Sound of OSC

skOre · 26 Jul 2006

I think achieving the fatness of the Minimoog is best to come by with waveshaping and/or a good filter. It might also help to accent the attack of either the Amplitude or Filter Envelope.

But I'm just guessing 😉

tecknoize · 26 Jul 2006

In fact, the SM saw oscillator sounds 'low-pass filtered' in the low notes, And that's probably caused by the interpolation. I Think.

I'm going to check what waveshaping is. Do you have any great link/osm that might help me out?

Thanks!

angstrom · 26 Jul 2006

I've wondered about this before but C0 16.35 Lowest note for tuba, large pipe organs (without a BIG speaker, you won't hear this) C1 32.70 cycles per second (thats a very low note!) C2 65.41 cycles per second (this is actually audible on my speakers, but still V low.)

there are 256 x 256 samples in the wavetables , so I think that makes 256 points for every cycle for our lowest audible note 256 * 32.70 = 8,371.2 points per second, or 8 points per millisecond.

There's no way I can discern interpolation in the sub-millisecond range! It could be that we don't need to 'discern' it of course but that there is something else going on harmonically which is very different from how a bunch of transistors make and affect harmonics at 30 hz.

I don't know though, perhaps it is something else?

PS - my maths is notoriously bad - so I await corrections on my understanding here 🙂

note: perhaps record the wave out and compare and analyse it visually in an appropriate app - it's the best way to see if there is anything strange in the wave compared to a moog wave.

sambean · 26 Jul 2006

Yup, there's 256 samples per wavetable, that means that at 44100/256 = 172 Hz the wavetable will be running with no interpolation, anything less than this though means that there will be interpolation used, meaning the 'sharp corners' of the saw wave will be smoothed off, giving a slightly dulled sound below this frequency.

Try plugging a standard saw into a linear FFT analyser, you will see that below 172 herzt the bandwidth of the saw only takes up a portion of the full spectrum.

Also, if you are using the 'multi-osc,' the sawtooth in that is only half-band for some reason, you can easily switch it for the full band version..

One way to get full bandwidth all the way down, it to below 172 hertz, switch to a standard un-bandlimited saw, there will be very slight aliasing..

Here's a quick module that uses a boolean switch to change oscs at 172 Hz, I havn't noticed any clicks at the transition, nor any aliasing, but I don't possess a pair of dog ears..

http://magalov.myweb.hinet.net/osc/CrossfadeOsc.osm

Sam

angstrom · 26 Jul 2006

hmm, it seems to me like we need to up the number of samples in the wavetable, so we can switch from a 256 wavetable to a larger wavetable at {samplerate/number of samples} as cool as your wavetable to codesaw might be, I think its will be shame if the default SM oscillators get known as 'muddy'

note that I sometimes work in my DAW at 88200 simply to raise the roof on my VSTi / VST - it helps out with any potential aliasing when doing stuff with filters / waveshapers etc. I now realise

88,200 / 256 = 344.53125hz

344hz is a bit high up the spectrum for me to start experiencing interpolation of this kind. I'm obviously losing quality on the oscilators where I am gaining with the filters!

😕

I know when Jorge introduced the wavetable osc (not the default oscs) he asked if a table larger than 256 was needed, I think I would say ' definitely yes' right about now. and to the defaults too!

an osc that did switching between 256 / 512 and 1024 would be the best solution. I guess we could make one from a chain of wavetable oscilators and some simple logic - but it seems daft to have the defaults be weak in this way ... lets face it - they are the basis of the programs audio and "you cant polish a turd" !

tecknoize · 26 Jul 2006

sambean wrote: Yup, there's 256 samples per wavetable, that means that at 44100/256 = 172 Hz the wavetable will be running with no interpolation, anything less than this though means that there will be interpolation used, meaning the 'sharp corners' of the saw wave will be smoothed off, giving a slightly dulled sound below this frequency.

That's what I had in mind. I don't think that changing the number of sample per oscillator is really a hard thing to do. I don't know if it's possible to make my own osc with the code module. I'll give it a try.

Thanks!

tecknoize · 26 Jul 2006

Well, I've managed to do a simple wavetable sin osc using the code module. The only problem is changing its pitch. For the moment, I play the wave faster/slower by modifying the read index increase. But you know, it's sounds like chimps.

Anyway, Take a look!

stefan · 27 Jul 2006

angstrom wrote: note that I sometimes work in my DAW at 88200 simply to raise the roof on my VSTi / VST - it helps out with any potential aliasing when doing stuff with filters / waveshapers etc. I now realise

88,200 / 256 = 344.53125hz

Actually I'd say you can still go down to 172 hz. Interpolation is active anyway, but it is below 172 hz that we start to loose "what-might-be-audible-harmonics". Even when running in 88.2kHz you don't want to have harmonics above 20kHz. That would counterfeit the gain you can get from working at a higher samplerate.

tecknoize · 27 Jul 2006

I've sampled a C1 from minimonsta. Well first of all, the shape of that 'saw' look quite wierd :

While the default Saw OSC in SM look like a normal saw.

How can a saw like that be created using the fourier principe? You know, sin(x) + 0.5*sin(2x) ...

[/img]

rl · 27 Jul 2006

tecknoize wrote: How can a saw like that be created using the fourier principe? You know, sin(x) + 0.5*sin(2x) ...

[/img]

-> make a FFT analysis over the sampled sound. Or go to http://www.falstad.com/fourier/ and try to reconstruct the signal (selecting saw and rising the 2nd cosine bin a little should look simillar)

Audesi · 29 Jul 2006

A wavetable with more points would be a nice thing to have, it would also be nice if it offered hard-sync. That's exactly what I'm looking for. 🙂 Or some good anti-aliased oscillators that can offer this, rather if they're wavetable generated or not.

I have an audio sample here for comparison. The 256 point wavetable sounds like crap on the low end. A lot like an audio sample that is transposed way too low and in need of another multi-sample. lol. You can hear how a plain saw wave becomes more and more muffled compared against a raw saw wave from OSC2 of a Roland JP-8000 synth.

http://www.audesi.net/audio/saw-wavetable_compare.mp3

Earl Dixon http://www.audesi.net

tecknoize · 29 Jul 2006

Sounds like water fart 😀

Sorry had to do it 😊

Audesi · 29 Jul 2006

haha. Oh yeah, it does. lmfao.

Earl Dixon http://www.audesi.net

Audesi · 30 Jul 2006

technoize, the waveform you sampled above, the C1 from minimonsta. It basically looks like a saw wave with less bass. So I suppose that would mean there is less amplitude in frequencies closer to the fundamental frequency of the waveform. Instead of sin(x)+(sin(2x)/2)+(sin(3x)/3)+(sin(4x)/4)+(sin(5x)/5)+(sin(6x)/6)+ ... ect. You might be able to just alter the first sin(x) by reducing the level of it, instead of having it full amplitude.. Or you could try putting a saw wave through different lowshelf filters. Or you can also subtract a sinewave of the same frequency from a saw wave. Put a volume knob to adjust the amplitude of the sine wave so that you can alter the shape of the saw wave.

Acrobat · 14 Jun 2007

tecknoize wrote: I've sampled a C1 from minimonsta. Well first of all, the shape of that 'saw' look quite wierd : While the default Saw OSC in SM look like a normal saw.

Not all saws are real saws. Hardly the ones in use on real gear of the past, especially in the 80ies. For example, Roland ones are really mutants in comparison, don't you think so?

I have an Oberheim (DCO) which goes very close to what we all have in mind as a "SAW" but even that ol' american monster has its special character, in that case, exalting the mids.

A.