LinuxMusicians

Hi guys!

Recently I found a very interesting YouTube channel: Adam Neely's channel:

https://www.youtube.com/channel/UCnkp4x ... 7sSM3xdUiQ

It is really interesting, full of good information about bass, composition and harmony. Today I saw this video:

https://www.youtube.com/watch?v=YfyNhfpLLZ4

I think that the most interesting part is here: https://youtu.be/YfyNhfpLLZ4?t=344

Adam reports that, according to various sources, the technology at the heart of music software is actually pretty old, significantly older than the bleeding edge technology being implemented in many other consumer devices and software in these very days.

I must confess that this impression really aligns with mine. Think about reverberation algorithms. Most of them are really decades old, but branded as bleeding edge in new plugin releases.

I believe that the reason behind this is a core of superstition that is engraved in the audio consumers brains. There has been a time period in which digital audio was inferior to analog, early in the computer music history. These times are long gone, and honestly I believe the opposite nowadays, but the attitude of many consumers is still overly positively biased towards analog style.

As a result, computers are often used as a mean to emulate analog devices, and plugins GUIs are almost always skeumorphic. Although I have nothing against simulation (which actually can involve a serious amount of research and bleeding edge technologies) and skeumorphic design (although knobs on computer GUIs are beyond awkward), it seems very weird to me that in 2018 we are still to unleash the true potential of computers by creating tools of musical expression that are inherent to the computer, without necessarily needing to simulate something else.

Maybe an exception to this is in our world though. We have excellent analog modelling software (Guitarix, for example (*)) but also fairly experimental concepts (like fastbreeder: http://www.pawfal.org/Software/fastbreeder/). I think that the real bleeding edge is being explored by people into algorithmic composition, making use (for example) of Faust, Csound and PureData, for which I think Linux as the best possible environment in my opinion.

As for finding new means of expression and human/computer interaction I always been fascinated by Onyx Ashanti work, which at some point made use of a Gentoo optimized laptop (if I recall correctly):

http://onyx-ashanti.com/

Adam mentioned Machine Learning and that brought me back to this thread:

viewtopic.php?f=48&t=18078

What if instead of trying to move knobs as on 70s radios we were trying to have synths and effects trying to learn our body language? Or be able to understand spoken instructions? What about trying to visualize sound generation and propagation models in a meaningful way and allow the user to manipulate them? Or what about a Machine Learning algorithm that silently watches you while you mix and master songs, so to learn how you do it, so to pre-mix and pre-master songs for you, and help your professional work by doing it on bulks of projects, leaving you to just finalize the result? I am starting being more and more intrigued by these concepts.

What do you think? Do you feel like modern day software is not as modern day as it should be?

(*) Guitarix may seem of the "analog simulator goodies" category, and it is for sure, but to me it feels more sci-fi tech due to how amazing and brilliant the process of modelling electronics works within its framework. I really think of Guitarix as a true gem.

Hmm. Is this a can of worms i see before me ?
The thing is there are a lot of contradictory factors involved.

First off, nobody who invented a thing was ever able to make the best use of it - It fell to the kids growing up with it - compare the way kids use a smartphone to an adult. It takes time to learn, along with a lack of preconceived ideas.

Also there's flexibility. Compare a hammer with a nail gun (a replacement in some circumstances). The nail gun is faster and more accurate for driving nails of a specific type into designated materials, but a hammer can knock anything into anything else (or even out). The hammer hasn't changed significantly in centuries, because it doesn't need to.

Also there are attention issues. Watch a professional mixing engineer 'playing' a bunch of faders with both hands - that's 10 input sources they are dynamically managing. However it isn't necessary to maintain contact all the time. You don't have to keep still for the faders to stay in place, but how do you manage such stability with something that is directly in contact with your body?

Finally, are you trying to create something that is impressive, or something that people will enjoy? I've seen and heard a lot of art forms where I thought "Wow! That's clever", but I wouldn't want them in my living room! This also links up to familiarity. People like what they know, so if you are creating for other people (not just yourself) you need to give them what they know, or at least lead them very gently into new ideas - this again takes time.

All-in-all, I think change here has actually been remarkably quick, but none of us are likely to see the best of it :/

P.S.
As far as I know, knobs were never considered to be the best way of controlling something - they were the easiest to make reliably, and took up the least amount of space - that's still true in the IT world (at least from the space point of view).

All of you say makes sense. Here a few points from the top of my mind.

