See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/316955785

Understanding MIDI: A Painless Tutorial on Midi Format

Article · May 2017

CITATIONS READS

0 7,711

2 authors:

Hélio Magalhães de Oliveira Raimundo Oliveira

Federal University of Pernambuco Universidade do Estado do Amazonas
197 PUBLICATIONS   602 CITATIONS    19 PUBLICATIONS   14 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Finite Field Transforms View project

Research on fast algorithms View project

All content following this page was uploaded by Hélio Magalhães de Oliveira on 18 May 2017.

The user has requested enhancement of the downloaded file.

 on midi files

Understanding MIDI: A Painless Tutorial on Midi Format

H. M. de Oliveira1
Departamento de Estatı́stica, Universidade Federal de Pernambuco, UFPE, Recife, PE
R. C. de Oliveira2
Depto de Engenharia da Computação, Universidade Estadual do Amazonas, UEA, Manaus, AM
arXiv:1705.05322v1 [cs.SD] 15 May 2017

Abstract. A short overview demystifying the midi audio format is presented. The goal is
to explain the file structure and how the instructions are used to produce a music signal,
both in the case of monophonic signals as for polyphonic signals.
keywords. Midi files, Computer Music, Monophonic and Polyphonic.

1 Introduction
Digital music is a broad and fascinating subject [2], [3], [12]. The Musical Instrument
Digital Interface (MIDI) protocol is an industry-standard defined in early 80’s to repre-
sent musical information [7], [9]. Even dealing with an old subject, it does not seem to
be available on the Internet enlightening texts (most shows only tables, commands and
technicalities). It is a the most spread binary protocol for communicating intended to
connect electronic musical instruments such as synthesizers, in computer for the purpose
of recording, editing and programming, and electronic music equipment in the electronic
home studios [10], [6]. MIDI System appeared for the need for standardization of the media
between the synthesizers. This standard was created when electronic music was developed
by a consortium of Japanese and American manufacturers Synthesizers (Sequential Sys-
tems, Roland Corporation, Yamaha, Kurzweil...). You can make free download millions
of popular songs in MIDI versions. It transmits data using serial ports. For many peo-
ple, still remains somewhat nebulous how to assemble a midi file from a music as well as
the interpretation of a file previously generated. Unlike other formats (such as .wav and
.mp3), a MIDI file does not contain audio itself, but the instructions to produce it, i.e. it
is basically a digitized score. That is to say that MIDI is not audio signal: it just con-
tains musical instruction. MIDI messages are comparable with those piano rolls, in which
data control are represented as opposed to audio waveforms. The MIDI protocol allows
sending messages over 16 independent channels, allowing 16 hearing instruments. MIDI
consists of both a simple hardware interface, and a more elaborate transmission protocol.
What MIDI specifies? Here is a list: hardware; driver; communication channels; messages;
modes; controllers; files standard MIDI; visual control (MIDI show control) among other
things. In this note we are just concerned with midi protocol [4], [7].
1
[email protected] also https://arxiv.org/a/deoliveira_h_1.atom
2
[email protected]
2

2 Structure of the MIDI File

Figure 1: Structure of a MIDI file.

Standard MIDI Files contain at the top level things called events. Each event consists
of two components: a MIDI time, and a MIDI message. These time/message pairs follow
each other one in a MIDI file as illustrated in Fig. 1. The time value is a measurement of
the time to wait before playing the next message in the stream of MIDI file data. This
method of specifying the time is called delta time that specifies the duration between
two events (Table 1). MIDI commands and data are distinguished according to the most
significant bit of the byte. If there is a zero in the top bit, then the byte is a data byte,
otherwise the byte is a command byte.

Table 1: Understanding delta times in a sequence of events.

