Re: [jOrgan-user] Thoughts on creating Soundfonts from Samples

[John R. <joh...@ex...>]

Hi Graham,

Thank you for starting this thread. The making of quality samples is basic
to having excellent jOrgan dispositions. Paul Stratman has already given
great advice about making synthesized samples, and I would suggest that
anyone interested in making samples should consult his website, where I
assume he has his advice somewhere. (I downloaded and printed out his
guidance quite some time ago, and have learned much from what he has
offered).

Here are my comments on various subjects you have raised:

"WHAT VALUES DO WE APPLY TO THE ATTACK AND DECAY?"
I don't apply any values to Attack and Decay, and am not aware that we need
to, if we are using properly-formed samples with the right characteristics
already built in.

"ANYTHING IN THE WAVE FILE THAT COMES AFTER THE LOOP DEFINITION CAN SIMPLY
BE DELETED"
If you are using Viena (one "n") as your sounfont editor, it is advisable to
leave in a few steps AFTER the end marker, especially if you are using
Zero-X Seamless Looper to find good looping points. (With that program,
there is an Save option which allows you to NOT truncate the saved file). I
use that program, and I have found that in Viena I always have to add one
step to the end marker when importing a sample. If the looped file has been
truncated (nothing left after the end marker), then it is not possible to
add that extra step. I suppose one could always move the start marker back
one step in order to achieve a good loop, but I haven't actually tried to do
it that way. I might say that the set-up Viena provides (Graphical User
Interface) for getting good looping points is excellent, very easy to use,
and I think that with it, not 
having the Zero-X Seamless is not a great disadvantage.

"ADSR RELEARSE DEFINITION"
I routinely use 0.2 seconds for the release parameter, regardless of the
pitch of the note, because I think it emulates the action of expansion
chambers in slider wind-chests, providing a slight carry-on of the note
after the playing key is released. In the expansion chamber, after the
pallet has closed, there is inevitably still some air pressure which can be
released only by passing through the pipe, and so it continues to speak very
briefly. I have no idea whether my practice corresponds closely to the times
involved.

"RELEASE SAMPLES"
I had great success with releases on my EO2 Krummhorn/Cremona stop, but only
when being played using the sfz sound engine. There were problems when using
Fluidsynth, and I didn't discover how to avoid them, other than lowering the
volume of the release samples for the Fluidsynth soundfont, and dispensing
with the use of any Attenuation parameters at either the Instrument or
Preset level. I regarded this as too cumbersome a solution. I would love to
hear from anyone who has been successful with release samples using
Fluidsynth, because this may show me what I was doing wrongly.

"THREE NOTES PER OCTAVE"
There is no simple answer to this question. It could be regarded as a
trade-off between quality and how much time you want to devote to making
samples. One of the principles by which I work is to minimise the time it
takes, but not at the expense of actual musical excellence. I have never
favoured fewer than one sample per octave, at least for flue pipes, because
anything fewer can't really cope with the changes to scaling across the
range of a rank. However, good results can be achieved with fewer than one
sample per octave, as evidenced in a number of jOrgan dispositions which
have met with high approval. In my dispositions (and I still regard myself
as a mere "starter" in this), my Earlwood Organ No.1 usually had two 
samples per octave. I have a good organist friend (Graeme R. in Sydney) whom
I set up with MyOrgan, on which he plays the C-C Mainz samples, which he
bought. I regard him as my chief "critic" or "consultant" in my organ-making
endeavours, and I take his comments very seriously indeed. Although he was
very pleased with the EO1 (his first jOrgan instrument), he has indicated
that only two samples per octave means that there is an inevitable blandness
across each group of six notes, whereas when he plays the C-C Mainz, he can
hear an individuality with each pipe sound (there is one sample per note). I
took this seriously with my EO2, and I used three samples per octave and
interleaved them across the octave in six pairs of notes, so there is no
suggestion of blandness. I think this has been successful. It is very
obvious with the Krummhorn/Cremona stop. However, I have just begun to work
on making recorded samples, at least hybrid samples where the attack section
is recorded and the looped section is synthesized and quite short. I am even
trying it with SIX samples per octave i.e. using six pairs of notes based on
recordings from actual pipe ranks. I think at six per octave I will be
drawing a line in the sand and saying "No more!). I think six pairs using
six samples per octave instead of twelve samples will be a good compromise
between "no blandess" and "too much work". (Each pair is one soundfont zone
(or "split") of two successive notes sharing the one sample. Those two notes
therefore have the identical attenuation and panning etc. parameters. I
don't see this as any disadvantage).

"A DEFINITION FOR EACH NOTE"
This can be done, of course, even when the one sample is shared over a few
notes, although see my previous comment. It all adds to the amount of time
and work involved for the soundfont creator.

MONO SAMPLES ARE BEST"
I agree totally. I have found panning to be very successful and satisfying.
The EO2 has 30 zones per rank as opposed to the 12 zones used in the EO1,
adn the panning is much more successful. I plan to upgrade the EO1 to follow
the interleaved and 30 zone pattern.

I might add that although creating good samples and soundfonts is very
demanding and time-consuming, it is also extremely satisfying as a creative
endeavour. I find great enjoyment as a full rank comes into existence.

Since writing all the above, two new posts have appeared on the subject of
mixtures. Mixtures are a much greater challenge, involving much
head-scratching, but it is the case that I have managed to preserve true
breaking-bank with the mixtures in the EO1 and the EO2.

John Reimer
-- 
View this message in context: http://jorgan.999862.n4.nabble.com/Thoughts-on-creating-Soundfonts-from-Samples-tp2551727p2551867.html
Sent from the jOrgan - User mailing list archive at Nabble.com.