<HybrLo>

As the results of the <Hybr> robot turned out very positive, and after building the top octave extension <HybrHi>, we decided to give it another extension into the very low range of notes. An extension in the low range was already conceived after finishing <Hybr> but due to the inherently large size of the pipes, we left the idea dormant for a few years. But as the need became more and more urgent to get easier to transport instruments in the very low register, we took up the idea again. Using thin PVC pipe material makes indeed transportation a lot less problematic, even though by necessity, it would have to be a rather large instrument. In order to minimize size, we decided to limit the polyphony to 12 voices, and have each pipe play either the low octave or the higher octaves. Thus octave doubling becomes impossible by design. However, this way the ambitus of the instrument becomes four octaves. It could even be further extended by adding the duodecimo's and other harmonics. The inharmonicity of the pipes however becomes quite noticeable and therefore we did not implement it.
. Learning some lessons from the <Hybr> design, here we strictly adhered to the design rule stating that the ratio of pipe length to internal diameter must be larger than 10. The pipes work acoustically as 1/4 lambda resonators.. As the required pipe lengths turned out too long to construct a transportable instrument, we decided to crop them such that the total height of the instrument stays smaller than 2 meters. So even fully mounted it will pass through a normal door. We first calculated the ideal sizes for the lowest pipe and came to the conclusion that this pipe ought to have a diameter of 300 mm if it should have a resonance at the octave. Clearly, this would lead to a way too large instrument, as even the diameter of the shortest pipe would become ca. 150 mm. However, It would make a great instrument, as with such sizing the use of very powerful drivers would become possible. If anybody wants to commission it, we are ready to accept an order...
<HybrLo> was designed at a time the Logos Foundation was highly endangered, as an advisory commission composed of complete idiots (however politically highly correct...) issued a negative advice about our governmental funding. In such an awkward context it became a priority to design the instrument such that it could be build from recycled materials and merely components we had in stock. Clearly this had severe repercussions on the visual appearance of this robot. Sorry for that. Blame the conservative puritan leftists that oppose to real experimental work in the arts for this.
As we used 16 bit microcontrollers (and not the 32 bit ARM controllers used for <Hybr> and <HybrHi>) the implementation of the controllers for this instrument turned out a bit different. Some controllers found on the other instruments (vibrato and tremolo) were dropped here, as for such low frequencies it didn't make much sense. However, in this instrument we thought about experimentally inclined composers and provided in individual ADSR control and waveshape for every single note. Thus the instrument can even be used to mimic percussion instruments or pizzicato bass lines. For compatibility with the other instruments, global controllers for these parameter are implemented as well. To make things even easier for users, we implemented a small set of presets as well. After been used for almost five years, we discovered none of our users actually ever used all the possibilities we had implemented. So, at the occasion of a firmware revision near the end of 2020 - a revision we undertook in the first place to remove a minor bug from the code causing some amplitude instability at unpredictable moments - we decided to simplify the controllers a bit. We kept the possibility to individually modulate the amplitude of every note as well as the individual modulation and modification of the wave shape. The ADSR-controllers however are now global, affecting the entire instrument.

The 16-bit microprocessors in this robot are working at the upper limit of their capabilities. All the code is based on a multitude of fast interrupt handlers running at up to ultrasonic frequencies which by their very nature had to be handled in a hierarchical way. All wave processing makes use of PWM modulation, nine channels for each processor. This 'on the edge of processor possibilities' approach is not without musical consequences and in a way the robot in the end has somewhat of a more human behavior: as the music gets faster and more complicated, it may give up on precision. It will not crash nor get out of tune, but as the number of controllers and key pressure commands send to it becomes very high, the envelopes will tend to become a bit irregular and shaky.In the firmware upgrade we undertook in 2020, a lot -but not all- of these problems were solved.

Here is an overview of the complete circuitry:

Definitive pipes as made, tuned and measured:

• freq = frequency wherefore the pipe is used and tuned
• f0 = fundamental resonant frequency of the pipe
• f1 = first octave frequency
• f2 = second octave frequency
• f3 = third octave frequency
• l/d = pipe length to internal diameter ratio
• L = pipe length in mm (1/4 lambda resonators) with end correction applied
• De = outer diameter of the tube in mm
  
• Di = internal diameter of the tube in mm
  
• SPL= Sound pressure level with a 1 Watt rms. sine wave drive, in dBA measured at 30 cm from open end.
• Spkr = loudspeaker type used (all speakers except the one for note 22 (USA), are made in Germany)
• L1 = first length of the pipe, when bend (generally 2/3 of L)
• U-lenght = length of the U bend.
• L2 = second length of the pipe when bend (generally 1/3 of L)
• Le = length of the single elbow on the end of the pipe
• The first column gives the MIDI-note number.

