It was stupid of the FluidSynth developers not to make all MIDI
numbers floating-point numbers, but they are far from alone. However,
you can still use alternative tuning systems with FluidSynth, though
it is a bit of a hack. You assign each of your pitches to a MIDI
channel and MIDI note number, then detune that channel and number to
the frequency that you need. You can play any tuning system at all
this way.

The conversions between MIDI key numbers and frequencies are simple
and can be found here: https://newt.phys.unsw.edu.au/jw/notes.html.

Both of these formulas will work with fractional MIDI note numbers.

In Csound, you can use cpsmidinn to obtain frequencies from MIDI key
numbers, including fractions. Here is the Csound code for the opposite
conversion:

opcode hertz2midinn, i, i
ihertz xin
print ihertz
; m = 12*log2(fm/440 Hz) + 69
ilog2 = log(2)
imidinn = 12 * (log(ihertz / 440) / ilog2) + 69
print imidinn
xout imidinn
endop


In my own compositions, I use exclusively MIDI note numbers to
represent pitch, but I sometimes use fractional MIDI key numbers to
obtain different tuning systems. Obviously this is much easier in pure
Csound code. Typically I use a ratio to represent a scale degree, then
I start with a root frequency and add octaves to get the actual note
frequency, which I convert to a fractional MIDI key number and send to
the Csound instrument. This works just fine.


-----------------------------------------------------
Michael Gogins
Irreducible Productions
http://michaelgogins.tumblr.com
Michael dot Gogins at gmail dot com


On Mon, Oct 19, 2015 at 7:50 AM, Peter Burgess
 wrote:
> Hi there, is there any way of playing fluidSynth without using midi note
> numbers? My project involves a whole range of different temperaments, and so
> all my instruments are played using Hz, not midiNN.
>
> Or are there any alternatives in csound (besides the old soundfont opcodes,
> which I will explore if I can't find any other way)? I'm assuming not, I'm
> guessing they would be listen in the quick reference if they were.
>
> I also wondered if I could convert from cps to midinn, but I can't find an
> opcode for that, I guess because it isn't as linear a conversion. That would
> also not be an ideal fix, because I would like to use the soundfonts in
> different temperaments, but if there were such an opcode and I've just
> missed it, then I would go with it until I found a better solution.
>
> Cheers!
>

Or use log2() function?

On Mon, 19 Oct 2015, Michael Gogins wrote:

> ilog2 = log(2)

"Full orchestral pieces" is a bit ambiguous. Do you mean emulations of
the traditional orchestra and its instruments, or music of comparable
complexity and depth but using synthetic textures that do not
necessarily sound like fiddles etc, or what?

FluidSynth is useful but it is not capable of producing textures of
truly orchestral quality and depth. This is an extremely demanding
standard, and most electroacoustic music, computer music, electronic
music also falls short of this standard. If you mean emulating the
traditional orchestra, look to the leading film and game music
composers for how to do this. It usually involves a room (literally)
full of hard disks full of samples of traditional instruments. You can
buy sample sets for this purpose, some of them are pretty good, those
are usually not cheap and take up a lot of storage. They're designed
to be used from Sibelius, Finale, Pro Tools, etc. Believe me these
things sound a lot better than FluidSynth. Duplicating this in Csound
on Windows would not be possible because they use memory-mapped sample
sets, which Csound does not currently support. It would also be a hell
of a lot of work. I think there are VST sample sets that would give
you a real leg up here. But on Windows you would still need a 64 bit
built of Csound, which doesn't exist yet, although it is probably not
far off. OS X would be the way to go because Csound is 64 bits and all
the good commercial music software and sample sets also runs on OS X.
Csound on Linux is also 64 bits.

If you mean synthetic sounds of equivalent depth and quality, among
those who have done best in this department, focusing on
'instrumental' textures, are James Dashow (mostly custom hardware
synths controlled by custom software), Robert Normandeau (mostly
working in Pro Tools with samples and processing, as far as I know),
later Wendy Carlos (hardware synths with custom tuning systems), Barry
Schrader (instruments deliberately not specified), Eric Lyon
(variously Csound, Max, etc.), Paul Lansky (Cmix, software synthesizer
written by him, same sort of beast as Csound but you have to write
instruments in C++).. I am sure there are many others, these are just
some of the ones I have listened to most.

The reason so many of these use hardware not software is partly a
historical accident, capable hardware came before capable software,
partly that the hardware usually comes with pretty good patches as a
starting point, and partly that to experienced ears hardware
oscillators and filters literally sound a bit different -- think
"snap" instead of "click."

