I wonder if what Tim is asking is whether there's likely to ever be a new _analog storage_ format? We went through wax cylinder, various discs, shellac, vinyl, wire and tape magnetic, optical on film...
@@Projacked1 Bezier and NURBs based 3d can't do sharp corners well. Tight bends at best. While polygonal has single point intersections. But can't do curves well. Uses lots of separate points to create a curve. Same with audio. DSD/ vector would follow a curve. Where PCM is hard sample points with each a stand alone sample.
@@glenncurry3041 hmm....you're right. Not so brilliant , maybe. You can change the functionality of the intersection points tough. I can make sharp points with addtional curving from 2 points, if you have the option like graphic/ 3D design. But the processing load will be bigger.
@@Projacked1 I sold GCI back in the '80's and '90's with Unix workstations like SGI and Symbolics. Some 3D software was poly, some NURBs. You just can't make a hard intersection with NURBs nor Bezier. But you can't make a true circle nor radiused corner with poly's. The same thing is true with digital audio. PCM is like Poly. Straight line from point A to point B. Straight line up from zero to the voltage of that sample. Then next sample point. And some outside system has to figure out how to transition between them/ how to make it round. The more samples the closer, but never round.
I've seen this used, but not in high end audio. I first saw it used in 1983 in a Pelton Electronics (Ponka City, OK) signal source used to drive vibroseis trucks for oil exploration. Vibroseis uses a sweep frequency vibration source, instead of dynamite explosives. The unit used slope and offset D-to-A conversion to make the low frequency audio sweep to drive the electro-hydraulic vibrator which shakes the ground. Seemed to work pretty well, at least from about 6Hz to 150Hz.
It should be noted that any vector format also has limited precision even if you were to use double precision floating point numbers everywhere. And it only really improves the sampling noise not the quantization noise (given that any digital number even 64-bit floats have limited precision). Though the sampling noise is also replaced by the noise made given the limited number of vectors you are willing to store. And this is probably easier to solve with just regular variable rate sampling. Since the x coordinate of your vector is already the sample time, the y is either zero for all of them or the same as your traditional PCM sample would have been. And the angle would be the direction to the next sample. So you either need to use bezier curves instead of plain vectors, or just end up with fancy PCM. And even then you would then need to sample this vector in the player again before sending it to the DAC.
Vector DAC is an interesting concept. It would take something as big as Sony to pull off a new audio format, so not likely to happen. Adobe use case was to become a ubiquitous file format for printing and archival. Compression and security seems to be what keeps it going. Not sure if that would drive a new audio format today as storage and bandwidth are becoming cheap. Sadly, there is no market for quality in audio recording as MP3 demonstrated.
I think we should face the fact that vinyl records are not going anywhere anytime soon. We should also recognize that there are several ways of presenting digital audio and I use both digital and analog in my listening experiences. I'm tired of technology moving on in every way shape and form because it just adds to the confusion and direction as where we're headed. It's not time yet for this to happen as you can see vinyl sales are doing very well in the marketplace.
as much as I understand vector graphics - it approximates complex shapes using a combination of basic ones. For signals - one could use straight lines in the first approximation - this way we'd quantize the slope rather than momentary absolute value. Another way, suitable for the sound could be a sine wave - where we'd record the amplitude and phase - MP3 and other compressed formats work that way. General curve approximations often use polynomials. Finally, one could record notes and instrument arrangements, like musicians did for centuries ;-)
Great idea! But it would only take [xyt] but it would require precise clock and something similar to the blanced signal tranfer - common mode rejection
More interesting than the explanation per se (that was very very good for the layman) is the brilliant idea of this “vector dac”. I’m eager to see the full research and results in a future PS Audio product as a closer to perfect solution for digital music audiophile ears because this is the philosophers' stone in today’s music. Greetings from Brazil.
Technically a multi-bit DSD would be vector based. A sample would determine 1: Change - Yes/No, 2: if Yes - Up? Yes/No. Next sample would determine if that pattern continues. If the pattern continues over "n" number of bits data compression could attach that tag. A byte might say "Yes UP for 7 samples". That would be a vector.