For the calculations the velocity of sound was taken as 344 m/s, for a temperature of 21 degrees Celsius. Note that with this mensure the cross section surfaces of the lowest and highest pipes relate to each other as 2:1. (11122 mm2 versus 6361 mm2). All speakers are used below their resonant frequency if the fundamental is sounded.

The electronic circuitry consists of a midihub board containing a MIDI parser, and the controls for the lights. Next we have four tone-generator boards each generating three notes. For these boards we used 16 bit PIC processors designed for steering motors in three phases. Thus we could implement three channels of independent pitch generation with full ADSR control on each board. This is the circuit:
The component values for the note filters are different for each opamp. They were calculated using an online filter calculator by Analog Devices. The -3dB frequency was calculated for 8-times the fundamental frequency. The -12dB point at twice this frequency. The envelope filters are dimensioned for a cut off frequency at 40Hz. All resistors are 1% precision, but for the capacitors we had to live with 5% due to unavailability of precision capacitors on the market. (For inductors, the situation is even worse, as common tolerances are no better than 20%). In the circuit drawing given above, the values shown are for the notes 22, 23 and 24. The resolution for the ADSR-channels is 16 bits and the envelope sampling rate 1800 samples/s. There is no audible leak-through of the sampling frequency in the output signal. The dynamic range is user selectable from 20dB to 60dB in 10dB steps. The mapping of the midi values (0-127) on dynamics are always logarithmic. The most recent source code for these four boards can be found under these links:

• Hybrlo_22_24.bas
• Hybrlo_25_27.bas
• HybrLo_28_30.bas
• HybrLo_31_33.bas

The hex-dumps are available as well. These can be uploaded in the microcontrollers using MPLAB and a PICkit-3 programmer (Microchip). If you really want to get some understanding as to how this instrument works, we strongly advise you to pay the effort to study the source code. We did an effort to document it pretty well. Notwithstanding this, a close reading of the Microchip programmers manual for the 24/33EPxxxMC202 microprocessors may be required before you attempt to change anything in the code.

As we wanted quite a powerful sound we used >20 W loudspeakers to drive the PVC tubes. We used recycled 135 mm chassis speakers as commonly used in quality cars produced between 1980 and 1990. (Hence the 4 Ohm impedance of most speakers). This made a twelve channel audio amplifier a requirement. We succeeded in designing these 12 channels on a single Eurocard board, cut in two halves. We used six TDA7264 stereo amplifier chips rated for 2 x 22 Watt each. Note that we are using all speaker drivers at frequencies below their resonant frequency. There is a risk of overloading the speakers, not because of too much electrical power but rather due to excessive mechanical excursion of the 110 mm cones at resonance with the air columns. At the one hand, using speakers below their resonant frequency lowers their efficiency a lot, but then at the other hand, using them with a tuned resonator makes up for the loss in efficiency. To finish it all up, we had to design the required power supply circuits as well. For the amplifiers we used a classical analog power supply with a 225 Watt toroidal transformer (2x 12 V). All amplifiers are mounted on a common heatsink. Note that the heatsink carries the -15 V power and is not insulated from ground! In order to limit power consumption and heat developed on the heatsinks, users are advised to always use the power-off controller (#66) when the instrument is not playing. This fully mutes all amplifiers without really turning the power off.

Midi implementation and mapping:

Midi channel: 8 (counting 0-15) [same channel as <Hybr> such that it can use a same track in sequencer software and be treated as a single instrument.]

Note Off: notes 22 to 69, note release implemented.
Note On: notes 22 to 69, velo implemented.

Lights:

The lights are mapped on the note range 0 to 3. The velocity byte steers the flashing speed. With velo=127, the lights will be steady ON. Flashing speed can be modulated with the key pressure command.

• Note 0: White LED spotlight left side
• Note 1: White LED spotlight right side
• Note 2: White LED strip underneath.
• Note 3: Yellow power LED on the hub PCB.

Controllers:

In 2020, we upgraded the firmware for this robot. The controllers implemented have undergone some changes. Here is the newest implementation. For reference, and for updating older midi-files and/or software using <HybrLo>, the previous implementation can be found in this archive file: hybrlo_V1.html

If <HybrLo> is replacing <Bourdonola> in existing midi files, setting controller #90 to 62 will do. If it is used as a mere bass extension of <Hybr>, the controller could be set to 33.

(*) If the tuning is set much different from 440Hz, the overall amplitudes will be affected, unless tuning follows the changes of resonance frequencies as a result of temperature changes.  The range is a quartertone up or down. Parameter value 64 equals A=440Hz)
(**) For detailed information, look into the PIC source code, in particular, lines 635 - 709.

