Fluidsynth Overflow Settings

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

Fluidsynth Overflow Settings

John Reimer
Administrator
This post was updated on .
Hi all,

I have never given any real thought to the Overflow Settings in the
Fluidsynth Properties View, but having just “downed my tools” after years of
making samples from recordings, I had enough time at least to compare the
Settings in my recent jOrgan dispositions with those which Paul Stratman has
in one of his dispositions (the Moller was the one I looked at). Paul has
settings which vary from 0.1 to 0.55, whereas all my settings are “0”.

This provoked me to make a search through the Nabble Archive, and also a
Google search. Those searches certainly did not answer all the questions I
had, but they did shed some light. They did suggest, for example, that the
use of “0” for all the settings can be traced to the suggestion which Dan
Dietzer made in a Mailing List post back in  2011, after some degree of
experimentation on his part. I am assuming that the values Paul used were
actually chosen by Sven.

Overflow is described in an email reply sent in 2013 (from memory) by a
Fluidsynth developer to some forum in these terms:  “In a voice overflow
situation, i e, when you must kill a previous voice to leave room for the
new one to start playing, all voices are given a score, and the one with the
lowest score is killed.”

I assume that by “voice” he means the same thing as we would mean by “note”.
I also gather that Overflow comes into operation when the player makes
demands upon Fluidsynth which exceed the Polyphony Setting currently in use.
For example, this would occur if the Polyphony Setting is 24, and 12 stops
are in use on a given keyboard, and more than 2 notes are played at the one
time (this simple example ignores the effect of long release times). It is
also likely that if a stop consists of a soundfont preset which has more
than one layer, then that actually increases the number of “notes” which is
operative as regards polyphony.

So far, so good. But I noticed that the possible scores used by Fluidsynth
range from from high negative values (in the thousands) to high positive
values (in the thousands). Fluidsynth default values range from -1000 to
+4000. jOrgan allows us to set a fractional number between 0 and 1
inclusive. Perhaps that number represents a multiplier of the Fluidsynth
default numbers  - I really don't know.

In conclusion, an Overflow situation can be avoided by making the polyphony
setting quite high, but I am sure that as far as the computer CPU is
concerned, this attracts a penalty. Perhaps someone can tell us what that
penalty is. And hopefully, they can give us some guidance as to what the
Overflow settings should be, for whenever we feel that Overflow should not
be avoided altogether.

There is perhaps another situation when Overflow operates - or if not, then
some other mechanism comes into play. That is, if we try to use a number of
stops where that number exceeds the "Number of Channels" Setting. Fluidsynth
(or jOrgan) allows us only specific numbers of channels to set, between 16
and 256. Some sound engines allow only 16 channels (or 15, if we exclude
MIDI Channel 10 - the percussion channel), but clearly Fluidsynth allows a
larger number. I don't know what it is. 256 seems a bit excessive, but
perhaps I'm just showing my ignorance.

John Reimer



--
Sent from: http://jorgan.999862.n4.nabble.com/jOrgan-User-f999863.html

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
jOrgan-user mailing list
jOrgan-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/jorgan-user
Reply | Threaded
Open this post in threaded view
|

Re: Fluidsynth Overflow Settings

greenfox
Hi John

Interesting looking into this, thank you.

I would suggest that the terminology "Voice" would translate to "Rank" for jOrgan users.
It seems to me this is a way to rate the importance of each rank and to guide the system on which to drop if needed.
Fundamental ranks and heavy reeds would be assigned higher priority and softer ranks lower priority as they would be less noticeable dropped out of an ensemble.

Regards
Rick
 

On Mon, 18 Dec 2017 at 15:03 John Reimer <[hidden email]> wrote:
Hi all,

I have never given any real thought to the Overflow Settings in the
Fluidsynth Properties View, but having just “downed my tools” after years of
making samples from recordings, I had enough time at least to compare the
Settings in my recent jOrgan dispositions with those which Paul Stratman has
in one of his dispositions (the Moller was the one I looked at). Paul has
settings which vary from 0.1 to 0.55, whereas all my settings are “0”.

This provoked me to make a search through the Nabble Archive, and also a
Google search. Those searches certainly did not answer all the questions I
had, but they did shed some light. They did suggest, for example, that the
use of “0” for all the settings can be traced to the suggestion which Dan
Dietzer made in a Mailing List post back in  2011, after some degree of
experimentation on his part. I am assuming that the values Paul used were
actually chosen by Sven.

