Novakordo Archives - James House Brighton

Revolutionising robotic wrist movement …

… with a compact, cowerful, and versatile differential system

Introduction

Robotics enthusiasts and engineers are constantly looking for ways to improve the efficiency, compactness, and power of their designs. In this blog post, we will discuss a novel approach to creating a robotic wrist using a differential system, incorporating two servo motors and three beveled gears. This method offers a simple, compact, and powerful solution for maneuverability, while also providing precise positioning capabilities.

The Design:

The innovative design consists of two back-to-back servo motors, with their shafts aligned in a straight line. Each motor is equipped with a large beveled gear (A and B) attached to its shaft. A third beveled gear (C), of the same size but positioned at a right angle to gears A and B, is meshed between them. Activating the servo motors in opposite directions causes gear C to rotate along with gears A and B, resulting in a differential-like mechanism that drives the robotic wrist’s movement.

Benefits of the Design:

Simplicity: The design is straightforward and easy to understand, which makes it accessible to hobbyists and professionals alike. By using only three beveled gears and two servo motors, this solution simplifies the complexity of creating a robotic wrist mechanism.
Compactness: Due to its minimal components and clever arrangement, this design is highly compact. The back-to-back positioning of the servo motors saves space and allows the mechanism to be integrated into various robotic applications without taking up a significant amount of room.
Maneuverability: The differential-like action achieved by moving gears A and B in opposite directions enables excellent maneuverability for the robotic wrist. This design allows for a wide range of motion and flexibility, making it suitable for multiple purposes, from delicate tasks to robust applications.
Power and Precision: Utilizing servo motors in this design offers two distinct advantages: power and precision. Servo motors are known for their high torque capabilities and accurate positioning control, making them an ideal choice for a robotic wrist mechanism. The combination of powerful motors and the efficient transmission of force through the beveled gears results in a robust and precise system.

Conclusion:

This innovative approach to designing a robotic wrist using a differential system with servo motors and beveled gears offers numerous advantages over traditional designs. It is simple, compact, and provides excellent maneuverability, power, and precision. This solution has the potential to revolutionize robotic wrist mechanisms, opening up new possibilities for various applications in robotics and automation. Give it a try and experience the benefits of this ingenious design for yourself!

The above description was written by GPT4, and with the help of the video explains what I am trying to do although not why. You may have read elsewhere on the site that piano keyboard assemblies will be attached to the robot wrists and move (with instructions from sensors) so that however I move my hands, the keyboards will always be under the fingers. The original design using stepper motors worked moderately well, but needed a lot of improvement to be able to be used in a live instrument. I thought I would try this aproach too, and it seems quite promising. In the next stage I will introduce more powerful motors, better printed gears, and a slightly different arrangement that will allow greater movement.

How to make the gloves

Tech

Shopping list:

2 x BNO055 https://www.adafruit.com/product/2472
Teensy 4.0 dev board https://www.pjrc.com/store/teensy40.html
Wrist support gloves such as https://amz.run/6cQV
Velcro, staples, flexible wire, cable ties

I also 3D-printed housings for the electronics. I used this box for the Teensy from Thingiverse and adapted and printed 2 of these for the BNO055s. Contact me if you want my versions – they are not particularly good, but good enough for my need. In a later post, I will describe a better version that will work with the hand straps that I am now using.

On boxing the electronics, I attached the boxes to the gloves using Velcro, and to stop the Velcro from peeling off the gloves, I added some staples. I mounted one sensor to each of the “back of hand” parts of the left and right gloves, and the Teensy to the wrist of the right glove. Make sure you have enough wire to connect to the left hand so you can wave your arms around! The instructions for wiring the sensors to the Teensy can be found on the Adafruit site (or you can ask ChatGPT). Just to note that they both connect to the I2C bus so you need to change the address of one of the boards by linking the ADR pin to the 3V pin. Other than that it is 4 wires in total for Vin and Gnd and SCL and SDA for the I2C bus.

If you are going to program your Teensy 4.0 via the Arduino IDE you will need to add “Teensyduino” into the mix. These are the instructions you will need to do it. You will also need to configure the Teensy for USB midi.

Having done all this, the program is actually quite trivial and could probably be improved hugely. You will need to add the sensor and midi libraries to your IDE first. Once programmed your computer will recognise the USB as a midi device and you can add it to your DAW. I use Ableton and I set up 6 different choir voices to 6 different midi channels. Moving the hand will send different notes to different channels. How often the notes are played will depend on your hand position, as will the volume of the notes.

Once you get the idea, you could send out all sorts of data. Let me know what you come up with. In the next post, I will show you my next experiments in a gloveless arrangement and using midi CC data.