(***) In many sequencer programs, controllers 98 and 99 are treated as 'NRPN': non registered parameter number. Both controllers are then handled together as a 16 bit controller, where ctrl.99 holds the MSB and 98 the LSB. If that is the case, the command is followed by ctrl. #6 for the parameter value and ctrl.#38 for the fine adjust value.
(****) In many sequencer programs, controllers 100 and 101 are treated as 'RPN': registered parameter number. Both controllers are then handled together as a 16 bit controller, where ctrl.101 holds the MSB and 100 the LSB.

To understand the different parameters and controllers available in this robot, a close look at the following graph might help:

All levels are scaled on a log scale selectable with controller #80. Followings scales are implemented: 20dB, 30dB, 40dB, 50dB, 60dB. .

Key pressure is implemented for all notes in the ambitus of the instrument and used to modify the waveshape of the drive signal. By default, key pressure it set to 127, corresponding to a 50% duty cycle wave. Key pressure commands can be sent at all times and do not require the note(s) they refer to, to be sounding. Key pressure commands are sticky: they are not reset with a note-off. Power off (ctrl.#66) will reset all settings to 127. The waveshape in use can also be globally controlled with controller #14. It will also change for different program change commands. For the lights (mapped on notes 0 to 3) key pressure commands change the flashing speed for the lights that are turned on.

Channel aftertouch: Since version 2.0, this is implemented for the notes within the ambitus. It applies to all notes sounding on the moment the command is issued. A 'standard' use is to obtain a general crescendo or diminuendo on sounding chords.

Pitch bend is not implemented on <Hybrlo>

Program Change: This command can be used to select a few different controller presets.

Note that sending program change commands whilst notes are playing, can lead to unpredictable results at times. So this should be avoided. Sending an all-notes-off command followed by the program change command for the selected preset will always work.

All level related controllers follow a logarithmic scaling from 0dB to -60dB (depending on the setting of controller #80).: The -60dB scale applies to a controller #80 setting to 127.

• Midi 127 = 0dB
• Midi 125 = -1dB
• Midi 123 = -2dB
• Midi 121 = -3dB
• Midi 119 = -4dB
• Midi 117 = -5dB
• Midi 115 = -6dB
• Midi 113 = -7dB
• Midi 111 = -8dB
• Midi 108 = -9dB
• Midi 106 = -10dB
• Midi 104 = -11dB
• Midi 102 = -12dB
• Midi 100 = -13dB
• Midi 98 = -14dB
• Midi 96 = -15dB
• Midi 93 = -16dB
• Midi 91 = -17dB
• Midi 89 = -18dB
• Midi 87 = -19dB
• Midi 85 = -20dB
• Midi 83 = -21dB
• Midi 81 = -22dB
• Midi 79 = -23dB
• Midi 76 = -24dB
• Midi 74 = -25dB
• Midi 72 = -26dB
• Midi 70 = -27dB
• Midi 68 = -28dB
• Midi 66 = -29dB
• Midi 64 = -30dB
• Midi 61 = -31dB
• Midi 59 = -32dB
• Midi 57 = -33dB
• Midi 55 = -34dB
• Midi 53 = -35dB
• Midi 51 = -36dB
• Midi 49 = -37dB
• Midi 46 = -38dB
• Midi 44 = -39dB
• Midi 42 = -40dB
• Midi 40 = -41dB
• Midi 38 = -42dB
• Midi 36 = -43dB
• Midi 34 = -44dB
• Midi 32 = -45dB
• Midi 29 = -46dB
• Midi 27 = -47dB
• Midi 25 = -48dB
• Midi 23 = -49dB
• Midi 21 = -50dB
• Midi 19 = -51dB
• Midi 17 = -52dB
• Midi 15 = -53dB
• Midi 12 = -54dB
• Midi 10 = -55dB
• Midi 8 = -56dB
• Midi 6 = -57dB
• Midi 4 = -58dB
• Midi 2 = -59dB
• Midi 1 = -60dB
• Midi 0 = off

As the processor used here has 16 bit PWM resolution, the 0dB reference corresponds to 2^16. Note that we use output voltage units (VU) here and not sound pressure levels according to a dBA scale. .

• sizes: width 1150 mm, depth 500 mm , height 2140 mm
  
• weight: 75 kg. (to be verified)
  
• power: 230 V. Power consumption: 230 Watt maximum..
  
• tuning: A=440
• Ambitus: 22 - 69
  
• Loudness level:  <= 96 dBA
• Microcontrollers: 4 PIC 24EP128MC202 (16-bit) and 1 Microchip PIC 18F2525 (8-bit)
• Insurance value:  18.000 Euro
  

• Mattias Parent (hardware assistance and maintenance)
• Peter Van Lancker (requisites)
• Bert Vandekerckhove (car loudspeaker donations)
• Moniek Darge (PVC pipe donation for the 125 mm pipes)
• Kristof Lauwers (preset research and evaluation)
• Lara Van Wynsberghe (PIC code evaluation)

Music composed for <HybrLo>:

Godfried-Willem Raes 'Bottom Boot' , a once only namuda performance for an unfinished HybrLo robot (15.09.2016, premiered by Emilie De Vlam and the author) Here is an MP3 of this piece...
Kristof Lauwers, 'Study #20 for HybrLo' (2016)

• MP3-recording of this piece
• Midi file for this piece

Godfried-Willem Raes 'Lithos' (act1 in Technofaustus) with a special interactive part for HybrLo.

