Musical
Robot
|
<Toypi>
a robotic toy piano
Godfried-Willem
RAES
2008
|
Definitely, I am not the first designer to build an automated toy piano. My
friend and colleage Trimpin has -as far as I know- been the first to deliver
a good working programmable acoustic toy piano. Our own design presented here,
however started completely from scratch. First we removed and saved the internal
harp (clamped rods mounted on a cast iron bar) from a 35 note chromatic toy
piano made by Antonelli (Italy). We designed a completely new soundboard, replacing
the original plastic construction. The new soundboard was made from hardened
brass, the same type as we used before in <aeio>,
our robotic cello. In contrast to the robotic cello design however, here we
did not clamp the soundboard on its circumference, but we mounted it free swinging,
using elastic material for mounting in the piano chassis. This lowers the resonant
frequency for the mimimal surface dictated by the design here. The soundboard
operates more or less as a Chladni plate. It also contributes greatly to the
damping of mechanical noises. To preserve the typical sound, we kept the original
design for the small wooden hammers. The keywork was completely replaced by
a tubular solenoid assembly, controlled by a couple of PIC microprocessors.
The maximum sound volume of the instrument is pretty limited. We could not change
this, since sound volume is inherently connected to the sizing of the rod assembly.
Louder sound would dictate thicker as well as longer rods. As to the electronic
hardware, we used the same printed circuit boards here as developed earlier
for <Xy>, our robotic quartertone xylophone. The
boards were mounted at the spot where you would normally expect the keyboard.
The power supplies found a place under the soundboard. The general shape of
the instruments chassis follows closely the typical shape of a normal grand
piano, although in this case, it was made entirelly using welded stainles steel.
It stands on three sturdy legs. The instrument listens to midi commands and
very precise velocity control is implemented.
The acoustics of toy piano's are quite a complicated thing to approach mathematically.
The acoustic laws to calculate the fundamental frequency of a rod clamped at
one end only apply in part. As one applies the usual formula for the fundamental
frequency f = k ./ (l^2) you will always get very low frequencies, apparently
unrelated to what you hear as a result [1]. This is because the apparent perceived
pitch corresponds in fact to one of the very inharmonic upper partials of this
fundamental. After the literature the partial series goes like this: f , 6.267
f, 17.55 f, 34.39 f etc... This explains why it is even for well trained musical
ears, quite difficult to judge the pitch unambiguously and it certainly explains
why it is pretty impossible to build a toy piano with a consistent ambitus of
say larger than three octaves.
At first sight, it may appear to be a bit silly to spend all the effort and
money to automate such a cheap instrument as the toy piano. The building costs
are about a hundreth times the cost of the toy piano itself. But at the other
end, there appears to be quite some serious music literature for the toy piano...
Margareth Leng Tan even devotes a large part of her career to concerts on this
instrument! After all, one must confess its sound is quite unique. Realizing
this, it is obvious that the toy piano is quite clumsy to play professionally:
not only the keys are undersized for normal hands, but also the mechanics are
pretty unreliable. By making a robot player toypiano, it becomes possible to
play the toy piano via the interface of a normal touch sensitive keyboard, a
hitherto unimaginable possibility. Of course, <Toypi> can also be completely
computer controlled and used in interactive applications. As an alternative
to MIDI control, we also implemented UDP/IP control.
A series of pictures shot during the construction (in chronological building
order):
Download high resolution picture
Power supply voltages and currents:
- +24 V dc / 5.6 A for the valve solenoids (SMPS)
- +5V/ 500mA for processor boards (analog)
Midi Mapping and implementation:
Midi channel : 10 (0-15).
Midi note range: 72-107. Note on, velocity is implemented and has a wide control
range.
Note Off commands are not required, but can be sent with no effect.
There is an intrinsic relation between the applied velocities and the maximum
note repetition rate. The table below clarifies this:
velocity patch 0 |
repetition rate
(note/second)
|
remarks |
1 |
120 |
limited only by the inertia of the mechanics |
2 |
120 |
limited only by the inertia of the mechanics |
4 |
120 |
limited only by the inertia of the mechanics |
8 |
80 |
resonances with the rods can occur |
16 |
40 |
|
32 |
20 |
|
64 |
10 |
|
127 |
5 |
|
Note that high repetition rates in combination with velocities higher than
what's strictly required, can be detrimental to the instrument. Moreover, it
sounds very bad also because the beaters will stick to the tone rods..
