Sound Font audio quality for Newbies - was: FluidSynth Sound

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

Sound Font audio quality for Newbies - was: FluidSynth Sound

DellAnderson
Sometimes stating the obvious basic principles is useful, especially when discussing an area of technology in which some users are so experienced that a high level of general knowledge is assumed but not present.

Case in point, my earlier post questioning the audio quality of FluidSynth.  My ears did not lie, yet it was not clear to me why SoundFont based dispositions sounded metallic or harsh or otherwise mechanical to me compared to note-by-note *wav sampled based synthesizers.  I had made some assumptions regarding FluidSynth which were unfounded.  

So in case anyone else is a foolish as me to think that FluidSynth had kept up with the times, I will summarizing what I have learned so far:

FluidSynth is simply open source software module to play SoundFont (SF2?).  SF2 is an published wavetable synthesis standard which otherwise requires special hardware, and can only be edited by Vienna software if one has that hardware, or the open source Viena hardware if one does not.   According to Wikipedia, the term 'wavetable' can be confusing because various hardware manufacturers started using it as a marketing term to apply to any sound card that used PCM samples as the basis of sound creation (hence my confusion with sample based synthesis).  However, wavetable synthesis is really based on periodic reproduction of an arbitrary, single-cycle waveform.  Therefore it cannot accurately reproduce a real pipe organ note by note with the subtle variability inherent in real life.   In addition, even the best hardware SF2 based systems cannot compete with sample based synthesis.  Wikipedia says

"Depending on the details of the actual implementation the sound produced by wavetable synthesis may also contain recognizable artifacts, especially aliasing, quantization, and phase truncation noise."
http://en.wikipedia.org/wiki/Wavetable

Anyone reading the above wiki page will find abundant evidence of nearly infinite variations on ways to recreate sound and explain why this topic is confusing to a non-electronic engineer, and anyone who thinks it is simple and should be basic fundamental assumed knowledge is probably underestimating the various possible ways to recreate sound.

Bottom line, this is sufficient information to explain the smaller size of SF2 based sound sets (compared to true full wave sample sets) and just as with graphical imagery, it is impossible to reconstruct the lost data without artificiality.

What I did not learn was whether the SF2 technology was limited to wavetable synthesis and small files or whether full note-by-note high resolution wav files could be used with that file format, thus allowing SF2 playing hardware or software such as FluidSynth to compete with programs like Grand Orgue or other wav sample based software.   It seems to me, in this day of less expensive hardware and increased audio discrimination and taste for realism, that this is the direction most of us will eventually go.

I hope this was a useful distinction for newbies like me as although I had a vague notion of what was lost with SF2, I had not realized that this was a nearly insurmountable technological limitation with the SF2 format.  

best,

Dell

Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies - was: FluidSynth Sound

MasterPerforator
Hello Dell. I'll take a shot here...

The SoundFont Format (SF2, there was at one time an SF1) is not really a
'wavetable' type playback (no matter what Wiki says). True wavetable
synthesis is more like our old first and second gen Allen 'Computer
Organs' (MOS 1 & 2). This method takes a single short (or very few)
'sample-in-hardware' sounds and manipulates them after the fact. The
hardware synth in most computers (there's one on Creative cards also) is
like this.

A SoundFont is a collection of true PCM wav files (many, or a few),
arranged in a similar method of standard letter fonts. There is a limit
to the physical size of the whole thing, but it can be split, and each
soundfont assigned separately (such as left chamber, right chamber, echo
chamber, etc.).

You can recreate 'wavetable' style by using only a few samples, spread
over many notes (be sure to cut off any attack, and use the ADSR
controls to voice it). If you open one of the soundfonts that Creative
supplies, you'll see very few instances of one sample spreads.

This is different than 'additive synthesis', which creats a sound by
manipulating individual frequency elements. Even if you create a sound
from scratch, it still needs to be recorded as a wav file, and
positioned within the soundfont.

The quality of the original sample relates directly to the quality of
the final sound. The longer the sample before the loop (within reason,
and best with real pipe samples) allows some 'breath' so it doesn't
sound like a looped single cycle. But, you CAN loop a single cycle if
you like. Or use 6-or-more second long samples.

FluidSynth's only major drawback seems to be the sound of the reverb.
Turn it off, and run the signal through an external reverb, or one of
the software choices. Alternately, turn up the damping to dull the high
end some.

Just for the record, I mostly use the Creative reverb that comes with
the cards, I almost always turn off the FluidSynth one.

But if you turn FluidSynth's damping off, and set the level to 50% or
more, you can play inside a large water tank (hehe)...

Rich, Rich, you Son-Of-A-Biscuit

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies - was: FluidSynth Sound

DellAnderson
Thanks, Rich.  Your statement is more in line with what I thought I had
read previously on the jOrgan list, yet there still seems to be some
confusion in my head about what is going on.  From what you said,
theoretically, there is no reason SF2 should not be able to sound every
bit as good as *.wav samples (as used in Grand Orgue) if recorded 'wet',
but I and others were previously chastised as 'absurd' for comparing
Fluidsynth sound with GrandOrgue, so I'm trying to clarify wherein lies
the absurdity of what seems to me a more than normal comparison.

