dr.Godfried-Willem RAES

Ghislain & Dierik POTVLIEGHE

Johannes TAELMAN


2005 - 2008

onderzoeksprojekt rond de ontwikkeling van nieuwe expressiemiddelen van de School of Arts, Hogeschool Gent

[Nederlandstalige versie]

Quartertone organ module: <Qt>

The research and development of this instrument building project is an intensive collaboration in the realms of artistic research under the auspices of the University College Ghent (Hogent), department of music and the Logos Foundation. On the practical level, this organ is the result of an intensive collaborative effort by Ghislain Potvlieghe (specialist in historic organ building) and dr.Godfried-Willem Raes. The development of the PIC firmware in assembly language used to control this robotic organ is in the hands of Johannes Taelman. Research into practical and historical repertoire for such a novel quartertone instrument, was done by Sebastian Bradt (Hogeschool Gent). The driver software developped within the GMT language was written by Kristof Lauwers (Logos Foundation).

The main purpose was to build a wide ambitus organ register tuned in equal temperament quarter tones. The instrument should be capable of automatic playing, controlled by computer software, as well as through a set of two conventional keyboards. The whole instrument had to be easily transportable, dictating the use of brass and or pure tin/antimony pipework instead of the traditional lead/tin alloy. The range covers 6 full octaves (from midi note 36 to 108) . Wind pressure modulation can be used to expressive purposes. Tremulant effects as well as accented notes are a possibility. Subtle and individual touch velocity control is implemented.

The pipes are constructed after the following early prototype, made of brass and sounding midi note 77:

For the actual realisation, we decided to make use of an alloy composed of 95% tin and 5% antimony, leading to mechanically quite strong and leadfree pipework that also will hold tuning very well. Experiments relative to the mechanical strength of Sn-Pb soldered brass pipes have shown that these are pretty sensitive to breaking under shock conditions. The heterogenity of the materials being at the origin of this weakness. The picture below shows the labium and pipefeet for the prototype pipes sounding notes 36 and 48.

Qt is internally controlled by thirteen PIC processors. Twelve processors (PIC 18F4620-I/P) are used to control each set of 72 solenoids used to operate the valves in the windchest, whereas the thirteenth PIC (18F2525-I/SP) steers the compressor motor functions and the windpressure modulation. For the latter two precize stepper motors are used to drive valves inside the conducts leading to the upper and the lower windchest. We opted for the MicroChip PIC18F4620 because of its 64kByte flash program memory (good for 32768 single word instructions) , 3986 bytes static RAM, 1024 bytes Eeprom and last but not least, its 36 available I/O pins. With these specs it became possible to use a single chip to control 14 notes, since each note requires 2 I/O pins. The PIC 18F2525 follows the same architecture but has only 25 I/O pins and 48kByte flash program memory. Enough in any case for the requirements of motor control and feedback in this design.

The upperplate of the windchests under which the solenoid valves are mounted, is made from very thick african obeche (or wawa) wood. This wood is very good for wood cutting and carving, an essential feature since we had to bring the wood to an exact and airtight matching with the curved metal work, using chivels.

The arrangement of the lowest octave of the pipes on the lower windchest is as follows: The pipes 47 and 47.5 are in the very front of the instrument. The arrangement for the high windchest is: The notes are numbered in fractional midi, note number 60 corresponding to middle C on the piano keyboard.

The entire Qt module is mounted on 4 large wheels and thus is, although pretty heavy (ca.220kg) , also very transportable. All metal work for the two windchests and the entire chassis was realized using the TIG welding process (manual) on AISI 304L stainless steel.

Technical specifications:

sizes: length: 2000mm, width: 700mm, height: 1760mm.
power: 230V - 950Watt (peak)
weight: 226kg
Control: midi protocol, using 2 channels. Note ON/OFF with velocity implemented. Volume controller (7) used for wind pressure. Controllers nr. 1 and 2 for wind valve control and tremulant effects. Other controllers may be added later. Direct control via UDP/IP network is under development.