Lights:
- yellow LED lights on the keyboard, from left ro right mapped op notes 60-67.
(On/Off only)
- blue LED lights underneeth, mapped on notes 48-53 (On/Off only)
Program change: implemented. Different velocity scalings are mapped on program change numbers 122 to 127.
The default scaling is obtained with program change 122. With this default scaling, the lowest velocities are only usefull for fast repetitions. For single notes the velocity should be at least 20 for the note to sound. Avoid using high velocities at high repeat rates!
Technical specifications:
- size: width: 50 cm , height: 30 cm, length: 50 cm (exact sizes to be determined)
- weight: 10kg. (without case)
- transportation: in flightcase (to be build)
- power: 230V ac / 250W (3-prong euro power connector mounted under the instrument!)
- Tuning: based on A = 440 Hz. The imprecision of the original tuning was
preserved.
- Ambitus: 72-107
- Polyphony: 10 notes (limited only by the peak power available from the power
supply)
- MIDI-in connector (DIN)
- 2 MIDI Thru connectors (DIN)
- Insurance value: 2900 Euro. (materials only)
Design and construction: dr.Godfried-Willem
Raes
Collaborators on the construction of this robot:
- Kristof Lauwers (GMT implementation)
- Johannes Taelman (PIC coding)
Music composed for <Toypi>:
- Godfried-Willem Raes, "Schroeders
dream", for <Toypi> and performer. (6')
[MP3 download]
- Sebastian Bradt, "Kallikantzaros" (2'40")
- Hye Kyung Lee, "Dream Play" (No-ri I, II, III, IV)
- Godfried-Willem Raes "Gestrobo
study for Toypi", for <Toypi> and invisible instrument. (naked
performer)
- Xavier Verhelst, in the works...
- Jonas Runa (10/2013)
- Kristof Lauwers 'PicRadar Study for Toypi' (2016)
Nederlands:
<Toypi>
In het bouwen van een automatische speelgoedpiano waren we beslist
niet de eerste. Mijn goede vriend en kollega robotbouwer Trimpin, was me ruim
voor op dit gebied. Mijn eigen ontwerp dat we hier verder toelichten, vertrok
evenwel geheel vanaf nul. We begonnen met het geheel demonteren van een 'Antonelli'
drie-oktaafs speelgoedpiano. Alleen de 'harp' hielden we integraal over uit
het origineel instrument. De klankkast en het chassis waren geheel uit plastic
en gooiden we in de vuilnisbak. Gezien onze positieve ervaringen met gehard
messing als klankbordmateriaal (cfr. <aeio>, lag
het voor de hand verder te bouwen op de daarbij opgedane ervaring en dus ook
hier een hard messing (MS58) klankbord te gebruiken. Deze keer evenwel geen
ingeklemd klankbord, maar wel een chladni plaat met vrije en elastische ophanging.
Daardoor kan de resonantietoon van het klankbord heel laag worden gehouden met
minimale oppervlakte. Voor de automatisering van de hamertjes gebruikten we
aanvankelijk kleine buisvormige trekmagneten voorzien van een terugkeerveer.
Dat vergde evenwel nogal wat tussenmechaniek en overbrengingen met alle bijgeluiden
vandien, zodat we reeds vrij vroeg in de bouw overschakelden op een aansturing
via omgekeerd werkende kleine duwmagneten, voorzien van een houten slagkogelje.
Een bijzonder ontwerpprobleem stelde de bijzonder kleine afstand tussen de toetsen
en dus ook tussen de klankstaafjes onderling: ca. 11mm slechts. Hierdoor was
het onmogelijk de vele spoelen naast elkaar op te stellen zoals nodig om de
staafjes op eenzelfde plek te kunnen aanslaan en zo een voor alle tonen gelijkblijvend
timbre te garanderen. De diameter van de kleinst vindbare degelijke elektromagneten
is ca. 14mm, wat ons noopte tot een geschrankte opstelling. Het pianochassis
volgt grotendeels het model van een grote vleugelpiano maar werd hier helemaal
in gelast inox uitgevoerd. Op de plaats waar men normaal het toetsenbord zou
verwachten, plaatsten we de beide mikrokontroller boards voor de automatische
besturing. Voor de voeding vonden we een plaatsje onder het messing klankbord.