 From your description, though, it sounds like there are usually fewer
actual samples in SF2's than in a normal one-per-note GO sample set, and
the interpolation perhaps is where the artifact is coming in.

Supposedly the SF2 standard is published, but even if I had found it, I
probably wouldn't understand what it was saying in plain English.

BTW, it's nice to have a Creative Card (I do have in a box somewhere),
but not everyone chooses to limit themselves to compliant hardware nor
has the opportunity to open up a laptop and insert one.

The question is: What advantage does Fluid Synth have over a Wave file
player (certainly there must be a free and open source version of
something like Kontakt or similar as it seems there is even less
intensive processing going on in serving up wav files than doing
additive synthesis etc, right?

Dell




On 3/3/2012 3:06 PM, R. Olsen wrote:

> Hello Dell. I'll take a shot here...
>
> The SoundFont Format (SF2, there was at one time an SF1) is not really a
> 'wavetable' type playback (no matter what Wiki says). True wavetable
> synthesis is more like our old first and second gen Allen 'Computer
> Organs' (MOS 1&  2). This method takes a single short (or very few)
> 'sample-in-hardware' sounds and manipulates them after the fact. The
> hardware synth in most computers (there's one on Creative cards also) is
> like this.
>
> A SoundFont is a collection of true PCM wav files (many, or a few),
> arranged in a similar method of standard letter fonts. There is a limit
> to the physical size of the whole thing, but it can be split, and each
> soundfont assigned separately (such as left chamber, right chamber, echo
> chamber, etc.).
>
> You can recreate 'wavetable' style by using only a few samples, spread
> over many notes (be sure to cut off any attack, and use the ADSR
> controls to voice it). If you open one of the soundfonts that Creative
> supplies, you'll see very few instances of one sample spreads.
>
> This is different than 'additive synthesis', which creats a sound by
> manipulating individual frequency elements. Even if you create a sound
> from scratch, it still needs to be recorded as a wav file, and
> positioned within the soundfont.
>
> The quality of the original sample relates directly to the quality of
> the final sound. The longer the sample before the loop (within reason,
> and best with real pipe samples) allows some 'breath' so it doesn't
> sound like a looped single cycle. But, you CAN loop a single cycle if
> you like. Or use 6-or-more second long samples.
>
> FluidSynth's only major drawback seems to be the sound of the reverb.
> Turn it off, and run the signal through an external reverb, or one of
> the software choices. Alternately, turn up the damping to dull the high
> end some.
>
> Just for the record, I mostly use the Creative reverb that comes with
> the cards, I almost always turn off the FluidSynth one.
>
> But if you turn FluidSynth's damping off, and set the level to 50% or
> more, you can play inside a large water tank (hehe)...
>
> Rich, Rich, you Son-Of-A-Biscuit
>
> ------------------------------------------------------------------------------
> Virtualization&  Cloud Management Using Capacity Planning
> Cloud computing makes use of virtualization - but cloud computing
> also focuses on allowing computing to be delivered as a service.
> http://www.accelacomm.com/jaw/sfnl/114/51521223/
> _______________________________________________
> jorgan-sound mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/jorgan-sound


------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies - was: FluidSynth Sound

MasterPerforator
Geez Dell, you're making me think about things I haven't had to since
the 90's (hehe)...

> theoretically, there is no reason SF2 should not be able to sound every
> bit as good as *.wav samples (as used in Grand Orgue) if recorded 'wet',
> but I and others were previously chastised as 'absurd' for comparing
> Fluidsynth sound with GrandOrgue, so I'm trying to clarify wherein lies
> the absurdity of what seems to me a more than normal comparison.

It should be 'FluidSynth Sound to GrandOrgue sound'. A soundfont
contains wave samples (maybe not as big and elaborate) just as GO. I can
export a sample from the soundfont, and it is a true PCM wav. I can
import into a soundfont any wave file, and it's still a wave file.

The difference here is the technology. The soundfont standard was
conceived when 512k ram was a lot of power. As such, there was a size
limit to how big a single soundfont could be. For the old SoundBlaster
Live cards, the limit was around 255 megs.

The Go setup has the advantage of better sample 'construction', such as
multiple loop points for clean releases, etc. And, the size limit is how
much memory and processor power one has.

> From your description, though, it sounds like there are usually fewer
> actual samples in SF2's than in a normal one-per-note GO sample set, and
> the interpolation perhaps is where the artifact is coming in.

The builder of the soundfont has the ability to use a unique sample per
note if they wish, within the size limits of the soundfont. When dealing
with the organs I work with (for) this isn't an issue. For an organ with
a few thousand pipes, it can easily become an issue of size, unless you
brake the soundfont into smaller pieces, each assigned to it's correct
place.

