's a great idea to use well known *analog* >> synthesizer designs as the basis for (partial) digital simulation, which I >> though already before people like Dave Smith were writing award winning >> software to that effect, and which interested me long before the advent of a >> number of software companies that occupy themselves with the subject. As the >> suggestion is from some of my quotes, it would be good to have a potent, 64 >> bit circuit simulator which allows audio output, and explicit (parts with >> curves for parameter changes) or implicit (driven sources in the network, >> OTAs in replacement circuits, etc) time dependencies, possibilities for >> storing/continuing network states, and a choice of accuracy feedbacks that >> I've been hinting at, and which clearly isn't understood by most, which >> doesn't make me continue. >> >> Also, I've suggested signal improvements, but they won't work without some >> fundamental changes to the most used algorithms, which I would prefer to be >> applied to some musical software, preferably Open Source. Those things are >> very audible, and it surprises me that people who may feel the need for >> improvements are so numb. Must be some limited musicians trying to rule the >> show, which in broader circles, which also can benefit from DSP for musical >> purposes is getting in demand. Just saying. Sounds like a simple statement >> of truth to me, and I think I'm qualified to judge that, so if you feel a >> bit humble, don't confuse that with feeling insulted. I do feel slandered, >> regulaly, and that *is8 a real issue for me, and the law. >> >> T.V. >> >> >> -- >> dupswapdrop
Re: [music-dsp] Analog versus digital systems (Ezra Buchla)
i'm sorry mr. vereslt, i do apologize for insulting you. to be honest, i made a mistake with the e-mail and thought this was still under that other thread. and so i tried to read your code, [ which if course is very interesting and informative to me (since i am not nearly at your level as an engineer), it shows some fundamental aspects of high-resolution sample reconstruction rendered in a highly-efficient way ] - but it is not so directly relevant to coefficient changing within the filter, is it? if it is, can you put the relationship a little more plainly so that dummies like me could understand? for example, would you bandlimit an SVF coefficient change signal using sinc or something? perhaps by offline-processing it and using buffer playback? is this stuff only relevant at 64 bits (high vertical resolution)? what if we obtain much greater processing efficiency at 32? switched capacitors is interesting to me too because i have been thinking a lot about tunable sampling rates. this schematic itself is old news though, is it the precise behavior of these integrators while changing the clock rate that you wand to point out? you see, i really respect your intelligence and erudition on this stuff, so i woud love to hear more explanation, just without being so angry. i have of course read the code on your website before now, but still need to know more specifically what part you refer to. i mean to try and starve troll "in all of us," i am glad that your pursue correctness aggressively, it is a benefit for all. and so telling you that i'm from a certain area of the world that you make fun of, and so is dave smith for that matter, and that it is easier to pay attention when the tone is more civil. thank you for reminding me of it as well. sincerely, and thanks, ezra b On Sun, Nov 10, 2013 at 3:31 PM, Theo Verelst wrote: > ... >>a) of course we all know about the sampling theorem and sync >>interpolation. truly! >> >>b) though i enjoy reading ASM synth code, i don't see anything here >>that is interesting. ... > > I'm sorry to say, but while of course I don't feel all too much of it, and of > course that isn't a reason to use my free speech necessarily for placing a > correction, but that little snippet is quite insulting, given the story thus > far. So outside of rethoric, that is a technical/scientific insult of the > first order that you're trying to force my direction. > > I don't really take the insult, and am glad there's serious discussion, and > people feeling inspired to share maxima code, etc. and apparently not > overwhelmed or something, so they post about what interests them, so on the > average, I am glad about the results. > > I don't feel like scientifically defending the quotes I few simple quotes I > posted. > > As a serious remark about the content of many of the musical and signal > processing subjects: it's a great idea to use well known *analog* synthesizer > designs as the basis for (partial) digital simulation, which I though already > before people like Dave Smith were writing award winning software to that > effect, and which interested me long before the advent of a number of > software companies that occupy themselves with the subject. As the suggestion > is from some of my quotes, it would be good to have a potent, 64 bit circuit > simulator which allows audio output, and explicit (parts with curves for > parameter changes) or implicit (driven sources in the network, OTAs in > replacement circuits, etc) time dependencies, possibilities for > storing/continuing network states, and a choice of accuracy feedbacks that > I've been hinting at, and which clearly isn't understood by most, which > doesn't make me continue. > > Also, I've suggested signal improvements, but they won't work without some > fundamental changes to the most used algorithms, which I would prefer to be > applied to some musical software, preferably Open Source. Those things are > very audible, and it surprises me that people who may feel the need for > improvements are so numb. Must be some limited musicians trying to rule the > show, which in broader circles, which also can benefit from DSP for musical > purposes is getting in demand. Just saying. Sounds like a simple statement of > truth to me, and I think I'm qualified to judge that, so if you feel a bit > humble, don't confuse that with feeling insulted. I do feel slandered, > regulaly, and that *is8 a real issue for me, and the law. > > T.V. > > > -- > dupswapdrop -- the music-dsp mailing list and website: > subscription info, FAQ, source code archive, list archive, book reviews, dsp > links > http://music.columbia.edu/cmc/music-dsp > http://music.columbia.edu/mailman/listinfo/music-dsp -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
Re: [music-dsp] Analog versus digital systems (Ezra Buchla)
... >a) of course we all know about the sampling theorem and sync >interpolation. truly! > >b) though i enjoy reading ASM synth code, i don't see anything here >that is interesting. ... I'm sorry to say, but while of course I don't feel all too much of it, and of course that isn't a reason to use my free speech necessarily for placing a correction, but that little snippet is quite insulting, given the story thus far. So outside of rethoric, that is a technical/scientific insult of the first order that you're trying to force my direction. I don't really take the insult, and am glad there's serious discussion, and people feeling inspired to share maxima code, etc. and apparently not overwhelmed or something, so they post about what interests them, so on the average, I am glad about the results. I don't feel like scientifically defending the quotes I few simple quotes I posted. As a serious remark about the content of many of the musical and signal processing subjects: it's a great idea to use well known *analog* synthesizer designs as the basis for (partial) digital simulation, which I though already before people like Dave Smith were writing award winning software to that effect, and which interested me long before the advent of a number of software companies that occupy themselves with the subject. As the suggestion is from some of my quotes, it would be good to have a potent, 64 bit circuit simulator which allows audio output, and explicit (parts with curves for parameter changes) or implicit (driven sources in the network, OTAs in replacement circuits, etc) time dependencies, possibilities for storing/continuing network states, and a choice of accuracy feedbacks that I've been hinting at, and which clearly isn't understood by most, which doesn't make me continue. Also, I've suggested signal improvements, but they won't work without some fundamental changes to the most used algorithms, which I would prefer to be applied to some musical software, preferably Open Source. Those things are very audible, and it surprises me that people who may feel the need for improvements are so numb. Must be some limited musicians trying to rule the show, which in broader circles, which also can benefit from DSP for musical purposes is getting in demand. Just saying. Sounds like a simple statement of truth to me, and I think I'm qualified to judge that, so if you feel a bit humble, don't confuse that with feeling insulted. I do feel slandered, regulaly, and that *is8 a real issue for me, and the law. T.V. -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
Re: [music-dsp] Analog versus digital systems
i don't want to feed the troll, but... a) of course we all know about the sampling theorem and sync interpolation. truly! b) though i enjoy reading ASM synth code, i don't see anything here that is interesting. what am i supposed to look at? a character display? a phasor? an in-place filter or chromatic tuning table in BASIC? confused. c) why are you posting a schematic of a switched capacitor filter to music-dsp? are you obliquely proposing a direct-modelling algorithm for this kind of circuit? thats an interesting thought but i'd love to see more elaboration, and what kind of advantage it would have over FIR/IIR structures. my degrees are in music, not EE, so actually i can often use a little help connecting theoretical dots and i certainly admit that. but i know how to hold the pen, and i can see the picture when it's done, so to speak. i would like to read your posts in the future, in case they are useful, but i'm no longer giving them the benefit of the doubt if they start off with vitiriol and end with ancient links of dubious relevance. - ezra buchla ( in Berkeley CA ) On Sun, Nov 10, 2013 at 10:08 AM, douglas repetto wrote: > > > Theo, please stop with the insults. > > > > On 11/10/13 9:55 AM, Theo Verelst wrote: >> >> Of course I'm aware of it this work probably won't give m a (bit late) >> YUP existence in SanFrancisco or a well paid Berkeley professorship that >> I like, but at least I don't really run the risk of looking like a >> dumb-*ss when playing the unpaid professor a bit in this territory, and >> hopefully cut down some Non-Giant Redwood trees that appear to create >> more pollution than oxygen. > > > -- > ... http://artbots.org > .douglas.irving http://dorkbot.org > .. http://music.columbia.edu/cmc/music-dsp > ...repetto. http://music.columbia.edu/organism > ... http://music.columbia.edu/~douglas > > > > -- > dupswapdrop -- the music-dsp mailing list and website: > subscription info, FAQ, source code archive, list archive, book reviews, dsp > links > http://music.columbia.edu/cmc/music-dsp > http://music.columbia.edu/mailman/listinfo/music-dsp -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
