Graham W. said, "So, if the sample set producer was careful to keep the
phase relationship between the looped and release sections as close as
possible to being synchronised, and the sound engine maintained sample
synchronisation between the looped and release samples, then there is no
significant phase change to occur. A good sample set would have its release
samples phases adjusted for best synchronisation to the loop section
phases."
Isn't the idea of "keep(ing) the phase relationship between the looped and
release sections as close as possible to being synchronized....." really a
near impossibility given the fact that the playback of a looped sample,
whether auto-looped or with fixed, start-and-end points, will cease at the
specific point-in-time of the release of the key which, then, triggers the
playback of the release transient?
While I was concerned about the issue of phase, and, more precisely or
correctly, the issue of the exact end-point of the looped portion of the
steady-state, and the start point of the release sample, I realized that it
is key release which determines the point in the looped portion of the wave
file at which its cessation occurs. This is not precisely determinable,
even with a fixed loop, to say nothing of an auto-looped, steady-state
sample. I can not conceive of any state-of-the electronic art, apart from
real pipes, which would make this a possibility. We are dealing with two
parts. The steady-state sample has the need for seamless looping. The
release sample has looped silence which plays during the variable length of
the playback of the looped steady-state sample. The former is a variable.
The latter is a constant.
Since the onset of this discussion concerning release samples and, in
particular, John R.'s assessment of the problem as well as the benefits to
be derived from release samples, I (who use synthesized sounds exclusively)
have redone all of the flute and diapason stops and many of the reed and
string stops in several of my largest soundfonts. Previously, I simply
discarded all
the recorded portion after four seconds, including the natural release
tails, and used the release parameters for specific footages, set in the
instrument pool of the soundfont editor. I need to say that I use Creative
Sound with my additive synth in jOrgan and that I do not use any reverb such
as the fluidsynth or Polyphone reverbs. So the release tails are formed by
the release of the key and the gradual decay of the sound over the course
of, approximately, 7/10ths to 1 second, after which there is edited silence.
The release wave files are the ends of the recorded steady-state, looped
wave files. Any subsequent reverb used in jOrgan or Creative Sound is
Impulse Response or electronic and artificial.
What I did, --a compromise---taking into account John R.'s statement that he
was content with the way things are with respect to ADSR parameter settings
in soundfont editors, and Graham G.'s observation that "GrandOrgue and
Hauptwerk use release markers, and so can have the release portion as part
of the same wave file as the attack and sustain, or have it as a separate
file. Giga has the release portion as a separate file that plays on Note
Off,"--- was to set the release parameter of the steady-state, looped
sample as usual for a soundfont instrument (1.6 for 16' flue, .8 for 8'
flue, etc.) which does not have release samples, then, globally, in the
release-sample instrument and preset of the soundfont, set the release
parameter for the specific footage of the stop.
In editing the release samples in Audacity, I "cut" the entire release tail
portion along with approximately a half-second of steady-state tone portion
before the release. I, then, use the Effects>Amplify, to reduce to silence,
the amplitude of the portion of the steady-state tone, leaving 4 to 6
complete cycles of the full-amplitude portion of the steady-state tone just
before
the point of the key release which triggers decay. I "cut" the silence
portion to about 1/8th inch in length before the start of the 4 to 6 cycles
of the full-amplitude portion before the release. In Polyphone, Samples,
I set the volume of the steady-state wave file relative to what it would,
comparably, be in a pipe organ and I set the volume of the release at 10%.
I auto-loop the steady-state wave file and I set the looped silence of the
release wave file, start=0, end=10. I realize that this is "rote," a kind
of arbitrary which recognizes that the release portion, while adding to the
overall quality of the sound, is, in fact, a transient, which, without the
steady-state, would be meaningless, a pizzicato string sound, indiscernible
as organ-tone timber. Therefore, I have concluded that the particulars of
phase relationship are inconsequential/inaudible using this technique. The
combination of both ADSR parameter settings, identical for Attack, in the
steady state Instrument and the release instrument effectively eliminate any
noticeable transition from the steady-state instrument to the release
instrument.
I do, in fact, notice a significantly improved sound output which makes the
wholly-synthesized, virtual organ stops more realistic and emulative of real
pipes. The benefits outweigh any
negative qualities. I don't find that Fluidsynth has any problem in dealing
with the reconstructed soundfonts with release samples.
John Beach