> BTW, it's nice to have a Creative Card (I do have in a box somewhere),
> but not everyone chooses to limit themselves to compliant hardware nor
> has the opportunity to open up a laptop and insert one.

This is true, that's why there's FluidSynth and it's brothers, to play a
soundfont (or other format) on a non Creative sound system. For my
needs, FluidSynth has an issue dealing with the bottom 13 or so notes in
the scale (staring at midi note 0), so I can't use it (I work down there).

> The question is: What advantage does Fluid Synth have over a Wave file
> player (certainly there must be a free and open source version of
> something like Kontakt or similar as it seems there is even less
> intensive processing going on in serving up wav files than doing
> additive synthesis etc, right?

FluidSynth is one of the easier programs to 'attach' to jOrgan. It's
been around a long time, and it's occasionally improved. That doesn't
mean it will sound spectacular, just that it works with little effort.

For a little more effort, Panos has figured out how just about any of
the alternatives work.

'Additive Synthesis'- just think Hammond drawbars...

Rich

p.s. GO/HW are great programs, until you want to change or add things.

p.s.s. I'm playing the midi files supplied with the Earlwood download,
through the big surround system. The E3b is a great sounding organ...

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies - was: FluidSynth Sound

DellAnderson
On 3/3/2012 7:09 PM, R. Olsen wrote:
> Geez Dell, you're making me think about things I haven't had to since
> the 90's (hehe)...

hehehe... Always trying to help!   ;-)

> It should be 'FluidSynth Sound to GrandOrgue sound'.

I guess I am missing the distinction somehow.  Loss of information is
going to affect the sound quality right?  Even MP3's don't sound as good
as as a CD or even lossless compression FLAC codec for example.

> A soundfont contains wave samples (maybe not as big and elaborate)
> just as GO.

That may be the problem - too simplified for the complexity needed for
realistic pipe sound.

> I can export a sample from the soundfont, and it is a true PCM wav. I
> can import into a soundfont any wave file, and it's still a wave file.
> The difference here is the technology. The soundfont standard was
> conceived when 512k ram was a lot of power. As such, there was a size
> limit to how big a single soundfont could be. For the old SoundBlaster
> Live cards, the limit was around 255 megs.

Even that large size is only about 1/2 of what a good GO sample set
runs. But with FluidSynth, wouldn't this limit go away since it is
emulating hardware in software (or that's my guess as to what is going
on)...


> FluidSynth has an issue dealing with the bottom 13 or so notes in the
> scale (staring at midi note 0), so I can't use it (I work down there).

What kind of issue does you have with the bottom 13 notes?  Audible
artifacts?  Note number limitations?

> FluidSynth is one of the easier programs to 'attach' to jOrgan. It's
> been around a long time, and it's occasionally improved. That doesn't
> mean it will sound spectacular, just that it works with little effort.
>
> For a little more effort, Panos has figured out how just about any of
> the alternatives work.

Yeah, Panos definitely has it figured out alright, but it's a steep
learning curve playing catchup with someone of his musical experience.  
My goal short term is to find some solid ground to build on.   In any
case, your post inspired me to search for the SF2 standard.  It wasn't
hard to find and came conveniently posted along with multiple other
documents and applications:

http://connect.creativelabs.com/developer/SoundFont/Forms/AllItems.aspx

> 'Additive Synthesis'- just think Hammond drawbars...
>
Maybe that's why I'm not a huge fan of additive synthesis!   Audio
compromises were made in Hammonds and it results in a sound some love,
others don't.   ;-)

> p.s. GO/HW are great programs, until you want to change or add things.
>
I was reading the GO help file the other day and it does look like it's
customizable, but not so much within the program itself, more by editing
the XML description.

> p.s.s. I'm playing the midi files supplied with the Earlwood download,
> through the big surround system. The E3b is a great sounding organ...
>
I have downloaded the E3b  and plan to give it more testing.  I think
there may be a difference in the way I set it up to play though speakers
and how it might sound via headphones, so I will explore that a bit
further before ruling it out completely.   I recently discovered that my
Allen's amp system might be doing uncomplimentary things to my computer
audio output and I find that headphones in general are much easier on my
ears.  But that said, the setup I tried had several stops that sounded
similar to older analog synthesis organs (with their modified sawtooth
and square waves) to my ear, which is not always bad, but simply could
not be compared with the real pipe sounds I get with the Pitea School GO
sample set for example.

I'm going to poke and prod the SF2 standard a bit more before throwing
it out with the bathwater, but my initial impression is that anything
designed for 512k RAM is unlikely to meet my audiophile needs long term.

best,

Dell

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies

MasterPerforator
> I guess I am missing the distinction somehow.  Loss of information is
> going to affect the sound quality right?  Even MP3's don't sound as good
> as as a CD or even lossless compression FLAC codec for example.

FluidSynth is a sound engine that plays the soundfont. Not sure what GO
uses as a sound engine, but it's obviously better.