• MP3-recording of this piece

Godfried-Willem Raes 'Namuda Study #61: Family Completed'

Back to Logos-Projects page : projects.html

Back to Main Logos page:index.html

To Godfried-Willem Raes personal homepage...

To Instrument catalogue

Construction diary:

01.07.2016-16.08.2016: Study of the PIC 24EP128MC202 microprocessor. We will use this type to generate 3 notes and 3 independent ADSR's. So for a 12 note instrument, we could have enough with 4 boards. Comments on the firmware development can be found in the header of the sourcecode files. For code development we are using the Proton24 compiler and a PicKit3 programmer/debugger.
20.08.2016: Circuit and PCB designed for the 12-channel audio amplifier to drive the speakers. The power supply for all these amps together should be rated +/- 15 V - 8 A. (120 VA should be enough). Here is the PCB: The chip amplifier used on this board is TDA7264. There is a separate volume potentiometer for each channel (Bourns 10 k log). Amplifier gain for each amp is 30 dB. We provided in a muting input for all amplifiers together.
21.08.2016: 12-channel power amp PCB finished: etched, drilled and soldered, apart for some missing components that we had to order. Test and evaluation board etched for a 3-channel synth using LC filters. The final version to be used for HybrLo ought to have active filters though. Start design for the PCB holding the tone generators.
22.08.2016: Further soldering work on the amp- and test/evaluation board. Test board using LC filters finished. Results evaluated: we have excessive noise stemming from the ADSR channel. First of all, the PWM should be disabled at the end of the release phase of a note. Second: we have to recalculate the LF filters with the actual values soldered in place. The eternal problem here is the impossibility to get precision capacitors in the range 1uF to 47uF... However, the ADSR implementation seems to perform well. There seems to be no way beyond the use of active filters.
23.08.2016: Test and evaluation board finished. Signal noise ratio is not too great with ca. -40 dB. Here is the circuit drawing with measurement results: The noise at the output stems from the AD633 multiplier. The noise level remains the same with the inputs shorted. Here is a picture of the finished board: There are better multipliers on the market, but they are a lot more expensive... This is a scope image of the audio output on two channels, with ADSR applied: Missing components soldered on the 12-channel amplifier board.
24.08.2016: First working version for the firmware ready and tested. This code serves notes 22, 23, 24 and their octaves 34, 35, 36. Continued work on the PCB design for these boards. The active filter components can be calculated with the on-line utility made available by Analog Devices. OP27 low noise precision opamps will be used for these filters.
25.08.2016: Measurements and experiments carried out to make a good sounding pipe resonator for the lowest note (midi 22, or 29.13 Hz) using 125 mm diameter sewer pipe. It would be a good idea to give the pipes a flared end.
26.08.2016:Test pipe made as follows: 1 straight tube 1440 mm , 1 elbow 200 mm, straight piece 70 mm, 1 elbow 200 mm, 1 straight length 969 mm, so total pipe length becomes 2879 mm. Taking into account the end correction this gives an acoustical length of 2956 Hz. The Q-factor of the tube is not very high, making it quite difficult to find a very exact tuning. In any case this works pretty well when driven with a 4 Ohm speaker at 29 Hz (Our tonegenerator has a readout with a resolution of only 1 Hz...). If we stick to 135 mm chassis loudspeakers, it looks like we will have all speakers in stock.
27.08.2016: PCB's for the midi hub board and a first Pic24 with active filters board etched. First tests with components for note 22.
28.08.2016: Some corrections made on the PCB for the active filter board. PCB etched, drilled and start soldering components. Running out of photosensitive Eurocard PC boards, so new order placed at Farnell.
29.08.2016: All filter values calculated and optimized for low noise using the online Analog Devices utility. Code for hub board v1.0 finished. Hybrlo channel decided to be 8, so same as Hybr and HybrHi. If we want to avoid the note overlap, we should use this robot on a different port. E96 series resistors are no common place at Farnell... Many essential values are missing in their assortment. Decided to use OP27 opamps for all the filters. OPA627 would be overkill taking into account the noise produced by the multipliers...
30.08.2016: Further design of the instrument. Waiting for components from Farnell. Test pipe made for note 23. Two tone generator boards finished now, with the components for the active filters. PCB design for the next two boards improved: the film capacitors take more space than we anticipated... The signal/noise ratio appears to be a lot better than what we feared. Two more corrected PCB's etched. Drilling will be for tomorrow...
31.08.2016: All drilling done. All components soldered in. The PCB's are all ready for testing and programming now. Midi controllers to be implemented worked out. There will be a wealth of them... We will implement key-pressure to modulate the waveshape. Continued work on the firmware for the sound generator processors.
01.09.2016: GMT code written for testing all the boards. The test code is integrated in the GMT-cockpit for <Hybr>. All note generator boards programmed and tested with the oscilloscope. Source code and hex-dumps for the microcontrollers uploaded to the Logos website. Start construction of the pipeholder -hardly to be called a windchest in this case- from 25 mm thick PVC. The holes for the pipes have to be sawn out with a jigsaw, as we do not have round saws large enough. Start design of the chassis and the wheel base.
02.09.2016: Continuing work on the windchest. Missing PVC pipe material ordered. Welding works on the chassis: stainless steel profiles 50x50x2 and 50x30x2.Wheels mounted on the corners: these wheels have also horizontal rollers. Two wheels have brakes. Running out of M10 x 100 bolts: up to MEA. Experiments with small woofers (90 mm cones) recycled from Aciko small loudspeaker enclosures performed. The cone excursion is too small and they produce a lot of distortion. Here are some pictures of the works progress: As you can see, quite a way to go still.. It would be a good idea to have the pipework completely removable. The amplifiers and their power supply ought to find a place underneath the instrument. Front row with five pipes drawn out on the windchest. Drilling and sawing is for tomorrow... Works in progress shown to Tim Deurinck, Xavier Verhelst and Kristof Lauwers. No further disclosure yet.
03.09.2016: All holes in the windchest sawn out. Start mounting of the speakers.
04.09.2016: All speakers mounted with 4 M4 x 50 countersunk bolts and nuts, just to test functionality. Spacers required for the 90 and 100 mm tubes as the cones touch the rims of the tubes. A workshop cleanup becomes mandatory now, as we run out of space to comfortably continue the works on this robot. Heatsinks mounted on the power amps. Note that these are not insulated and carry the -15 V voltage! Bottom plate holding the amps finished. There will be enough space underneath to place the tone generator boards as well. Thus only the midi-hub board ought to be placed on the upperside. Suffering a lot from my backbone hernia, as much of the work is so low to the ground and the construction is getting too heavy now to lift it up easily...
05.09.2016: Start mounting of the tone generator boards on the bottom plate. Assembly of the power amp power supply. Mains entry with IEC sockets mounted on the back side. First wiring up.... As it comes out, we brought ourselves into trouble with the central mounting bar for the windchest. With the wiring, it would be very tricky to remove it. We might have to mount connectors for the speakers and the power supply such that it can be taken apart without unsoldering.
06.09.2016: First tests: The power supply is working as it should. Open voltage without load is -18 V and +18 V. The speakers will be wired with individual connectors for each loudspeaker pair: 4-pole IMO or Weidmueller. Color coding will be applied to the wiring.