De hele besturingselektronika kan uit het hoofdchassis gegleden worden nadat
de vier M4x10 boutjes aan de onderzijde van het 'klavier' zijn losgemaakt. Net
zoals bij een grote piano, steunt ook <Toypi> op drie stevige poten voorzien
van (gerecikleerde) stalen wieltjes.
De stemming van speelgoedpianos is altijd een beetje een heikel
punt. In het laagste oktaaf klopte de stemming van ons exemplaar vrij goed,
maar in het middengebied was ze ongeveer een kwarttoon te hoog. Het bijstemmen
van de staafjes is een vrij riskante onderneming, hoewel het principe erg eenvoudig
is: om de toon te verhogen moet het staafje gewoon een beetje worden ingekort.
Om te verlagen, moet aan het inklempunt in de gietijzeren staaf met een vijltje
rondom rond wat materiaal worden weggenomen. Bij de konstruktie worden de staafjes
overigens klem ingedreven in de gietijzeren drager, maar het iets verder indraaien
ervan leidt onherroepelijk tot een breuk. Het grote risiko is eigenlijk steeds
dat de staafjes gemakkelijk afbreken, want aan het inklempunt zijn ze erg dun
(ca.1.6mm). Ze zijn bovendien vervaardigd uit gehard staal en dat is van nature
uit erg broos. Overigens is de speelgoedpiano nooit helemaal 'goed' te stemmen,
en dit vanwege de extreme inharmoniciteit van het voortgebrachte spektrum. [1]
Op het eerste gezicht is het wel wat belachelijk om een automatische
speelgoedpiano te bouwen, immers niet alleen kruipt er toch bijzonder veel tijd
in de bouw en het ontwerp, maar ook zijn de kosten van zo'n automaat heel snel
zowat het honderdvoudige van de kost van een nieuwe speelgoedpiano op zich...
Maar, niet alleen bestaat er wel behoorlijk wat eigentijdse literatuur voor
het instrumentje, er zijn zelfs solisten die er het grootste deel van hun pianistieke
karriere aan gewijd hebben, zoals bvb. de onvolprezen Margareth Leng Tan. John
Cage schreef reeds in 1948 zijn 'Suite for toy piano'. George Crumb gebruikt
het instrument in 'Ancient voices of Children' uit 1970 en ook Mauricio Kagel,
John White, Christopher Hobbs, Karlheinz Essl, Hugh Shrapnel schreven er muziek
voor. Vanuit dat perspektief biedt ons automaatje wel degelijk perspektieven,
niet in het minst omdat het nu mogelijk wordt om aan de kreupele en praktisch
voor normale handen onspeelbaar kleine toetsen en de gebrekkige onbetrouwbare
mechaniek helemaal te ontsnappen. Dankzij de automaat is het nu immers mogelijk
de speelgoedpiano menselijk te bespelen middels een gewoon aanslaggevoelig klavier!
Dat komputerbesturing ook mogelijk is ligt daarbij natuurlijk voor de hand.
Hierbij kunnen trouwens de meest onwaarschijnlijke klanken aan het instrument
worden ontlokt. Zo is het bijvoorbeeld mogelijk kontinu aangehouden tonen te
spelen, door de repetitie van de noten op te drijven tot 64 en meer aanslagen
per sekonde. Uiteraard kan dit uitsluitend wanneer heel kleine velocity waarden
worden gebruikt.
Construction Diary:
- 2000-2007: Collecting different brands and types of chromatic toy pianos.
First ideas about toy piano robots. By the end of 2007 we had some 20 toy
pianos in our collection.
- 24.01.2008: First steps taken to implement toypi in our GMT programming
language: resource files prepared.
- 04.04.2008: Toypi support added in the menu structures for GMT. Declared
constants decided.
- 31.07.2008: First practical sketches and choice of components. Experiments
with hammer weights and different types of solenoids.
- 01.08.2008: Start construction of the hammer mechanism. We finaly decided
to go for small size Lucas Ledex push type solenoids with wooden mallets.
- 02.08.2008: Design and realization of the brass soundboard. This
soundboard ought to be suspended elastically in the main chassis. We could
use either strong helical springs or rubber mounts. At least, that is the
first experiment to be performed. In a 'normal' toy piano the rod bar is screwed
tightly either on the soundboard (if one is present) or on at least one of
the case panels.