>>A soundfont contains wave samples (maybe not as big and elaborate)
>>> just as GO.
>
>
> That may be the problem - too simplified for the complexity needed for
> realistic pipe sound.
>
It really depends on how big things are. The soundfont supports stereo
(distinct left-right) samples, but these take up a lot of real estate
within the soundfont.

> Even that large size is only about 1/2 of what a good GO sample set
> runs. But with FluidSynth, wouldn't this limit go away since it is
> emulating hardware in software (or that's my guess as to what is going
> on)...

The soundfont itself has a practical size limit, which has nothing to do
with FluidSynth. Later formats such as GIG or SFZ got around this limit.
But, GIG is dead, and soundfonts live on. You can get a lot of organ
into a soundfont, if you know how.

>>FluidSynth has an issue dealing with the bottom 13 or so notes in the
>>> scale (staring at midi note 0), so I can't use it (I work down there).
>
>
> What kind of issue does you have with the bottom 13 notes?  Audible
> artifacts?  Note number limitations?

This is a little hard to explain. As you go down the scale, the pitches
of sounds in those positions need to be upped by two or three steps. If
I remember correctly, medi note #0 has to be tunes up by 25 or so whole
steps. I don't wish to do this with multiple samples sets, where
different samples use those notes. My 'scale' starts at midi note 0 or
1, and goes straight up the notes with no gaps. The target organ may
have only all white keys, but the scale has no gaps.

> your post inspired me to search for the SF2 standard.  It wasn't
> hard to find and came conveniently posted along with multiple other
> documents and applications:

I went and reread some of that stuff, it's been a long time since I
reviewed it. The soundfont is what it is, a method of getting high(er)
quality sound with the ability to adjust it within a reasonable file size.

>>'Additive Synthesis'- just think Hammond drawbars...
>>>
>
> Maybe that's why I'm not a huge fan of additive synthesis!   Audio
> compromises were made in Hammonds and it results in a sound some love,
> others don't.   ;-)
>
The problem here is not having 8 more drawbars between (e.g.) 8 and 4...

> I was reading the GO help file the other day and it does look like it's
> customizable, but not so much within the program itself, more by editing
> the XML description.

But you can't add (for instance) a Quint into the mix.

> I think
> there may be a difference in the way I set it up to play though speakers
> and how it might sound via headphones, so I will explore that a bit
> further before ruling it out completely.   I recently discovered that my
> Allen's amp system might be doing uncomplimentary things to my computer
> audio output and I find that headphones in general are much easier on my
> ears.

This is the difference between PRODUCING sound, and REPRODUCING sound.
You will get much better results by using an audiophile type stereo amp
with good audio speakers. The Allen amps/speakers are designed like
'musical instrument' style equipment, meant to 'produce' the sound. I
have two Allen S100 amps, one labeled 'Main', the other 'flutes'. There
is a resister in line with the input that acts as a pass filter, and the
values are different between the amps. They kick butt when it comes to
bass though.

> But that said, the setup I tried had several stops that sounded
> similar to older analog synthesis organs (with their modified sawtooth
> and square waves) to my ear, which is not always bad, but simply could
> not be compared with the real pipe sounds I get with the Pitea School GO
> sample set for example.

Samples of real pipes, when recorded well, will always sound more 'real'
than a synthesized sound. But sometimes a sound must be synthesized, or
you may never have that sound at all.

Try the Pitea School for jOrgan disposition, it's very close.

> I'm going to poke and prod the SF2 standard a bit more before throwing
> it out with the bathwater, but my initial impression is that anything
> designed for 512k RAM is unlikely to meet my audiophile needs long term.

It's the player medium, not the soundfont that is the ultimate issue
here. If it sounds good in headphones (designed to reproduce sound), try
other speaker/amp combinations.

And put the speakers in another room (like the bathroom), then crank it
up...

Rich

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies

MasterPerforator
In reply to this post by DellAnderson
Just for grins and giggles, and so you can see what I'm up against, here
are the scale charts (with midi note positions) for the smallest and
largest organs I'm currently arranging for. The soundfont layout allows
me to set up samples to directly match these note positions, and jOrgan
allows automagic stop control (when applicable, the smallest scale has
no auto stops) in real time, directly from the appropriate control
positions. A file in this layout can be directly punched into rolls/books.

Without the flexibility within the soundfont, I'd have a little problem,
unless I arrange in gm and convert it, which is a pain for this stuff.
So, you see why I'm a big proponent of soundfonts. A lot of sound for a
little work.

The small organ starts on midi note 0, the large one on midi note 1...