Home HiFi VCRs (VHS and Beta), and also LaserDisc, have high quality analog audio recorded as frequency modulation . While none of those were fully as good as CD, further development might have made worthwhile improvements. Of course, when LaserDiscs got CD-style digital audio the difference was immediately obvious.
Quantum computing will change this. Paul is correct only because [historically] digital only uses 1s and 0s. Quantum can use various methods, but electron spin state is probably the most recognized. This means that more information is contained within each bit essentially. Eventually, the new technology will be indiscernible to analog. Who knows if we’ll see it in our lifetimes though.
If one cannot come up with a solid theory or, alternatively, a demonstration of a workable such thing in practice, then I think it will remain at the stage of fantasy/sci-fi etc. It may sound like a platitude, but I am not even sure this would be possible. The universe itself is quantized. Matter is quantized (in the form of atoms and their subcomponents). Energy is quantized as individual photons. Trying to represent a signal of any form as bits of something other than discrete quantities sounds fictitious. This video reminded me of the concept of vector displays, which exist but are way more limited in capabilities than the standard raster displays. Actually putting into practice the concept of a vector signal is way more limited than the results you may be looking for. It's nice to dream of workarounds to the intensive requirements of accurately representing real-world signals, be they sound or images, but ultimately I think we have to find practical solutions for managing much higher-resolution signals of the classic-type (i.e. raster in both images and sound) than we have been so far.
Paul left off optical storage. Even the films of old - film as in photographic film - recorded sound. Always thought it was a miscalculation that the big electronics companies did not develop music on film, analog recording via photographic film.
First thing that comes to mind is What you describe as "vector encoding" is very similar to what MIDI does. In a real life recording you would have infinite vectors.
You know Paul I've heard of a saying once that "If you Build it They will come". 👍 I wonder... 🤔 Ya never know 🤷♂️ Great Ideas should have a chance ✌️
Vector format is an interesting idea. With AI becoming more sophisticated, the conversion from analogue source to vector data could be made through a trained AI that matches the original source as close as possible. So theoretically the reproduction would be the best you can realistically get.
Saving audio as vectors with properties of shape similar to vector based graphics is sort of a data compression concept. For example, an audio saw tooth wave would need very few data. But I’m not convinced such approach would make sense for audio waves when we want the finer details. Audio is not zoomed like graphics is and for example, if that saw tooth is not a perfect saw tooth, perhaps it’s modulated with another wave, the concept won’t work well. Paul is not stupid so I’m curious what he envision will be solved with audio represented as vectors.
The results he's looking for are probably this: Being able to record, store and reproduce real-world audio signals at much higher precision and detail but without a correspondingly high increase in the resources consumed in the process (bits to store the signal, processing power, transfer speeds etc.). I think it's unnecessary really, because I think the requirements to produce fantastic sound digitally are not that high compared to what is achievable today (think Octave Records). They are significantly higher than with CD-PCM, but not impossibly high. I am an avid albeit low-budget searcher of super hi-quality sound who is enamored with DSD and its sound, and listening to DSD-128 and 256 stuff already I think we're not hopelessly far from reaching sound perfection in the digital domain. I think the industry should still explore one day higher resolution levels, but we're already at a point where you don't need to invest bank-breaking money in a decent hi-fi system to get a level of sound quality in the personal home that most people may not think is possible. I mean live, concert/opera levels of quality as if the listener would be hearing the real thing, all of it from the comfort of one's home. That is already almost if not fully achievable with technology we have today.
I need to send my clients to this video, after explaining for the 100th time that I can’t use their tiny PNG logo for a billboard. Send me an EPS instead! 😂
I think the questioner *might* have been referring to physical formats, like LP, cassette, reel-to-reel, etc., not an "information format". Didn't someone have an analog laser that could read LPs? An analog format with something not rotating but still read from a laser such that it is never degraded by playing might be interesting.
@@PlatypusPerspective I wonder if modern manufacturing techniques and the march of technology could make such a laser analog turntable viable these days?
