This page is read from MIDI Instrument Definitions, a Python library used in the author's other projects, in which every number is traced to a document its maker published. Its source is on GitHub.

Yamaha reface CP MIDI implementation

The MIDI implementation of the Yamaha reface CP: what it answers to, and the 18 controls it offers, on firmware 1.30.

At a glance

Firmware

This page describes firmware 1.30.

MIDI channel

Any one of 1 to 16.

Voices

Not recorded.

Note range

Not recorded.

Velocity

Sent and received. Release velocity is ignored.

Aftertouch

None, sent or answered.

Pitch bend

How far it bends is not recorded.

Program change

Not received or sent.

MIDI clock

Neither sent nor received.

Start and stop

Neither sent nor received.

Control change

18 controls, in the table below.

NRPN

None.

System exclusive

Supported.

Controls

In the order the definition gives them, by group. Under each control is the name a program calls it by, once pymidiinstrumentdefs.load("yamaha/reface_cp") has loaded them.

Where these facts come from

Source documents

• reface CS/DX/CP/YC Data List, b0, its MIDI implementation, read on 5 October 2026 (the file read).
• reface Version 1.30 Supplementary Manual, its user manual, read on 5 October 2026 (the file read).
• reface Reference Manual, its user manual, read on 5 October 2026 (the file read).

How they were read

reface CS/DX/CP/YC Data List, which carries this instrument's MIDI data format, its MIDI data table and its implementation chart on sheets 11 to 14: the data format and the controllers outside the table at sheet 11, the system overview at 12, the control change table and the system exclusive parameter tables at 13, and the implementation chart at 14. The Version 1.30 Supplementary Manual is cited for what that firmware added, and the reface Reference Manual for nothing at all - see below.

THE PARTITION IS WHAT MAKES A FOUR-MODEL DOCUMENT READABLE FOR ONE MODEL. Its contents give the reface CS sheets 2 to 5, the DX 6 to 10, the CP 11 to 14 and the YC 15 to 19, each with its own format, table and chart, and the running head of every sheet names exactly one model. The test a machine can apply is the system exclusive Model ID, which is a value rather than a layout: all four share the group number 7FH 1CH and the model byte separates them. This instrument's is 04H, and Notes/4591_extract_reface_cp.py asserts both halves - that it appears on this instrument's sheets, and that none of the other three models' bytes appears inside 11 to 14. No material of this instrument's lies outside those four sheets and no other model's lies within them.

THE TABLE WAS READ TWICE BY TWO METHODS IN ONE SCRIPT, AND THEN A THIRD TIME BY A FRESH READER. Once in the text layer's own order, with each value entry consumed as a three-line group; once from the position of every word on the page, grouping into rows by y-origin and into cells by x-origin. The two agree on all twelve parameters, every controller number, every name and every value entry. Two methods in one script catch a mistake in one method; they do not catch a mistake in one reading of the document, which is what the second reader is for.

THE TRAP THE reface DX's ACCOUNT RECORDED IS HERE AND IT IS QUIETER. A parameter's controller number and a setting's transmitted value are both bare integers on their own line. On the DX four of the algorithm's transmitted values are themselves controller numbers of other parameters in the same table, so a careless reader gives two parameters somebody else's number - a loud failure. On this instrument no transmitted value collides with any controller number: the values are 0, 25, 51, 64, 76, 102 and 127 and the numbers are 17 to 19, 80, 81 and 85 to 91. So the same careless reader invents seven controllers here and misattributes none, which is harder to notice rather than less wrong. The script prints the overlap so that a revision creating one says so.

WHERE THE REAL MISATTRIBUTION RISK LIES IS BETWEEN THE FOUR MODELS, AND IT IS WORSE THAN THE BARE INTEGERS. All four of them use controller 80 and all four mean something different by it: it is this instrument's electric piano TYPE, the CS's OSC TYPE, the DX's ALGO and the YC's WAVE. Controller 19 means three things in three models - this instrument's tremolo-or-wah rate, the CS's EFFECT RATE and the YC's ROTARY SPEED - and 17 and 18 mean two apiece, this instrument's switch and depth against the CS's EFFECT TYPE and EFFECT DEPTH. And 85 to 90 are this instrument's chorus, delay and reverb and the DX's first operator's six parameters. Nothing in the numbers themselves says which model a reading came from; only the sheet range and the Model ID do.

THE NUMBERING IS NOT CONTIGUOUS AND THAT LOOKS LIKE AN ERROR AND IS NOT. The table runs 80 and 81 for the voice, then backwards to 17, 18 and 19 for the first effect, then forwards to 85 through 91 for the other two and the reverb, leaving 82, 83 and 84 unassigned - and the CS uses 82 for its oscillator modulation. Both readings confirmed 17, 18 and 19 by their place in the number column, and the second reader confirmed them again off the page rendered at 300 dots to the inch. The likely reason is the CS, whose effect block is on the same three numbers, so a player moving between the two instruments finds the effect controls where they left them. That is a conjecture about intent and the maker states nothing.

AND THIS SHEET IS NOT LAID OUT LIKE THE DX'S, WHICH COST THE SECOND METHOD ITS FIRST RUN. On the DX's sheet 8 the control change table is the whole left column and the system exclusive tables are the right one. Here the table is the top of the left column and the address tables carry on below it in the same column - and their High, Mid and Low columns put bare hexadecimal pairs exactly where a controller number sits, so a reader bounded only by the column runs on and harvests 00 and 30 as two more parameters. It did. The bound is the maker's own heading, Parameter Base Address, and it is in the script.

TWELVE PARAMETERS IN THE TABLE AND SIX CONTROLLERS OUTSIDE IT, WHICH IS EIGHTEEN. The table is the four effect sections and the voice: the electric piano type, the drive, and then a switch, a depth and a rate or speed or time for each of tremolo/wah, chorus/phaser and delay, with a reverb depth. The six outside it are in the data format instead, and the difference between the two groups is not what they do but whether a setting can turn them off.

THE TWELVE ANSWER ONLY WHILE MIDI CONTROL IS ON AND THE SIX ALWAYS ANSWER. The table's own heading says so - "Transmitted and recognized Control Change Number and Value, when MIDI control is on" (dl p. 13) - and the data format puts it the same way: "When MIDI Control Mode is turned on, Control Change numbers are assigned in order that voice parameter changes made using controllers on the front panel can be controlled via MIDI" (dl p. 11). The implementation chart carries the qualification as a footnote, "Transmitted and recognized if MIDI control mode is on" (dl p. 14), against the three rows that are the table's numbers and against no others. MIDI Control is a system exclusive parameter at 00 00 0E, off, ON, reachable from the panel and by system exclusive and by no controller at all - so nothing in this file can turn on the thing that most of this file needs.

THE SIX OUTSIDE THE TABLE, AND FIVE OF THEM ARE RECEIVED AND NEVER SENT. The data format lists one transmitted control number, c = 64 SUSTAIN SWITCH, and six received - modulation 1, volume 7, expression 11, sustain switch 64, sostenuto 66 and soft pedal 67. The chart confirms each cell. Sostenuto and the soft pedal are this instrument's alone in the family: the CS, the DX and the YC receive 1, 7, 11 and 64 and nothing else.

AND WHAT THE SUSTAIN SWITCH SENDS DEPENDS ON THE PEDAL PLUGGED INTO IT. A footnote: "When Sustain is set to 'FC4/5,' operating the foot switch transmits only values of 0 (off) or 127 (on)" (dl p. 11) - and Sustain Pedal Select, FC3, FC4/5, is a system parameter at 00 00 0B. So with an FC3 the controller is continuous and with an FC4 or FC5 it is a switch, which is a fact about the pedal rather than about the instrument and is why no kind is declared for it below.

SOSTENUTO AND THE SOFT PEDAL ARE SWITCHES AND THE DOCUMENT SAYS WHERE THEY TURN OVER: v = 0-63:OFF , 64-127:ON for each (dl p. 11). Recorded as bands below, because a consumer that drew them as continuous would be drawing something this instrument does not have.

PROGRAM CHANGE IS A CHECKED ABSENCE AND THIS IS WHERE THE FAMILY MOST INVITES A MISTAKE. The chart marks it absent in both columns, and the data format has no program change section at all. The reface series does not do program change is true of this instrument and of the CS and the YC, and false of the DX, which does it over four banks of eight. That sentence is nobody's but mine, written as the belief a reader of this family most easily forms. It is true three times out of four, and the one place it fails is the one place a reader would carry it from.