The MIDI messages are sent as a sequence of one or more bytes. The first byte is a com-
mand (STATUS) byte, often followed by data (DATA) bytes with additional parameters.
The command byte determines the type of the command. The number of DATA bytes
that follow depend on the type of the message. For instance, the following two commands
have different number of data bytes: FF 58 04 04 02 30 08 and 90 3C 28. The main com-
mands in midi files are Note-on (0×90) and Note-off (0×80) that allows starting/stopping
playing a single musical note. There are at least two playing modes, namely mono and
poly. Monophonic: the start of a new “note-on” command implies the termination of the
previous note. Polyphonic: multiple notes may be sounding at once, until the notes reach
the end of their decay envelope, or when explicit “note-off” commands are received.
 3

3 About Midi Messages

The main messages are the NOTE ON and NOTE OFF messages. The NOTE ON message
is sent when the performer hits a key of the music keyboard. It contains parameters to
specify the pitch of the note as well as the velocity (i.e. intensity of the note when it is
hit). When a synthesizer receives this message, it starts playing that note with the correct
pitch and force level. When the NOTE OFF message is received, the corresponding note
is switched off by the synthesizer. The channels are numbered 1-16, but their actual
corresponding binary encoding is 0-15. MIDI commands can be further decomposed into
a command type nibble (four bytes) and a channel number nibble (Fig. 2).

Figure 2: Command type nibble (four bytes).

command (hex 0×80) ⇒ channel nibble (1000 0000)

The NOTE ON (NOTE OFF, respectively) message is structured as follows:
note on note off
command byte : 1001 cccc 1000 cccc (MIDI channel, from 0 to 15)
data byte 1 : 0ppp pppp 0ppp pppp (pitch value from 0 to 127)
data byte 2 : 0vvv vvvv 0000 0000 (velocity/intensity value from 0 to 127)
For the midi sequence: 0×90 0×3c 0×28, we have: command: play a note in channel 0
data: note is C4 (pitch, midi code 60=hex 0×3c), its velocity ppp=20 (hex 0×28).

9 0 3 c 2 8
1001 0000 0011 1011 0010 1000

The velocity value normally goes from 1 to 127, covering the range from a practically
inaudible note up to the maximum note level. It basically corresponds to the scale of
nuances found in music notation, as follows. Figure 3 shows different standard note level.
In the beginning of the file, it is quite common to use the following commands:

Figure 3: Parameter (data byte) associated with the note-on command.

midi time signature: FF 58 nn dd cc bb (e.g. FF 58 4 4 128 18)

nn ⇒ compass identification, e.g. for 4/4 set nn=4 dd=4
cc ⇒ number of ticks per metronome clock (reference note).
In this case, the “reference note” is 1/4 (crotchet), setting 128 ticks/crotchet, and the hex
code associated with note value can be computed to state the delta time.
4

Another command used in the beginning encompasses the definition of the metronome:
FF 51 03 tt tt tt set number microseconds per quarter note. For instance: FF 51 03 07
A1 20 for 120 bpm; FF 51 03 0f 42 40 for 60 bpm.
End of Track: FF 2F 00. This event is not optional. It is included so that an exact ending
point may be specified for the track, so that an exact length is defined, which is necessary
for tracks which are looped or concatenated. For looking inside a midi file, we suggest [11].
For example, here is initial data from a monophonic MIDI file (in hex): 00 FF 58 04 04
02 30 08 00 FF 59 02 00 00 00 90 3C 28 81 00 90 3C 00 00 90 3C 1E 81 00 90 3C 00 00 90 43
2D 81 00 90 43 00 00 90 43 32 81 00 90 43 00 00 90 45 2D 81 00 90 45 · · · The way of reading
this file is separating {time/message} pairs as {00 FF 58 04 04 02 30 08} {00 FF 59 02 00
00} {00 90 3C 28} {81 00 90 3C 00} {00 90 3C 1E} {81 00 90 3C 00} {00 90 43 2D} {81 00 90
43 00} {00 90 43 32} {81 00 90 43 00} {00 90 45 2D} {81 00 90 45 00} · · · Black characters
are delta time information, and blue characters are messages. Blue boldface type part is
command and ordinary blue characters are data associated with the previous command.
Final remark: many times the command note-off 80 nn 00 is replaced by the command
90 nn 00 (note-on, pitch nn, null intensity), which is equivalent.