Wiring overview:

Electronic circuitry:

For <Qt> either the use of pallet valves or conical valves in the windchest were first under investigation. We were after a maximum touch sensitivity for this organ, such as to make it sound like an ancient mechanical organ positive, where the player can in fact control a little bit the onset of the pipe sounds by modification of the speed wherewith the keys are depressed. Conical valves, in combination with softshift solenoids do indeed work extremely well to this end. (cfr..the use of a single conical valve in our automated saxophone). However, each valve has to be made on the lathe and the cones have to match perfectly well with the conical holes in the windchest upper plate itself. Not an easy matter, particularly since wood was proven to be unsuitable. A massive upperplate from metal was investigated but rejected due to the extraordinary weight it would add to an instrument that was meant to be transportable. Synthetic materials such as glassfiber reinforced epoxy were tried, but are not commercially available in the sizes we needed. Thus we decided to go back to our well known pallet valves. They can be seen on the picture showing the internal view of the upper windchest (notes 48-108). To get a better sensitivity to velocity, we modified the spring tension. Normally a spring is used to keep the valves just closed at rest. So the force is just a bit larger than the downwards force exerted by the weigth of the valves, since they are mounted upside down. By increasing this force above this minimum required for keeping them closed at rest, the valves gain a more gradual opening traject, at the detriment however of some response speed. The circuit we developed for achieving a bit of velocity sensitivity from these solenoid operated valves looks like this:

This circuit is a further development of the circuits we used in the design of our <player pianos> and <Harma>. The PNP transistor we used in these earlier designs is here replaced with a p-channel mosfet (BS250 or similar type). The gate goes directly to ground and the zenerdiode, essential in the bipolar design, could be left out. In this new design the negative pulse driver mosfet has to be a logic level type, identical to the hold-driver mosfet. Note that, due to the gate capacitance of the negative pulse mosfet (ca. 1.5nF), there is a inherent timeconstant in the order of 7 microseconds. Since the PIC's minimum pulse duration is 19.2 microseconds, this does not lead to a performance penalty. A single PIC controller (Microchip PIC 18F4620 - I/P) can steer 14 notes. For the low windchest we thus use two boards, leaving four outputs unused for notes (they can be used for light effects). For the high windchest, ten boards in total. The prototype PC board looks like:

It was developed in Eagle software for PCB design and produced from the datafiles at Eurocircuits in Hungary. (Plot & Go). The first batch had a wiring bug, repaired in our second (gold layered) production batch.

The midi input board is designed after the following schematics:

Here we have merely the compulsary optocoupler input with reverse polarity protection diode. The selection of resistors around the optocoupler is quite critical if you want to achieve optimum speed and no data loss even with midi back-to-back data streams. In the practical realisation we substituted the 6N138 type originally specified, by a 6N137, leading to a fivefold improvement in speed. Next the inverted output signal, on TTL levels, is fed to 10 inverter buffers which will drive the many note valve steering boards:

To allow users easy star-like connections to many more robots and automats, we provided this circuit also with no less then 5 real midi-thru drivers and corresponding 5pole din connectors:

Still on the very same PC board, we found place for a PIC microcontroller to be used for the control of the motor functions, lights, tremulant and other optional effects:

The board offers buffering for the midi signals to be fed into the very many pulse/hold driver boards as well as a single PIC controller (PIC18F2525-I/SP) , in charge of wind control for the radial compressor. Extra bit outputs are provided for the expression control: two stepping motor driven tremulants etc. Two pins are configured as inputs and used to read the output of the MLX90316 Hall Effect based rotary position sensor IC's. These are mounted on the axis of the wind valves.

At last, almost as a reminder, however an essential one, DC voltage buffering and decoupling with ample 100nF caps has been foreseen in the design:

The actual external wiring for this PC board looks like:

The specifications of the pallet solenoids used in the windchest are: 12/14V dc, 75 Ohm resistance, 190mA current. The Laukhuff order numbers are given in the list of materials at the end of this document. Cost: 3800 Euro. The original springs in the valves (0.4 x 3.4 x 26) were replaced by custom made stronger types by us (0.5 x 4.5 x 25) , this to allow for note aftertouch control.