Paul, you really lost me with this one. What are analog tape and LP if they are not "analog formats"? An analog format is any method of storing or transmitting audio where at no time is the signal converted into a stream of digits. I thought this was pretty basic stuff.
Come on now, Paul...if you had the staff, and if you had the money...ok...anyway...this vector dac sounds like a legacy piece of equipment...something that would revolutionize the industry...food for thought. I think your mail was asking more about an analog source...not a format as such. I remember recording back in the 80/s when DAT was the newest greatest thing. Using vhs tape. Frankly, it was fantastic. Wonder why it disappeared so quickly. I think that was the first product to combine digital and analog in one product. Something like that....
Digital PCM recording (even lossy) disappeared mainly because music industry feared perfect digital copies would destroy their golden mine. DATs were lossless, as opposed to mini Discs and DCC that were lossy.
Paul, what you just described is (to me) the holy grial of sound. You should at least pursue the theory of it in a more detailed working fashion, and take it to a point that the only thing to do next would be to build such a DAC. That seems to leave DSD, PCM and even "analago recording" way behind. I wish you the best on this project.
PCM theoretically has 100% precision, it is mathematically proven. All remaining issues are in the engineering not the method and current DAC do a quite good job much better then any "analogue recording" that we have.
Interesting. You are arguing from the digital side what I have been arguing for decades in analog. You are comparing using the standard instantaneous amplitude approach of PCM to a delta time change approach. Just like how in analog I argue we need to switch away from the old static sinewave/ frequency measurements to time based. Yes I know about Fourier Transforms. But those require accumulation of information over time to determine frequencies. They do not measure the instantaneous changes that is the musical signal.
Your Vector DAC dream sounds like a good project for DARPA , with full government funding . They are bound to find some military use for it . Throw 6 Billion Dollars at the project .
Call me radical, but I think in the 21st century the new analogue format should be real bands, playing real music, singing real songs( and absolutely no auto tune).
You are always going to be hit by the limitations of ANY encoding/decoding process and the limitations of the storage media itself. Then there's the hilariously poor human audio perception and neural system, I mean come on, it's made out of meat .
That you think this is something that would even be needed indicates that you do not actually understand PCM audio recording and reproduction. But, alas, this is not the first thing you’ve said that indicates a misunderstanding of digital audio. You possess a large amount of great audio information, but you also think some bogus things are true, unfortunately. As I’ve told you before, you need to watch and rematch the Digital Show & Tell video until you understand it all. It isn’t that complicated. You just refuse to do so and adjust your beliefs accordingly. Which is too bad because it means you also continue to spread misinformation about digital audio.
I wonder if what Tim is asking is whether there's likely to ever be a new _analog storage_ format? We went through wax cylinder, various discs, shellac, vinyl, wire and tape magnetic, optical on film...
Nah, it's digital online streaming all the way nowadays.
Yes, that’s exactly what Tim was asking. It is an interesting question. Too bad Paul misunderstood it and didn’t answer the real question.
@@tamstutz921 A number of other people scattered through the comments have also had the same thought.
As a graphic designer I have to say that's quite brilliant; the Bezier function for perfect curve reproduction.
As opposed to PCM's polygonal.
@@glenncurry3041 I wonder where/ what the differences would be, since it all is digital 🤔
@@Projacked1 Bezier and NURBs based 3d can't do sharp corners well. Tight bends at best. While polygonal has single point intersections. But can't do curves well. Uses lots of separate points to create a curve. Same with audio. DSD/ vector would follow a curve. Where PCM is hard sample points with each a stand alone sample.
@@glenncurry3041 hmm....you're right. Not so brilliant , maybe. You can change the functionality of the intersection points tough. I can make sharp points with addtional curving from 2 points, if you have the option like graphic/ 3D design. But the processing load will be bigger.
@@Projacked1 I sold GCI back in the '80's and '90's with Unix workstations like SGI and Symbolics. Some 3D software was poly, some NURBs. You just can't make a hard intersection with NURBs nor Bezier. But you can't make a true circle nor radiused corner with poly's.