Re: [music-dsp] Analog versus digital systems
Theo, please stop with the insults. On 11/10/13 9:55 AM, Theo Verelst wrote: Of course I'm aware of it this work probably won't give m a (bit late) YUP existence in SanFrancisco or a well paid Berkeley professorship that I like, but at least I don't really run the risk of looking like a dumb-*ss when playing the unpaid professor a bit in this territory, and hopefully cut down some Non-Giant Redwood trees that appear to create more pollution than oxygen. -- ... http://artbots.org .douglas.irving http://dorkbot.org .. http://music.columbia.edu/cmc/music-dsp ...repetto. http://music.columbia.edu/organism ... http://music.columbia.edu/~douglas -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
[music-dsp] Analog versus digital systems
Hi all, Of course I'm aware of it this work probably won't give m a (bit late) YUP existence in SanFrancisco or a well paid Berkeley professorship that I like, but at least I don't really run the risk of looking like a dumb-*ss when playing the unpaid professor a bit in this territory, and hopefully cut down some Non-Giant Redwood trees that appear to create more pollution than oxygen. First, a repeat of what I've tried to communicate a number of times, as it were to discourage the idea of taking interesting mathematical truths from (lame or interesting) digital signal processing effects in music, let's first consider the theoretical basics that defy all ignoration: (A) {Digital System} --> {Digital Sample stream} --> {Digital to Analog} --> {Analog signal} , versus: (B) {Analog System} --> {Analog Signal} The main differences are on a short list: (1) The only way in which the two analog output signals of graph (A) and (B) are going to be (almost or perfectly) the same is when somehow the digital system creates very well made samples (which *CAN* come from simply playing back accurately sampled form of some frequency limited signal), and the Digital to Analog convertor is very high quality (or to achieve actual mathematical perfection: is a perfect reconstruction type) (2) The digital system implemented as a filtering of any kind of combination of FIR/IIR tap-connections is normally not coming close to making analog-equivalent signals, by far, unless it is big, and there explicit measures being taken (extremely high sampling frequencies and vertical resolution, tuning of the DA-convertors always-present transient behavior, medium long averaging effects control (hard problem) seriously long sinc-based integral corrections (computationally intensive)). (3) Reconstructing an analog signal from samples that isn't a retarded subset of all possible signals, will require a DA convertor design which has a serious signal delay, for all known normal and industrial Audio convertors. So to prevent some very measurable (by over see-ably simple traditional measurement techniques at the level of the THD of a very moderate transistor radio) distortion, serious measures would have to be taken, like outside the scope of this list. Even making sure those distortions don't become multi-fold ugly and even a potential danger to the hearing of the customers isn't easy (and thus far never has been discussed, even though these distortions are almost incredibly ugly, and host of unrealistic monitors have been "invented" which are supposed to smooth some of this over, apparently through lack of awareness of the impossibility to approach per-sample sinc functions by any resonance or other mechanical or switching amp trick). (4) It is quite possible to create a computer simulation of an electronics circuit, like a Moog filter, even with serious accuracy, and to state the output of such simulation in the form of a sequence of equidistant digital samples with accurate vertical quantization. Even this does not preclude you from having to take equal relevant care of the above, except for point (2). There, that's a few "New" things, apparently for those not blessed with either the intelligence, means or geographical or time opportunity to follow a good EE university (or for most of this: bachelor level) Sophomore year equivalent. Of course going a bit further in the better EE education (say second or third year of a serious education), you may want to practice yourself in creating computer models of interesting non-linear electronics circuits, and see if you computer simulations on the basis of these models and some form of circuit-to-signal strategy, be it based on the frequency domain or not, turn out to be accurate, and maybe invent some fun games with this, like a "Virtual Prophet-5" that everybody can run on their home computer for free, or things equally thrilling and educational! I had done some (extremely low budget) preparatory work because of my much longer standing personal interests for this (like owning various synthesizers and samplers with digital filters like the TG500 in the 80s), see eg http://theover.tripod.com/so1.html and http://theover.tripod.com/switch.html , written before the year 1999. Ir. T. Verelst http://www.theover.org/Synth -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp