Analog summing emulation idea

[funkadil]

Het kingston, do you juggle? And im suprised to see a juggler om the internet outsind of a juggleing forum. Im drunk that may be why. Do you mean in you sig the physical juggling of objects? just a small wonderance,,,, may be a waste of time. sorry.

[Kingston]

Het kingston, do you juggle?

The sig is actually a loose reference to Alastair Reynolds space operas (featuring alien species called patterns jugglers). It's not meant to make any sense.

I *wish* I could juggle.

[jens]

I ride a unicycle - do I win?

[Aleksey Vaneev]

voidar, where you've got that? 22050 Hz sinewave is sinewave. It just lacks phase information. But it won't unfold into squarewave after interpolation.

[Sascha Franck]

kritikon, that was a nice post of yours on page 4 and it's IMO adressing the problem quite a bit more than the "pure" technical issues that may or may not get in our way when working with digital audio.

Just as another example: I use to fiddle around with the knobs of my (guitar) amp when playing live, even during a song. Things such as getting closer to the stack, raising the drive to get into controlled feedback, etc.
I can't recall doing this even once when using programmable devices (which I am doing as well).
Couldn't imagine fiddling with, say, GR2s gain knob while tracking either (again something I just do while tracking with analog amps). That's why I'm building up some sort of analog control setup for these sorts of things.

As said, to me this is much more something related to the sonic limitations of working digitally than some bussing or filtering issues.

Another thing, probably getting OT, but it might fit a bit: We've got too much things to chose from.
Even the cheapest folks among us have a choice of instruments and FX that just 10 years ago not even some of the most expensive studios could've dreamed about.
They had like 2-5 reverb units to chose from (unless you rented something else) and probably like 10 synths hanging around - ok, some had more, some had less, but that's beyond the point. Today even the freeware only folks have 50 reverbs, 100 modulation FX and 2000 synths to chose from.
Getting the most out of one thing? Hah, no way. Instead, you just chose another synth/effect.
Impossible back then, you just had to live with the limitations of what was available.
Admittedly, having to use, say, only a 480L and some PCR90 might not be all too much of a limitation, but you couldn't call up more than one instance of them either.
When I started with home-recording, I've even been using my entire guitar amp top as a reverb unit, simply because there's been nothing else available at my budget. Did it sound "good"? I don't know. Did it add an "analog" touch? Certainly yes. Have the results been "somewhat useful" or "interesting" at least? Certainly yes as well. I'd never do such a thing these days though...
Heck, I even dampened the reverb by applying some tissue to the reverb springs because the damned thingies wouldn't stop ringing.
My parents also had a piano. When I started I had no mic to record it, so I used some headphones instead. These days you'll most likely be getting a handful of prepared piano libraries to get there and still be not happy. No, I'm not saying I was happy with the results either, but I used things nonetheless because I simply had no choice.

Of course, just as about anybody I am absolutely enjoying the possibilities of the digital world, my current main (mobile) setup fits in a backpack, delivering better quality than anything I've ever used before, along with total recall and the likes...
But it surely has some limitations by not being limited.

Btw, as a last thing, "total recall" being one of the other questionable things in the digital world. Back in the days you just had to decide that a mix was finished (unless you had some studio assistant taking care of writing down any setting, doing SysEx dumps, saving sampler patches to floppies, etc). These days you don't even have to commit a guitar sound to a track anymore as you could allways wait for the next shebang in the amp simulation business.

As said, I'm enjoying things enormously, but even if I never owned something that could've been considered a reasonable studio in the analog days, it had me going creative in quite some aspects that seem to be entirely missing today. "Getting the most out of it" probably being the main issue.

[C00kie]

Great posts, Sascha & Kritikon!

With analog gear it's easier to have slight variations in your settings which keeps it interesting. Same reason why I ditched my digital multi-effects unit for guitar, and reverted back to a small set of stomp box pedals.

On the digital unit I only used my 8 fav presets I made in the first week. On the analog units you just swing some knobs to change the sound, and thus it remains more organic.

