Musical Instrument Digital Interface

Over recent years the term M.I.D.I. has become increasingly familiar to musicians and computer users. There was a time when it was only really of interest to those who made music, but with the increaing proliferation of the personal computer, and its growing multimedia capabilities M.I.D.I is becoming a tool for consumers of music too, particularly across the internet.

Read on to learn everything you would ever want to know about M.I.D.I or return to my frontpage by clicking HERE

What do you want to know?

What Is M.I.D.I?
A Bit Of History
Connecting M.I.D.I. Equipment
Is This Message For Me?
So Where Does The Computer Fit Into All This Then?
Never Mind The Quality, Count The Notes.
Don't All Talk At Once!
Is There Anything Else You Need To Tell Me?
M.I.D.I. Files, Schmiddy Files
Do I Need A Synth Then?

What Is M.I.D.I.?

Having asked the question its probably easier to start by looking at what M.I.D.I. isn't!
The first possible surprise is that M.I.D.I. isn't a sound file format. It doesn't carry the same sort of information as a .WAV or .AIFF file for instance.

.WAV files contain digital recordings of real sounds, similar, in a way, to recordings stored on a CD. M.I.D.I. files don't contain this sort of information at all.

M.I.D.I. is actually a control system. A M.I.D.I. file contains instructions to perform particular commands.
Back to the top
Frontpage

A Bit Of History

The M.I.D.I. protocol was developed by musical instrument manufacturers in order to make the connection of different models and makes of instrument easier. Until this point it was difficult, for instance, to play one synthesiser from the keyboard of another. This meant that layering sounds was difficult, and that in order to have access to a wide variety of sounds the keyboard player had to be surrounded by synthesisers.

M.I.D.I. was devised as a Universal Control System, in other words, a way of one synthesiser controlling another, irrespective of the make and model of either. At its simplest level, when a note is pressed on the master keyboard in a M.I.D.I. setup, it will transmit an instruction to the instruments that are connected to it saying "play this note", and the receiving instrument(s) will respond.

So, a M.I.D.I. file simply contains the instructions for a receiving instrument. It transmits these instructions in the form of numbers, and in this way it allows musical instruments to talk to each other. It is an interface. This, of course, is where the term M.I.D.I. comes from, since it stands for:

Musical Instrument Digital Interface
One of the first results of the development of M.I.D.I. was the removal of keyboards from synthesisers. This began the proliferation of the M.I.D.I. Module. These usually come in the form of 19" rack mountable boxes (such as the Emu Vintage Keys, or Novation Bass Station) so that they can be easily integrated into music studios, but can actually take all sorts of shapes and sizes.
Back to the top
Frontpage

Connecting M.I.D.I. Equipment

So, how do the master keyboard and another keyboard or M.I.D.I. module fit together?
M.I.D.I. information is transmitted down M.I.D.I. cables. These are fitted with M.I.D.I. plugs, which come in the shape of the 5-pin DIN plug. Chosing these as connectors wasn't a particularly bright idea since, unlike XLR's for instance, input and output connectors are identical, and there is no standard for 5-pin DIN wiring. A 5-pin DIN lead is therefore not always necessarily suitable for use as a M.I.D.I. lead. On top of this the connectors themselves aren't always robust.

A M.I.D.I. master keyboard will have M.I.D.I. sockets - usually hidden on the back where you can't see them during normal use. The M.I.D.I. instructions leave the keyboard from the M.I.D.I. Out socket, travel down the M.I.D.I. cable, and are received at the module's M.I.D.I. In socket.

This enables one keyboard to send instructions to another keyboard, or module, but what happens if there is more than one device which needs to be controlled?

The M.I.D.I. data needs to be passed from the first receiving module to the second. This is done via the M.I.D.I. Thru socket which simply echoes what is received at the M.I.D.I. In. Note that the information isn't passed along the chain via the M.I.D.I. Out socket which only transmits information that is originated by the instrument to which it belongs. This method of connecting M.I.D.I. devices is called Daisy Chaining:

Back to the top
Frontpage

Is This Message For Me?

Let's suppose that the first receiving module in the chain is set to play a piano sound, and the second is set to play a brass sound. It would be helpful if the devices could distinguish between information which is targetted at each other, so that the piano sound could be played independently of the brass.

This is done through the use of M.I.D.I. Channels. Every instruction transmitted by the mother keyboard is identified as being on a particular M.I.D.I. Channel. Meanwhile the receiving modules are set to respond on their own M.I.D.I. Channels (these can be the same if desired). When an instruction arives at the module, the device will only follow the instruction if the message's transmit channel (TX) corresponds to its receive channel (RX).