No software synthesizer would be any better than Csound in this
department. It would be a lengthy discussion, how best to do this,
starting with what you really mean -- do you have some example pieces,
from you or anyone else, that would represent what you are aiming for?

Regards,
Mike

-----------------------------------------------------
Michael Gogins
Irreducible Productions
http://michaelgogins.tumblr.com
Michael dot Gogins at gmail dot com


On Mon, Oct 19, 2015 at 8:52 AM, jpff  wrote:
> Or use log2() function?
>
>
> On Mon, 19 Oct 2015, Michael Gogins wrote:
>
>> ilog2 = log(2)

OK, much clearer now.

You will not get realistic emulations of every orchestral instrument
with physical models, but using them would really simplify your
project, so if you just need placeholders use the STK opcodes,
physical models, and other Csound instruments.

I have not researched whether sample sets have good support for
alternative tunings, but that might be worth researching.

You are aware that Csound can host VST plugins (and, on OS X and
Linux, other kinds of plugins)? But this would problematic on mobile
devices.

Regards,
Mike

-----------------------------------------------------
Michael Gogins
Irreducible Productions
http://michaelgogins.tumblr.com
Michael dot Gogins at gmail dot com


On Mon, Oct 19, 2015 at 10:55 AM, Peter Burgess
 wrote:
> Cheers mike for that truly thorough answer! You're pretty good at answering
> questions I didn't realise I'd asked but clearly need answers to all the
> same ;) I do really appreciate you taking the time to give such rich
> explanations.
>
> I'll start with my "orchestral" goals. To put a quick end to my ambiguity,
> I'm going to give a rough figure of 10 instruments. After some thought,
> there probably isn't much need for me to stretch far beyond 10 instruments
> to make my orchestral pieces. This would allow me to do microtonal tunings
> with up to 25 divides per octave with 10 instruments over the 256 channels
> on one fluidEngine. I guess my only question then is how demanding this
> would be on memory and cpu?
>
> As for the quality of the instruments, I hope to achieve relatively
> realistic emulations eventually, however for the time being, I am happy to
> work with place-holders as there is alot of work to be done on the c++ side
> of things yet, so I really just need to hear the instruments working rather
> than sounding great. Hopefully, I am eventually going to be in a position to
> purchase things like good soundfonts, sample packs and extra software for
> this project if needed. In fairness, I haven't tried any of the physical
> modelling opcodes out yet, maybe I should start with them as place-holders
> instead? I had hoped that soundfonts would be the final answer, which is why
> I've gone straight for them, but maybe I was being overly optimistic.
>
> Next let's talk envirenment. I am building my app in linux, but I want my
> app to be available to all major desktop and mobile platforms. The mobile
> platforms are particularly important for my project. It's worth mentioning
> that all audio in csound is currently happening behind the scenes.
> Currently, it is all being exported to wav and then the wav is played in
> SDL_mixer for "realtime" playback. There is no user control of the
> instruments, only c++ control. This buys me some extra headroom in terms of
> performance on mobile devices, but as for storage space for samples, I will
> need to be careful.
>
> Thanks again for your help,
>
> Pete
>
> On Mon, Oct 19, 2015 at 2:52 PM, Michael Gogins 
> wrote:
>>
>> "Full orchestral pieces" is a bit ambiguous. Do you mean emulations of
>> the traditional orchestra and its instruments, or music of comparable
>> complexity and depth but using synthetic textures that do not
>> necessarily sound like fiddles etc, or what?
>>
>> FluidSynth is useful but it is not capable of producing textures of
>> truly orchestral quality and depth. This is an extremely demanding
>> standard, and most electroacoustic music, computer music, electronic
>> music also falls short of this standard. If you mean emulating the
>> traditional orchestra, look to the leading film and game music
>> composers for how to do this. It usually involves a room (literally)
>> full of hard disks full of samples of traditional instruments. You can
>> buy sample sets for this purpose, some of them are pretty good, those
>> are usually not cheap and take up a lot of storage. They're designed
>> to be used from Sibelius, Finale, Pro Tools, etc. Believe me these
>> things sound a lot better than FluidSynth. Duplicating this in Csound
>> on Windows would not be possible because they use memory-mapped sample
>> sets, which Csound does not currently support. It would also be a hell
>> of a lot of work. I think there are VST sample sets that would give
>> you a real leg up here. But on Windows you would still need a 64 bit
>> built of Csound, which doesn't exist yet, although it is probably not
>> far off. OS X would be the way to go because Csound is 64 bits and all
>> the good commercial music software and sample sets also runs on OS X.
>> Csound on Linux is also 64 bits.
>>
>> If you mean synthetic sounds of equivalent depth and quality, among
>> those who have done best in this department, focusing on
>> 'instrumental' textures, are James Dashow (mostly custom hardware
>> synths controlled by custom software), Robert Normandeau (mostly
>> working in Pro Tools with samples and processing, as far as I know),
>> later Wendy Carlos (hardware synths with custom tuning systems), Barry
>> Schrader (instruments deliberately not specified), Eric Lyon
>> (variously Csound, Max, etc.), Paul Lansky (Cmix, software synthesizer
>> written by him, same sort of beast as Csound but you have to write
>> instruments in C++).. I am sure there are many others, these are just
>> some of the ones I have listened to most.
>>
>> The reason so many of these use hardware not software is partly a
>> historical accident, capable hardware came before capable software,
>> partly that the hardware usually comes with pretty good patches as a
>> starting point, and partly that to experienced ears hardware
>> oscillators and filters literally sound a bit different -- think
>> "snap" instead of "click."
>>
>> No software synthesizer would be any better than Csound in this
>> department. It would be a lengthy discussion, how best to do this,
>> starting with what you really mean -- do you have some example pieces,
>> from you or anyone else, that would represent what you are aiming for?
>>
>> Regards,
>> Mike
>>
>> -----------------------------------------------------
>> Michael Gogins
>> Irreducible Productions
>> http://michaelgogins.tumblr.com
>> Michael dot Gogins at gmail dot com
>>
>>
>> On Mon, Oct 19, 2015 at 8:52 AM, jpff  wrote:
>> > Or use log2() function?
>> >
>> >
>> > On Mon, 19 Oct 2015, Michael Gogins wrote:
>> >
>> >> ilog2 = log(2)
>> >> imidinn = 12 * (log(ihertz / 440) / ilog2) + 69
>