It's just too easy to get "perfect" in digital which kinda neglects the "soul".

[voidar]

voidar, where you've got that? 22050 Hz sinewave is sinewave. It just lacks phase information. But it won't unfold into squarewave after interpolation.

I said, represented digitaly it is a perfect square. From 1Hz and up to 22050Hz, at 44.1KHz sampling, sines will progressively get worse.
Think about it. A 1Hz wave is [theoretically] sampled 44100 times in a second. That gives great precision on how the wave actually looks. Especially after the D/A-stage.
A 22050Hz wave however is only sampled once. You can't represent it any other way than as a square/pulse/tick. And yes, naturally, phase-information would be lacking.

Now, it doesn't sound like it after D/A because of the lowpass-filtering. A perfect square is afterall constructed of an unlimited amount of frequencies. Now, a lowpass-filter set to cut off at 22050Hz will shave most of those off, thinning out the square.

I've got Kingston to back em up on that one, I believe.

[Kingston]

Now, it doesn't sound like it after D/A because of the lowpass-filtering. A perfect square is afterall constructed of an unlimited amount of frequencies. Now, a lowpass-filter set to cut off at 22050Hz will shave most of those off, thinning out the square.

I've got Kingston to back em up on that one, I believe.

The easiest way to live with that knowledge is to think the digital 22050Hz square as a sinewave, because that's what it IS (ideally, but that's beside the point).

I'm not backing up anything, Nyquist is. I'm absolutely sure Aleksey has a much more thorough understanding of the theory than most of us here. Afterall, he coded one of THE best resamplers in the world.

[DWb]

voidar, where you've got that? 22050 Hz sinewave is sinewave. It just lacks phase information. But it won't unfold into squarewave after interpolation.

I said, represented digitaly it is a perfect square. From 1Hz and up to 22050Hz, at 44.1KHz sampling, sines will progressively get worse.
Think about it. A 1Hz wave is [theoretically] sampled 44100 times in a second. That gives great precision on how the wave actually looks. Especially after the D/A-stage.
A 22050Hz wave however is only sampled once. You can't represent it any other way than as a square/pulse/tick. And yes, naturally, phase-information would be lacking.

Now, it doesn't sound like it after D/A because of the lowpass-filtering. A perfect square is afterall constructed of an unlimited amount of frequencies. Now, a lowpass-filter set to cut off at 22050Hz will shave most of those off, thinning out the square.

A lowpass filter set to 22050Hz like the one that is an inherent limitation of the human ear, for instance.

[voidar]

While true, the filtering of A/D/A stages is needed for anti-aliasing. You don't want artifacts to show up below the nyquist threshold. Any human hearing will not help that. It will just sound bad.

[voidar]

The easiest way to live with that knowledge is to think the digital 22050Hz square as a sinewave, because that's what it IS (ideally, but that's beside the point).

I'm not backing up anything, Nyquist is. I'm absolutely sure Aleksey has a much more thorough understanding of the theory than most of us here. Afterall, he coded one of THE best resamplers in the world.

You seem to know your nyquist, so that is what I meant in essence when saying what I said.

And Aleksay is right in his way, like I am. The 44100Hz/2 square is ideally meant to be a sine, like all the other frequencies below that. Digitally, it is not. But I guess the important part is that information on phase gets progressively worse at higher frequencies.

This will result in a strange representation of frequencies strong in harmonic content in that range.
Filtering and oversampling is all about making such a scenario sound "good" within the limiting design.

[friteuse]

Taking an computer generated 1 kHz @ 44.1 kHz sample rate will make it impossoble to delay 10 µs without interpolation (as the smallest possible delay @ 44.1 kHz is 1/44100 ~= 25 µs).

Maybe I missed the point completly, but I see no direct relation to nyquist here... of course, nyquist is extremly important when dealing with AD/DA, but wasn't at all the topic I was talking about

You are right, in a sense. 44.1khz can't do it without an allpass filter (oversampled), but this is where nyquist theory (and discrete digital signal representation) has a very significant importance. It is DIRECTLY related to the issue, and if you studied it a bit more you'd find out why.