Rich
-------------------------------------------------------------
[Wurlitzer 125 Tracker] Paper Roll

0=1 Shut Off
1=2 Snare Drum
2=3 Bass G
3=4 Bass C
4=5 Bass D
5=6 Bass E
6=7 Bass F
7=8 Accomp G
8=9 Accomp A
9=10 Accomp B
10=11 Accomp C
11=12 Accomp C#
12=13 Accomp D
13=14 Accomp E
14=15 Accomp F
15=16 Accomp F#
16=17 Melody G (A#)
17=18 Melody A (G)
18=19 Melody A# (A)
19=20 Melody B
20=21 Melody C
21=22 Melody C#
22=23 Melody D
23=24 Melody E
24=25 Melody F
25=26 Melody F#
26=27 Melody G
27=28 Melody A
28=29 Melody B
29=30 Melody C
30=31 Trumpet E
31=32 Trumpet F
32=33 Trumpet F#
33=34 Trumpet G
34=35 Trumpet A
35=36 Trumpet B
36=37 Trumpet C
37=38 Trumpet C#
38=39 Trumpet D
39=40 Trumpet E
40=41 Trumpet F
41=42 Trumpet F#
42=43 Trumpet G
43=44 Bass Drum
44=45 Rewind

-------------------------------------------------------------
[81 key Marenghi] Folded Cardboard Book

0=0 blank
1=1 Shut Off
2=2 Mel Vox ON
3=3 Triangle
4=4 C/M Unda Maris ON
5=5 Crash Cymbal
6=6 Snare Drum R
7=7 Woodblock
8=8 Snare Drum L
9=9 Bass F
10=10 Bass G
11=11 Tambourine
12=12 Bass Trombone ON
13=13 Bass A
14=14 Bass A#
15=15 Bass B
16=16 Bass C
17=17 Mel Xylophone ON
18=18 C/m Sax ON
19=19 Bass D
20=20 Bass E
21=21 Accomp F
22=22 Accomp F#
23=23 Mel Piccolo ON
24=24 Mel Violins ON
25=25 Accomp G
26=26 Accomp A
27=27 Accomp A#
28=28 Accomp B
29=29 Accomp C
30=30 Accomp C#
31=31 Accomp D
32=32 Accomp D#
33=33 Accomp E
34=34 Bass Drum
35=35 Melody F
36=36 Melody F#
37=37 Melody G
38=38 Melody G#
39=39 Melody A
40=40 Melody A#
41=41 Melody B
42=42 Melody C
43=43 Melody C#
44=44 Melody D
45=45 Melody D#
46=46 Melody E
47=47 Melody F
48=48 Melody F#
49=49 Melody G
50=50 Melody G#
51=51 Melody A
52=52 Melody A#
53=53 Melody B
54=54 Melody C
55=55 Melody C#
56=56 Melody D
57=57 Melody D#
58=58 Melody E
59=59 Melody F
60=60 Counter Melody C
61=61 Counter Melody C#
62=62 Counter Melody D
63=63 Counter Melody D#
64=64 Counter Melody E
65=65 Counter Melody F
66=66 Counter Melody F#
67=67 Counter Melody G
68=68 Counter Melody G#
69=69 Counter Melody A
70=70 Counter Melody A#
71=71 Counter Melody B
72=72 Counter Melody C
73=73 Counter Melody C#
74=74 Counter Melody D
75=75 Counter Melody D#
76=76 Counter Melody E
77=77 Counter Melody F
78=78 Counter Melody G
79=79 General Cancel
80=80 Melody Bells ON
81=81 C/M Flute Celeste ON


------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies

DellAnderson
In reply to this post by MasterPerforator
On 3/3/2012 10:30 PM, R. Olsen wrote:
> Samples of real pipes, when recorded well, will always sound more
> 'real' than a synthesized sound. But sometimes a sound must be
> synthesized, or you may never have that sound at all. Try the Pitea
> School for jOrgan disposition, it's very close.

I suppose if one were extremely limited by computer power, yes, but
powerful computers are fairly cheap now (mine was around $500 with 8Gb
Ram, i3 laptop) so there is little excuse to use synthesis for 'real'
pipe organ sound.  Other than that, I believe the day of SF2 is past
(1998 per the specifications).  Furthermore, since I began playing the
Pitea School organ (in the original GrandOrgue version) earlier this
week, I cannot go back to synthesized sound.   I plan to learn
everything I can about GO sound reproduction.  As you know, it is an
open source, so the only thing holding me back is my own mental
limitations.   And this does not rule out use in conjunction with jOrgan
as a customizable front-end, I just have not yet gotten around to
reading up on how to do it.

Whenever I open GO, it spends about 2-3 minutes loading an orderly
sequence of *wav files which appear to be rather unambiguously named
with note number, note name, and ending in suffix  *.wav (e.g
039-D#.wav).  Each *.wav  is filed in a folder indicating division and
stop name (e.g. SVGamba8).   All are filed in a general folder with the
name of the organ (e.g. PiteaMHS).   I don't know how GO 'generates' the
actual sounds, but I presume it first loads all these wave files into
RAM, then uses program pointers from the *.wav (or sample set organ
definition file(ODF) ?) specs to indicate loop points in the individual
loaded wave files for attack, sustain, and release, adds them all
together, and squeezes them out through a D/A converter in a manner
analogous to playing back a CD recording, only with nearly infinite
flexibility.  Apparently the main work of creating a sample set is
generating these precious wave files, a tedious job beyond imagination
where loop points are found where almost none exist in nature, and
blower noise is removed (to prevent cumulative build up of vacuum
cleaner or tornado sound effect as more notes are played!) and attack
and release points defined.  Done well, you have an organ of immense
realism, such as the Pitea School.   Done poorly, you have clicks, pops,
and other artifact.   I don't know how much real time interpolation is
done in general, but I do know that different temperament tunings are
available, so it must have at least limited capabilities of frequency
shift, something I am always a bit nervous about just out of principle,
not from any real first hand experience with the subject.