There are 16 M.I.D.I. channels within MIDI protocol, however, some hardware devices allow multiple streams of M.I.D.I. data, and the affect of hundreds of M.I.D.I. channels can be achieved. Be aware though that the number of actual channels is still just 16 - the hardware (in combination with the Software) simply redirects data to more than one M.I.D.I. This allows (for instance) two devices to each receive a full compliment of 16 channels, giving the impression of 32 channels at work, but no instrument can be set to receive on M.I.D.I. channel 17!
Back to the top
Frontpage

So Where Does The Computer Fit Into All This Then?

Although the original idea was to allow one keyboard to control a number of sound sources (in the form of other synthesisers or M.I.D.I. modules) it wasn't long before computers started to be incorporated into the M.I.D.I. chain.

It was bound to happen, since M.I.D.I. involves instruments communicating by using a string of binary digits, and the one thing that computers can do is store and manipulate binary digits.

Although there are all sorts of computer programmes designed to store and manipulate M.I.D.I. data, the most common is the Sequencer. This is a programme which records and manipluates M.I.D.I. instructions, commonly called Events. Note that a M.I.D.I. sequencer doesn't record sounds (although the issue is getting somewhat confusing with the arrival of programmes which integrate M.I.D.I. sequencing with Hard Disk Recording).

Ok. so, remembering that the sequencer only records the instruction eminating from a M.I.D.I. instrument will enable you to avoid one of the most common difficulties afflicting M.I.D.I. newbies. Here's the scenario:

Fledgling M.I.D.I. musician Jon-Mitchell Jam-Jar buys himself a computer, complete with a M.I.D.I. sequencer. Having read up on some of the classic synthesisers of the past 20 years he equips himself with a Yamaha DX7. He hooks up the M.I.D.I. cables correctly, and boots up his sequencer. He's ready to go.

He starts by recording a hypnotic repetitive bass part which he's very happy with. He now wants to add a string part over the top. He changes the sound on the DX7 so that its now playing a lush approximation of the London Symphony Orchestra, but ...... Oh No! The bass part is now being played by violins. "I've recorded my bass part into the computer" he says, "but its not playing back properly".

Of course, now that you know what a M.I.D.I. sequencer actually is you'll realise the mistake he's making. The computer and sequencer are simply recording M.I.D.I. instructions and not the sounds produced by the keyboard. When the sequencer plays the recorded data back it re-transmits the instructions, just as the original keyboard would when it was played in the first place.

The computer is doing the job of a mechanised keyboard player. In fact, in many ways it is simlar to the old pianolas, or player-pianos, which played music according to holes punched in a roll of paper. The sequencer punches the computerised equivalent of holes in the paper. Without the sound producing mechanism the holes are just ....uhm... holes. If Jon wants a bass sound and a string sound playing at once he'll either need two modules/synthesisers, with one sound each, or a device that is capabale of playing more than one sound at once, on more that one M.I.D.I. channel (more of this in a bit).

There are a couple of things to notice from the diagram above. Firstly, the master keyboard is connected so that it can send M.I.D.I. data to the computer, but is no connection to its M.I.D.I. In. This means that the computer can't "play" the master keyboard. This is a common scenario since many master M.I.D.I. keyboards are in fact "dumb", featuring no sound generating hardware themselves, but rather being devoted solely to producing M.I.D.I. instructions. The alternative is to use a genuine synthesiser for both roles.

The second thing to note is that the computer only has two M.I.D.I. connections. One is the M.I.D.I. In which receives instructions from the master keyboard. The other is a M.I.D.I. Out/Thru, which covers both functions. This is because when the master keyboard is being played it is necessary to retransmit the instructions arriving at the M.I.D.I. In to the following modules, but when the M.I.D.I. data is being replayed from the sequencer the computer itself needs to originate the data.
Back to the top
Frontpage

Never Mind The Quality, Count The Notes.

As the use of M.I.D.I. has developed so have the demands placed on the sound modules and synthesisers themselves. As described earlier, in order to construct complete arangements using a M.I.D.I. sequencer you'd need to provide enough sound sources for the sequencer to play. When M.I.D.I. was first introduced this would mean having a separate instrument for each sound that you wanted. However things have thankfully moved on since then.