- 03.08.2008: Construction of the beater heads. Here we use wooden balls,
turned to 10mm diameter. They are drilled with a centered 3mm hole. The stems
for the solenoid archors have 1.5mm diameter. We filled the 3mm hole with
a 10mm long red copper rod of the same diameter, and then drilled the 1.5mm
holes, using a precision Dremel drill in a stand.
- Forming and bending of the required stainless steel parts. TIG welding of
the main chassis in the shape of a concert grand piano, with three legs.
- 04.08.2008: Elastic mounting of soundboard proves to lead to an inferiour
sound as compared to a soundboard clamped at the two points going through
the rod harp. Thus we welded two 20mm x 10mm x 50mm supports on the innersides
of the chassis to support the soundboard.
- 05.08.2008: Construction of the PC board holder at the place of the keyboard.
This part too is made from bent stainless steel, thickness 2mm. It accomodates
the small transformer for the logic power supply and the mains entry. The
switch in mounted in top of the keyboard. The midi in and out connectors come
fully frontal. Icon added in the GMT programming context. Resource files recompiled.
Kristof Lauwers wrote a first test module for Toypi in GMT.
- 06.08.2008: Soldering of the required components on the two PIC microcontroller
boards. On the first board, the outputs 1 to 18 go to the note beater solenoids
72 - 89. The outputs 19 to 28 remain free for other functionalities. On the
second board, he outputs 10 to 28 connect to the note beater solenoids 90
- 107. Here the outputs 1 to 9 remain free. Board 1 is wired for midi input,
with optocoupler. Board two is a slave and gets the MIDI signal as TTL data
stream. The two midi thru sockets on the front panel have current limiting
resistors, according to the midi standard.
- 07.08.2008: Design and welding of the carrier chassis for the 24V supply.
This is mounted under the keyboard with four M3 screws and forms a functional
unit with it. In case we have to reject the use of a SMPS, we left enough
space to accomodate a 30W toroidal transformer and an analog supply with a
large buffer electrolytic capacitor. This would make the robot a lot heavier
however. Wiring of the solenoids started. We have already 18 notes that -electrically-
are ready to play. The microprocessor boards have to be placed on 5mm standoffs,
to avoid short circuits with the chassis. A piece of epoxy board cut to the
same size as the PC boards can be used as well.
- 08.08.2008: Finishing of the solenoid wiring. Construction of the wheels
under the instrument. The axe holders are turned on the lathe from stainless
steel 10x8 tubing cut to 25mm lengths, and then welded on the legs. After
welding we redrilled the holes to exactly 7mm diameter. We still have to turn
a groove in the wheelstems to prevent the wheels from falling out when the
instrument is lifted up. These grooves can then take small omega rings.
- 09.08.2008: PIC microprocessors (PIC18F4620 I/P type) programming session
with Johannes Taelman. Basically, the code is a copy of the firmware developped
for our <Xy> robot. The velocity scalings are
a bit lower however. The complete specifications for the PIC firmware are
in an ascii textfile. Design and construction
of a (removable) polycarbonate cover to prevent people from touching the tone
rods. This lid is connected to the main chassis with hinges. Toypi can play
its first notes. It has excellent sensitivity for very low dynamic levels
and nuances. The repetition speed is excellent: we can get up to 60 repetitions
a seconds at very low velocities, musically this sounds like a sustained note.
However, polyfony -as predicted from the power supply specs- is limited. A
remedy would be to use a buffer capacitor of 47mF, but there is just not enough
space in the design to place such a bulky component. Again we encounter here
the apparently inherent problems related to commercial SMPS designs. Even
with a load of only a single solenoid (drawing 650mA), we get in trouble as
soon as this load is switched at a fast rate. Fast, here is not even really
fast: the trouble starts already at about 6 transitions a second! So we will
have to look out for an alternative for our badly behaving Sunpower supply.
Probably not a bug, but in any case a somewhat strange behaviour: at switch
off, invariably note 76 sounds. Pin 16 of the PIC chip (output 24 on the board)
must make a high transition at power off. We noticed the same effect on <Xy>.
- 10.08.2008: Wheels secured with splitpins, size 1.6 mm. Omega rings did
not work, for they have too large a diameter. GMT test code for Toypi updated.
