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.

Oberheim TEO-5 MIDI implementation

The MIDI implementation of the Oberheim TEO-5: what it answers to, and the 198 controls it offers.

At a glance

Firmware

Not recorded.

MIDI channel

Any one of 1 to 16.

Voices

5.

Note range

Not recorded.

Velocity

Sent and received. Release velocity is ignored.

Aftertouch

Channel pressure and polyphonic, each answered. It sends aftertouch too.

Pitch bend

How far it bends is not recorded, and the range can be changed.

Program change

Received and sent, with 512 presets.

MIDI clock

Sent and received.

Start and stop

Sent and received.

Control change

110 controls, in the tables below.

NRPN

Supported, and preferred to CC. 186 of the controls below have one.

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("oberheim/teo_5") has loaded them.

Performance

Oscillators

Glide

Filter

Envelopes

Envelope 1

Envelope 2

LFO 1

LFO 2

Effects

Reverb

Unison

Arpeggiator

Keyboard

Sequencer

Clock

Mod slot 1

Mod slot 2

Mod slot 3

Mod slot 4

Mod slot 5

Mod slot 6

Mod slot 7

Mod slot 8

Mod slot 9

Mod slot 10

Mod slot 11

Mod slot 12

Mod slot 13

Mod slot 14

Mod slot 15

Mod slot 16

Global settings

Where these facts come from

Source documents

• TEO-5 MIDI Implementation Document, v2, its MIDI implementation, read on 2 October 2026 (the file read).
• TEO-5 User's Guide, Version 2.0, dated June 2025, its user manual, read on 2 October 2026 (the file read).
• TEO-5 Synthesizer - Oberheim, its product page, read on 2 October 2026.

How they were read

MIDI Implementation Document v2, 16 pages: the global settings that govern MIDI response on pp. 1-2, received and transmitted messages on pp. 3-4, the control change table on pp. 5-6, how an NRPN travels on p. 7, the program parameter NRPNs on pp. 8-10, the global parameter NRPNs on p. 11, the one control NRPN and the system exclusive messages on pp. 12-15. The User's Guide Version 2.0 (116 pages) states what the instrument is and names the states some parameters step through; its pages are cited as file pages, because its front matter is unnumbered.

NRPN IS THE MAKER'S PREFERRED WAY IN, AND WHICH ONE IT USES IS A SETTING. "MIDI Param Send: Off, CC, NRPN" (p. 1) and "MIDI Param Receive: Off, CC, NRPN - Sets the method by which parameter changes are received via MIDI. As with transmission, NRPNs are the preferred method" (pp. 1-2). So a control below with both numbers is one parameter reachable either way, and which way depends on how the instrument is set.

SIXTEEN PARAMETERS GET A DIFFERENT RANGE IN EACH TABLE, and nrpn_range carries the NRPN's. Thirteen of those are NRPN being finer, which is the maker's own reason for preferring it - the two filter figures at 0-1024 and 0-512, and ten envelope and LFO figures at 0-256, where the control change gives 0-127. Three are not finer and are worth knowing about before trusting either number: Noise Type is a switch at 0-1 by control change and a value at 0-127 by NRPN; Scale is 0-65 by control change and 0-64 by NRPN, so the NRPN is the narrower of the two; and BPM is 15-127 by control change against 30-250 by NRPN, which do not even overlap at the top, so the control change cannot reach the tempos the NRPN can.

AND READ EVERY ONE OF THOSE WIDE RANGES AS A COUNT RATHER THAN A TOP, which is the one thing that makes them make sense. 0-256, 0-512 and 0-1024 all name one value more than the bit width they imply, while the forty-eight modulation amounts in the same table are 0-255, which is right. So the maker appears to have written the number of values where the highest value belongs, in some rows and not others. Both readings found this and neither corrected it: the figures below are the document's.

ONE RANGE IS THEN ALMOST CERTAINLY A TYPO, recorded as printed for the same reason. Filter Res is given 0-265 (p. 9) where every other wide parameter is 256, 512 or 1024. 265 falls out of no bit width and is one digit away from 255.