The windvalves in the lower and upper windchest can be used both as windpressure expression controls and as tremulants. These valves are controlled by stepping motors and controllers wired after this schematic:

If we take the static windpressure in the windchest as 150mm H2O (15mB, 1.5kPa or 1500N/m2), then the worst case static force on the windvalves (100mmx150mm = 1.5dm2 in surface) is 22.5Newton. This force however only applies if the other side of the valve is fully opened.. Such circumstances could only exist under full cluster conditions, where the wind pressure will collapse anyway. Under normal conditions, the pressure difference between both sides of the valves will be limited to ca. 5%. In addition to this consideration it must be noted that in our design, very much on purpose, the valves do leak. Otherwize, they would fully interrupt the tones produced rather than cause a smoorth amplitude (and a bit of frequency-) modulation. So the stepper motor we selected should under all normal circumstances be powerfull enough to rotate this valve. The manufacturer gives following force curves for the given (Sanyo) controller and stepping motor combination:

Since the current through the motor winding will fall back to 50% of the normal current within 200ms after the last pulse on cw or cww is received by the controller, the slowest possible tremulant frequency will be ca. 0.06Hz. The pulsing frequency would than be 6Hz. This is also the slowest possible fluent speed of any valve movements. The pulse timing requirements and constraints are summarized in the graph below:


The specifications for the motor and compressor are:

Note that the operational pressure is higher. Qt was intonated and tuned on a pressure of 140 mmH20 (14 mBar).

MIDI-Implementation table for <Qt>

Midi command status byte byte1 byte 2 remarks
Note Off 128 + 5 note (36-108) 0 release not implemented
  128 + 6 note(36-107) 0 release not implemented
Note On 144 + 5 note (36-108) velocity velo=0 = note off
  144 + 5 note (36-107) velocity velo=0 = note off
Poly Aftertouch 160 + 5 note (36-108) note pressure only if ctrl.69 is ON
  160 + 6 note (36-107) note pressure only if ctrl 69 is ON
Program Change 192 + 5

velo lookup selection
122 = default
120-127: tables

- look ups are sysex programmable
  192 + 6 velo lookup selection, as above - look ups are sysex programmable
Controller 176 + 5 or + 6 1 low windvalve position 0= closed, 127 = opened
  176 + 5 or + 6 2 high windvalve position 0 = closed, 127 = opened
  176 + 5 or + 6 7 wind pressure motor normal value = 102
  176 + 5 or + 6 8 light strenght blue LED spotlites
  176 + 5 or + 6 11 low tremulant speed normal value = 96
  176 + 5 or + 6 12 high tremulant speed normal value = 96
  176 + 5 or + 6 66 0 or not 0 wind motor on off switch
  176 + 5 or + 6 67 0 or not 0 low valve motor power off
  176 + 5 or + 6 68 0 or not 0 high valve motor power off
  176 + 5 or + 6 69 0 or not 0 enable/ disable poly aftertouch
  176 +5 or + 6 70 0 or not 0 enable/disable hold (stacc. mode)
  176 + 5 or + 6 123 0 all notes off
Channel pressure 208 + 5 low windvalve position - same as ctrl 1
  208 + 6 high windvalve position - same as ctrl 2
Lights: 144 + 5 35 0 or not 0 lights lower windchest K-side
  144 + 6 35 0 or not 0 lights lower windchest Q-side
  144 + 5 110 0 or not 0 light tail upper windchest
  144 + 5 111 0 or not 0 light Kside upper windchest
  144 + 6 111 0 or nor 0 light Q side upper windchest
  144 + 5 or +6 120 0 or not 0 light BIM box 5V supply
SysEx       to be documented later
        used to program different velo scalings

Music composed for <Qt>:

Back to Logos-Projects page : projects.html >Back to Main Logos page:index.html To Godfried-Willem Raes personal homepage... To Instrument catalogue Naar Godfried-Willem Raes' homepage


Kwarttoons orgel : <Qt>

Qt is de roepnaam van een postdoktoraal experimenteel artistiek onderzoeksprojekt, officieel gestart op 1 oktober 2005, onder de auspicien van Hogeschool Gent in samenwerking met Stichting Logos. De wetenschappelijke ondersteuning en praktische leiding was in handen van dr.Godfried-Willlem Raes, docent kompositie, akoestiek, klankonderzoek en onderzoeksmetodiek aan het departement muziek en drama. Dit geautomatiseerd orgel kwam praktisch tot stand als een samenwerkingsprojekt van Ghislain Potvlieghe (specialist historische orgelbouw), Dierik Potvlieghe (orgelbouwer), Johannes Taelman (ontwikkeling en research inzake mikrokontrollers voor de besturing), Sebastian Bradt (onderzoek naar aktueel en historisch repertoire op het gebied van de kwarttoonsmuziek en komponist van nieuwe muziek voor deze automaat), Kristof Lauwers (ontwikkeling aansturingssoftware ten behoeve van komponisten) en dr.Godfried-Willem Raes.