NO MIDI CLOCK AND NO TRANSPORT, ALSO CHECKED AND ALSO UNLIKE THE DX. The chart's System Real Time rows - Clock, and Commands - are crossed in both columns, as are Song Position, Song Select and Tune Request. The DX's chart marks Clock and Commands present both ways, so this is a real difference between two instruments in one document rather than an omission here. The only system real time message this instrument has at all is active sensing, "Transmitted at every 200 msec" (dl p. 11), which this format does not record.

PITCH BEND IS RECEIVED, NEVER TRANSMITTED, AND CANNOT BE SET. The data format heads its section "PITCH BEND CHANGE (Receive only)" (dl p. 11) and the chart agrees, crossed in the transmitted column and present in the recognised one. And there is no parameter for its range anywhere: no controller, no registered parameter number, and the system address its sibling uses for exactly that is reserved here. The CS has Pitch Bend Range at 00 00 0F, 28 - 58 in hex, minus 24 to plus 24 semitones; this instrument's 0F is reserved. So writing the CS's pitch-bend-range address to a reface CP targets a reserved byte, which is the concrete form of the warning the reface DX's account left for this rank.

AND FOUR MORE OF THIS INSTRUMENT'S RESERVED BYTES ARE A SIBLING'S PARAMETER, which is the general shape of that warning and worth having whole. The four models share one system table layout with a different set of holes punched in it: 00 00 08 and 09 are Tempo MSB and LSB, 30 - 300, on the CS and the DX and reserved here; 0A is the DX's LCD Contrast and reserved here; 0F is the CS's pitch bend range and reserved here; 14 is the CS's Foot Volume / Sustain switch and reserved here. So five addresses that do something on a sibling do nothing on this instrument, and every one of them is a byte a reader could send in good faith.

THE MISSING TEMPO IS NOT A SEPARATE FACT FROM THE MISSING CLOCK. This instrument has no tempo parameter, no transport message, no MIDI clock either way - and no looper, where the CS's and the DX's system overview diagrams carry a recorder and this instrument's sheet 12 is a keyboard, a local control switch and a tone generator. Four absences that are one absence: it is the only member of the family with nothing in it that runs in time.

AFTERTOUCH OF BOTH KINDS IN BOTH DIRECTIONS IS CROSSED IN THE CHART, which is four cells and a checked absence rather than a silence. So are the registered and non-registered parameter numbers, whose controllers are among the numbers the table leaves unused.

THE KEYS SEND A NARROWER RANGE THAN THE INSTRUMENT ANSWERS TO. The chart gives note number transmitted as 24 to 108 and recognised as 0 to 127, with Transpose in the remarks - and Master Transpose, minus 12 to plus 12 semitones, is a system parameter at 00 00 07 and a new function of this firmware. So the transmitted span is the 37 keys plus what transposing them reaches, and nothing restricts what the instrument will sound, which is why no note range is recorded below. Velocity goes both ways on note on and neither way on note off: the chart prints o 9nH,v=1-127 transmitted against o v=1-127 recognised, and for note off x 9nH,v=0 against x, while the data format says of the note off message that velocity is ignored (dl p. 11).

IT HAS LOCAL CONTROL AND DOES NOT ANSWER THE CONTROLLER FOR IT, which is the reface DX's shape and the Leviasynth's one rank earlier. Local Control, off, ON, is a system parameter at 00 00 06; the chart crosses Local ON/OFF in both columns, so controller 122 does not reach it. Three instruments in five ranks where the feature exists and the standard controller for it does not work.

WHAT IS NOT RECORDED, AND WHY. mode - the chart gives two different defaults, 3 transmitted and 1 recognised, and the field holds one number. This is roland/juno_106's case exactly and yamaha/reface_dx's, and is settled the same way. It is also where this instrument differs from that one again: the DX's chart recognises modes 1 - 4 with a footnote that the mono count is always treated as one, and this instrument's recognises 1,3 alone and carries no such footnote - omni on and omni off, and no mono-or-poly change. Its own channel mode section matches, saying of mono and poly only that each does what All Sound Off does. polyphony - none of the three documents states a voice count, and the owner's manual that would is not among them. The receive channel's All setting is omni as a setting rather than a mode: the system table gives MIDI receive channel as 00 - 0F, 10 for 1 - 16, All and the transmit channel as 00 - 0F, 7F for 1 - 16, off, and the two sentinels are different numbers for the same idea, which is easy to transpose. The Supplementary Manual gives the factory settings: the receive channel starts at All and the transmit channel at 1, so out of the box this instrument answers to every channel.