Today's synthesisers and modules effectively consist of a number of individual sound sources within one box, each of which is capable of playing a different sound source on a different M.I.D.I. channels. These devices are known as Multi-Timbral, literally capable of playing many tones. Devices will vary in the number of different tones that they can play at once, having different numbers of Multi-Timbral Parts. My Ensoniq ESQ1 synthesiser, which is rather dated in terms of facilities, is 8-Part Multi-Timbral. These days it is most common for devices to be provided with 16-part multi-timbrailty. There is a logic behind this number, since if a module has one M.I.D.I In socket then the maximum number of individual M.I.D.I. channels that can be addressed is 16. Further multi-timbrality would therefore be wasted.

Devices also vary in the number of notes or voices that can be played at once. Note that is different from multi-timbrality. Just consider this scenario. An 8-part multi-timbral device has a piano sound assigned to part 1, on M.I.D.I. channel 3 (it is often possible to assign part numbers and M.I.D.I channels independently), and a string sound to part 2 on M.I.D.I. channel 5. You record an 8 note piano chord which is transmitted by the sequencer to the module on M.I.D.I. channel 3. You then record a string melody which is transmitted to the module on M.I.D.I. channel 5, but the device actually only has 8 voices. Two things may happen, depending on how the device assigns voices. Either the strings will not sound, or they will steal notes from the piano part.

The number of notes that are available overall is given by the unit's Polyphony. My ESQ1 is not only 8-part multi-timbral, it is also 8-note Polyphonic. Current devices are generally at least 16-note polyphonic, though 28 and 32 notes are very common, and 64-note polyphonic devices such as the Alesis QS6 are increasingly available. Early multi-timbral devices, like the Yamaha FB01 required the user to specify the number of notes that each part was to have available, but today's units tend to assign voices dynamically, changing the number of voices used depending on what is being played at any one time.
Back to the top
Frontpage

Don't All Talk At Once!

In the diagram on the previous page you'll notice I've shown the master keyboard with a cable attached to its M.I.D.I. In. This is required if the master keyboard is in fact a synthesiser with its own in built sounds rather than a dumb controller.
With the cable attached to the M.I.D.I. In the sequencer will be able to send instructions back to the master keyboard in order for it to generate its own sounds in response.

This causes a problem though.
If the keyboard isn't connected to any other M.I.D.I. devices, and a key is pressed on its keyboard, we want the key to cause the sound generator in the synthesiser to play the corresponding note. If this didn't happen then it wouldn't actually be possible to play the synthesiser as a stand-alone instrument, which of course it is.

When the keyboard is connected to the computer, the sequencer programme will re-transmit the received information, which will eventually be received back at the synthesiser's M.I.D.I. In.

In short, the synthesiser will be receiving instructions from two places at once. This can result in two identical notes being played using the same sound, or if the targets for the two instructions are different, then unwanted sounds may play inappropriate parts. Either result halves the polyphony.

Just in case the problem isn't quite clear, here's a diagram which may help:

There are is a solution to this problem. In fact there is more than one, but I'll only look at the best, which breaks the connection between the sound generating part of the synthesiser and the keyboard. This is acheived by turning Local Off on the synthesiser. This may be easy or difficult depending on the complexity of the synthesiser concerned. The Alesis Quadrasyth for instance has a reference to Local Off in its manual, but the controls used to acheive it aren't called by that name, and are rather confusing. On the other hand the Kawai KC20 has a simple page where Local can be turned on or off

Of course, leaving Local Off makes things simpler when working with a sequencer, but you'll need to remember to turn it back on again when you use the synthesiser as a stand-alone instrument. If you find the keyboard doesn't generate any sound in this situation, its probably worth checking that Local has been turned back on.
Back to the top
Frontpage

Is There Anything Else You Need To Tell Me?

When a key is pressed on the master keyboard, and instruction is sent to any receiving device to play the corresponding note, but there are other sorts of information besides the note numbers that need to be transmitted and received. For instance, it is possible for Programme Changesto be sent via M.I.D.I.

In addition, the speed, or velocity at which the key is pressed is used to convey how hard the note was struck, whilst movement of the pitch bend and modulation wheels can be transmitted via M.I.D.I.

This sort of information comes under the heading of M.I.D.I. Controller Data, some of which is pre-defined as being volume data, or pan data, and some which can be defined by the user so that the receving instrument may map the data to an unspecified parameter.

What the receiving instrument will do with the incoming controller data is determined by that instrument, not by the transmitting device, so that if you were playing a keyboard on which full downwards pitch bend only produced a shift of a semi-tone, you may find that the receiving instrument shifted its pitch by an octave, unless it was set up to respond in the same way.

When the transmitting device sends a M.I.D.I. message it begins by saying what sort of message it is. Each of the controller types is numbered, so that if the receiving isnstrument receives a message labelled as Controller 7 it knows that this will be intended to set the instrument's volume.