Overflow is described in an email reply sent in 2013 (from memory) by a
Fluidsynth developer to some forum in these terms:  “In a voice overflow
situation, i e, when you must kill a previous voice to leave room for the
new one to start playing, all voices are given a score, and the one with the
lowest score is killed.”

I assume that by “voice” he means the same thing as we would mean by “note”.
I also gather that Overflow comes into operation when the player makes
demands upon Fluidsynth which exceed the Polyphony Setting currently in use.
For example, this would occur if the Polyphony Setting is 24, and 12 stops
are in use on a given keyboard, and more than 2 notes are played at the one
time (this simple example ignores the effect of long release times). It is
also likely that if a stop consists of a soundfont preset which has more
than one layer, then that actually increases the number of “stops” which is
operative as regards polyphony.

So far, so good. But I noticed that the possible scores used by Fluidsynth
range from from high negative values (in the thousands) to high positive
values (in the thousands). Fluidsynth default values range from -1000 to
+4000. jOrgan allows us to set a fractional number between 0 and 1
inclusive. Perhaps that number represents a multiplier of the Fluidsynth
default numbers  - I really don't know.

In conclusion, an Overflow situation can be avoided by making the polyphony
setting quite high, but I am sure that as far as the computer CPU is
concerned, this attracts a penalty. Perhaps someone can tell us what that
penalty is. And hopefully, they can give us some guidance as to what the
Overflow settings should be, for whenever we feel that Overflow should not
be avoided altogether.

There is perhaps another situation when Overflow operates - or if not, then
some other mechanism comes into play. That is, if we try to use a number of
stops where that number exceeds the "Number of Channels" Setting. Fluidsynth
(or jOrgan) allows us only specific numbers of channels to set, between 16
and 256. Some sound engines allow only 16 channels (or 15, if we exclude
MIDI Channel 10 - the percussion channel), but clearly Fluidsynth allows a
larger number. I don't know what it is. 256 seems a bit excessive, but
perhaps I'm just showing my ignorance.

John Reimer



--
Sent from: http://jorgan.999862.n4.nabble.com/jOrgan-User-f999863.html

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
jOrgan-user mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-user

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
jOrgan-user mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-user
greenfox - Brisbane Queensland Australia
Reply | Threaded
Open this post in threaded view
|

Re: Fluidsynth Overflow Settings

jbeach2646
In reply to this post by John Reimer
John R., for the reason that I did not know what the "Overflow" settings
actually referred to or effectuated, I have never varied the values of any
of them in any disposition,  made by others which I have downloaded,  or
which I have made myself.   The "default" values, therefore, are Overflow
Age =.55, Overflow Released=.5, Overflow Sustained=.1, and Overflow
Volume=.5.  Since you stated that Paul Stratman's settings varied from .1 to
.55, I am assuming that his are also a jOrgan "default" value.   I believe
the default value for Polyphony in jOrgan must be 512, since I have never
altered that setting either.   Back when the Portaudio backends were first
presented by Bernd and Graham G., it was stated that the value for the
Sample Rate had to be 48,000 when using Portaudio, regardless of the sample
rate at which the wave files were recorded.  I, purposefully, made a
comparison using wave files recorded at 44,100  and 48,000 to see if the
there was any noticeable difference in quality or functionality in jOrgan
with the recommended setting of 48,000 for Portaudio and I can not say that
I notice any difference.  I recently changed that setting in one of my
dispositions to 96,000 (also changing the "Default Format" value of
Advanced, Speaker Properties to 24-bit, 96,000 under Playback Devices of the
Windows Volume Control, the Speaker Icon in the lower right-hand corner of
the launchbar on the Desktop).  I do believe there is a significant
enhancement of audio quality at this setting vis-à-vis 16-bit, 44,100.

Is the subject of "Overflow"  values addressed in any of the four volumes of
jOrgan tutorials by Bill Skees?

