What makes analog so analog?

[Robert Randolph]

Robert Randolph quoth

Anything derived from a clock can be predictable given the seed (the time in point where the clock is sampled) and the algorithm...

Ummm, are you sure? Just the fact that your machine can never boot at the same point in space/time twice in a row means that sampling the realtime clock for the first time on any given boot-up is non-predictable from boot to boot.

such routines are pseudorandom simply because they lack complete unpredictability....

For the reasons above, Im not sure about that. The random-number generation is pseudorandom in and of itself, but when provided with an unpredictably non-repeating seed, I would hazard that they do lack predictability. And if one resamples that seed (or another, say the current onscreen raster position of your display) at pseudorandom intervals I'd really like to know where the predictablity comes from...

though in use it's of course it's perceptibly and practically random. just semantics again really...

True.

It's not random because you can predict what the value will be if given the seed and algorithm. Since using a clock of anykind is a fairly stable and predictable seed, it's not really presumptious to assume you could predict the 'random number' if one really so desired. The problem is you will always know (or can know) all the variables as they are generated or assumed... which means theorectically you could take those numbers and predict the outcome before the generation completely occurs. You can go through many chains of "randomness" but as the author of it, you will always know the possibilities and the bounds. Which means there is a finite set of solutions from a single algorithm with known variables.

Practical? not at all... possible? certainly. And the possibility makes it not "true" randomness... as incredibly semantic and silly as it is.

[Kraku]

Funny enough It seems people are talking about feel than sound now...

wood, plastic, glass.. you cant hear any of those Ironically those happen to be materials most synths are made of... wood/metal = most "succesful" analogs, plastic = most digital synths and VSTi boxes (cpu or actual box). It's a weird coincidence if it is indeed one.

Sounds harsh, deep, pleasant, distorted, noisy... At least most people can relate to those worst slightly better than a physical reference to "feel"

OK I'll try better.

My VSTi MS-20 sounds something like when you take the sound of a dry wood been sawed. It's not the sawing itself, but something in the sound's character the material makes when being sawed which tells you that "it's wood, not metal, not plastic, not bread". Don't get me wrong, I'm not saying that anyone would mistake the MS-20 for the sound of a wood being sawed

In other words, it's very dry, a bit crispy and... wood like

[Joxer the Mighty]

lets just embrace them both...

That's probably the best thing to come out of this thread so far.

[valley]

For the reasons above, Im not sure about that. The random-number generation is pseudorandom in and of itself, but when provided with an unpredictably non-repeating seed, I would hazard that they do lack predictability. And if one resamples that seed (or another, say the current onscreen raster position of your display) at pseudorandom intervals I'd really like to know where the predictablity comes from...

The problem is that to be truly random there needs to be no repeating patterns. Even given a random seed from the system clock, and modulating your random seed from the raster line, left generating random numbers for long enough your system will begin to show discernible patterns.

If the system was entirely random then your MI value should be truly apporaching zero. I'm not sure that would be the case using the method you describe.

If there are any cryptography experts here, I'm sure they'll have something to say on this matter.

[meister eder]

[]

[peejunk]

b.t.w. when some kid tries to tell the bloke who wrote FL Studio that "you seem to be quite a noob to what is considered pretty basic DSP knowledge" you can safely ignore anything else he has to say.

I'd like you to expand on this a bit. This "kid" does not have his DSP books here currently so I cannot throw any quotes at you, but what I said about aliasing & D/A is what I have learned from those books (one of them being Roman Kuc's 'Introduction to DSP', and the other is by Miodrag Popvic, then there are other resources) and lectures in Signals & Systems.

So, while my sentence seems arrogant, it's merely me being surprised (no insult aimed at gol, no I don't think he's an idiot or whatever). So, please... enlighten me. I'm wrong (on the subject), how?

[Rabid]

If you want to program a true analog sound, you have to come up with a way for the VSTi to monitor the environment. That way the VSTi can detune the oscillators when the air conditioner kicks on.

Robert

[foosnark]

(edited 'cause my post got screwed up by a DNS error)

The problem is that to be truly random there needs to be no repeating patterns. Even given a random seed from the system clock, and modulating your random seed from the raster line, left generating random numbers for long enough your system will begin to show discernible patterns.

The Mersenne Twister algorithm has a period of 2^19937-1. Which means if you sample it at a rate of 96kHz, then in about 1.4244537493174394625671376560697e+5989 years the pattern will repeat.

Considering that the universe is theoretically 1.37e+10 years old, I don't see this as a major issue.

[valley]

I wasn't talking about audio. The issue was whether a computer can create a truly random number sequence.

I frankly doubt that a sequence that repeated every minute would be easily discernible if it was just being used to simulate oscillator drift, but that wasn't my point.

[kybernaut]

I think the "airiness" has to do with loads of subtle distortions that happen in the signal path of analog synths. Add to that the higher audio bandwidth of all components...


--kybernaut

[snooky]

that thing about analogue being affected by the enviroment and perhaps also the randomnes of the user (how often do you get the exact same patch when editing an analogue synth) together with the fact that complete randomnes and infinite resolution in the audio is what makes analogue so dif. from that "static sound" that is digital.

[Lunch Money]

*sigh*

Somehow the thread has become predictable and boring again.

<shrug>

Greg

[foosnark]

The issue was whether a computer can create a truly random number sequence.

Not entirely with a software algorithm. Just a highly chaotic number sequence, determined by a variable (time) which is never the same from one sequence to the next.

There are random number generation hardware devices though, and even online true random number generation services that use them.

Online random number generation is a fun idea, assuming you're not the only one requesting numbers.

[tony tony chopper]

It's not random because you can predict what the value will be if given the seed and algorithm.

randomness just doesn't exists (do you believe in it?), everything has a cause. Here random rather means unpredictable for humans, and you can easily make an unpredictable random number generator on a PC. It would only be predictable if you had the algo, the random seed, and could, for ex, analyze 44100 random values per second in your analog head.

[no_signal]

Digital can not and will never be able to produce real randomness

What kind of randomness are you talking about ?
Temperature dependancy of electronic components ? Tolerance ? Semiconductor nonlinearity ? Noise ?

What if i take some components let's say 3 transistors and compare them. I will have 3 different set of diagrams. This is because each has it's own imperfection and impurity. Maby you cannot modell every detail of a circuit, but DSP designers already discovered a lot of important things that make analog sound analog, and they can modell it using math.

Analog oscillators : people thought a long period that the analog specific timbre comes from the waveform, and they were trying to capture those waveforms and use them with samplers.
Than when even powerfull pc's came out they started to modell circuits, and a new synthesis type came out Virtual Analog.

Analog oscillators don't have exact frequency. If you would test 2 analog oscillators with exactly the same parameters, and you would start them up syncronised, you would notice that they go out of phase. This is easy to emulate in digital, you just have to detune one of the oscillators, and maby add some FM. Same goes to filters and other modules.

It's funny that in the 80's people were trying to get sound to mathematical perfection, and now they are trying to get rid of it.
As long as it sounds good, ie. it's not disturbing to our ears, it's good.