[quote="folderol"]
Also there's flexibility. Compare a hammer with a nail gun (a replacement in some circumstances). The nail gun is faster and more accurate for driving nails of a specific type into designated materials, but a hammer can knock anything into anything else (or even out). The hammer hasn't changed significantly in centuries, because it doesn't need to.
[/quote]

Yes, but has someone tried hard enough to invent the equivalent of a nail gun in computer audio world? (legit question). If so, what do you think the nail gun is (are)?

[quote="folderol"]
Finally, are you trying to create something that is impressive, or something that people will enjoy? I've seen and heard a lot of art forms where I thought "Wow! That's clever", but I wouldn't want them in my living room! This also links up to familiarity. People like what they know, so if you are creating for other people (not just yourself) you need to give them what they know, or at least lead them very gently into new ideas - this again takes time.
[/quote]

What if enjoyment comes after an adaptation time to what it was impressive at the beginning? Or, on another line, what would happen if all suddenly you show a modern day piano to a middle ages psalterium player? Would he recognize its potentialities and capabilities, or it would seem an overly complicated impressive machine to him?

Probably it also worth to consider how music instruments evolved in the past. Slowly and steady during centuries. So, baby step after baby step, the psalterium (and similar instruments) evolved into the piano. The computer is radically different from everything existed before, so it would allow a jump discontinuity in expression. This, would be hard to adapt for artists, so culturally used to different things. However, evolution in visual arts seems faster (I am thinking about 3D and VR), is there something more holding audio back to older paradigms? (another legit question).

CrocoDuck wrote:All of you say makes sense. Here a few points from the top of my mind.

folderol wrote: Also there's flexibility. Compare a hammer with a nail gun (a replacement in some circumstances). The nail gun is faster and more accurate for driving nails of a specific type into designated materials, but a hammer can knock anything into anything else (or even out). The hammer hasn't changed significantly in centuries, because it doesn't need to.

Yes, but has someone tried hard enough to invent the equivalent of a nail gun in computer audio world? (legit question). If so, what do you think the nail gun is (are)?

Good question... Oh you wanted an answer? Sorry I don't have one

folderol wrote: Finally, are you trying to create something that is impressive, or something that people will enjoy? I've seen and heard a lot of art forms where I thought "Wow! That's clever", but I wouldn't want them in my living room! This also links up to familiarity. People like what they know, so if you are creating for other people (not just yourself) you need to give them what they know, or at least lead them very gently into new ideas - this again takes time.

What if enjoyment comes after an adaptation time to what it was impressive at the beginning? Or, on another line, what would happen if all suddenly you show a modern day piano to a middle ages psalterium player? Would he recognize its potentialities and capabilities, or it would seem an overly complicated impressive machine to him?

If it was guildmaster Jordan of the Rudess he would have probably jumped on it with glee
Anyone else would have run to the priest demanding an exorcism
This is probably not so different from people today.

Probably it also worth to consider how music instruments evolved in the past. Slowly and steady during centuries. So, baby step after baby step, the psalterium (and similar instruments) evolved into the piano. The computer is radically different from everything existed before, so it would allow a jump discontinuity in expression. This, would be hard to adapt for artists, so culturally used to different things. However, evolution in visual arts seems faster (I am thinking about 3D and VR), is there something more holding audio back to older paradigms? (another legit question).

I've seen a number of interfaces that look quite interesting. All of them very different with no apparent common ground. Learning to use any of these is an investment of time and effort, neither of which I have enough of to spend on something that might just be an overnight fad. I suspect a lot of musicians would take the same view.

For a comparison look at the anguished complaints when a certain other OS keeps changing its GUI.

folderol wrote:Good question... Oh you wanted an answer? Sorry I don't have one

I am sure there actually are few examples, but I cannot come with one either.

folderol wrote: Learning to use any of these is an investment of time and effort, neither of which I have enough of to spend on something that might just be an overnight fad. I suspect a lot of musicians would take the same view.

That's probably a pretty good point. Especially for professional musicians time is a huge factor, and there isn't materially the time to switch to a completely different paradigm, with a long slow learning curve.

42low wrote:To me it's rather clear.
We are no robots. We are analoque. So to me it's rather clear that analogue is the goal to reach.

As always, you touch very good points. Here some things we can expand upon.

Yes, we are no robots. But we are no vacuum tubes or germanium transistors either. I think that analog and digital are two abstract models that come from technology which are completely arbitrary when it comes to relate them to human perception.

That is, human perception is very complex, and it actually discards most of the physical signal information impinging on our senses. Just think about auditory masking. Auditory masking removes from a physical signal much more than what sampling and digitization do.