LED light design: blue LED's to shine underneeth, eigth bright yellow LED's
(5mm) on top of the keyboard panel. We cannot use LED strips, since we do
not have a 12V power source available here. The power supply needs to be recalculated.
If we have 36 solenoids at 25% duty cycle, this is equivalent to 9 solenoids
at 100%. Thus the power requirements become 9 x (24V/38 Ohm) = 5.68A. Thus
the power rating for the supply must be at least 24V x 5.68A = 130Watt. Thus
we ordered a new SMPS module from Farnell. Seems the design for the yellow
led's in the keyboard panel is not workable. It just cant fit. So we redesigned
it with the led's mounted on the Weidmueller connectors. Files prepared with
Bach preludes and fugues, adapted for Toypi. Midi-thru connections functionally
tested.
- 11.08.2008: Further work on the LED light assemblies.
- 12.08.2008: Placement of the new power supply module.
It can clearly be seen on the picture, sticking out. Tests and adjustment
of the beaters and the rods. Slight corrections of the tuning.
- 13.08.2008: Composition of "Schroeders dream", for Toypi and performer.
- 14.08.2008: Further work on 'Schroeders dream'. Sebastian Bradt also is
making a piece for Toypi. Implementation of Toypi in GMT further improved.
- 19.08.2008: Demonstrations of Toypi for Barbara Buchowiec, Hans Roels and
Xavier Verhelst.
- 20.08.2008: Construction of the two definitive sound transfer poles for
the rod assembly from 14x14mm machine steel. We drilled the required very
long 6mm hole on the lathe, using the square plate.
- 21.08.2008: Composition code for "Schroeders dream" finalized.
- 22.08.2008: Sebastian Bradt works on his solo piece for <Toypi>: "Kallikantzaros".
- 27.08.2008: Xavier Verhelst works on a piece with/for <Toypi>.
- 31.08.2008: Tests and evaluation of sonar control code for Toypi: Gestrobo
studies:'Toypi' chapter.
- 03.09.2008: Preliminary demonstration of the Gestrobo Study for Toypi at
the occasion of ISSSM.
- 04.09.2008: Left square steel standoff for tone rods ground to exactly 72mm
(was 73.5) for better sound production in the low notes. End surfaces machined
to good flatness and polished.
- 05.09.2008: Murphy's law did it again: the tone rod for note 76 (E) broke
off in an attempt to get a somewhat better allignment of the beaters on the
rods... It took us about 2 hours to carefully drill out the broken off end
of the tone rod from the transversal soundbar in cast iron thus freeing the
hole. We replaced the broken rod with a length of stainless steel M3 , grade
A4, threaded rod. After shaping one of the ends more or less conical and cutting
the rod to 215mm length (the original full length of the rod, diameter 2.5mm),
we screwed it tightly into the hole using a spare drill chuck. We got again
the original pitch, but lost quite a bit in resonance on that note. We will
have to go hunting for a replacement rod assembly. If there is anyone in outer
space owning an identical 'Antonelli' 3-octave toy piano in good shape, please
sell it to me, come to rescue!
- 13.09.2008: Murphy did it again: this time the tone rod for the highest
note (107) broke off during the rehearsals of 'Schroeders Dream'. We replaced
the tone rod with a rod of slighty larger diameter (3.2mm) removed from an
old broken Schoenhut toy piano, hammered it in, and tuned it correctly. This
note now sounds actually better than the original. When we get in trouble
again with note 76, we might use this approach again for the repair.
- 16.09.2008: Toy pianos are clearly no professional instruments... In the
first piece of the tonites concert, two tone rods broke off.. (72 and 77 this
time)
- 17.09.2008: Repair of tone staves for 72 (C) and 78 (F#) by replacing them
with staves from an old french made upright toy piano ('Opera', marque deposee).
Note 76 (E) redone and also replaced. We made spare staves already for notes
73,74,75, 76 in case these also break... The 'Opera' tone rods have a slightly
larger diameter (3mm) and thus the sound color is a bit different.
- 09.10.2008: Sysex velocity lookup tables calculated and uploaded by Kristof
Lauwers. These lookups can be used with program change 122.
- 21.11.2008: <Toypi> leaves Logos for the first time: it goes to the
Dorkbot presentation held at Vooruit in Gent. Animators are Kristof Lauwers,
Sebastian Bradt and the author.
- 13.02.2009: We are considering the construction of a second -improved- version.
