Subsystem Search

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.

Behringer TD-3 MIDI implementation

The MIDI implementation of the Behringer TD-3: what it answers to.

At a glance

Firmware
Not recorded.
MIDI channel
Not recorded.
Voices
One, so it plays one note at a time.
Note range
Not recorded.
Velocity
Not recorded.
Aftertouch
None.
Pitch bend
Not recorded.
Program change
Not received.
MIDI clock
Received.
Start and stop
Received.
Control change
None.
NRPN
None.
System exclusive
Not recorded.

Where these facts come from

Source documents

How they were read

Quick Start Guide (55 sheets, 107 printed pages, V 5.0), in every language Behringer publishes: the hook-up diagrams at pp. 14-17, the rear-panel connectors and the USB port at pp. 20-21, the clock-source setting at pp. 38-39, the Poly Chain system mode at pp. 92-93, the MIDI message table at p. 94, and the specifications at p. 103.

Two pages are printed to a sheet, so pages_per_sheet: 2 with an offset of one, and a citation to p. 94 is the left half of the 48th page of the file.

THERE IS NO FULL MANUAL. This maker publishes one for its Model D and does not for this instrument, which was established by reading what each of three product pages offers rather than by failing to find one: the TD-3-RD's page serves this guide in all languages plus Japanese and Chinese editions, the TD-3-BK's serves only a Japanese and a Chinese one under different file names and no English at all, and the TD-3-MO's serves its own. So the one English-bearing document for this instrument sits on one colour variant's page, filed under a second variant's product code, and is absent from a third variant's page.

THE MAKER'S OWN MIDI MESSAGE TABLE IS ITS WHOLE PUBLISHED IMPLEMENTATION, AND THAT IS WHY THIS FILE SAYS none RATHER THAN SAYING NOTHING. The table at p. 94 is headed "MIDI message" (p. 94), gives five columns - Status, Second, Third, Parameter, Description - and groups its rows as Channel Message and SysRT. The channel messages are 8n Note Off, 9n Note On, Bn 7B All Notes Off and En Pitch Bend; the real-time messages are F8 Timing Clock, FA Start, FB Continue and FC Stop. That is every row.

So the table's Channel Message group reads as a complete list of the channel messages this instrument answers to, and three are missing from it: Bn with anything but 7B, Cn program change, and An or Dn aftertouch. Everything recorded below as absent rests on that one reading of that one table, stated here once rather than argued three times.

THE READING GOES NO FURTHER THAN THAT GROUP. The table has no column for direction and no word saying it is complete; the page's one word on direction is a footnote to the examples printed with the table, "Note: 1, MIDI input channel 1." (p. 94). And the guide sends the owner to an application for one setting, "Please use the 'SynthTool.exe' to configure the Poly Chain mode." (p. 92), without saying what that application sends. The TD-3-MO's guide gives that job for its own instrument, one of five that include setting it up and updating its firmware, to an application it calls Synthtribe (mo_guide p. 60), and no document says the two are one program. So the table is not a list of everything the instrument takes. The only Bn the table gives is All Notes Off, which is a channel mode message belonging to the MIDI specification rather than a controller this instrument publishes, so there is no control to carry and control_change: none is a checked absence.

THE TABLE WAS READ THREE TIMES AND THE THREE AGREE ON EVERY CELL. Once by cutting lines at their column positions; once independently, without sight of the first, by taking the table's ruled grid out of the page's vector drawings - the stroked verticals through the dark header bar giving the column dividers, the horizontals giving the row boundaries - and dropping each word into the cell its own bounding box centres in, then checking the result against the page rendered at four times size and read by eye; and a third time by comparing the English copy against the nine language copies the guide prints of the same table, which agree with it row for row and cell for cell on Status, Second, Third and Parameter, only the Description differing.

THAT THIRD CHECK HAS A TRAP IN IT WORTH PASSING ON: the Dutch copy's section heading is printed in English, identical to the real English one, and the Italian and Swedish copies leave their column headers in English while the Italian and Dutch leave the group label as "Channel Message" (guide pp. 99, 100). So a nine-language document cannot be navigated by searching for its headings, and the English block has to be found by position.

TWO THINGS ABOUT HOW THE TABLE IS PRINTED, RECORDED BECAUSE THEY LOOK LIKE EXTRACTION FAULTS AND ARE NOT. Within the SysRT group, F8 and FA fill their Second, Third and Parameter cells with em dashes while FB and FC leave them genuinely empty - true in both documents and in all nine language copies, so it is the artwork and not a slip, and no reason for it is given. And All Notes Off is printed as two bytes, Bn 7B with Third dashed, where a control change carries three; the missing value is not inferred here.