Problematic in that as far as I know, mobile software does not support
loading VST plugins. Perhaps I'm wrong. There are enough resources on
high-end devices.

The plugins themselves would have to be compiled for Android or iOS,
of course. As far as I know this is not being done.

Regards,
Mike

-----------------------------------------------------
Michael Gogins
Irreducible Productions
http://michaelgogins.tumblr.com
Michael dot Gogins at gmail dot com


On Mon, Oct 19, 2015 at 11:50 AM, Peter Burgess
 wrote:
> Hmm, ok, that's definitely worth thinking about! Is that problematic from a
> resource perspective or a compatability perspective?
>
> On Mon, Oct 19, 2015 at 4:29 PM, Michael Gogins 
> wrote:
>>
>> OK, much clearer now.
>>
>> You will not get realistic emulations of every orchestral instrument
>> with physical models, but using them would really simplify your
>> project, so if you just need placeholders use the STK opcodes,
>> physical models, and other Csound instruments.
>>
>> I have not researched whether sample sets have good support for
>> alternative tunings, but that might be worth researching.
>>
>> You are aware that Csound can host VST plugins (and, on OS X and
>> Linux, other kinds of plugins)? But this would problematic on mobile
>> devices.
>>
>> Regards,
>> Mike
>>
>> -----------------------------------------------------
>> Michael Gogins
>> Irreducible Productions
>> http://michaelgogins.tumblr.com
>> Michael dot Gogins at gmail dot com
>>
>>
>> On Mon, Oct 19, 2015 at 10:55 AM, Peter Burgess
>>  wrote:
>> > Cheers mike for that truly thorough answer! You're pretty good at
>> > answering
>> > questions I didn't realise I'd asked but clearly need answers to all the
>> > same ;) I do really appreciate you taking the time to give such rich
>> > explanations.
>> >
>> > I'll start with my "orchestral" goals. To put a quick end to my
>> > ambiguity,
>> > I'm going to give a rough figure of 10 instruments. After some thought,
>> > there probably isn't much need for me to stretch far beyond 10
>> > instruments
>> > to make my orchestral pieces. This would allow me to do microtonal
>> > tunings
>> > with up to 25 divides per octave with 10 instruments over the 256
>> > channels
>> > on one fluidEngine. I guess my only question then is how demanding this
>> > would be on memory and cpu?
>> >
>> > As for the quality of the instruments, I hope to achieve relatively
>> > realistic emulations eventually, however for the time being, I am happy
>> > to
>> > work with place-holders as there is alot of work to be done on the c++
>> > side
>> > of things yet, so I really just need to hear the instruments working
>> > rather
>> > than sounding great. Hopefully, I am eventually going to be in a
>> > position to
>> > purchase things like good soundfonts, sample packs and extra software
>> > for
>> > this project if needed. In fairness, I haven't tried any of the physical
>> > modelling opcodes out yet, maybe I should start with them as
>> > place-holders
>> > instead? I had hoped that soundfonts would be the final answer, which is
>> > why
>> > I've gone straight for them, but maybe I was being overly optimistic.
>> >
>> > Next let's talk envirenment. I am building my app in linux, but I want
>> > my
>> > app to be available to all major desktop and mobile platforms. The
>> > mobile
>> > platforms are particularly important for my project. It's worth
>> > mentioning
>> > that all audio in csound is currently happening behind the scenes.
>> > Currently, it is all being exported to wav and then the wav is played in
>> > SDL_mixer for "realtime" playback. There is no user control of the
>> > instruments, only c++ control. This buys me some extra headroom in terms
>> > of
>> > performance on mobile devices, but as for storage space for samples, I