#include <Adafruit_BNO055.h>
#include <MIDI.h>

MIDI_CREATE_DEFAULT_INSTANCE();

Adafruit_BNO055 bno1 = Adafruit_BNO055(1, 0x28);
Adafruit_BNO055 bno2 = Adafruit_BNO055(2, 0x29);

// Set the base note and interval
const int notes[] = {43, 45, 46, 47, 49};

const int sopNote = 60;
const int interval = 7;

void setup(void)
{
  Serial.begin(115200);
  Serial.println("Dual BNO055 and Midi Test");  Serial.println("");

  // Begin MIDI
  usbMIDI.begin();

  if(!bno1.begin() || !bno2.begin())
  {
    Serial.print("Ooops, no BNO055 detected ... Check your wiring or I2C ADDR!");
    while(1);
  }
}

void loop(void)
{
  sensors_event_t event1;
  bno1.getEvent(&event1);
  sensors_event_t event2;
  bno2.getEvent(&event2);

   // Map the sensor data to MIDI values. The sensor data comes out between -180 and 180 
  // degreens and is mapped to one of  24 positions - my hands don't twist that much but it 
  // was convenient to do like this. To get one of 5 notes, I took modulo 5 so no other notes 
  // would be produced. Similar for the midi chanels

  int lhnote = notes[map(int(event2.orientation.x), -180, 180, 0, 24) % 5];
  int lhmidiChannel = map(int(event2.orientation.y), -180, 180, 0, 24) % 3 + 1;
  
  int rhnote = notes[map(int(event1.orientation.x), -180, 180, 0, 24) % 5] + 12;
  int rhmidiChannel = map(int(event1.orientation.y), -180, 180, 0, 24) % 3 + 4;

 int volume = map(event1.orientation.z, -180, 180, 0, 127);
 //   the z axis varies the delay between when notes are played in this loop
 int tempo = map(event2.orientation.z, -180, 180, 100, 3000);
 
  // Calculate the note a fifth apart
  int lhfifth = lhnote + interval;
  int rhfifth = rhnote + interval;

  Serial.print(F("Left: "));
  Serial.print(lhnote);
  Serial.print(F(" "));
  Serial.print(lhfifth);
  Serial.print(F(" "));
  Serial.print(lhmidiChannel);

  Serial.print(F("   Right: "));
  Serial.print(rhnote);
  Serial.print(F(" "));
  Serial.print(rhfifth);
  Serial.print(F(" "));
  
Serial.print(rhmidiChannel);
  Serial.print(F("    V: "));
  Serial.print(volume);
  Serial.print(F(" T: "));
  Serial.print(tempo);
  Serial.println(F(""));

  // Send the notes

  usbMIDI.sendNoteOn(lhnote, volume, lhmidiChannel);
  delay(10);

  // Send the note a fifth apart
  usbMIDI.sendNoteOn(lhfifth, volume, lhmidiChannel);

  usbMIDI.sendNoteOn(rhnote, volume, rhmidiChannel);
  delay(10);

  // Send the note a fifth apart
  usbMIDI.sendNoteOn(rhfifth, volume, rhmidiChannel);
  delay(tempo);


  // Turn off the note
  usbMIDI.sendNoteOff(lhnote, 0, lhmidiChannel);
  delay(10);

  // Turn off the note a fifth apart
  usbMIDI.sendNoteOff(lhfifth, 0, lhmidiChannel);
  delay(10);
  
  // Turn off the note
  usbMIDI.sendNoteOff(rhnote, 0, rhmidiChannel);
  delay(10);

  // Turn off the note a fifth apart
  usbMIDI.sendNoteOff(rhfifth, 0, rhmidiChannel);
  delay(500);

}

The gloves are on …

News

… and then they are off again! Last week, I gave myself 5 days to make a set of musical gloves that I would demonstrate at the weekend at UX Camp Brighton and with the help of ChatGPT I was able to pull off a fairly reasonable demonstration, even though I say so myself.

This week, the gloves came off and I made a Mark 2 version that had the sensors and the microcomputer attached to wristband replacements as used for fitness trackers. These seem like a better option for what I am hoping to achieve. In separate posts I will give all the info needed for one to build these yourself. In this post I am just going to talk about the rationale for doing this in the first place.

Anyone that knows anything about musical gloves will immediately think of Imogen Heaps’s Mi.Mu gloves which have orientation and gesture sensors, and also a means to detect finger motion. 10 years in development, these can be purchased for around £2500.