So...after the rather quick scanning of the SoundFont specification,
with generous mention of LFO and modulation etc, I was reminded more of
the schematics for my 1970's Rodgers organ, only on steroids, with far
more options to tweek (inverse response to modulator wheel etc), but
less like true sampled recordings.   The latter would be my preference
for an ideal 'sound engine':   In other words, I like the idea of having
one CD quality recorded *.wav file for each and every single note,
loopable but with real attack and release codings of equal fidelity
ready to splice on the beginning and ending of each note played
(respectively) with zero crossings and perfect fade in/fade out.   So
far, this seems to be what Grand Orgue is doing, although I admit to
having had some issues with occasional popping at certain driver
settings (possibly due to other windows programs hogging the CPU).

So while the SF2 specification appears to have more options to tweak and
play, they aren't really needed for what I want, which is a fairly
accurate CD quality reproduction of a pipe sound.   From what I have
learned of real pipes, the whole idea of synthesizing them artificially
(either with physical modeling or wavetable synthesis, whatever that may
mean now) would make Cavaille Coll or even Skinner roll over his their
graves.    Perhaps John Reimer has closest to emulating the normal
variability with his famous Earlwood Wobble, but I have yet to hear any
synthesized sound of any kind that successfully competes with high
fidelity recorded audio based sample synthesis.    It's like the
difference between watching ToyStory 3D vs. live action film.   The
realism is in the imperfections, but not random imperfections; real life
imperfections that our subconscious can perceive.   Computers are
getting much better I will grant you, but I have yet to be fooled by
anything artificially generated that didn't at least start out as either
a live photo or digital recording.  Even Hollywood with it's millions
can't do it without photographs or real audio, so I have no hope of
creating verisimilitude from digital scratch.

best,

Dell
PS watched Invictus (Matt Damon movie about Nelson Mendela (played by
Morgan Freeman)) tonight.  Very highly recommended.

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies

DellAnderson
In reply to this post by MasterPerforator
On 3/4/2012 12:03 AM, R. Olsen wrote:
> Just for grins and giggles, and so you can see what I'm up against, ....
Thanks!
> The small organ starts on midi note 0, the large one on midi note 1...
>
>
I can't say that I understood much of it except that one was transposed
from MIDI note by one.  Which reminds me, the SF2 specs seemed to obsess
way too much about what MIDI was, stuff like banks, and so on.   All
that seems like just so much bookkeeping to me.   Important, but best
left to the accountants.   Shouldn't we have unique notes for each key
to play?  Leave the smooth, electronically perfect interpolations to the
Hammonds and instead relish the liveliness of a  pipe organ vibrantly
full principal chorus?

best,

Dell

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies

MasterPerforator
> I can't say that I understood much of it except that one was transposed
> from MIDI note by one.

It shows the actual playing notes available, and their midi note positions.

> Shouldn't we have unique notes for each key
> to play?

I do. Some sets are of the actual target organ, some are put together
from pipe samples from other organs, but all have a unique sample in
every position, and all are a recording of a real pipe.

That's not to say that all my sample sets sound particularely good. The
large organ samples were sent to me at 22050 sample rate, less than ideal...

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies

MasterPerforator
In reply to this post by DellAnderson
> I suppose if one were extremely limited by computer power, yes, but
> powerful computers are fairly cheap now (mine was around $500 with 8Gb
> Ram, i3 laptop) so there is little excuse to use synthesis for 'real'
> pipe organ sound.

Yes, but not always practical. I use a P3 running Win98 for most of my
work. The main reason is that there are a number of programs I use that
require a pure DOS shell. They will not work above 98, no way, no how.
Better minds than I have tried.

>  I believe the day of SF2 is past
> (1998 per the specifications).

But it's still usable. Look at all the players that still support it.

> I plan to learn
> everything I can about GO sound reproduction.

Go to Lars' site. He has a good explanation of how to do things. It's
exactly the same recording methods that one would use to record the
samples for a soundfont. Samples are samples, it's the sample player.

The GO samples are at 16 bit 41000 rate, considered 'cd quality'. A
soundfont can handle the same samples, up to it's physical size limit.

Maybe GrahamG can enlighten us on the sample player side of things, then
again, that may be a trade secret...

> Apparently the main work of creating a sample set is
> generating these precious wave files, a tedious job beyond imagination
> where loop points are found where almost none exist in nature, and
> blower noise is removed (to prevent cumulative build up of vacuum
> cleaner or tornado sound effect as more notes are played!) and attack
> and release points defined.

This also applies to recording samples for soundfonts. You have to be
somewhat of a recording engineer.