- 13.04.2009: Three sound rods broke off: notes 79 (G), 83 (B) and 93 (A).
For note 93 we have a replacement in the old opera toy piano, but for the
other two we will have to search further... We finaly made and tuned completely
new rods on the lathe. Toypi is healed again. Spare notes in stock now: 85,86,87,89.
- 16.08.2009: Two broken rods repaired: 90 and 96. Murphy being the way he
is, of course none of the spares... New rods turned on the lathe. One of the
beaters got lost, so we had to make a new one.
- 13.09.2009: Again two tone rods broken. This time notes 76 and 97. New rod
made and put in place. Tuning of rod 96 corrected.
- 13.10.2009: Again 3 tone rods broken. This time notes 77, 88 and 89. New
rod made for note 77, for 88 and 89 we recycled two rods from the former Opera
toy piano. Toypi is operational again for tomorrows concert with the robot
orchestra.
- 30.11.2009: Robody pictures selected and added at the end of this webpage.
- 05.12.2009: Note bar 86 broke off and needs to be replaced... We ought to
have a spare in stock.
- 09.12.2009: Note 86 repaired. M6 long bolts holding the harp, replaced with
stainless steel bolts 120mm long.
- 08.04.2010: Four rods broke off ( 74, 90, 91, 92). New rods made and replaced.
Beaters replaced and spares made.
- 15.04.2010: Barely repaired, and again one of the rods broke...
- 22.06.2010: Again two rods broken (notes 72, 84). New rods made and replaced.
Toypi should now be ready again to join the orchestra in the ancient music
concert at Alden Biesen.
- 14.08.2010: Four broken rods replaced again. Do users forget to send the
program change 122 command???
- 18.08.2010: Tone rod for note 94 broken off.
- 25.03.2011: Three tone rods broken and replaced.
- 07.10.2011: Again three tone rods broken and in need of repair... We may
have to use martensic stainless steel rods for future replacements.
- 19.12.2011: All broken tone rods repaired using stainless steel, 3mm diameter.
No narrowing applied this time.
- 30.05.2012: And again, two rods broke off...
- 25.07.2012: Toypi in need of repair again...
- 08.08.2012: Four tone rods repaired and replaced with 4mm thick ferritic
stainless steel rod material. The rods ought to be hardened though, as this
would greatly improve the sound. Anyone out there that can help us out on
this? We performed tests with different materials for rod diameters of 4 mm:
in fact hardened carbon steel sounds better than any of the stainless steel
materials we have in stock. Brass also sounds quite good, but requires shorter
staves for the same pitches. For equal length, brass sounds about a fifth
lower. We wrote a new rod length calculation utility and the results can be
read in this table. We are considering to
change all staves to 4 mm material now.
- 09.03.2013: Mount constructed for keeping the top wing opened. Sofar this
has always been a loose part. This one is made from 2 mm thick stainless steel
and a small recycled hinge in plated brass. All bolts are M3 , countersunk.
There are no more loose parts on Toypi anymore. Realigment
of the solenoid bar.
- 14.03.2013: Construction of a flightcase accomodate both <Toypi> and
<Snar>. There is even place for a laptop and interactive electronics.
The tone rods in
<Toypi> must be protected against excessive vibration during transportation
with a piece of soft foam.
- 10.04.2013: Eight tone rods replaced with 4mm diameter steel.
- 20.04.2013: <Toypi> survived the trip to Glasgow very well! We expect
it back in Ghent on monday 22nd of april.
- 22.04.2013: No damage to Toypi. It just works fine after its return from
Glasgow.
- 26.04.2013: Again one of the tines broken on Toypi... We ordered new 4mm
diameter stock from Demar-Lux.
- 04.10.2013: Tines for notes 93 and 99 broken. Start replacement of all old
tines with new ones cut from 4mm staff material. <Toypi> goes on the
road to the Venice Biennale with Jonas Runo. Mattias Perent, Katie Couck,
Moniek Darge and Kristof Lauwers are also joining. Other robots that will
form the orchestra in Venice are <Asa>, <Snar>, <Bono>,
<Temblo>.
- 06.10.2013: Complete rebuild of the tone rods. This is the newly made tines-harp
for <Toypi>
All rods are 4
mm thick steel now.
- 09.10.2013: <Toypi> left this morning for its appearance on the Venice
Biennale on october 12th...