Waiting for the M3 shock absorbers to come from Farnell. (Ettinger, Farnell order code 1466989)
06.09.2016: Start wiring of the loudspeakers. Finishing the board assembly of the underplate as the shock absorbers came in today. Missing PVC pipes and pieces ordered from Audenaert.
07.09.2016: Construction of the holding chassis for the midi-hub and parser board. This board will get a protective polycarbonate cover plate as the board carries mains voltage. A bug found and killed in the synth board firmware: the mute logic was inverted! Half of the wiring of the board holding chassis finished. We forgot to mount connectors for the mute logic...
10.09.2016: Continuing wiring on <HybrLo>. As we forgot to provide connectors for the mute inputs on the amplifier PCB's we have to use wires leaving the boards directly... Holes drilled for the wires going to the midi hub board. There would still be some space left for a small 12V SMPS to power some lights... The mosfets for these are as yet not mounted on the hub board.
11.09.2016: Wiring in principle finished. First tests become possible... A smoke stack appeared on power up. Amplifier power supply appears shorted, fuse melt down. The third TDA7264 exploded... Careful examination of the soldering of the third amplifier (notes 26-27) revealed a short between an input pin and the +15V power line. We replaced the TDA7264. The board now draws (unmuted) 240 mA running off a +/-15 V bench power supply. This means 80 mA per amplifier which seems normal. The data sheet gives no data with regard to this parameter. All amps appear working now. Making a setup for the tests with MIDI control... The mute polarity needs an inversion. Furthermore there must be some wiring errors as note 22 does not come from the speaker we wanted... work to clear out. However, the first notes did sound already!
12.09.2016: Continued debug work. PicKit3 installed in the workshop for ease of programming. Wiring of the connectors was all wrong but we got it all corrected now. All PIC microcontrollers reprogrammed. First sounds demonstrated for Moniek Darge and Kristof Lauwers. Wave shape control seems not to be working for some reason. The six highest pipes are fully tuned and working now. Waiting for 110 mm and 125 mm diameter PVC pipe to flow in... The pipes in the picture are made from recycled material. The gray ones (90 mm and 100 mm diameter) are definitive already. Fortunately our workshop has quite high ceilings... The volume controllers for the individual notes are absolutely not meant to be operated by users. They are trimmers and changing their settings will endanger the instrument as well as bringing it completely out of balance. Hub board PIC reprogrammed: now key pressure is working fine.
13.09.2016: Hybrlo implemented in the GMT coding for the robot orchestra. Controller #15 added for global waveshape control, overriding individual settings with the key pressure command. Test code for scales added in the GMT test application for <Hybr>.
14.09.2016: Sawing of 10 mm thick standoff rings for the loudspeakers, as the cones tend to make contact with the windchest plate. First coding experiments using Namuda technology.
15.09.2016: Preliminary presentation of <HybrLo> to the audience in our 'Cold Boot' concert. Performance by Emilie De Vlam.
16.09.2016: Robot disassembled again, to mount the 10 mm loudspeaker standoff's. A hole day of work... The ordered PVC pipes still did not come in.
17.09.2016: White LED strip mounted on the underside of the chassis. This appeared to be a big mistake, as the strips we used are not insulated on their underside... Thus we had to stick them to the edge of the polycarbonate plate protecting the potentiometer axes. 12 V Vicor SMPS (0.83 A) glued to the underside (PVC glue). LED 12 V spotlights mounted on de upperside, pointing upwards. All lights decided to be white, mapping on notes 0,1 and 2. Noticed a strange behavior on startup: all speakers give an ugly signal for about 2 seconds on cold boot. We will have to find a remedy in the PIC24 firmware. Two speakers got a small bump on the inner dome of the cone. We repaired them with a needle and than closed the punched hole with some red Loctite silicon compound: This slightly lowers the resonant frequency of the speaker.