> Done well, you have an organ of immense
> realism, such as the Pitea School.   Done poorly, you have clicks, pops,
> and other artifact.

This is solely the responsibility of the sample 'engineer' to clear
these up. The clicks are not always so noticeable. Configuring samples
for any playback format is a real challenge.

> So...after the rather quick scanning of the SoundFont specification,
> with generous mention of LFO and modulation etc, I was reminded more of
> the schematics for my 1970's Rodgers organ, only on steroids, with far
> more options to tweek (inverse response to modulator wheel etc), but
> less like true sampled recordings.   The latter would be my preference
> for an ideal 'sound engine':

A soundfont is not a 'sound engine', it's a collection of real wav
samples. The sound engine is the part that plays back those samples.

> So while the SF2 specification appears to have more options to tweak and
> play, they aren't really needed for what I want, which is a fairly
> accurate CD quality reproduction of a pipe sound.

What you want is a plug-n-play organ with good sound. That's GO. If I
want to construct an organ to my specific needs, I don't think the GO
developers will cater to my personal whims, so jOrgan is my option. It's
up to me what quality my samples are, and how good (or bad) they sound.

My advice is use your Creative card (since you have one) with the
Creative Element, and stop using FluidSynth. You'll get much better
results sound wise, and the reverb is superior in so many ways.

Rich

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies

DellAnderson
On 3/4/2012 1:51 AM, R. Olsen wrote:

> A soundfont is not a 'sound engine', it's a collection of real wav
> samples. The sound engine is the part that plays back those samples.

Perhaps I should have said that the SF2 requires a specific, dedicated
sound engine to reproduce it.  Hey, you cannot even edit SF2's with the
official sound font editor (Vienna) unless you have a SF2 compatible
hardware installed.  Right?   This seems like an artificial limitation,
esp. given the limited availability of such hardware and concerns
regarding FluidSynth audio quality previously expressed.   If SF2 allows
a standard *.wav  41k 16bit audio quality, wouldn't these limitations
all go away since almost any audio card can play those?

>> So while the SF2 specification appears to have more options to tweak and
>> play, they aren't really needed for what I want, which is a fairly
>> accurate CD quality reproduction of a pipe sound.
> What you want is a plug-n-play organ with good sound. That's GO. If I
> want to construct an organ to my specific needs, I don't think the GO
> developers will cater to my personal whims, so jOrgan is my option. It's
> up to me what quality my samples are, and how good (or bad) they sound.

GO is not necessarily plug-n-play.  I had to do some minor parameter
tweaking.   I am thinking more about the flexibility of the SF2 standard
to allow on-the-fly playback modification of the samples etc (if indeed
they are samples -- I am taking your word for this because based on the
SF2 specs that I read, there are VERY few samples (maximum of about 8)
per rank if I understood correctly - the rest are played back at various
frequencies to simulate the in-between notes.   Your mention that you
can have one actual recorded sample for each and every note is good news
to me and wasn't clear from the specifications (which I admittedly only
scanned quickly noting things like LFO's and modulation etc which
sounded more like FM synthesis or something)..

>> I suppose if one were extremely limited by computer power, yes, but
>> powerful computers are fairly cheap now (mine was around $500 with 8Gb
>> Ram, i3 laptop) so there is little excuse to use synthesis for 'real'
>> pipe organ sound.
> Yes, but not always practical. I use a P3 running Win98 for most of my
> work. The main reason is that there are a number of programs I use that
> require a pure DOS shell. They will not work above 98, no way, no how.
> Better minds than I have tried.

> My advice is use your Creative card (since you have one) with the
> Creative Element, and stop using FluidSynth. You'll get much better
> results sound wise, and the reverb is superior in so many ways.

That's OK for someone with a Windows 98 desktop computer.   Won't work
with laptop.  And remember your need for Win 98 applications is a
particularly unique situation for most audio and synthesizer users these
days I would guess (although there are several still using Creative
cards with more modern OS's on this list no doubt).  My main issue with
them is the obsolescence of interfaces.  I have a several old Creative
cards, some ISA, some VESA, etc...Don't even have a motherboard that
they will plug into, and if I did they would still have the memory
limitations you mentioned.  I did order a couple online with I think PCI
interfaces, but then I have the opposite problem than you:  To make them
work, I would have to buy another computer because my current computers
are loaded with stuff that wouldn't work with Win 98, no way, no how.  
I've already purchased a couple external USB sound cards which are quite
excellent, but have their own quirks and idiosyncrasies (Cakewalk
Roland) but I actually had to drive an hour special to pick them up the
closeout outdated inventory of a store which no longer stocks them.  
This gives me the feeling that this is not a viable direction for long
term development or broad user base of  VPO software.    Certainly, VPO
sound is not so complicated that modern computers could not be
programmed to reproduce it right out of the box with the generic, hohum,
sound cards -- most of which are far and above the capabilities that
even the Creative Cards had at one time.   Except for the unique SF2 EMU
processor capabilities that is.    Over and over I have learned in
computers --- stay in the middle of the road!  If you try to go cutting
edge, you bleed eventually (incompatible drivers, etc).   I suppose
playing  a VPO could be considered cutting edge, but I don't think it
has to be, not with current CPU and audio card speeds.