Similarly, no examples with mic placement (even when moved a millimeter), and perception of direction do not work here. It's simply the way nyquist works. 44.1khz DOES capture even the smallest changes in mic placement.

It's not the easiest of things to grasp, and there's no easy way of doing the "nyquist 101", unfortunately.

It's not just a simple issue of nudging 1 sample either way. It's the complex interaction of time vs. amplitude vs. the resulting nyquist sine formula (basically sinc) that deal with both frequency and phase.

'Any bandlimited sine wave can be represented by drawing only two points in discrete time' (yes, that's only two samples needed for one perfect sine wave)

Understand that, you're a good way off of the placebo land.

Yes, I understand your point. I know about nyquist theory (though it's not my daily work), sinc functions and how to perfectly reconstruct a bandlimited signal with a sinc convolution (the ideal lowpass function).

This works perfect for stationary sinus signals (below nyquist), signals that have no beginning and no end. These are very rare in real music signals. Even in long played notes, i.e. from a cello, you find lots of very small structural signal changes within the sound (="lots of little impulses"). Those starts and ends of in itself "low frequencies" can contain very high frequencies, if they are impulsive. Here, bandlimitation brings some problems, as it "blurs" impulses more or less.

So you'd more have a look on these impulses than on stationary sinus waves. Any bandlimitation of an impulse means loss of accuracy. The higher the nyquist, the better representation of impulses. In other words: the higher the sample rate is, the better impulses can be processed digitally. The higher impulse accuracy has the mentioned side effects to direction perception, as I understood the article.

And I think that's the point when working with higher sample rates, you gain, easy spoken, on fine elements, that, like the above mentioned tests, can be well percepted and lead to a more transparent, natural and 3-dimensional sound.

Finally, I want to give up this discussion at this point - as I'm not a god-like expert on this, and don't wanna be one. I have some knowledge which is sufficient for my daily work and, really, to see what's placebo and what's not. I just tried out what happened when I switched to 96 kHz and was very positively surprised. Everyone can try this out, it's just a mouse click in your DAW, and if one don't hear any difference to 44.1 kHz - no matter, I won't prevent him to go on working like before. I'm happy with 96k so far I use it and if it's placebo, oh my god, then I'm very happy with placebo.

[camsr]

Just remember, phase is dependent on samplerate when working with digitally recorded samples. Record a signal from a synth, or drum machine, into your daw, and notice two seperate hits will not be the exact same sample. This is why analog sounds different, even in 44100hz recordings. Our DAWs don't interpolate to do timings in between samples, they just round their value to the nearest sampling interval, and play the exact sample that was recorded. Maybe one day when we have enough processing power, we could make a sampler that does this. It would be very cool in my opinion

[aciddose]

you guys dont seem to understand the concepts of pcm.

every sample is to be regarded as a single scalar value with an infinitely small period at an infinitely perfect instant in time. when this scalar is replaced with a sinc function, the original signal is reproduced.

current DACs rather than reproducing the signal correctly, simply insert a flat dc voltage starting from the pulse time (which should be the center of the pulse, not the beginning) and lasting until the next pulse. this means the output is first incorrectly delayed by half a sample period, and also the output pulse is a simple change in dc voltage (possessing infinite harmonics) rather than a sinc pulse.

before arguing about sampling and analog reproduction of a sampled signal, please first take the time to understand the basics of the process.

[Aleksey Vaneev]

voidar, I think you do not understand Fourier analysis. If you knew it enough, you would not say that 'sines get worse with approaching Nyquist frequency'. They do not.

The fact that some second order (biquad) filters do not behave well near Nyquist does not mean sampling itself has alike limitations. It is true that *filters* of low orders can be hard to design near Nyquist frequency, but *signal* you are going to filter is always precise up to Nyquist frequency.