Het opzet bestond erin een zuiver getemperd kwarttoonsinstrument te bouwen met een grote tessituur: minstens 6 oktaven (midi 36 - 108). Verder gaat het in dit ontwerp om een zuiver akoestisch instrument, met pijpen dus. De luchtdruk is ten dienste van de expressieve mogelijkheden op dynamisch vlak goed en snel moduleerbaar, terwijl ook aksenten en tremulant mogelijk zijn. De tremulant werd in twee onafhankelijke systemen uitgevoerd, waardoor de bas een andere tremulant frekwentie kan krijgen als de diskant. De akoestisch opgebrachte geluidsdruk diende bij dit ontwerp voldoende groot te zijn om in een orkestrale onversterkte bezetting probleemloos te kunnen worden ingezet. Bovendien diende de stemming zo vast te zijn, dat regelmatig bijstemmen van het instrument niet nodig zou zijn. Daarom gingen we bij het ontwerp uit van aan de bovenkant dichtgesoldeerde pijpen. Transportbestendigheid en verplaatsbaarheid waren een belangrijke eis. Deze eis sluit bij voorbaat het gebruik van klassieke legeringen voor orgelpijpen (60% tin, 40% lood in het beste geval) uit. Deze zakken immers bij blootstelling aan trillingen, onvermijdelijk bij verplaatsingen, door onder de druk van hun eigen gewicht. Ook louter ekologische (en sedert kort ook Europees wettelijke) overwegingen pleiten trouwens tegen het gebruik van lood. Proeven werden gedaan met messing enerzijds en legeringen bestaande uit 95% tin en 5% antimoon anderzijds. De met Sn-Pb gesoldeerde messingpijpen bleken erg gevoelig aan breuk op de soldeernaden. (cfr. <Puff>). Hardsolderen of brazeren ware ook mogelijk, maar technisch erg lastig uitvoerbaar gezien de vereiste precizie. Lassen van messing, de meest superieure techniek, bleek -althans voor ons- zo goed als ondoenbaar omwille van de onstabiliteit van de legering (het zink oxydeert namelijk). Uiteindelijk opteerden we dan maar voor de volledig loodvrije pijpen uit tin en antimoon, een legering waarmee ook traditionele orgelbouwers goed overweg kunnen. Omwille van de gewenste hardheid, opteerden we in ultimo nog voor de toevoeging van 1% koper in de legering: 95% tin, 4% antimoon, 1% koper werd het dus uiteindelijk.

Omwille van de gewenste mobiliteit van het instrument, werd het integraal op goed stuurbare grote wielen met massief rubberen banden gemonteerd en werden de afmetingen toch zo kompakt mogelijk gehouden. Deze voorwaarde dikteerde welhaast automatisch het gebruik van gedekte pijpen evenals een betrekkelijk hoge winddruk (100 tot 150 mm waterkolom, of 10 tot 15 mBar).

Het chassis en alle metalen delen voor de beide windladen en de kondukten werd uitgevoerd in minstens 3 mm dik roestvast staal AISI304L en AISI316 onder gebruikmaking van het volledig manuele TIG lasproces. Als toevoegmateriaal bij het lassen werd uitsluitend AISI316 gebruikt. Het snijwerk werd uitgevoerd met een hoogfrekwent plasma brander. Voor de afwerking werd volstaan met een grove borsteling met een inox komstaalborstel. Aan de binnenzijde werden alle lasnaden, omwille van het beperken van ongewenste turbulenties in de windstroom, zorgvuldig gladgeslepen. Gezien deze opties in materiaalkeuze, moet het instrument kwa duurzaamheid zeker aan alle stelbare eisen kunnen voldoen.