Also it should be noted that while analog devices emit signals in a continuum, the ear does not appear to perceive it like that. The ear has both time and frequency finite resolution. At the core of the ear there might be "sampling" processes. This is what evidence from physiology suggests. If ears were pure continuum time systems, we would have infinite resolution.

I think that the word "digital" is almost never used properly. Digital does not exactly mean "sampled". Sampled means that a time signal is, indeed, sampled at regular (or not) steps. Well, you can do the same with analog devices. All analog flanger and chorus and delay pedals that are not based on tapes use a circuit known as "bucket brigade" which creates delay lines by sampling the signal in analog domain. That is, they are analog effects that do sample the signal, but they do not use any digital technology.

Digital is a signal that other than being sampled is also expressed as a sequence of bits, for computers to do maths with them.

I see where you come from: If I reproduce the signal exactly as it is at my ear, no changes, then I have the maximum fidelity. So, digital cannot offer the maximum fidelity, as sampling and quantization can only reduce information.

That in theory is not properly correct, as band limited signal can be converted back to their analog "fathers" without any mathematical error, but is true in practice for many reasons. However, there are two main points:

1) Are the errors audible? And, more importantly:
2) Analog devices do also alter the signal very drastically, and might not yield any more fidelity than good digital equipment.

Point 2 is what makes me think that digital VS analog in terms of fidelity is an arbitrary contraposition when it comes to compare properly designed devices. Every analog device has finite bandwidth, noise rejection characteristics and linearity characteristics that reduce fidelity, which are actually not present in many digital systems. Many praise vinyl disks. Those might sound indeed good to many, but the process by which they are printed overly compresses the sound, not to mention the background noise. To me, used to modern day audio, they sound awful. They are a good example of analog device that has very little to do with fidelity.

As a final note from the above, culture has a very significant impact. I remember a psychoacoustcs experiment being discussed while I was at uni. I cannot find the sources right now. However, it was about perception quality of analog VS digital formats.

There was a very significant difference between subjective scores by age groups. People older than 35 usually consistently assigned higher scores to analog formats in double blind tests.

People younger than that the opposite. Remarkably, among the youngest, MP3 scored the highest quality in double blind tests.

What you are most used, or what you developed your musical listening skills and taste on, ends up dictating your judgment of it. Unconsciously. An interesting result, which I will link if I can find it somewhere (hopefully I remember the facts right).

CrocoDuck wrote: What you are most used, or what you developed your musical listening skills and taste on, ends up dictating your judgment of it. Unconsciously. An interesting result, which I will link if I can find it somewhere (hopefully I remember the facts right).

This touches on the point of how easily we can be misled.

In the 1970s I got a record of a song I really liked, and played it many, many times. Eventually I lost it.

I couple of years ago, on a whim, I did a search for it on Youtube. However the first copy I found sounded absolutely wrong, and I commented to a friend that it wasn't bad but not as good as the original. My friend laughed and told me the 'wrong' one was being sung by the writer and original artist.

To this day, the original still sounds wrong to me.

42low wrote:That's what i meanth. I think that we experience music outside our hearing limits. We feel that unhearable deep base in hour stomach and that unhearable high with hour neck hair.

Yes, evidence that we perceive music also through other senses is started growing. Which makes sense, all what we experience is multi-sensorial (if that's an actual word).

42low wrote:Digital is simply limited within the hearing range so lacks those 'outside' low and high. Therefore we can't experience those with digital.

This isn't quite correct. The process of sampling only imposes a high frequency limit (half the sample rate). The low frequency limit usually comes from the analog pre amplifiers, due to the fact that they use DC blocking capacitors, as most analog circuits do. This are just needed to ensure the devices operate properly.

42low wrote:A digital curve comes near an analogue signal, and within the analogue amps that curve get's straigtened, but at the end it is a calculated/guessed curve which never ever will be the same as the original. You can't get back (exactly) what's gone.

I think it is better to be clearer on this. A sampled signal can really exactly be converted to the analog signal from which it was sampled. Thinking of a sampled signal as an approximation is misleading, it is more like expressing a signal in another domain (discrete time domain), which is linked to continuum time domain by well defined transformations (unless aliasing happens). Now, in digital world this exact DA conversion process (sinc interpolation) is not typically done as there are more efficient (even if less exact) techniques for it that end up being more convenient. However, no guesses here: the sinc interpolation is an exact reconstruction: https://en.wikipedia.org/wiki/Whittaker ... on_formula.

