This doesn't really answer your question, but all I can say
is that the cubic interpolation that was introduced in V3.50
is doing a very good job indeed. I did some testing of my own
with sine waves, and examined the spectrum of the interpolated
results - no noise spectra appeared above the -100dB level,
and the overall THD (A weighted) was about 88dB, which 
is SIGNIFICANTLY better than the linear interpolation.  (I was
creating an output waveform with about a -1dB amplitude)

I think perhaps rather than looking at the shape of the
waveform produced by Gen06 and Gen08, it would be better
to look at the spectra. This will make it much easier to 
determine whether those fgens are *really* doing a better
job than oscil3/loscil3 etc - they may LOOK "smoother", but
are they really producing waveforms that are closer to an ideal
sine wave?  I am using Sound Forge 4.5 and sometimes SpectraPRO 
to look at the spectra.

Of course it is possible to look at the actual values
and compare with an ideal sine wave, but I find it easier
to just examine the spectra.

Greg.

-----Original Message-----
From: Gabriel Maldonado [mailto:g.maldonado@agora.stm.it]
Sent: Monday, 18 January 1999 4:33
To: J P Fitch
Cc: Csound Mailing List
Subject: oscil3: bug or feature?


I noticed that oscil3 produces waves less sharp than oscili, but not enough
smooth as the
smoothness of cubic interpolation routines provided by GEN06 and GEN08. I
tested it with
four-point and two-point sine tables, and compared it with tables generated
by GEN06 and
GEN08. I opened a wave editor and zoomed the waveform to few cycles to see
the sharpness.
Is it normal or we have to deal with a wrong formula?

I tested it with this .csd file (with DirectCsound 2.502, which implements
the new opcodes
of 3.50 canonical version).
I attached the example to this message. Any idea?

-- 
Gabriel Maldonado