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.

Synthstrom Audible Deluge MIDI implementation

The MIDI implementation of the Synthstrom Audible Deluge: what it answers to, on firmware 4.1.0.

At a glance

Firmware
This page describes firmware 4.1.0.
MIDI channel
1 to 16.
MPE
Sent and received: its voices are spread across its MIDI channels, a note to each.
Voices
Not recorded.
Note range
Not recorded.
Velocity
Sent and received.
Aftertouch
Channel pressure and polyphonic, each answered. It sends aftertouch too.
Pitch bend
12 semitones, and the range can be changed.
Program change
Sent.
MIDI clock
Sent and received.
Start and stop
Sent and received.
Control change
Assigned by the player, so there is no fixed chart.
NRPN
Not recorded.
System exclusive
None.

Where these facts come from

Source documents

How they were read

Deluge Official Guidebook, 338 pages, read in both of the editions Synthstrom publishes: the OLED edition whose cover says OS 4.1.0, and the numeric-display edition whose cover says OS 4.0. The MIDI chapter is chapter 12 at printed pages 235-264, and everything numeric in this file comes from four of its pages and one page of chapter 14: the MIDI Implementation Chart at p. 241, the MIDI settings table at p. 240, the program change and bank select ranges at p. 253, the output channel range at p. 254, the controller range it can be set to send at p. 255, and the system settings table at p. 276 for the pitch bend range and the default velocity. MPE is pp. 258-263. Plus the product page, the printed manual's own page, the open-source page, the news listing and one forum thread, each named below for what it settles.

THE TWO EDITIONS AGREE ON EVERY MIDI NUMBER, WHICH IS A MEASUREMENT AND NOT AN ASSUMPTION. A maker that publishes two editions can print two different numbers for the same row, which is what the OP-XY's did. So both guidebooks were read by the same scripted method and compared: the implementation chart comes out identical, row for row, all 24 rows in all 9 groups, and the whole MIDI chapter differs only in how the display is quoted - the OLED edition writes a menu as TEMPO MAGNITUDE MATCHING where the numeric one writes MAGN, forty-odd times over. Two differences are not about spelling and neither touches a number: the OLED edition's MIDI sound editor has a PLAY / DIRECTION row the numeric one does not list, and the worked example on p. 257 uses channel 1 and note 0 where the numeric one uses channel 10 and note 4 - with both printing the same ranges, 1-16 and 0-127, which is two independent printings of the ranges this file records.

THE CHART'S TWO COLUMNS DISAGREE ON SEVEN OF ITS ROWS, SO IT WAS READ BY POSITION. Its columns are Transmitted and Recognised, and they differ wherever the instrument is a sequencer rather than a synth: program change and both bank select rows go out and are not answered, polyphonic aftertouch is answered and not sent, All notes off is answered and not sent. A reading that took the words in the order the file stores them would have been right about the seventeen rows where the two columns agree and silently wrong about exactly the seven that carry information. So each cell was placed by its horizontal position between the table's own ruled column edges, and the group column was checked by arithmetic: a merged label like System real time is drawn once, centred on the rows it covers, so matching the nine labels to nine consecutive blocks of rows is a sum that can be repeated rather than a judgement about which label looks nearest.

THE SECOND READING FOUND A CELL THE FIRST HAD WRONG, WHICH IS WHAT IT IS FOR. A second reader was given the document, the table's heading and the two warnings above, was not shown the first reading, and was required to use a different library. It reached the same 24 rows, the same nine group labels with the same spans - 2, 2, 1, 1, 2, 1, 4, 5, 6 - the same column order and the same page, by a different route: it recovered the grid from the ruled lines in the content stream, where a full-width rule cuts through the group column and a short one divides rows inside a merged group cell, and reported the centring relation separately as a confirmation rather than as its method. And it disagreed about one cell. The footnote under the chart, "*Only if MIDI Clock output is enabled" (guidebook_oled p. 241), wraps onto a second line, and the first reading skipped lines beginning with an asterisk - so it skipped the first line of that footnote and attached the second to the chart's last cell, which came out as System real time / Reset transmitting No output is enabled. That is corrected and the script now stops a column at its first asterisk. Nothing in this file ever depended on that cell, which is luck rather than diligence.

TWO THINGS THE SECOND READER ESTABLISHED THAT THE FIRST HAD NOT. This PDF contains no space glyphs at all: a word space is a kerning adjustment inside the text-showing operator, so what any extractor hands back depends on whether it rebuilds spaces and at what threshold - the plain reading of one row is MIDItimecodequarterframe. The two readings used different thresholds and agree on every cell's spacing anyway, which is why the comparison measures letters and digits first and reports spacing separately. And the chart arguably has 25 rows rather than 24: the footnote sits inside a ruled band of its own between the last row and the table's bottom frame, so a reading that counts ruled bands gets 25. It is a footnote - no group label, no sub-label, no cell in the other column, and it opens with the asterisk four cells above it carry - but that is a judgement and is recorded as one.