18.09.2016: Finalisation of the soldering of the midi-hub board: placement of the power Mosfet's. Bright yellow power LED added on the board. The lights should be working now. Firmware for the hub board adapted. Milling the polycarbonate protective plate over the hub board. Test code added in GMT. It all works fine now.
19.09.2016: An easy way to stabilize the pipes could be something like this: Five frontal pipe feet glued for good using PVC glue.
20.09.2016: Searching for possible replacement parts for the speakers. Four Dayton DS115-8 4" speakers ordered from Parts Express. Visaton SC13-8 (Reichelt) could also be a candidate. We will add a table with possible alternative speakers in the maintenance section at the bottom of this webpage. Further work on the firmware for the note generator boards. Release-not-working bug found and killed. Board for notes 22-24 reprogrammed. Code modified such that waveform modulation during a note is now possible using the key pressure command.
21.09.2016: All synth boards reprogrammed. Start implementation of the MIDI program change command, to simplify presets for the users: we could have a Pizz bass, a string bass, an organ bass, a sfz bass etc... preset. Still waiting for PVC to flow in... Stuck on deliveries.
25.09.2016: Missing 100 mm PVC pipe fitting found on the flea market. Base for pipe 28 finished. Pipes 22 to 27 remain to be done...
05.10.2016: Users manual for the robot lights updated with instructions for HybrLo. Still stuck on deliveries to continue work.
07.10.2016: All pipe holders glued to the windchest plate. Tubes diameter 110 mm sawn to size. Still no supply of gray 125 mm diameter PVC pipe and MF-elbows. These are required for notes 22, 23, and 24. Trying to find someone with a car to get more tubes at Deschacht, Antwerpse Steenweg... Start making of the U-bends in the pipes. Only the seven lowest pipes will get a U-bend. Inventory of missing parts: 5 elbows diameter 125 PVC or 6 MF elbows, glue mount. 3 meters of diameter 125 gray PVC pipe. All pipes finished so far, labeled with Dymo labels for ease of mounting.
08.10.2016: Up to Deschacht to get the missing PVC pipes and elbows, with Peter Van Lancker. Pipe work in principle finished now, apart from some minor aesthetic issues. First tests with all pipes cut to size and tuned. Some U-turns may need a bracket to keep them perpendicular. Oddly enough, standard PVC elbows do not make right angle degrees but rather 87 degrees angles. The picture was taken in our workshop during construction.
09.10.2016: Fine tuning and minor corrections on pipes 29 and 30. Pipe 29 lengthened a bit and pipe 30 given a single elbow on its end. Firmware for all 24F PIC's upgraded: now pwm changes are no longer immediate but performed after each PWM cycle. Program changes implemented in this firmware as well.
10.10.2016: Firmware greatly improved. Now modulation of volume during a sounding note is made possible. All levels now make use of a 40dB log scale. Pipes for notes 31, 32, 33 given an elbow on the top. It just looks better...
11.10.2016: Considering to extend the note range with 3fo and 4fo pitches... Tried out. We implemented every note now in 4 octaves, thus the range of the instrument becomes 22 to 69. However, keep in mind the instrument remains limited to 12 note polyphony and octaves are mutually exclusive. With this range <HybrLo> can be used as a full and extended replacement for <Bourdonola>. Controller #90 implemented on the hub board, to allow easy limiting the ambitus. <HybrLo> placed in the robot orchestra for first evaluations and tests: Awaiting feedback...
12.10.2016: All levels and volumes now have a 40 dB range and full log scaling. Elbow added on pipe 29. The 40 dB scale after all does not fulfill our expectations. So we changed it back to 60 dB. All level related parameters now make use of this logarithmic scaling.
13.10.2016: Frequency uncertainly introduced in the firmware and implemented for controller #1. This models acoustical sounds much better than pure periodicity. In fact we modulate the duration of the off phase in the oscillators. Thus the on-pulses are kept constant for any given frequency. As a consequence the spectrum stays more stable than in the case we would have randomized the entire period as such. Scaling for attack and decay reduced to 206 ms. Release time kept with a range of 0-512 ms on request of Kristof Lauwers. Pleased with the result now...
14.10.2016: Finalizing the five presets in the firmware after extensive testing of the parameters by Kristof Lauwers. Firmware upgraded again.