De volledige MIDI-implementatie, ten behoeve van eenvoudige gebruikers, publiceerden we aan het eind van de engelse tekst hierboven. Deze implementatie maakt ook een manuele bespeling via twee elektronische midi-keyboards eenvoudig mogelijk. Voor het speelkomfort is het wel aangewezen keyboards te gebruiken met een tessituur van zes oktaven en met individuele touch sensitivity. De beide keyboards worden via een midi-merger rechtstreeks op <Qt> aangesloten. Een rechtstreekse netwerkbesturing onder gebruikmaking van UDP/IP is eveneens voorzien. Dit is vele malen superieur aan Midi, maar in gebruik alleen weggelegd voor technologisch onderlegde komponisten en musici. Voor wie met hetzij PD hetzij GMT overweg kan, is dit het aanbevolen besturingssysteem.

Gedetailleerde berekeningen van de mensuren en dimensionering voor de bouw van het Qt register:

Eerste proef, berekening op grond van proefpijp uit messingbuis. Mensurering rekening houdend met handelsdiameters en korresponderende materiaaldiktes van bestaande messing buis.

link to look up table - naar overzichttabel

Tweede proef, berekening op grond van proefpijpen gesoldeerd uit een legering 95% tin en 5% antimoon. De breedte van het labium werd hier bepaald op grond van 1/4 van de omtrek van de cilindrische pijpen.

link to look up table - naar overzichttabel

Derde proef, berekening op grond van 5 proefpijpen (36,48,60,72,84) gesoldeerd uit een legering 95% tin en 5% antimoon. Winddruk 85mm H20. De breedte van het labium werd hier bepaald op grond van 1/4 van de omtrek van de cilindrische pijpen. Alle pijpen krijgen hier een individueel verschillende mensuur.

link to look up table - naar overzichttabel

Berekend door dr.Godfried-Willem Raes op 13.03.2006 naar prototypes gebouwd begin maart 2006. Het volledige register werd in elkaar gezet en gesoldeerd door Ghislain Potvlieghe in diens atelier. De eerste verzameling pijpen (120) was klaar op 22 mei 2006. Het kompleterende oktaaf 96-108 op 29.06.2006. De stemming en intonering vonden plaats in verschillende werkfazen na de montage van de pijpen op de afgewerkte windladen. (Oktober 2006- Maart 2007) De gebouwde pijpen werden, na het stemmen (met dichtsolderen aan de bovenzijde) en intoneren opnieuw exakt opgemeten en vergeleken met de berekeningen.

link naar definitieve en samenvattende tabel van de mensurering voor Qt.

Bouw- en research dagboek:



Nog verder af te werken komponenten en nog uit te voeren research:

(Terug) naar logos-projekten:


Terug naar Logos' index-pagina:


Naar Godfried-Willem Raes personal homepage... Naar katalogus instrumenten

gebouwd door

Godfried-Willem Raes

Terug naar <Gorgel>

Technical data and circuit drawings:

Ordering data and parts list for essential components used in the construction:


Wiring diagram for the midihub board in QT:

Wiring diagram for the windvalve controls:

Analog 5V power supply:

Power supplies:

Motor Controller signals:


Motor controller parameters (parameters not mentioned here retain their default value)

Wiring tables and color coding for the lower and upper windchests

Power supply requirements:


Lay out of the pipes on the windchests:

Assembly and disassembly instructions and remarks:

If the high windchest has to be opened, always first remove all pipes. Next remove the connectors from the boards K2-K6 and Q2-Q6. Now, using a hex key unloosen the 80mm long M6 bolts (never turn the nuts!). Number the nuts and put them in a safe place: they are individually different (long stainless steel bus-connection nuts with one end cut under a variable angle between 30 and 45 degrees) and have to be reassembled the same way. Remove the tending rim and put it aside. Now loosen the windchest gently. Dont insert screwdrivers in the gap since that will lead to permanent damage and leaks. If a tool seems required, use a 40cm long, 50mm wide and 2mm thick blade of stainless steel with rounded edges. The windchest only has to be opened in the following cases: to repair persistent leaking pads, to get access to the tremulant valve, to repair solenoid valves that burnt out as a result of an external short circuit or a software failure. It takes two people to lift the windchest vertically and flip it over. Take care not to touch the pads. They get very easily misalligned.

All pipes are marked on the right side of the labium. C0 corresponds to pipe 36. All quartertone pipes are marked with an additional + sign.. Thus c''+ means pipe 60.5. If pipes are removed they have to be handled very carefully. Put them in low crates on a soft woolen surface. Do never stack them in top of each other. Never touch the tuning ears nor lean the pipes against them.

Last update: 2021-04-26 by Godfried-Willem Raes