>> > will
>> > need to be careful.
>> >
>> > Thanks again for your help,
>> >
>> > Pete
>> >
>> > On Mon, Oct 19, 2015 at 2:52 PM, Michael Gogins
>> > 
>> > wrote:
>> >>
>> >> "Full orchestral pieces" is a bit ambiguous. Do you mean emulations of
>> >> the traditional orchestra and its instruments, or music of comparable
>> >> complexity and depth but using synthetic textures that do not
>> >> necessarily sound like fiddles etc, or what?
>> >>
>> >> FluidSynth is useful but it is not capable of producing textures of
>> >> truly orchestral quality and depth. This is an extremely demanding
>> >> standard, and most electroacoustic music, computer music, electronic
>> >> music also falls short of this standard. If you mean emulating the
>> >> traditional orchestra, look to the leading film and game music
>> >> composers for how to do this. It usually involves a room (literally)
>> >> full of hard disks full of samples of traditional instruments. You can
>> >> buy sample sets for this purpose, some of them are pretty good, those
>> >> are usually not cheap and take up a lot of storage. They're designed
>> >> to be used from Sibelius, Finale, Pro Tools, etc. Believe me these
>> >> things sound a lot better than FluidSynth. Duplicating this in Csound
>> >> on Windows would not be possible because they use memory-mapped sample
>> >> sets, which Csound does not currently support. It would also be a hell
>> >> of a lot of work. I think there are VST sample sets that would give
>> >> you a real leg up here. But on Windows you would still need a 64 bit
>> >> built of Csound, which doesn't exist yet, although it is probably not
>> >> far off. OS X would be the way to go because Csound is 64 bits and all
>> >> the good commercial music software and sample sets also runs on OS X.
>> >> Csound on Linux is also 64 bits.
>> >>
>> >> If you mean synthetic sounds of equivalent depth and quality, among
>> >> those who have done best in this department, focusing on
>> >> 'instrumental' textures, are James Dashow (mostly custom hardware
>> >> synths controlled by custom software), Robert Normandeau (mostly
>> >> working in Pro Tools with samples and processing, as far as I know),
>> >> later Wendy Carlos (hardware synths with custom tuning systems), Barry
>> >> Schrader (instruments deliberately not specified), Eric Lyon
>> >> (variously Csound, Max, etc.), Paul Lansky (Cmix, software synthesizer
>> >> written by him, same sort of beast as Csound but you have to write
>> >> instruments in C++).. I am sure there are many others, these are just
>> >> some of the ones I have listened to most.
>> >>
>> >> The reason so many of these use hardware not software is partly a
>> >> historical accident, capable hardware came before capable software,
>> >> partly that the hardware usually comes with pretty good patches as a
>> >> starting point, and partly that to experienced ears hardware
>> >> oscillators and filters literally sound a bit different -- think
>> >> "snap" instead of "click."
>> >>
>> >> No software synthesizer would be any better than Csound in this
>> >> department. It would be a lengthy discussion, how best to do this,
>> >> starting with what you really mean -- do you have some example pieces,
>> >> from you or anyone else, that would represent what you are aiming for?
>> >>
>> >> Regards,
>> >> Mike
>> >>
>> >> -----------------------------------------------------
>> >> Michael Gogins
>> >> Irreducible Productions
>> >> http://michaelgogins.tumblr.com
>> >> Michael dot Gogins at gmail dot com
>> >>
>> >>
>> >> On Mon, Oct 19, 2015 at 8:52 AM, jpff  wrote:
>> >> > Or use log2() function?
>> >> >
>> >> >
>> >> > On Mon, 19 Oct 2015, Michael Gogins wrote:
>> >> >
>> >> >> ilog2 = log(2)
>> >> >> imidinn = 12 * (log(ihertz / 440) / ilog2) + 69
>> >
>> >
>