Most instruments should conform to the controller assignments shown here:

M.I.D.I. Files, Schmiddy Files

Once people started creating sequencer files it became obvious that they would want to exchange them, either because they were working with other M.I.D.I. musicians, or because they simply wanted to play their work to other people.

There were problems though. Firsltly, a file recorded using one sequencer program couldn't be played by another sequencer program, and secondly, although it was possible to embed intructions to change programmes within sequencer files, the sounds that were attached to various programmes within different synthesisers didn't necessarily bare any relation to each other.

The first difficulty was overcome by the development of M.I.D.I. files. These are sometimes referred to as MIDIs, or .MID files. These allowed sequences to be transported between different software setups.

The second problem was more difficult though.
It was solved by the development of a standard by which sounds were assigned to the 128 available programme numbers, so that, in short, when the original composer determined that a part should be played by a trumpet, it would be played by a trumpet when loaded into someone else's system.

The standard allocation of sounds is called General M.I.D.I., or GM.

This standard is somewhat frowned upon within the professional music making world, since it is seen partly as a way of dictating to M.I.D.I. musicians what sounds they should use, but also because it is lumped together as a close relative of the one finger auto-accompaniment found on many home keyboards.

Having said that, it is undoubtedly a very useful way of ensuring that M.I.D.I. Files sound close to the way they were intended. Of course, an Akai piano won't sound the same as an Ensoniq piano, but at least piano parts will be played by pianos when GM is used.

The standard sound allocation for GM sounds applies to all M.I.D.I. channels ecxept M.I.D.I. Channel 10. More of this in a minute. On all other channels the sounds are allocated to programme numbers as follows:

Back to the top
Frontpage

Under the GM standard, M.I.D.I. Channel 10 is reserved for percussion sounds, and a system for allocating sounds to particular notes is specified too:

General M.I.D.I. also specifies that:

The receiving instrument should be 16 note polyphonic across M.I.D.I. channels 1 to 9, and 11 to 16.
The receiving instrument should be 8 note polyphonic on M.I.D.I. channel 10.
All 128 program numbers should be implemented.
Controllers 1, 7, 10, 11, 64, 121 and 123 must be implemented (these are described on the previous page).
Back to the top
Frontpage

Do I Need A Synth Then?

Because they simply contain instructions to play a musical note, rather than digitised recorded audio, M.I.D.I. files are relatively small, and therefore provide a relatively easy way of providing a soundtrack to the World Wide Web, or CD ROM based multimedia software.

Many web pages now include embedded M.I.D.I. files, and computer games provide a musical accompaniment through M.I.D.I.

Of course, not everyone who wants to listen to music in this way has a M.I.D.I. equiped keyboard at their disposal. Luckily they don't need one.

Although M.I.D.I. files which are played via an internet connection can be routed through a M.I.D.I. interface to an external device, it is also possible to have them played by a M.I.D.I. device housed inside the computer itself. The quality of the device and its method of operation will depend on the computer, along with the additions that have been made to it, but generally, M.I.D.I. based music is an accepted facility that computers now provide.

In the case of PCs (IBM compatibles/Windows and Intel based machines), the M.I.D.I. processing hardware is usually housed along with the audio hardware on the computer's soundcard. This may simply offer the sounds themselves along with an output port to enable connection to a stereo system, or to external speakers. Alternatively, the soundcard may offer M.I.D.I. Ins and Outs, enabling the connection of a keyboard and receiving devices.

Apple Macs are already equiped with sound generating hardware, and include a software addition called Quicktime Musical Instruments, as part of the system software. This is effectively a basic GM synthesiser which will not only play M.I.D.I. files from within other programs, but may also be accessed by a sequencer as an instrument.

If you visit a web page with an embedded M.I.D.I. file you'll be able to play it as you look at the page, as long as you have a Plug-In which can process the M.I.D.I. data and send it to a M.I.D.I. equiped soundcard or to Quicktime Musical Instruments.

If you download and save a M.I.D.I. file you'll be able to play it with a variety of software when you're off-line. There are many shareware M.I.D.I. file players available, my Mac favourite is Arnold's M.I.D.I. Player, or if you have a sequencer program you could load it onto there and play it on a GM synthesiser.
Back to the top
Frontpage

And Finally...

If you're a musician M.I.D.I. will provide you with a way of getting your ideas to your ears (and hopefully to other people's), and if you're a music consumer M.I.D.I. will provide you with a way of hearing other people's work. There will be times when you won't know whether you're listening to a M.I.D.I. or an audio file. That's the way it should be. Just enjoy the music. But at least you now know something about how it all works.
Back to the top
Frontpage