THE PRINTED PAGE IS SIX LESS THAN THE PAGE OF THE FILE, CHECKED ON EVERY PAGE THAT HAS ONE. 338 pages by pypdf and by PyMuPDF, which agree; file offers no count for either edition. 312 of them print a folio and all 312 are out by exactly six, in both editions - so the offset is measured rather than sampled, which is what the Muse's table of contents taught. The 26 without one are the front matter and the chapter dividers. That 312 is also the maker's own figure for the book - "A5 size, full colour, 312 pages" (manual_page) - while the product page calls both editions "This 300 page, full colour, step-by-step manual" (product_page), so the round number is the one to disbelieve.

THE OFFICIAL GUIDEBOOK'S LAST CHAPTER IS THE USER-WRITTEN COMMUNITY GUIDE, REPRINTED. Chapter 15 is headed COMMUNITY GUIDE and runs from printed p. 295. Set beside the separately published "Deluge Community Guide" (product_page) - which the product page links in the same list as the two guidebooks, described as "A quick reference guide to the Deluge, adapted from contributions from our users" (product_page) - its pages 3 to 23 map one to one, in order, onto pp. 296-316, at 91% to 98% of their words in common, and its pages 24 to 27 map one to one onto pp. 317-320; on all 26 reprinted pages the guide's own running head and folio are replaced by the guidebook's running foot and folio. Its page 2, which begins "The Following people have contributed to the making of this community guide" (community_guide) and names six of them, is reprinted at the chapter's end, p. 321, without its heading. Only the guide's cover and its last page, p. 28, are not reprinted.

THAT CHAPTER ALSO CONTRADICTS THE OFFICIAL CHAPTER BY ONE, WHICH IS WHY IT IS CITED AND NEVER READ FROM. The system settings table says of the default velocity "Sets the default velocity for newly entered notes between 1-127. Default is 64" (guidebook_oled p. 276); the community chapter says of the same setting "Default is 64, Range is 0-127" (guidebook_oled p. 299). Velocity 0 is a note off, so the official chapter is right and the reprinted one is wrong - and the disagreement is inside one document, between two chapters with one cover. No number in this file comes from printed pages 295 to 322. The only MIDI figure in that chapter, "Select MIDI Channel 1-16" (guidebook_oled p. 313), is one chapter 12 states anyway, so nothing was lost by refusing it.

NO DOCUMENT STATES WHAT NOTES THIS INSTRUMENT SOUNDS, so there is no note_range. The guidebook prints 0-127 twice and neither printing is about the engine: p. 255 gives it as the value range of a controller the gold knobs send, and p. 257 as the note number a kit row can be set to send to an external device. That is the Iridium's and the Blofeld's trap in a third document - a range that bounds where something may be put, read as what the instrument answers to - and it was searched for properly before being given up: octave, transpose, note number, lowest, highest, C-2, C-1, G8 and G9 across all 338 pages of both editions.

NEITHER IS POLYPHONY A NUMBER HERE, AND THE MAKER PUBLISHES TWO CEILINGS. The guidebook says "Deluge's maximum voice count relies on the CPU loading and processing power available, but around 64 is the limit for most basic synth sounds" (guidebook_oled p. 99), and the product page claims "Up to 64 synth voices may play simultaneously" (product_page) in one line and "Upto 90 unaffected sample voices may play simultaneously" (product_page) in another. Two engines, two CPU-bound ceilings, and one field that holds one count: recording either would turn a limit that depends on the patch into a figure a consumer could draw a voice meter from. So polyphony is absent. The polyphony menu's own options - POLYPHONY, MONOPHONIC, AUTO, LEGATO, CHOKE (p. 99) - are modes rather than counts, so voicing_modes does not fit them either. Unison is a separate setting, "Number of voices from the maximum of 8" (p. 100), and is not polyphony.

THE FIRMWARE IS OPEN SOURCE AND THE COMMUNITY BUILD IS NOT WHAT THIS DESCRIBES. "The firmware for the Deluge is now open-source" (open_source_page), says the maker's own page, with "Community version 1.0.0 (Amadeus) is live now!" (open_source_page). Nothing here was read from that build or from its documentation, and a community release may add a MIDI answer this file does not carry; model.firmware says which firmware this is, which is the only honest way to hold that open.

WHAT THE CHART SAYS NO TO, recorded here because a no is a reading too: no system exclusive in either direction, no MIDI time code quarter frame, no song select, no tune request, no active sensing, no reset, and of the channel mode messages only All notes off, which is recognised and not sent. The word SysEx and the phrase System Exclusive appear nowhere in either guidebook outside that one chart row. The chart prints no Basic Channel row and no Mode row, which a full implementation chart would, so mode is absent rather than guessed at; and it has no NRPN or RPN row either. NRPN is named once in 338 pages and about other people's instruments - "Some synths and devices may use the bank or sub bank to change the bank, They also may be set for NRPN instead of CC messages" (guidebook_oled p. 253) - so nrpn is absent as well. An abridged chart's silence is not a statement of absence.