Related to sound quality, I noticed that in your most recent release, the
"Sydney Schuke" organ, your reverb settings are Room=6, Width=10, Damping=2,
Level=1.  These settings are excellent.
I find the subject and effects of reverb settings to be really problematic
since they affect other aspects of sound quality, loudness, articulation,
the proportionalities of loudness in ensembles of
footages in choruses, and reverb, alone, having these effects, makes it
difficult to know whether or how the individual parameter settings within
the soundfont are going to be treated or affected
by alterations of the four reverb values.  If "room" is perceived to be the
border lines of a square or rectangle,  "Width" is the length of those
border lines.  "Damping" is the value causing the least amount of noticeable
change of the four, which is probably logical, given the factor of physical
elements in a given acoustic space, (such as wall medium, rugs, columns or
partitions, etc., which would "dampen" sound).  In fact, using your
settings, given above, if the "Damping" is varied, leaving the other
settings the same, there is no noticeable difference between 0
and 10.  In one of my dispositions, using the same settings, there is only a
marked change noticeable between the 9 and 10 (highest) level of the Damping
slider.  "Level"  has the effect of
"grandiosity".  Seemingly, it should not even be a consideration since the
acoustic ambience would consist of the size of the space in which a sound
source (organ) is located
and the physical objects which would dampen the sound from that source.  If
"Level" represents some "cumulative" of the relative values of the other
settings, it also negates specificity of
their values.  Reverb seems to be more "art" rather than "science."  What I
found is that your settings give excellent articulation of the sound quality
of wave files used.  Settings which approach
a "Cathedral" reverb lose this kind of articulation, yet the
"cathedral-reverb" sound is probably that with which most people identify
the sound of a pipe organ.

I have often read that the space in which an organ is located is the most
significant aspect of the organ itself.   I have always tended to believe
that if the sound source, i.e., instrument, was not
of a good quality, particularly as this applies to electronic instruments,
the space in which it is located would make little difference.  On the other
hand, a good instrument would sound good
in any space.

What is problematic about the Reverb in jOrgan, (and I expect, digital
reverb, in general) is that we are creating an artificial space and varying
the breadth of sound in that space, through speakers that are located in a
much smaller space.

If the original, recorded source of the sound of an organ pipe were the
chamber in which it is located, for purposes of purity (non-ambient reverb),
what would the equivalent values
of the jOrgan Reverb elements Room, Width, Damping and Level be?  And,  if
the recorded wave files of synthesized sounds are done with NO digital
reverb at all, isn't there a kind of parity
between the two?

Thanks.

John Beach



------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
jOrgan-user mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-user
Reply | Threaded
Open this post in threaded view
|

Re: Fluidsynth Overflow Settings

John Reimer
Administrator
This post was updated on .
In reply to this post by greenfox
greenfox wrote
> I would suggest that the terminology "Voice" would translate to "Rank" for
> jOrgan users.

Rick,

Thank you for your contribution to this thread. The use of the word,
“voice”, is certainly not clear to us, hence our differing opinions on what
the term means as used in the Fluidsynth documentation. It turns out that on
this occasion my assumption is the correct one: it does correspond to what
we mean by the word, “note”, and not “stop” or “rank”. This is clear from
what one of the developers replied when asked about the meaning of the term:

“A voice represents a single audio sample being played back. A single
note-on event may trigger several voices, if an instrument has layered
samples.”

So you see that he has also referred to the situation I mentioned in my
post, namely multi-layer samples (the Preset having more than one
Instrument), except that I made a mistake in using the word “stop” there. I
have edited it in my original post, so as to have the correct word.

John Reimer




--
Sent from: http://jorgan.999862.n4.nabble.com/jOrgan-User-f999863.html

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
jOrgan-user mailing list
jOrgan-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/jorgan-user
Reply | Threaded
Open this post in threaded view
|

Re: Fluidsynth Overflow Settings

John Reimer
Administrator
In reply to this post by jbeach2646
jbeach2646 wrote
> Back when the Portaudio backends were first presented by Bernd and Graham
> G., it was stated that the value for the Sample Rate had to be 48,000 when
> using Portaudio, regardless of the sample rate at which the wave files
> were recorded.
>
> I noticed that in your most recent release, the "Sydney Schuke" organ,
> your reverb settings are Room=6, Width=10, Damping=2, Level=1.  These
> settings are excellent.

JohnB,

Thank you for your post. On the first matter, it is not the case that a
48,000 Sample Rate has to be used with Portaudio. 44,100 works very well, I
can assure you. There is a condition when WASAPI is being used, and I
believe it requires a consistency between what is set in any soundcard in
use, and what is set in the Fluidsynth Properties. GrahamG can confirm
whether this is so.

On the Fluidsynth Reverb settings, I agree that those settings give an
excellent, if short, reverberation effect, with very good clarity - which I
why I use them. The settings were arrived at by our late friend, Panos,
except that he had a Level=2.

John Reimer




--
Sent from: http://jorgan.999862.n4.nabble.com/jOrgan-User-f999863.html

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
jOrgan-user mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-user