- 03.11.2013: <Toypi> presented at Bozar , Brussels, for Zonzo.
- 01.12.2013: <Toypi> presented at the Lille Opera.
- 23.12.2013: Start construction and design of <Rodo>,
like <Toypi> based on rods clamped at one side. Rodo is a complete redesign
along the basic principles of the toy piano.
- 05-08.11.2015: <Toypi> played four days continuously at 30CC in Leuven,
Rode Hond Festival.
- 28.10.2016: <Toypi> on the road to Brugge for Iedereen Klassiek
- 30.10.2016: Safe return from Bruges.
- 29.04.2019: Toypi on the road to Matrix, Leuven for Heritage Day.
- 07.11.2019: <Toypi> on the road to Tallinn (Estland).
- 02.07.2021: <Toypi> on the road for Luxemburg
- 15.07.2021: Innundations in the abbey where <Toypi> is installed...
Awaiting news.
- 26.07.2021: <Toypi> returned from Luxemburg where it got innundated.
The water clearly didn't reach the power supply module mounted on the underside.
The iron wheels were fully rusted. Traces of mud all over the internal guts
of the instrument. Humidity caused a layer a rust on the cast iron bar holding
the tines as well as on the tines themwselves. The brass resonator plate was
also found to be seriously corroded.
- 27.07.2021: Thorough cleaning session. Thorough electric tests and measurements.
After booting up again, it looked and sounded functional again. Relief.
- 14-15.05.2022: <Toypi> on the road again, to Rennes (France), where
it is part of a radar controlled installation with some more of our robots.
- 22.07.2023: <Toypi> returned safely from the tango production in the
Ghent Feasts.
- 17.09.2023: <Toypi> doet mee aan de opstelling met het interaktief
robotorkest in de Hnita hoeve in Heist-op-den-Berg. Begeleider en animator
daar is dr.Hans Roels. Ook <Ake>, <3Pi> en <Steely> zijn
van de partij.
Last update: 2023-09-17
by Godfried-Willem Raes
Technical drawings and component
data sheets:
Firmware specifications and mapping.
PC-board: designed by Johannes Taelman, version rev.2, march 2007. Production:
Europrint, with soldermask. gold plated finish.
Power supply: Original designed to use a Sunpower, 100-240V, 47-63Hz AC input,
24V DC / 2.5A output. (adjustable). Removed on 11.08.2008 because of the very
bad behaviour under switching load circumstances. Replaced with XP power supply,
SDR series, type SDR250AS24. (Peak output current: 25A, continuous without cooling:
5.63A, cooled with fan: 10.42A) (website: http://www.xppower.com). Farnell order
number:1436100.
Solenoids: Lucas Ledex STA 13x27 Tubular Push types 195223-2-34. Gelijkstroomweerstand:37.4
Ohm. Uitvoering met 60 graden anker, vrij roterend. Diameter:13mm, bouwhoogte:
27mm. Gatmaat voor de bevestiging: 10mm. Schroefdraad: M10 x 0.75. Toelaatbare
bedrijfsspanning bij 100% duty cyle: 12.2V - kracht: 0.27-0.58N, bij 50% duty
cycle: 17.3V - kracht: 0.62-1,11N, bij 25% duty cycle: 24V, kracht: 1.38 - 2.14N,
en bij 10% duty cycle 39V, kracht 3.07 - 4.18N. ( prijs in 2004: 32$ per stuk,
ca. 12€ per stuk in grotere aantallen). Te bestellen bij Saia-Burgess Benelux
(Mechelse Steenweg 277, 1800 Vilvoorde)
Detailed schematic for the PIC microcontroller boards:
Wieltjes: vastgezet met splitpennen, diameter 1.6 mm. Het gat in de wielas
is geboord op 2 mm diameter.