AND THAT READING HAS TO EARN ITSELF, BECAUSE A QUICK START GUIDE THAT LISTS NO CONTROLLER IS NOT THE SAME AS AN INSTRUMENT THAT HAS NONE. The PWM Malevolent in this corpus is the case that settles the general rule the other way: its only document is a quick start guide, it lists no controllers, and its definition deliberately does not say none, because silence in a short document means only that it is not in the guide.

What makes this instrument different is not that the guide is longer - it is that the guide carries a table of MIDI messages at all, and that the same table in the sibling document does carry a controller. The TD-3-MO's copy prints one extra row, Bn 4A xx Filter Cutoff (mo_guide p. 62), in the same table of the same guide by the same maker. So this is demonstrably the place where Behringer lists a controller when the instrument has one, and its absence here is therefore informative rather than merely silent. Two documents are what turn a silence into a statement, and without the MO's copy this file would say nothing instead of none.

One caution about that evidence, since everything above leans on it: the MO prints its MIDI section once, in English only, where this guide prints nine translated copies of its own. So the extra row exists in exactly one place in the two documents, and was read there three times by three methods rather than corroborated by a translation.

THE TD-3-MO IS NOT THIS INSTRUMENT. Its own quick start guide (V 3.0) prints the same table with one extra row, Bn 4A xx Filter Cutoff (mo_guide p. 62), and nothing else differs in it. The maker names the difference on the MO's cover as well, which describes an "Analog Bass Line Synthesizer with VCO, MIDI-Controllable VCF and Sub-Harmonics Oscillator" (mo_guide p. 1) where this instrument's cover says VCO, VCF and a sequencer. Its product page's feature list carries a line about the filter frequency being controllable over MIDI CC which this instrument's does not - described rather than quoted, for the reason the channel note below gives. So the TD-3-MO has one control, and this definition must never be read as describing it.

THE COLOUR VARIANTS ARE FINISHES. The TD-3-BK's and TD-3-RD's product pages carry the same eleven feature lines, word for word apart from a comma and a missing space (product_page_html, bk_product_page), and neither carries a specification. Until 2026-10-13 this said nine lines, one difference and the same specifications, and the TD-3-BK's page was not among the sources. Nothing distinguishes them in MIDI, and the guide that covers them is named after a third, the TD-3-SR.

ONE THING THIS DOCUMENT SAYS TWICE AND CONTRADICTS ITSELF ABOUT, so neither statement is recorded. The controls section calls the rear jack MIDI OUT/THRU and says it is "used to output MIDI OUT data, and pass through MIDI data received at the MIDI INPUT" (pp. 20-21); the specifications call the same jack MIDI In/Thru and name "MIDI In and MIDI Thru, 5-pin DIN" (p. 103), with no output at all. One of them is wrong and the document does not say which, so nothing below claims anything about what leaves the DIN socket. The USB port is not in doubt: "The TD-3 will show up as a class-compliant USB MIDI device, capable of supporting MIDI in and out" (pp. 20-21), with "USB MIDI OUT - sends MIDI data to an application" (pp. 20-21) - which says that it sends, and never says what.

NOT RECORDED: the MIDI channel, because no document cited here prints a range or names a setting for it. The table's worked examples end "Note: 1, MIDI input channel 1" (p. 94), which is an example and not a statement that 1 is the only one.

THE MAKER DOES CLAIM THE CHANNEL CAN BE SELECTED, IN A PLACE A PAGE'S VISIBLE TEXT DOES NOT REACH, WHICH IS WORTH KNOWING ABOUT THIS SITE AND NOT ONLY THIS INSTRUMENT. Every one of these product pages carries a list of feature lines, and one of them is "MIDI and USB implementation with MIDI channel and Voice Priority selection" (product_page_html). That list is delivered inside the page's script payload rather than its markup, so the visible-text copy of the page does not contain it - and stripping scripts is the right rule for rendering a page, not a defect to work around. Until 2026-10-13 the visible text was the only copy cited, so the line was described and not quoted; the page as served is a source now, and the line can be looked up in it. The channel stays unrecorded all the same: the line says a channel can be selected, gives no range and names no setting. The feature line about the TD-3-MO's filter above is still described and not quoted, because that page is not a source here.

ALSO NOT RECORDED: velocity, since the note-on row carries a vv byte in [0, 7F] and no page says the instrument does anything with it; the note range; and how far pitch bend bends, which the table gives as a 14-bit value in [0, 3FFF] without naming a depth in semitones or saying whether one can be set.

THE FORMAT CANNOT YET SAY THAT PITCH BEND IS RECEIVED WHILE NOBODY HAS ESTABLISHED ITS RANGE, which is the one fact this instrument publishes that nothing below holds. voice.pitch_bend carries a depth and whether the depth is settable, and both are unknown here while reception is printed in the maker's own table. It is recorded in this sentence instead.