Let's remember that sampled and digital are not exactly the same thing. Sampling per se actually does not yield a massive loss of information, if done properly. And, as I mentioned by referring to the bucked brigade circuits, it can be done in analog world too. The biggest limitation of digital, which is sampled + conversion to bits, is due to dynamic range. The number of bits you use to encode a sampled value is where actually most of the information gets lost, and while it is hard to tell the difference between 44.1 kHz and 48 kHz audio, it is easy to tell the difference between 16 bit and 24 bit.

As for comparing digital equipment to properly designed analog equipment, you can for sure measure differences. However, both technologies are actually band limited. If you input a square wave into an analog circuit you will observe Gibbs oscillation:

https://en.wikipedia.org/wiki/Gibbs_phenomenon

Which is due to finite bandwidth. Bandwidth of analog equipment is not infinite. It is bad idea to do: you would make your equipment susceptible to radio noise interference, that can be demodulated and become audible in your analog chain. Other than that, bandwidth is limited by dynamic behavior of components. It is much higher than computer audio bandwidth sure, but not much higher than that of dedicated DSP chips, that sample at MHz range, exceeding the audible range a lot, their bandwidth aligns with that of most analog equipment.

Other than that, discrete analog components work -great- for small signals. However, they tend to loose linearity for larger signals. If you want to make distortions, that's perfect. I think that nonlinear devices is where analog still shines over digital for real (tons of distortion, no aliasing). That and band limited signal generation. If you want fidelity, however, non-linearity is a problem. And this is the reason why high power high fidelity audio amplifiers are actually pretty damn complex. Also, linearity and bandwidth don't scale well for integrated circuits or smaller components sizes. All of these are not issues for digital chips. Finally, reasonably complex analog circuits allow usually only for low order linear filtering (but pretty crazy nonlinear behavior if that is what you are after). If you need complex filters, they become very complex soon, and with that all the side effects (unwanted non-linearity, background noise susceptibility, very high industrial product variation).

You are right mentioning that loudspeakers are analog devices too, but they are dynamic systems, and their dynamic dictates their bandwidth. When the frequency becomes too high the inertia of the diaphragm (if we talk about dynamic speakers) prevent them from working, and they stop radiating sound. Other physical phenomenon link the radiated low frequency energy to the size of the radiator, meaning that they cannot radiate very low frequency either (wavelength bigger than their size). I cannot came up with a typical frequency range now, but this is why speakers for reproducing very low frequency and very high frequency are completely different and purposely designed. The question is, how much useful is to have huge analog bandwidths with analog stuff while the bandwidth of the speakers is still limited?

I think that bandwidth isn't really the problem of digital audio.

Coming back to the original topic, I think that what comes from this is the following:

Digital and analog are 2 different technologies, both with weaknesses and strengths. That is, they have different capabilities. That is why I don't believe that digital has to exists to imitate analog, as it has many limitations that analog does not have, and it has many limitations when it comes to emulate analog behavior instead, especially nonlinear behavior (in a computationally efficient way). I think it is best to embrace what digital audio is, and try to came up with new means of exploit it, rather than trying to carry out with analog paradigms in a world that opens up to many different possibilities.

CrocoDuck wrote:What if instead of trying to move knobs as on 70s radios we were trying to have synths and effects trying to learn our body language?

We've some kind of that already, the Hot Hand from source audio:
https://www.sourceaudio.net/hot-hand.html

here a bit in action:
https://www.youtube.com/watch?v=Ph3ymLurfHI

CrocoDuck wrote:What do you think?

Sometimes computer based equipment solved problems you may have with analogue gear.
Here for example:
http://arre234.blogspot.de/2018/02/linu ... mp-on.html

My final comments before moving back on track Thank you for your input. I think it will give context to who is reading (if they feel like reading all of our text )

42low wrote:At first. If analogue amplifiers cut of that siginificant then it's shitty gear. I'm talking about good gear with good sound qualities.

How significant? To what figure you are relating when you say significant? As significant as consumer digital audio equipment? In the low frequency range the low frequency cutoff is well below 20 Hz, and usually the very same for both analog and digital equipment, as it is indeed dictated always by analog stages, also for digital equipment. Digital stuff has always an analog input stage to it, and that stages dictates the low frequency cutoff.

For analog, it is typical to never go beyond 1 MHz upper frequency limit, unless you plan to hear AM radio through your gear (if someone is still broadcasting AM nowadays). Yes, larger bandwidth of most computer interfaces. In par with many dedicated DSP chips, though. But that is slightly out of topic, as we are referring mostly to normal soundcards.

42low wrote:Although new techniques appear, the majority of digital sound for instance still is based on Nyquist and this cuts off at the borders of hearing (from the simple thought, you don't hear those so not needed cause to avoid extra dataspace).