26.10.2016: Photosession with HybrLo...
16.09.2018: HybrLo is replacing Bourdonola in all robot orchestra concerts up to the end of October, as Bourdonola is off to Manchester. Midi files adapted accordingly.
07.12.2020: a bug discovered in the PIC 24EP code: reading 32-timers with TMR2 and TMR3 must use TMR3HLD to read the msb, not TMR3 directly. This bug was found in all projects using this chip since 2016: <So>, <Autosax>, <Flut>, <Hunt>, <Bug> , <Pi> and <2Pi>. Fortunately we can reprogram the PIC's without disassembling the instrument!
14.12.2020: Revision of the IC 24EP firmware. Controller 100 added to change the dynamic lookup tables. Later changed to controller 80, for compatibility with many other robots.
15.12.2020: We see further improvements possible... however, if we implement these fully, <HybrLo> becomes a version 2 robot as we will loose full compatibility with existing midi-files and applications.
16.12.2020: A new version of the firmware ready now to be tested on the hardware...
17.12.2020: Everything seems to work again glitchfree. Documentation adapted and brought in tune with the firmware implementation again. Berekende waarden: bij CC15 @ 127 is Timer2=ADSR_period= 112 + ((127+1) <<4) = 2160 ==> 9.1238 ms. Bij aan adsr-setting van 127, geeft dit een ADSR-sektie duur van 1.1678 s. Minima: bij CC15 = 0 is Timer2=ADSR_period = 112 + 16 = 128 ==> 0.54 ms. Bij adsr set to 127, geeft dit dan een ADSR-sektie duur van 69 ms
18.12.2020: Alle PIC's opnieuw geprogrammeerd. Testkode in GMT voor <HybrLo> afgesplitst van die voor <Hybr>. Eerste tests met bestaande midi bestanden met een partij voor <HybrLo>: in het bestand voor 'La Cambiada' (Di Giusto) dienden we juist de velocities (ze stonden allemaal op 40) te verhogen tot 120. De ADSR-period kontroller (#15) stelden we in op 32.
19.12.2020: Channel aftertouch implemented. However, this entailed an upgrade of the firmware on the midi-hub board. At that occasion we discovered a bug in this firmware: it did'nt pass program change commands, thus explaining why the presets did not have any effect on HybrLo... This problem is by now solved.
20.12.2020: Light automation added on the hub board.
21.12.2020: Six different light shows added, selectable with controller #69.
10.01.2021: Extra 24EP board gemaakt voor research doeleinden. Polo-film van het ontwerp 30.08.2016 hiervoor gebruikt.
24.03.2021: Note 25 (C#) failure. We presume the loudspeaker (LPB128-4 Ohm) to be burned out. It was a 20W type.
25.03.2021: Search for a replacement speaker: JBL Club 5000C, or Pioneer Woofer?
26.03.2021: Two pioneer speakers TS-130CI came in. We thought we ordered only one, but apparently they come as a pair... Advertised as 130W ('max. music power', you know...), they look like they can handle at the very most 10W sine wave... DC resistance of the voice coil is 3.33 Ohm. Nominal impedance: 4 Ohm.
27.03.2021: JBL Club 5020 speakers ordered. Advertised as 120W... but 40W would be great. Impedance is only 3 Ohms. They should come in monday 29th...
23.06.2021: Note 25 repair session: the speaker was found to be burned out. We replaced it with a Pioneer speaker (TS-130CI) .. We would have preferred to use the LBL Club 5020 speakers, but these have cones that stick out way too much such that mounting them became impossible. They are designed for mounting on the backside, something only possible if mounted on thin plate material such as in a car. This repair session was done with Mattias Parent and Bert Vandekerkhove. Godfried-Willem Raes just supervised the repair.
27.09.2021: After playing uninterrupted for 25 days in the exhibition at the SMAK museum, <HybrLo> returned to the robot orchestra in perfect condition.
02.08.2023: <HybrLo> is doing fine, but we upgraded the midi-specs as we found them not clear enough. Found out that ctrl.14 does not work during and for playing notes. Key pressure works as documented.
14.09.2023: <HybrLo> transported to Berlin for the performances of 'Zeroth Law' by Gamut Inc. at the Deutsche Oper, Tischlerei.
02.10.2023: <HybrLo> reurned from Berlin. Note 27 found to be defective... Damage through transportation or overload in Berlin? We will have to find out...
11.10.2023: Checking out <HybrLo> to find out nothing went wrong, except that someone must have turned down the trimpots on the frontside... Readjusting the volumes of all pipes is indicated now. Done.