The same thing is true with digital audio. PCM is like Poly. Straight line from point A to point B. Straight line up from zero to the voltage of that sample. Then next sample point. And some outside system has to figure out how to transition between them/ how to make it round. The more samples the closer, but never round.
Wow amazing to hear this! Thank you Paul, it's a new way of seeing. I hope and pray this would be achieved within your lifetime! Bless you!
That’s an incredible idea. I immediately understand the benefit of mapping digital music in that way.
I've seen this used, but not in high end audio. I first saw it used in 1983 in a Pelton Electronics (Ponka City, OK) signal source used to drive vibroseis trucks for oil exploration. Vibroseis uses a sweep frequency vibration source, instead of dynamite explosives. The unit used slope and offset D-to-A conversion to make the low frequency audio sweep to drive the electro-hydraulic vibrator which shakes the ground. Seemed to work pretty well, at least from about 6Hz to 150Hz.
It should be noted that any vector format also has limited precision even if you were to use double precision floating point numbers everywhere. And it only really improves the sampling noise not the quantization noise (given that any digital number even 64-bit floats have limited precision). Though the sampling noise is also replaced by the noise made given the limited number of vectors you are willing to store.
And this is probably easier to solve with just regular variable rate sampling. Since the x coordinate of your vector is already the sample time, the y is either zero for all of them or the same as your traditional PCM sample would have been. And the angle would be the direction to the next sample. So you either need to use bezier curves instead of plain vectors, or just end up with fancy PCM. And even then you would then need to sample this vector in the player again before sending it to the DAC.
Can you repeat that? 😂
A vector-based DAC technology seems like a brilliant idea, Paul. Possibly the next big breakthrough in DAC s!
I'm going to say DSD is how you would practically encode the vector. The change in pulse density represented the angle.
Fascinating! I'm inclined to believe that, if ever realized, this DAC technology would come the closest to reproducing the original analogue signal
Vector DAC is an interesting concept. It would take something as big as Sony to pull off a new audio format, so not likely to happen. Adobe use case was to become a ubiquitous file format for printing and archival. Compression and security seems to be what keeps it going. Not sure if that would drive a new audio format today as storage and bandwidth are becoming cheap. Sadly, there is no market for quality in audio recording as MP3 demonstrated.
mp3 quality is fine, the crappy mp3's are due to crappy source or bad encoding options.
I think we should face the fact that vinyl records are not going anywhere anytime soon. We should also recognize that there are several ways of presenting digital audio and I use both digital and analog in my listening experiences. I'm tired of technology moving on in every way shape and form because it just adds to the confusion and direction as where we're headed. It's not time yet for this to happen as you can see vinyl sales are doing very well in the marketplace.
Don’t Audiophiles have enough to complain about without creating something new to complain about.
as much as I understand vector graphics - it approximates complex shapes using a combination of basic ones. For signals - one could use straight lines in the first approximation - this way we'd quantize the slope rather than momentary absolute value. Another way, suitable for the sound could be a sine wave - where we'd record the amplitude and phase - MP3 and other compressed formats work that way. General curve approximations often use polynomials. Finally, one could record notes and instrument arrangements, like musicians did for centuries ;-)
I was having flashbacks to college engineering vector calculus😀 thanks Paul for sharing your dreams of a possible future product.
Paul gives his million dollar idea to his fans. Now who’s going to take the idea and run with it?
Great idea! But it would only take [xyt] but it would require precise clock and something similar to the blanced signal tranfer - common mode rejection
The Onkyo DX-C390 uses "Vector Quantizer Audio DAC Technology". Wonder if it's the same
More interesting than the explanation per se (that was very very good for the layman) is the brilliant idea of this “vector dac”. I’m eager to see the full research and results in a future PS Audio product as a closer to perfect solution for digital music audiophile ears because this is the philosophers' stone in today’s music.
Greetings from Brazil.
Technically a multi-bit DSD would be vector based. A sample would determine 1: Change - Yes/No, 2: if Yes - Up? Yes/No. Next sample would determine if that pattern continues. If the pattern continues over "n" number of bits data compression could attach that tag. A byte might say "Yes UP for 7 samples". That would be a vector.