Yes, that is almost correct, but the Nyquist theorem prescribes to band limit the signal to half the sample rate, not to the nominal audible range. For example:

96 kHz sample rate -> you can record signals with bandwidth from 0 Hz to 48 kHz.
8 kHz sample rate -> you can record signals with bandwidth from 0 Hz to 4 kHz.

Sampling imposes only a high frequency limit, not a low frequency one. As a final example, my new sound-card can operate at 192 kHz. It means I can record signals up to 96 kHz. That is not a small bandwidth, but yes, it is smaller than most analog gear.

42low wrote:Btw. If you approach it from that angle your topic immediately is answered as "right, the music Industry is 10 to 15 years behind".
It's already proven that 24bits/32float on somewhere around 80-90khz are the best rates to capture most sound for our ears, so the industry should (and could) adapt their techniques to somewere around those to get by.

I am referring to the software methods by which the sound is generated and modified, not the sampling variables. That part of the technology is OK.

42low wrote:There's another test that states it. If a song ends with a deep base there's a moment that you stop hearing that bass.
If you feel (or measure) the speakers cone there will be some movement slightly after the sound stops. And if you measure the wires (done that too) you will notice that's no expelee of the already gone sound, but there actual is a signal to be measured (oscilloscope). So that proves outside hearing ranges there still are unhearable active sounds (which are also produced by the amplifier).

Yes, the amplifier will push the signal through the coils, and the speaker will react. Usually, by being able to radiate much less pressure with respect its nominal design bandwidth. To radiate low frequency sound efficiently, you need a well designed speaker for that, like a subwoofer. Otherwise, the low frequency pressure might be so small that ambient noise at the same frequency might mask it.

tramp wrote:Sometimes computer based equipment solved problems you may have with analogue gear.
Here for example:
http://arre234.blogspot.de/2018/02/linu ... mp-on.html

That's quite an hack!

tramp wrote:We've some kind of that already, the Hot Hand from source audio:
https://www.sourceaudio.net/hot-hand.html

here a bit in action:
https://www.youtube.com/watch?v=Ph3ymLurfHI

Reminds of my project (which stayed on paper forever) about creating a sort of ebow, but mounted on a glove, and being able to induce in the pickup coils of a guitar any synthesized signal, generated by hand posture.

Two interesting workshops of machine learning applied to music and sound were announced at CCRMA:

https://ccrma.stanford.edu/workshops/deep-mir-I
https://ccrma.stanford.edu/workshops/de ... ing-mir-ii

Adam also commented further in another interesting video:

https://www.youtube.com/watch?v=B0c4UBWFW-w

timestamap: https://youtu.be/B0c4UBWFW-w?t=349

On a more grounded level, do you know how demanding is bandlimited synthesis? My 4 GHz CPU can't render a single low note in realtime using one thread. I have to use GPU.

Lyberta wrote:On a more grounded level, do you know how demanding is bandlimited synthesis? My 4 GHz CPU can't render a single low note in realtime using one thread. I have to use GPU.

Yep, I got a feeling of it trying to synthesize band limited chirps. Either you are lucky enough that you can get a simple time domain algorithm for your special waveform (unlikely) or you could use wavetables of pre-calculated waves (not a fan, but probably one of the most efficient ways) or you can get a crazy oversampler and cranck it up until your CPUs explode. Or use Fourier expansion and probably many more methods, which usually are pretty much the opposite of cheap.

I did not mean that imitating analog is easy or trivial. Actually it is pretty interesting, especially when you are trying to identify and model a physical system. Think about pianoteq, for example. That is extremely interesting and for sure really high tech and bleeding edge. The same goes for guitarix, as I said.

Many applications in DSP have been unlocked only in recent years really. The ideas behind many algorithms might actually be decades old, but only nowadays computers can deal with them.

I think I am mostly thinking about the user interaction. I think that most old style paradigms are really there. We usually stick to the idea of manipulating parameters of a system, which is for sure powerful, but perhaps a bit aged with respect new kind of interfaces popping up nowadays and I am wondering what kind of impact this would have on music creation.

CrocoDuck wrote:… according to various sources, the technology at the heart of music software is actually pretty old

it actually is.

DAWs are still using the only metaphor of a multitrack reel-to-reel tape recorder.

software synths are still focused on emulating ancient TR-808 / MS-20 / stuff like that.

nearly all new hardware synths are monotimbral and analog again (why?! why nearly all of the virtual analog polytimbral DSP-powered synths that were perfect for live use are discontinued now?! are they forbidden now?!)

etc etc etc.