4 Monophonic Audio on MIDI Files

Time value are expressed in terms of the number of ticks. However, the hex-value to
record the time value is not obtained just converting the number of ticks. A special format
is used, which is referred to Variable Length Values (VLV, see [7] for details). Notice that
time values sometime are two bytes in length, and sometimes they are one byte in length.
Using more than one byte for the delta time implies a longer time value. For example,
here is some data from a simple monophonic MIDI file (in hex): 00 FF 58 04 04 02 30 08 00
FF 59 02 00 00 00 90 3C 28 81 00 90 3C 00 00 90 3C 1E 81 00 90 3C 00 00 90 43 2D 81 00 90
43 00 00 90 43 32 81 00 90 43 00 00 90 45 2D 81 00 90 45 00 00 90 45 32 81 00 90 45 00 00 90 43
23 82 00 90 43 00 00 90 41 32 81 00 90 41 00 00 90 41 2D 81 00 90 41 00 00 90 40 32 40 90 40 00
40 90 40 28 40 90 40 00 40 90 3E 2D 40 90 3E 00 40 90 3E 32 40 90 3E 00 40 90 3C 1E 82 00 90
3C 00 00 FF 2F 00 Do have then to look at the header of the MIDI file to understand what
the units mean. For this example, the time units are 128 ticks to the quarter note, so 128
is a quarter note duration, 256 is a half-note, and 64 is an eighth-note duration (Fig. 4).

Figure 4: Number of ticks for the relative duration of a note.

The sequence of played notes is C4 1/4, C4 1/4, G4 1/4, G4 1/4, A4 1/4, A4 1/4, G4 1/4, F4
1/4, F4 1/4, E4 1/8, E4 1/8, D4 1/8, D4 1/8, C4 1/2.
 5

5 Polyphonic Audio on MIDI

We are particularly interested in the midi perspective for musicians [1], [7]. Let us
take a straightforward example. It is easier to build the MIDI coding generating the
corresponding notegram (Fig. 6) to excerpt of score displayed in Fig. 5. The list of midi
events is 00 90 40 40 00 90 43 40 81 00 80 43 00 00 90 45 40 81 00 80 45 00 00 80 40 00 00 90
3C 40 00 90 47 40 81 00 80 47 00 00 90 48 40 81 00 80 48 00 00 80 3C 40.

(a) score (b) colored score

Figure 5: Excerpt of a score.

Figure 6: Excerpt of a score presented in Fig. 5.

6

00 90 40 40 (Start of E4 note, pitch = 64)

00 90 43 40 (Start of G4 note, pitch= 67)
81 00 80 43 00 (End of G4 note, pitch=67)
00 90 45 40 (Start of A4 note, pitch=69)
81 00 80 45 00 (End of A4 note, pitch=69)
00 80 40 00 (End of E4 note, pitch=64)
00 90 3C 40 (Start of C4 note, pitch = 60)
00 90 47 40 (Start of B4 note, pitch= 71)
81 00 80 47 00 (End of B4 note, pitch= 71)
00 90 48 40 (Start of C5 note, pitch= 72)
81 00 80 48 00 (End of C5 note, pitch= 72)
00 80 3C 40 (End of C4 note, pitch = 60)

Figure 7: Playing different instruments: how midi works.

