Sine Waves

[declassified]

I want to create a soundfont with a multisampled, looped sinewave mapped across the keyboard.
Why, you might ask. Well because I've tried various synths' sine wave oscillators and they don't produce one peak in the spectrum but one peak and a lot of other stuff. You don't really hear the difference, I rather need this for experimenting purposes. I work at 44.1Khz, btw.

So I can create sine waves using GoldWave. Is it enough to have one single-cycle at 1Hz and map this across the keyboard? It would cause aliasing, wouldn't it?
So what's the best way to do it? Have the sine .WAV files at 88200Hz sample rate? or even 176000? Would it change something?

Thank you for any help

PS: If you can link me to a VST that actually generates perfect (well, as good as possible at 44100Hz) sine waves, I'd be equally grateful

[BertKoor]

http://home.casema.nl/bertkoor/Angelic.sf2
That's a soundfont I created with AWave Studio (by fmjsoft) based on just one sine wave. I used one with a pretty low frequency, so it's always upsampled (less inaccuracies)

What you see is aliasing and other resampling artefacts. Does it also happen with SFZ in the highest quality mode?

Sampling rate doesn't matter. Data of 100Hz @ 44.1kHz is exactly the same as 200Hz @ 88.2kHz.

Oh, did you try MDA TestTone??

[mistertoast]

>>Well because I've tried various synths' sine wave oscillators and they don't produce one peak in the spectrum but one peak and a lot of other stuff.

I'm not sure I believe that. What I mean is that it's true that some oscs are garbage, but there are plenty of quality sines out in the free realm.

I've looked very closely at this issue (with freakoscope) while I was doing my own synth. My recollection is that Arguru's Voyager, among others, is essentially perfect. I know mine has no peaks within maybe 140dB of the fundamental. You may as well just use Voyager.

A sine wave is the one thing you don't need to multisample. Since it has no harmonics, you can take it all the way up to Nyquist. The key is the interpolation, but if you have enough samples, even linear interpolation is sufficient for sine. At 4096 samples with linear interpolation, you should be gold.

[mistertoast]

http://www.audiomulch.com/~rossb/code/sinusoids/