> Go to Lars' site. He has a good explanation of how to do things. It's
> exactly the same recording methods that one would use to record the
> samples for a soundfont. Samples are samples, it's the sample player.
> The GO samples are at 16 bit 41000 rate, considered 'cd quality'. A
> soundfont can handle the same samples, up to it's physical size limit.
> Maybe GrahamG can enlighten us on the sample player side of things,
> then again, that may be a trade secret...

Good idea!  I'll take another look.   Can SF2 play 61 notes/keyboard at
16bit 41000 rate for each stop?

Thanks,

Dell

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies

grahamg
Hi,
Sorry for not taking part in all of this, I will make time to reply
more in detail but currently in bed trying to prevent a chest
infection from getting worse.

>Can SF2 play 61 notes/keyboard at 16bit 41000 rate for each stop?

Yes. The soundfont format can handle everything that GrandOrgue 0.2
could give you.... BUT one GO sample becomes four samples in the SF2
file.

The SF2 specification can only handle MONO samples, so in order to get
'stereo' your Stereo wave is split onto two mono samples, the left
channel panned to the left and the right channel panned to the right.
The release portion also becomes two samples, and these have to have
silent loops added to the beginning portions and a fade in.

So your 61 wave files become 244 samples.

Polyphony quickly becomes a big issue ...

GrahamG

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Graham Goode
Location: Johannesburg, South Africa
VPOs with jOrgan, LinuxSampler, Fluidsynth, SFZ, GrandOrgue, NI Kontakt, and Hauptwerk
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies

DellAnderson
Thank you GrahamG, although my question was definitely not worth your getting out of bed to answer.  Please take good care, I don't want to be responsible for extending the illness of the illustrious GrahamG!

I would have responded earlier except that I just returned from an Isabelle Demers concert.  I wish you all could have been there - quite an interesting collection of pieces.

best,

Dell
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies

Brian Sweetnam-2
In reply to this post by MasterPerforator
Get wel soon!

Brian




Sent from Samsung tablet



Graham Goode <[hidden email]> wrote:


Hi,
Sorry for not taking part in all of this, I will make time to reply
more in detail but currently in bed trying to prevent a chest
infection from getting worse.

>Can SF2 play 61 notes/keyboard at 16bit 41000 rate for each stop?

Yes. The soundfont format can handle everything that GrandOrgue 0.2
could give you.... BUT one GO sample becomes four samples in the SF2
file.

The SF2 specification can only handle MONO samples, so in order to get
'stereo' your Stereo wave is split onto two mono samples, the left
channel panned to the left and the right channel panned to the right.
The release portion also becomes two samples, and these have to have
silent loops added to the beginning portions and a fade in.

So your 61 wave files become 244 samples.

Polyphony quickly becomes a big issue ...

GrahamG

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound


------------------------------------------------------------------------------
Try before you buy = See our experts in action!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-dev2
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies

MasterPerforator
In reply to this post by DellAnderson
Hello GrahamG, thank you for chiming in. Get better fast.

Well Dell, I suppose we'll have to agree to disagree on the viability of
the lowly soundfont. It's sort of like vinyl, or reel tapes, they never
really go way. They work for me, your milage may vary.

>>My advice is use your Creative card (since you have one) with the
>>> Creative Element, and stop using FluidSynth. You'll get much better
>>> results sound wise, and the reverb is superior in so many ways.
>
>
> That's OK for someone with a Windows 98 desktop computer.   Won't work
> with laptop.  And remember your need for Win 98 applications is a
> particularly unique situation for most audio and synthesizer users these
> days I would guess (although there are several still using Creative
> cards with more modern OS's on this list no doubt).

I should be clear, I have more than one pc, and all have a Creative card
of some sort. The 'arranging' pc is set up to run all the old programs I
need to do my work. The 'internet' pc can run 98 or XP, it used to be my
main working setup. The 'recording' (formally 'organ') pc has  a
multitrack card in it for my studio, plus an Audigy 2, and runs XP only.
The laptop is the new 'organ' pc, and can also run enough old programs
for me to do most of my work on it if needed.

That's all, back to work for me...

Rich

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
jorgan-sound mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/jorgan-sound
Reply | Threaded
Open this post in threaded view
|

Re: Sound Font audio quality for Newbies

DellAnderson
MasterPerforator wrote
...Well Dell, I suppose we'll have to agree to disagree on the viability of
the lowly soundfont. It's sort of like vinyl, or reel tapes, they never
really go way. They work for me, your milage may vary.
..
Not sure what you are referring to when you say we disagree about the viability of sound fonts.  I'm mostly interested in determining the possible parameters (which GrahamG clarified) and hardware requirements (which you have clarified).

Nothing to disagree about really.  

best,

Dell