LED's: Yellow: TLCY5100, 5 mm, 9 degrees. Farnell order code:491-2470
Berekening en herstelling van de klankstaafjes:
In dit ontwerp gingen we uit van een bestaand speelgoed instrumentje. Daarbij
kan het gebeuren -en die kans is zelfs vrij groot- dat we in de loop van het
bouwproces een staafje breken. Omdat die taps lopen naar de vierkantige inklemstaaf
toe, is de fragiliteit erg groot. De diameter van de staafjes is op die plaats
immers gereduceerd tot minder dan anderhalve millimeter. Daarom moeten we ook
in staat zijn het mechanisme zelf te herstellen en waar nodig, nieuwe staafjes
te monteren. Aangezien de formule voor de grondtoon van een staaf ingeklemd
aan een uiteinde f = (0.5596 / L^2) SQR(Q.k^2 / rho) is, kunnen we alle materiaal
en vormkonstanten reduceren tot een enkele konstante die te bepalen is aan de
hand van een gegeven staafje waarvan we juist de trillende lengte moeten meten
en de toonhoogte. Voor een gegeven speelpiano'tje maten we zo voor de noot Mi
(midi 76, frekwentie 659Hz), een staaflengte van 227mm. Hieruit volgt de voor
dit instrumentje te gebruiken materiaal-konstante als het produkt f x L^2 ,
in dit geval dus 33.96. Andere staafjes kunnen nu berekend worden uit L^2 =
33.96 / f. Het staafje voor de noot Fa (midi 77, frekwentie 698Hz) wordt dan
220mm lang. Het overeenkomstige staafje in het werkelijke instrumentje bleek
na opmeting 218.6mm lang te zijn, een resultaat dat dus binnen de meetfoutenmarge
valt. De moeilijkheid bij het vervangen van een afgebroken staafje bestaat erin
eerst en vooral een stukje hard staal te vinden van de juiste diameter. We hebben
zelf pogingen gedaan om bij gebrek aan vervangmateriaal inox AISI316 lasstaaf
te gebruiken (diameter 3.2mm), wat op zich goed klinkt maar toch veel minder
resoneert dan het originele materiaal. De tweede lastige karwei bestaat in het
verwijderen van het afgebroken uiteinde in het gietijzeren draagstuk. Uitboren
is erg moeilijk omdat het staafmateriaal harder is dan het gietijzer, waardoor
de boor steeds wegslipt. Uitkloppen met een scherpe drevel levert in onze ervaring
nog het zekerste rezultaat. Bij het inkloppen van het nieuwe staafje moet het
uiteinde wat konisch worden geslepen of gedraaid op de draaibank. Inschroeven
kan ook (met een losse boorkop vasthouden) maar dan moet het tapse uiteinde
van enkele gangen schroefdraad worden voorzien. Gebruik daarbij alleen de grofste
voortap-plaat. Het is uiteraard ook goed mogelijk de gehele staafjes konstruktie
zelf te bouwen. Het ingeklemd deel moet dan met een scherpe platte beitel van
enkele diepe inslagen voorzien worden. De staafjes worden dan vanuit de vlakke
kant ingebracht en met de hamer aan het uiteinde ingedreven. De gietijzeren
balk moet heel stevig in een zware bankschroef worden vastgezet bij deze operatie.
Onderstaande schets toont de konstruktie.
Bij het vervangen van staafjes bevelen we aan het nieuwe staafje steeds iets
te lang te dimensioneren. De toon kan immers steeds verhoogd worden door het
staafje (goed ingeklemd!) wat af te slijpen. In theorie kan de toon ook verlaagd
worden door aan het taps lopend deel wat materiaal weg te nemen. Praktisch werkt
dit echter niet, omdat de hele boventoonstruktuur van het klinkende staafje
dan gaat verschuiven: de grondtoon verlaagt meer dan de boventonen, die uiteindelijk
de toonhoogte bepalen zoals we die horen.
Aangezien we ons verplicht zagen heel regelmatig staafjes die afgebroken waren
door materiaalmoeheid te vervangen, vervingen we alle originele staafjes door
nieuw gemaakte met een diameter van 4 mm en zonder verjonging aan het ingeklemde
uiteinde. Dit maakt de klank iets minder brilliant, maar de grotere duurzaamheid
is nu wel een bonus.
Robody pictures with <Toypi>:
performer: Dominica Eyckmans
pictures by Godfried-Willem Raes, 2009
References:
[1] The formula is taken from Harry F.Olson "Music, Physics and Engineering",
2nd edition, ed. Dover Inc, NY 1967, pages76 and 77. The constant k as mentioned
in our text above can be calculated as 0.5596 *( (Q*(a^2)/4)/ Rho)^0.5. With
following definitions:
- rho = density in g/cm3
- Q = Youngs modulus in dyne/cm2
- a = radius of the bar in cm
- The formula applies for solid rods of circular section