THE CHART WAS NOT REDATED FOR THIS FIRMWARE AND A SIBLING'S WAS. This instrument's implementation chart is version 1.0, dated 23 March 2015; the CS's is version 1.1, dated 31 March 2016. The same is true of the DX's, version 1.0 of 20 February 2015. So two of the four charts in one file predate the firmware this definition names, and the Supplementary Manual adds no MIDI table, no controller and no message - what it adds for this instrument is master tuning, master transpose, and the transmit and receive channels, which is why channels can be stated at all. The documents are consistent with the charts being complete and with their having been left unrevised, and this definition does not choose.

THE SUPPLEMENTARY MANUAL IS PARTITIONED TOO AND ITS RUNNING HEAD WILL NOT TELL YOU SO. Every sheet from 2 to 28 carries "reface CS / reface DX / reface CP / reface YC New Functions" (supplement p. 15) as its head, so searching it for a model name matches every page and proves nothing. The partition is in the section headings instead: New reface CS Functions at sheets 3 to 12, the DX's at 13 to 14, this instrument's at 15 to 20, the YC's at 21 to 26, and an appendix at 27 to 28. Recorded because the Data List's own partition can be found by the running head and this one cannot, and the two documents sit beside each other.

FOUR DEFECTS IN THE DATA LIST, FOUND BY THE SECOND READER AND RECORDED RATHER THAN TIDIED. **Two footnotes on facing sheets are both numbered *1 and mean different things** - on sheet 11 it is the sustain pedal's conditional, on sheet 14 it is "Transmitted and recognized if MIDI control mode is on" (dl p. 14) - so a marker carried across the page boundary picks up the wrong text. The chart's controller 64 row carries no marker at all, a bare present-present with nothing to say the pedal setting changes what it sends, where the CS's chart does flag its equivalents; the chart read alone therefore promises a continuous sustain in every case. The three effect switches' first state is named twice and differently on one sheet - OFF in the control change table and thru (middle position) in the system exclusive table - and the second and third states likewise, TREMOLO against tremolo (TREMOLO). The names below are the control change table's, because that is the table a controller number comes from. And thru (middle potision) is printed at one of the three addresses and correctly at the other two, which is the maker's typing and is left alone.

AND TWO IN THE SUPPLEMENTARY MANUAL, NEITHER OF WHICH CHANGES A FIELD HERE. Its appendix marks the transmit channel as the CS's and the DX's while its own pages for this instrument give this instrument one; the two are reconcilable, because the appendix lists only what a single key held at power-on reaches and this instrument's channels are set in an advanced mode instead, but nothing on the page says so. And the setting-confirmation display for the sustain pedal is inverted between the CS and this instrument - lit for FC4/5 on the one and for FC3 on the other, under the same parameter name. One of the two is likely wrong and these documents do not say which, so it is recorded and not resolved.

ONE DISAGREEMENT BETWEEN THE TWO READERS, AND THE DOCUMENT SETTLES IT. The second reader reported that controller 81 is this instrument's drive and the DX's first operator level. It is not: the DX's table puts OP1 LEVEL at 85 and uses no 81 at all, its numbers being 80, 85 to 90 and 102 to 119. Checked in the DX's own number column, and it agrees with that instrument's definition and account. The CS is where 81 does double duty, as OSC TEXTURE. Said here because the reader's other comparisons were right and this one was not, which is the reason a comparison gets checked rather than accepted.

AND THE REFERENCE MANUAL IS THE OPPOSITE TRAP TO BOTH OF THEM. Its title names the family and its content is the reface DX's alone - the edit mode reference for the four operators the DX's controllers address. It names the DX on three of its eleven sheets and no other model anywhere. So one title over four models can mean the content is partitioned, as the Data List is; or that it is all one model's, as this is - and neither can be inferred from the title. It is cited for nothing here and recorded so that nobody has to open it to find that out.