pouët.net

I'm just curious...

Some intro soft-synths sound "harsher" than others. I'm also of the understanding that analog modeling seems to be the most common method of intro sound synthesis.

Do these soft-synths use theoretical/pure oscillator waveforms, or something that simulates what analog hardware might generate?

I would guess the filters and their characteristics make up 80% of the final sound, that is if you use subtractive synthesis (removing harmonics with filters)

look into lowpass, higpass and bandpass filtering algorhythms to spice up the waveforms's harsh sounding basic shapes

oh yeah : most of the classic analog synths use subtractive synthesis, so using "analog sounding" resonant filters will probably be what you are referring to :)

cant we just blame it on bad coding and antialiasing instead?

I'm not talking about filters. I'm talking about function generators.

No, you are not. You don't know what you are talking about.

Harshness of the sound has to do with the aliasing of the waveforms and how 'clean' your osc's are. If you use a mathematically 'correct' sawtooth for your saw, it will be very harsh sounding. Analog synths use different ways of generating a sawtooth waveform, for instance, so that's why they tend to sound different.

That, and the filters.

ah okay, don't know if people actually model the waveform generators

I guess you're right in suggesting that it would influence the sound

urgh, ^^ meant for thom

Makes me think... One needs a sort of spline editor anyway? Then why not reuse it for many things?

e.g. generate a set of m=2^n values using splines. Then IDCT it et voila, you have a waveform of ground tone and overtones, carrying m frequences 1/m, 2/m, ..., 1. I wonder how it sounds. Perhaps OK.

Then you can also make splines control progression of sound, which can be volume for multiple waveforms, allowing mixing them in various proportions...

no, you definitely DON'T need a spline editor.

Uh, "splines controlling progression of sound"... these things are called "envelopes", you may want to check them out.

Shame on you, eye/midiclub, etc :)

eyes idea is not all bad when you think about it. Splines can easily be facilitated to generate filters characteristics in frequency space. You just multiply it with sinc, 1/f or whatever waveform you intended to use as a starting point.

After all the common filters boil down to simple polynomials in frequency space. However, a direct representation in poles and zeros is probably less confusing.

Less confusing to YOU is probably to follow the old fashioned way of subtractive synthesis in "time" space, taking the code for some simple 6/12db HP,LB,BP filters from the web.

Sag: Maybe they were referring to something that originates from outer space. Ned Flanders-style.

superplek: are you referring to "the space", perchance?

I know of envelopes, and i'm not ashamed of that. Just that splines are a generalization of that. Plus you need decay...

You can also make periodic splines etc. like in Blender's IPO editor.

Ryg: You don't. I do. ;) Or what do you have against them? They are advantegous to specify camera paths etc.

Stelthz: you apparently did not understand the IDCT idea. The idea is, to draw not waveform, but frequency response. I would think it's not enough to generate really good sound, but it might be a good start. I have to try it out someday. Or actually try out anything. ;)

But hey, drawing waveforms is another funny idea. ;)

nah, who need samples anyway, just play noise and call it art music!

I've heard of really good paper describing sound theory and filter and such stuff and it was made by some finnish guy, perhaps someone knows about it and has a link or something?

eye: I did understand that, however I think it is easier just to use splines to model the filter characteristics and apply this to a standard signal. Using splines to define the entire signal in the frequency domain is probably just good for noise. Notice that the frequency response of most signals with "music" characteristics is actually not a continuous function.

But anways, there is another slight problem with that idea. You cannot define the frequency of the ground tone easily, since you are working with discrete frequencies. Shifting your spline by an offset will not only increase the frequency but also distort the signal..

ekoli: if you want to learn about filters try www.dspguide.com or www.dspguru.com

No, drawing waveforms is probably more or less a nonsense. It must be somehow made sure that at least the values at the left and right are (almost) the same, just as 1st and possibly 2nd derivation. Or is 2nd derivation inconsistencies too weakly audible? Besides, would it be harmonic? I would believe if it's consistent, then it's periodic, and as such also only consists of harmonic frequencies? OK, Waveform drawing is also worth giving a chance but make sure it loops well.

ekoli: you probably mean the top article here:
http://www.biochem.oulu.fi/~oniemita/dsp/
Published sometime in Hugi19, then updated.

Hahaha :)

eye: many old amiga programs supported drawing waveforms, when samples were still a curiosity. However this function is not really useful as it is way too abstract to use.

stelthz: no, i don't want to define the frequency of ground tone in terms of splines. I want to precalculate short samples (e.g. 256 entries long) and morph between them in a sample-based player. And i'm definately not getting the complete frequency spectrum, but the wavelength of the sample, and then 255 of its harmonics. ;) So it's suitable for music.

eye: Ok I see, that removes the frequency shifting problem. However, the others problems still remain. Part of it may be solved by defining even and odd harmonics separatly...

Even and odd harmonics? Uh - oh, i think i have some problem understanding the issue. Gotta read something smart on it.