You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Linux binaries and instructions. Thanks, Oman395!
|1 year ago|
|LICENSE.txt||2 years ago|
|Makefile||5 years ago|
|README.txt||1 year ago|
|miditones.c||2 years ago|
|miditones.exe||1 year ago|
|miditones_linux||1 year ago|
|miditones_scroll.c||2 years ago|
|miditones_scroll.exe||1 year ago|
|miditones_scroll_example.txt||5 years ago|
|miditones_scroll_linux||1 year ago|
MIDITONES: Convert a MIDI file into a simple bytestream of notes
MIDITONES compiles a MIDI music file into a much simplified compact time-ordered stream of
commands, so that the music can easily be played on a small microcontroller-based synthesizer
that has only simple tone generators. This is on github at www.github.com/LenShustek/miditones.
Volume ("velocity") and instrument information in the MIDI file can either be
discarded or kept. All the tracks are processed and merged into a single time-ordered
stream of "note on", "note off", "change instrument" and "delay" commands.
MIDITONES was written for the "Playtune" series of Arduino and Teensy
microcontroller software synthesizers:
This original version of Playtune, first released in 2011, uses a separate hardware timer
for each note to generate a square wave on an output pin. All the pins are then combined
with a simple resistor network connected to a speaker and/or amplifier. It can only play
as many simutaneous notes as there are timers. There is no volume modulation.
This second vesion uses only one hardware timer that interrupts periodically at a fast
rate, and toggles square waves onto any number of digital output pins. It also implements
primitive volume modulation by changing the duty cycle of the square wave. The number of
simultaneous notes is limited only by the number of output pins and the speed of the processor.
The third version also uses only one hardware timer interrupting frequently, but
uses the hardware digital-to-analog converter on high-performance microntrollers like
the Teensy to generate an analog wave that is the sum of stored samples of sounds for
many different instruments. The samples are scaled to the right frequency and volume,
and any number of instrument samples can be used and mapped to MIDI patches. The sound
quality is much better, although not in league with real synthesizers. It currently
only supports Teensy boards.
The fourth version is an audio object for the PJRC Audio Library.
It allows up to 16 simultaneous sound generators that are internally mixed, at
the appropriate volume, to produce one monophonic audio stream.
Sounds are created from sampled one-cycle waveforms for any number of instruments,
each with its own attack-hold-decay-sustain-release envelope. Percussion sounds
(from MIDI channel 10) are generated from longer sampled waveforms of a complete
instrument strike. Each generator's volume is independently adjusted according to
the MIDI velocity of the note being played before all channels are mixed.
The fifth version is for the Teensy 3.1/3.2, and uses the four Periodic Interval
Timers in the Cortex M4 processor to support up to 4 simultaneous notes.
It uses less CPU time than the polling version, but is limited to 4 notes at a time.
(This was written to experiment with multi-channel multi-Tesla Coil music playing,
where I use Flexible Timer Module FTM0 for generating precise one-shot pulses.
But I ultimately switched to the polling version to play more simultaneous notes.)
This is a much simplified version that fits, with a small song, into an ATtiny
processor with only 4K of flash memory. It also using polling with only one timer,
and avoids multiplication or division at runtime for speed. It was written
for the Evil Mad Scientist menorah kit:
(Imagine what you can do with the $1 8-pin ATtiny85 with a whopping 8K!)
MIDITONES may also prove useful for other simple music synthesizers. There are
various forks of this code, and of the Playtune players, on Githib.
*** THE PROGRAM
MIDITONES is written in standard ANSI C and is meant to be executed from the
command line. There is no GUI interface.
The output can be either a C-language source code fragment that initializes an
array with the command bytestream, or a binary file with the bytestream itself.
The MIDI file format is complicated, and this has not been tested on all of its
variations. In particular we have tested only format type "1", which seems
to be what most of them are. Let me know if you find MIDI files that it
won't digest and I'll see if I can fix it.
There is a companion program in the same repository called Miditones_scroll
that can convert the bytestream generated by MIDITONES into a piano-player
like listing for debugging or annotation. See the documentation near the
top of its source code.
Additional binaries have been compiled and tested on arch linux, with the files
having an additional _linux appended to their name; please contact @Oman395 on
github if your distrobution has issues running them.
*** THE COMMAND LINE
To convert a MIDI file called "chopin.mid" into a command bytestream, execute
Or on linux, execute
It will create a file in the same directory called "chopin.c" which contains
the C-language statement to intiialize an array called "score" with the bytestream.
The general form for command line execution is this:
miditones <options> <basefilename>
Or on linux:
./miditones_linux <options> <basefilename>
The <basefilename> is the base name, without an extension, for the input and
output files. It can contain directory path information, or not.
If the user specifies the full .mid filename, the .mid or .MID extension
will be dropped and the remaining name will be used as <basefilename>.
The input file is <basefilename>.mid, and the output filename(s)
are the base file name with .c, .h, .bin, and/or .log extensions.
The following commonly-used command-line options can be specified:
-v Add velocity (volume) information to the output bytestream
-i Add instrument change commands to the output bytestream
-pt Translate notes in the MIDI percussion track to note numbers 128..255
and assign them to a tone generator as usual.