THE TWO TABLES ARE NOT SYMMETRICAL, and three absences are the maker's rather than this file's: there is a control change for Osc 2 Detune and none for Osc 1 Detune, though the NRPN table has both; there is an NRPN for Env 1 Dest and none for Env 2; and the whole modulation matrix - sixteen slots of source, amount and destination, forty-eight parameters - is reachable by NRPN only.

EXPRESSION IS CLAIMED AT TWO DIFFERENT NUMBERS IN THIS ONE DOCUMENT, which is the sharpest thing here for anybody building a controller. The control change table gives "11 Expression 0-127" (p. 5) and the global table agrees, naming the pedal assignment "2 = Exp CC11" (p. 11) - but the transmitted controller table sends Expression on 0000 1101, which is 13 (p. 4). expression below is CC 11, because two of the three statements say so and because 11 is where the MIDI specification puts expression. A host that listens on 11 alone may still miss what this instrument sends. AND THE PERFORMANCE CONTROLLERS DO NOT GO BOTH WAYS. Breath Controller and Expression are in the transmitted table and not the received one; All Notes Off and Reset All Controllers are in the received table and not the transmitted one (pp. 3-4). direction is not set on any control below, because the control change table on pp. 5-6 - which is where the numbers come from - says nothing about direction at all, and the two message tables are about a different and shorter list. Volume appears twice in the transmitted table, once as the pedal assignment and once as "Volume knob", both on CC 7.

THE DOCUMENT CANNOT AGREE WITH ITSELF ABOUT BANKS AND PROGRAMS, in five places. Received program change is "0-127 for Programs 1-127 within current Bank" (midi_impl p. 3) and transmitted is "0-127 for Programs 00-128 within current Bank" (pp. 3-4) - one is 127 programs for 128 numbers and the other is 129. Received bank select is "1 - 4 select user banks 1 - 4; 5 - 8 select factory banks 1 - 4; 9 - 12 select add-on banks 1 - 4 others ignored" (midi_impl p. 3) and transmitted bank select is "0 - 9" (pp. 3-4). The dump request takes "Bank Number, 0 - 15" (midi_impl p. 13) and "Program Number, 0 - 127" (midi_impl p. 13) while the dump itself carries "Bank Number: 0 - 9" (midi_impl p. 14) and "Program Number: 0 - 99" (pp. 13-14). presets below is the guide's plain count of the instrument instead: "The TEO-5 contains 512 programs. These are divided into two sets of 256 programs - a Factory set and a User set" (guide p. 13), which the four user and four factory banks of the received bank select agree with at sixty-four programs a bank.

THE GLOBAL TABLE CONTRADICTS THE PROSE TWO PAGES EARLIER, twice. MIDI Clock Mode is given six settings in words - "Off", "Out", "Out No S/S", "In", "In/Thru", "In No S/S" (p. 1) - and four in the table, 0-3 as Off, Out, In and In Thru (p. 11), so the two that suppress start and stop have no number. MIDI Out Select is "Off, MIDI, USB, All" (p. 2) in words and 0-2 as MIDI, USB and All in the table, so Off has none.

THE SISTER DOCUMENT AGREES WITH THE PROSE BOTH TIMES, which is as close to an answer as anything here gets. The Take 5's implementation document, numbering its globals identically, gives MIDI Clock Mode 0-5 at 4099 and MIDI Out Select 0-3 at 4107 - six settings and four, exactly the counts this document's own prose gives. So this table is very likely the stale one. It is still a deduction rather than a statement, so choices below carries the four and the three the table prints, and this paragraph is where the doubt lives.

ONE GLOBAL'S DESCRIPTION IS ANOTHER'S, copied and not corrected: "MIDI Sysex On: Off, On - When set to On, the synth will respond to MIDI controllers, including pitch wheel, mod wheel, pedal, and volume" (p. 2), which is word for word what the line above says about MIDI Control. What MIDI SysEx On does is said instead by the table, which calls 4105 "MIDI SysEx Enable" (midi_impl p. 11).