The MIDI message used to specify the instrument is a “program change” message:
command byte=1100 CCCC (MIDI channel 0 to 15); Data byte 1=0XXX XXXX (instru-
ment number from 0 to 127). The sax, piano and bass drum will use channels 1, 2 and
10, respectively. An illustrative example of MIDI sequence is shown in the sequel. First
send the program changes to specify the instruments on each channel. In Fig. 7 there is
a score example with three instruments. Suppose that the sax, piano and bass drum will
use respectively channels 1, 2 and 10. The MIDI message sequence is the following.
Specifying the instruments to use on each channel: 00 C0 41 (Alto Sax=66 ⇒
coded 65 =0×41), 00 C1 00 (Piano=1 ⇒ coded 0), 00 C9 00 (Standard Drums Kit=1 ⇒
coded 0)
Notes inserting: 00 90 48 40 (Start sax C4, pitch=72=0×48), 00 91 3C 40 (Start piano
C3, pitch=60=0×3C), 00 91 43 40 (Start piano G3, pitch=67=0×43), 00 91 4C 40 (Start
piano E4, pitch=76=0×4C), 00 99 23 40 (Start Bass Drum=35=0×23), 81 00 90 48 00
(Stop sax C4, pitch=72=0×48), 00 99 23 00 (Stop Bass Drum=35=0×23), 00 90 4A 40
(Start sax D4, pitch=74=0×4A), 81 00 90 4A 00 (Stop sax D4, pitch=74=0×4A), 00 90
4C 40 (Start sax E4, pitch=76= 0×4C), 00 99 23 40 (Start Bass Drum =35=0×23), 81 00
90 4C 00 (Stop sax E4, pitch=76=0×4C), 00 99 23 00 (Stop Bass Drum=35=0×23), 00 90
4F 40 (Start sax G4, pitch=79=0×4F), 81 00 90 4F 00 (Stop sax G4, pitch=79=0×4F), 00
91 3C 00 (Stop piano C3, pitch=60=0×3C), 00 91 43 00 (Stop piano G3, pitch=67=0×43),
00 91 4C 00 (Stop piano E4, pitch=76=0×4C). Therefore, the midi sequence correspond-
ing to the score of Figure 7 is 00 C0 41 00 C1 00 00 C9 00 00 90 48 40 00 91 3C 40 00 91 43
40 00 91 4C 40 00 99 23 40 81 00 90 48 00 00 99 23 00 00 90 4A 40 81 00 90 4A 00 00 90 4C 40 00
99 23 40 81 00 90 4C 00 00 99 23 00 00 90 4F 40 81 00 90 4F 00 00 91 3C 00 00 91 43 00 00 91 4C 00
 7

6 Summary
This brief explanatory report revises one of the representations of discrete mathemati-
cal structures used in music. It is intended for midi beginners. The motivation for writing
it came from the difficulty finding a simple text that allows understanding midi files. The
presentation tries to throw some light in the way as a midi is built from a musical score.
It also allows to understand the commands of a hex file (midi) generated to build the
corresponding score. Fig. 6 provides insight into the way the time events are defined. It is
hoped that this note help ordinary users who deal with digital music. A final remark (just
for readers who have never attempted to understand midi): “when trying to learn midi
by Internet, the reader could get a feel of the hardship in having clarity on the subject.
Nevertheless, after reading this paper he may find it a trivial matter. Is that The mind
that opens to a new idea never returns to its original size (A. Einstein).”

References
[1] S. Airy, and J. M. Parr. MIDI, music and me: Students’ perspectives on compos-
ing with MIDI, Music Education Res., Vol.3,N.1, 41-49, 2001 http://dx.doi.org/10.
1080/14613800020029941

[2] B. Arnell. McScope: System. Music, Computers, and Software, April 58-60, 1988.
[3] Z. Cataltepe, Y. Yaslan, and A. Sonmez. Music genre classification using MIDI and
audio features. EURASIP J. on Adv. in Signal Processing Dec. 036409, 1-8, 2007.
http://dx.doi.org/10.1155/2007/36409

[4] J. Cooper, Mind over MIDI: Information sources and system-exclusive data formats.
Keyboard October, 110-111, 1986.
[5] B. Enders, and W. Klemme. MIDI and Sound Book for the Atari ST. Redwood: M&T
Books, 1989. ISBN:155851043.
[6] R. Guérin. Midi power- The comprehensive guide, Thompson Course Tech., 2001.
[7] J. Heckroth, A tutorial on MIDI and wavetable music synthesis. Application Note,
Crystal division of Cirrus Logic, 1998.
[8] T. Holman. Surround Sound: Up and Running, CRC Press, 2014. ISBN:0240808290
[9] J. Matzkin. A MIDI musical offering. PC Magazine 29 Nov.229+, 1988.
[10] MIDI 1.0 Detailed specification, document version 4.2, revised 1995.
[11] online hex editor at http://www.onlinehexeditor.com
[12] K. Pohlman. Principles of Digital Audio, McGraw-Hill, 2011. ISBN:9780071663465

View publication stats