-d Generate a self-describing file header that says which optional bytestream
fields are present. This is highly recommended if you are using later
Playtune players that can check the header to know what data to expect.
-b Generate a binary file with the name <basefilename>.bin, instead of a
C-language source file with the name <basefilename>.c.
-t=n Generate the bytestream so that at most "n" tone generators are used.
The default is 6 tone generators, and the maximum is 16. The program
will report how many notes had to be discarded because there weren't
enough tone generators.
The best combination of options to use with the later Playtune music players is:
-v -i -pt -d
The following are lesser-used command-line options:
-c=n Only process the channel numbers whose bits are on in the number "n".
For example, -c3 means "only process channels 0 and 1". In addition to
decimal, "n" can be also specified in hex using a 0x prefix.
-dp Generate Arduino IDE-dependent C code that uses PROGMEM for the bytestream.
-k=n Change the musical key of the output by n chromatic notes.
-k=-12 goes one octave down, -k=12 goes one octave up, etc.
-lp Log input file parsing information to the <basefilename>.log file
-lg Log output bytestream generation information to the <basefilename>.log file
-n=x Put about "x" items on each line of the C file output
-p Only parse the MIDI file, and don't generate an output file.
Tracks are processed sequentially instead of being merged into chronological
order. This is mostly useful for debugging MIDI file parsing problems.
-pi Ignore notes in the MIDI percussion track 9 (also called 10 in some documents)
-r Terminate the output file with a "restart" command instead of a "stop" command.
-sn Use bytestream generation strategy "n". Two are currently implemented:
1:favor track 1 notes instead of all tracks equally
2:try to keep each track to its own tone generator
-h Give command-line help.
-showskipped Display information to the console about each note that had to be
skipped because there weren't enough tone generators.
-noduplicates Remove identical notes played on identical instruments for the same time
that come from different tracks. This can reduce the number of tone
generators needed, and make the file smaller.
-delaymin=x Don't generate delays less than x milliseconds long, to reduce the number
of "delay" commands and thus make the bytestream smaller, at the expense of
moving notes slightly. The deficits are accumulated and eventually used,
so that there is no loss of synchronization in the long term.
The default is 0, which means note timing is exact to the millisecond.
-releasetime=x Stop each note x milliseconds before it is supposed to end. This results
in better sound separation between notes. It might also allow more notes to
be played with fewer tone generators, since there could be fewer
simultaneous notes playing.
-notemin=x The releasetime notwithstanding, don't let the resulting note be reduced
to smaller than x milliseconds. Making releasetime very large and notemin
small results in staccato sounds.
-attacktime=x The high-volume attack phase of a note lasts x milliseconds, after which
the lower-volume sustain phase begins, unless the release time makes it
too short. (Only valid with -v.)
-attacknotemax=x Notes larger than x milliseconds won't have the attack/sustain profile
applied. That allows sustained organ-like pedaling. (Only valid with -v.)
-sustainlevel=p The volume level during the sustain phase is p percent of the starting
note volume. The default is 50. (Only valid with -v.)
-scorename Use <basefilename> as the name of the score in the generated C code
instead of "score", and name the file <basefilename>.h instead of
<basefilenam>.c. That allows multiple scores to be directly #included
into an Arduino .ino file without modification.
Note that for backwards compatibility and easier batch-file processing, the equal sign
for specifying an option's numeric value may be omitted. Also, numeric values may be
specified in hex as 0xhhhh.
*** THE SCORE BYTESTREAM
The generated bytestream is a series of commands that turn notes on and off,
change instruments, and request a delay until the next event time.
Here are the details, with numbers shown in hexadecimal.
If the high-order bit of the byte is 1, then it is one of the following commands,
where the two characters are a hex representation of one byte:
9t nn [vv]
Start playing note nn on tone generator t, replacing any previous note.
Generators are numbered starting with 0. The note numbers are the MIDI
numbers for the chromatic scale, with decimal 69 being Middle A (440 Hz).
If the -v option was given, the third byte specifies the note volume.
8t Stop playing the note on tone generator t.
Ct ii Change tone generator t to play instrument ii from now on. This will
only be generated if the -i option was given.
F0 End of score; stop playing.
E0 End of score, but start playing again from the beginning. This is
generated by the -r option.
If the high-order bit of the byte is 0, it is a command to delay for a while until
the next note change. The other 7 bits and the 8 bits of the following byte are
interpreted as a 15-bit big-endian integer that is the number of milliseconds to
wait before processing the next command. For example,
would cause a delay of 0x07d0 = 2000 decimal millisconds, or 2 seconds. Any tones
that were playing before the delay command will continue to play.
If the -d option is specified, the bytestream begins with a little header that tells
what optional information will be in the data. This makes the file more self-describing,
and allows music players to adapt to different kinds of files. The later Playtune
players do that. The header looks like this:
'Pt' Two ascii characters that signal the presence of the header
nn The length (in one byte) of the entire header, 6..255
ff1 A byte of flag bits, three of which are currently defined:
80 volume information is present
40 instrument change information is present
20 translated percussion notes are present
ff2 Another byte of flags, currently undefined
tt The number (in one byte) of tone generators actually used in this music.
Any subsequent header bytes included in the length are currently undefined
and should be ignored by players.
Len Shustek, 2011 to 2021; see the change log.