AND THE TWO SISTER DOCUMENTS SWAP TWO RANGES. Here Master Coarse Tune is 0-100 and Master Fine Tune is 0-24 (p. 11); the Take 5's document, at those same two numbers, gives 0-24 and 0-100. Twenty-four semitones of coarse tuning and a hundred steps of fine is the reading that makes musical sense, so this document is probably the one with them the wrong way round - but it is the document for this instrument, so its figures are what is below. The sibling is not the better document either: it spells the parameter "Master Course Tune".

THIS DOCUMENT IS AN EXTRACT FROM THE GUIDE, WHICH IS WHY ITS HEADERS ARE RIGHT AND ITS BODY IS NOT. Its own running footer reads "TEO-5 User's Guide" on every page. So the furniture was updated when Oberheim made it from Sequential's and the sentences inside were not, which is the same fault as the guide's nineteen Take 5 headers seen from the other side.

FIVE MORE THINGS IN IT DO NOT ADD UP, and none changes a number below. The program table is introduced as "TEO-5 program parameters 0-352" (p. 8) and then runs 1 to 187 in 159 rows with no parameter 0 and nothing above 187. The page describing how an NRPN travels says "The parameter number can be found in the two tables below" (midi_impl p. 7) where there are three, and says the number's high byte "will be zero or one, and never more than two" (p. 7) where every global is 4096 or above and so has a high byte of 32. MIDI Param Send and MIDI Param Receive carry an asterisk and a dagger in the global table and no footnote exists anywhere in the sixteen pages for either. The system real-time table lists MIDI Timing Clock and has no row for Start or Stop (p. 3), although the clock settings two pages earlier send and receive both - which is why transport below is both on the strength of the prose rather than the table. The dump request's bank byte is written 0000 00vv, which holds two bits and four values, under a stated range of 0 to 15 (p. 13).

THE GLOBAL TABLE'S OWN SPACING IS ERRATIC and the choices below do not reproduce it. The states are printed "1= MIDI", "0 =All Off", "1 =Passthru", "2= Off" and so on, with the space on either side of the equals or on neither. Both readings found the same unevenness, so it is the document's rather than a tool's; the names below are keys and carry none of it.

AND THE GLOBALS ARE NAMED TWICE, differently. The panel names on pp. 1-2 and the table names on p. 11 do not match for four of them: MIDI Control becomes MIDI Control Enable, MIDI Sysex On becomes MIDI SysEx Enable, MIDI Out Select becomes MIDI Cable Out, and MIDI Prog Send/Receive become MIDI Program Send/Recv. The labels below are the table's, because the table is what carries the numbers.

NOT RECORDED: which number means which named state for any program parameter. The implementation document gives those as bare ranges, and the guide names the states in words without numbering them - "Shape: Triangle, Sawtooth, Reverse Sawtooth, Square, S&H" (guide p. 42) against a range of 0-4 (pp. 6, 9) tells you there are five and not which is which. The globals are different, and choices below carries their states because that table does number them.

NOT RECORDED: the note range and the key numbers. The guide gives the keyboard as "its 44-note keyboard" (p. 17) and names no note number for either end of it; the implementation document says only "Note number 0-127" (p. 3) for what it receives.

NOT RECORDED: any system exclusive layout. The messages are published - device inquiry, three dump requests, three dumps, and the packed-data format - but a program dump is "4096 bytes expanded to 4690 MIDI bytes in 'packed MS bit' format" (p. 14) with no map of what those bytes mean. 4690 does not follow from the packing the same document describes: seven bytes to eight makes 4096 into 4682, which both readings worked out separately. Three more details in that section are worth a reader's attention: the instrument's own identity byte is written 0b1011010 where every other byte in the document is two binary nibbles; the manufacturer byte labelled "Oberheim ID" is 0000 0010, the same byte as the "Inquiry Reply" (midi_impl p. 12) label directly above it in the same table; and the global dump says it carries "50 nibbles (LS then MS) for 25 Global parameters" (p. 15) where the table on the page before lists twenty-seven.

NOT RECORDED, DELIBERATELY: parts. The instrument splits its keyboard and stays one instrument while it does - "The program is the same for both key ranges" (guide p. 17) - so Key Split, Key Split 2 and Key Split Note are controls below and not parts.