TO DO:

• Construction of alternative pipe sets.
• Adding more program changes to implement more presets.
• Adding tuning control: quartertone up and quartertone down.
• Construction of a protective top for transportation.

Maintenance information:

Circuit diagrams:

General overview:

Power supply:

Transformer: MCTA225/12, Farnell order nr.9532790.

Note generator and filter boards:

Amplifier boards:

Midihub and parser board:

• White LED spotlight on the left connected to RC5 and mapped on midi note 0.
• White LED spotlight on the right connected to RC4 and mapped on midi note 1.
• White LED strip underneath connected to RC3 and mapped on midi note 2.
• Power relay connected to RA4 and mapped on controller 66.
• Watchdog LED (red) connected to RB5.
• Blue LED connected to RB2.

PCB's:

Note generator boards with active filters and multipliers (Eurocard size: 100 mm x 160 mm, single sided PCB):

These boards are mounted op the bottom plate with M3 male-male shock absorbers as stand off's.(Ettinger, Male-Male, Farnell order code 1466989).

Amplifiers PCB's: (at 200%)

Loudspeaker sizing:

Critical evaluation:

• The use of 32-bit processors would have been a better choice for this design.
• Larger diameter pipes (300 mm to 150 mm) and more powerful speakers/amplifiers would greatly improve the bass feel. Obviously that would make the robot more than twice as large.

by Godfried-Willem Raes

Further reading on this topic (some in Dutch):

Audsley, George Ashdown 'The Art of Organ-Building', ed. Dover Inc, NY,1965, (first edition: 1905)  ISBN 0-486-21314-5

D'Appolito, J. 'Luidspreker-meettechniek', ed. Segment BV, Beek , Nederland, 2000, ISBN: 90 5381 116 8

De Keyser, Ignace, 'Challenging von Hornbostel & Sachs', in: 50 years at Logos, 2019

Raes, Godfried-Willem , Expression control in musical automates

Raes, Godfried-Willem, <Hybr>

Raes, Godfried-Willem, <HybrHi>

Raes, Godfried-Willem, 'Logos @ 50, het kloppend hart van de avant-gardemuziek in Vlaanderen' (ed. Stichting Kunstboek, Oostkamp 2018)

• Omzettingtabel midi noten naar frekwenties

• Power Supply specs:

 Maintenance and disassembly instructions:

Before undertaking any repair on this robot we strongly advise to first read through our construction diary, as this will clarify fully how the machine was originally designed and assembled.
If the machine has to be taken apart, perform following steps:

1.- remove all the pipes. On replacing them, apply a thin layer of Vaseline to the bases.
2.- Place the machine upside down on two stands. Make sure it doesn't rest on the midi-hub board.
3.- Loosen the four long M10 bolts that hold the bottom plate together with the windchest.
4.- Loosen all four 3-pole Weidmueller connectors carrying the 5 V and MIDI lines.
5.- Loosen all six 4-pole Weidmueller connectors that connect the speakers to the amplifier board. Note the order! The wires are color coded.
6.- Loosen the wires that feed the mains voltage to the MIDI hub board.
7.- Now the underplate can be fully removed and serviced. The processor boards and amplifier boards can now be taken apart.

Loudspeaker replacement possibilities:

Lights:

• 
  White LED strip: Farnell order nr. 2214000, 12 V smd, 4.5 W. Ref. 9009032. Note that these strips have an adhesive layer but are not insulated. They cannot be glued to metal parts therefore.
• White LED spotlights: 12 V , 2 W. Do not replace with halogen bulbs as the power supply cannot deliver the required power for such bulbs.

Robody picture with <HybrLo>:

Candidates? Welcome!

[EOF]