So would a "no change" or a "no change in vector velocity" be opportunities for lossless compression?
Home HiFi VCRs (VHS and Beta), and also LaserDisc, have high quality analog audio recorded as frequency modulation .
While none of those were fully as good as CD, further development might have made worthwhile improvements.
Of course, when LaserDiscs got CD-style digital audio the difference was immediately obvious.
How about a QAC? Qubit to Analog Converter?
Quantum computing will change this. Paul is correct only because [historically] digital only uses 1s and 0s. Quantum can use various methods, but electron spin state is probably the most recognized. This means that more information is contained within each bit essentially. Eventually, the new technology will be indiscernible to analog. Who knows if we’ll see it in our lifetimes though.
If one cannot come up with a solid theory or, alternatively, a demonstration of a workable such thing in practice, then I think it will remain at the stage of fantasy/sci-fi etc.
It may sound like a platitude, but I am not even sure this would be possible.
The universe itself is quantized. Matter is quantized (in the form of atoms and their subcomponents). Energy is quantized as individual photons.
Trying to represent a signal of any form as bits of something other than discrete quantities sounds fictitious.
This video reminded me of the concept of vector displays, which exist but are way more limited in capabilities than the standard raster displays.
Actually putting into practice the concept of a vector signal is way more limited than the results you may be looking for.
It's nice to dream of workarounds to the intensive requirements of accurately representing real-world signals, be they sound or images, but ultimately I think we have to find practical solutions for managing much higher-resolution signals of the classic-type (i.e. raster in both images and sound) than we have been so far.
Paul left off optical storage. Even the films of old - film as in photographic film - recorded sound. Always thought it was a miscalculation that the big electronics companies did not develop music on film, analog recording via photographic film.
Vector is for graphics not for photography. I dont think so you can describe standard photography by vectors.
First thing that comes to mind is What you describe as "vector encoding" is very similar to what MIDI does.
In a real life recording you would have infinite vectors.
I like the idea. I shall give it some thought...
You know Paul I've heard of a saying once that "If you Build it They will come". 👍 I wonder... 🤔 Ya never know 🤷♂️ Great Ideas should have a chance ✌️
Vector format is an interesting idea. With AI becoming more sophisticated, the conversion from analogue source to vector data could be made through a trained AI that matches the original source as close as possible. So theoretically the reproduction would be the best you can realistically get.
Wow! I have also thought of such a concept. Now I know either (A) I am not crazy, or (B) we are both crazy. Let's assume "A"!
NTSC, PAL, etc. are analog "formats".
Saving audio as vectors with properties of shape similar to vector based graphics is sort of a data compression concept. For example, an audio saw tooth wave would need very few data. But I’m not convinced such approach would make sense for audio waves when we want the finer details. Audio is not zoomed like graphics is and for example, if that saw tooth is not a perfect saw tooth, perhaps it’s modulated with another wave, the concept won’t work well. Paul is not stupid so I’m curious what he envision will be solved with audio represented as vectors.
The results he's looking for are probably this: Being able to record, store and reproduce real-world audio signals at much higher precision and detail but without a correspondingly high increase in the resources consumed in the process (bits to store the signal, processing power, transfer speeds etc.).
I think it's unnecessary really, because I think the requirements to produce fantastic sound digitally are not that high compared to what is achievable today (think Octave Records). They are significantly higher than with CD-PCM, but not impossibly high.
I am an avid albeit low-budget searcher of super hi-quality sound who is enamored with DSD and its sound, and listening to DSD-128 and 256 stuff already I think we're not hopelessly far from reaching sound perfection in the digital domain. I think the industry should still explore one day higher resolution levels, but we're already at a point where you don't need to invest bank-breaking money in a decent hi-fi system to get a level of sound quality in the personal home that most people may not think is possible. I mean live, concert/opera levels of quality as if the listener would be hearing the real thing, all of it from the comfort of one's home. That is already almost if not fully achievable with technology we have today.
Vector DAC, cool idea :)
VECTOR 1, OVER AND OUT!!
I need to send my clients to this video, after explaining for the 100th time that I can’t use their tiny PNG logo for a billboard. Send me an EPS instead! 😂
On a side there is a patent for a floating point dac...
The "vector" version for music is... musical notation! Been around for centuries.
Or MIDI.
Paul is on to something!!!
Interesting idea. FFT is another.
so..neither on/ff, but somewhere inbetween?
I think the questioner *might* have been referring to physical formats, like LP, cassette, reel-to-reel, etc., not an "information format". Didn't someone have an analog laser that could read LPs? An analog format with something not rotating but still read from a laser such that it is never degraded by playing might be interesting.
I think that is what the question is actually asking. The ELP Laser Turntable was the hugely expensive unit that could play LPs with a laser head.
@@PlatypusPerspective I wonder if modern manufacturing techniques and the march of technology could make such a laser analog turntable viable these days?
Tda1541😊old but still sounding great.
Paul, you really lost me with this one.
What are analog tape and LP if they are not "analog formats"?
An analog format is any method of storing or transmitting audio where at no time is the signal converted into a stream of digits.
I thought this was pretty basic stuff.
Come on now, Paul...if you had the staff, and if you had the money...ok...anyway...this vector dac sounds like a legacy piece of equipment...something that would revolutionize the industry...food for thought. I think your mail was asking more about an analog source...not a format as such. I remember recording back in the 80/s when DAT was the newest greatest thing. Using vhs tape. Frankly, it was fantastic. Wonder why it disappeared so quickly. I think that was the first product to combine digital and analog in one product. Something like that....
Digital PCM recording (even lossy) disappeared mainly because music industry feared perfect digital copies would destroy their golden mine. DATs were lossless, as opposed to mini Discs and DCC that were lossy.
Whatever Paul said sounds good - but does it have to cost so much !
Paul, what you just described is (to me) the holy grial of sound. You should at least pursue the theory of it in a more detailed working fashion, and take it to a point that the only thing to do next would be to build such a DAC.
That seems to leave DSD, PCM and even "analago recording" way behind. I wish you the best on this project.
PCM theoretically has 100% precision, it is mathematically proven. All remaining issues are in the engineering not the method and current DAC do a quite good job much better then any "analogue recording" that we have.
Didn't Paul used to have a white-board or is that too old tech now? LOL nevertheless I love his videos.
A whiteboard is analog technology - perhaps it's been relegated to a niche application?
Interesting. You are arguing from the digital side what I have been arguing for decades in analog. You are comparing using the standard instantaneous amplitude approach of PCM to a delta time change approach. Just like how in analog I argue we need to switch away from the old static sinewave/ frequency measurements to time based. Yes I know about Fourier Transforms. But those require accumulation of information over time to determine frequencies. They do not measure the instantaneous changes that is the musical signal.
Your Vector DAC dream sounds like a good project for DARPA , with full government funding . They are bound to find some military use for it . Throw 6 Billion Dollars at the project .
James Clerk Maxwell 😉
I’m high but I know there is something here
perhaps you are talking about Tensor, not Vector!
That's the 4 ch version.
NURBS DAC would be even better 😉
better go and take a patient out on vector dacs now...
Patiently waiting to get a patent?
Call me radical, but I think in the 21st century the new analogue format should be real bands, playing real music, singing real songs( and absolutely no auto tune).
You are always going to be hit by the limitations of ANY encoding/decoding process and the limitations of the storage media itself. Then there's the hilariously poor human audio perception and neural system, I mean come on, it's made out of meat .
That you think this is something that would even be needed indicates that you do not actually understand PCM audio recording and reproduction. But, alas, this is not the first thing you’ve said that indicates a misunderstanding of digital audio. You possess a large amount of great audio information, but you also think some bogus things are true, unfortunately. As I’ve told you before, you need to watch and rematch the Digital Show & Tell video until you understand it all. It isn’t that complicated. You just refuse to do so and adjust your beliefs accordingly. Which is too bad because it means you also continue to spread misinformation about digital audio.
en.wikipedia.org/wiki/SVG