My gloves, in contrast, only have orientation sensors (BNO055) although it would be possible to detect gestures too. But the intention for mine is that whilst being able to send MIDI data for XY movements of the hand, that data will also be used to control piano keyboards attached to robot wrists that dastardly keep the keyboards under the hands in whichever orientation they might be. The rationale behind this is to allow accordion-like bellow movements without actually having bellows, and to control sound qualities other than volume (i.e. beyond simulating the pressure of air going through reeds).

I chose the Teensy 4.0 as the microcontroller for the project as I had used it before when I made a joystick to midi controller. The Teensy is an Arduino-like board with the advantage that it configures very nicely for USB midi and can be programmed from the Arduino IDE. Despite the fact that I had used the technology before, I got ChatGPT to write me a program for testing the midi interface.

For the position sensors, I chose BNO055 boards from Adafruit. Using a chipset from Bosch these boards give 9 degrees of freedom (9DOF) with XYZ outputs for position, acceleration and magnetic field. I was just going to be using the position data. I had some experience with these sensors too, having played with them for getting motors to move (in advance of building the robotic wrists). But once again I found it useful to get ChatGPT to write me some test routines. (It literally is just asking ChatGPT to write the program, pasting it into the Arduino IDE, compiling and uploading the program to the Teensy and then checking the output in the “Serial Monitor”.

Having established both the midi and the sensors were working, it was now just a question of mounting the sensors onto gloves (elasticated and fingerless orthopaedic £5 purchase from Amazon) using Velcro. I didn’t want a “beepy” electronic noise for my demo so I chose a choir plugin from Spitfire Audio. The left hand was going to play one of 5 notes from D, E, F, G and A from the tenor and bass samples, plus a note a 5th above. The notes chosen would be dependent on the x-axis rotation of my hand, and similarly the alto and soprano samples with my right hand. Choosing this combination would hopefully give a relatively pleasing sound whatever the combination. To get some tonal variation, the choir plugin allows you to choose between short and long sounds and I put these on different midi channels, which were selected by the y-axis position of my hands. I should add that this was done in Ableton Live. I should also add that ChatGPT pretty much wrote the whole program for this too – although my commands here were to add routines to what I already had

To add more variation, I was able to control the volume with the z-rotation of my right hand, and the frequency with which the midi notes were sent out with my left-hand z- position. Finally, I added a little random percussion using Dillon Bastan’s “Inspired by Nature” free sound pack. I had just been watching this demoed on this excellent YouTube video by ELPHNT.

Surprisingly, it all worked rather well as you will see in the video below, and also at my session at UXCamp. To say that I was completely in control of what I was playing would be wrong – let’s just say that the slight randomness of sound to hand gestures rather added to it’s appeal.

This is all getting a bit to TL:DR so I will save the bit about the mark 2 gloveless version for the next post, but here is the gloves-on version.

A grand piano attachment

News

What do you do with a piano that refuses to stay in tune and is too expensive to repair? My piano tuner said it would cost around £5000 to recondition and the best option was to replace it. I was tempted, but I was more than happy with my Roland FP7 and the Pianoteq plugin. I decided to strip the piano out and use the “box” as a box – a container for the electronics associated with my synth setup.

Things of possible interest are the foot – made to look as though it runs right through the case, it is, in fact, a bit below and a bit above. The tubes came from Jaspers, from where I got most of the other bits, and I 3D printed the black plastic bits into which the tubes are inserted. The number of people who come around and instantly want to lean against the case – a bad idea as the case is just resting on the foot. Possibly should change this at some point.

I like to have the lower keyboard raked down slightly and this worked out well as I was able to make a joint that both allowed for this and provided a way of attaching the piano box to the rig. The weight of the box is counterbalanced by the weight of the keyboards etc. In the picture below, you can see the adjustable joint I created, and also the point of attachment to the box – basically just a big nut and bolt.

Whilst not actually to do with the Novakordo, I thought it might be of interest, partly because it is sort of another musical invention, and because I was watching a Maker’s Muse video on attaching anything to anything else.

The year of the water rabbit

News

Must be time for version 3: this could be the final one, but don’t hold your breath.

The rabbit is a symbol of prosperity but if ever one was serious about becoming successful, one wouldn’t be inventing variations on the accordion. There are 3 areas that I am playing around with at the moment, although through the latter part of 2022, I became more serious about improving my actual accordion and vocal skills and now have a live set that I have been intimidating audiences with. Along with percussionist Andy Fraser, we are creating a very full sound in a very small space – you can wedge an accordion and cajon in almost anywhere!

And talking of cajons, I have a preliminary design for a tunable drum that is also based on wrist movement – put simply, one or two playing surfaces are moved out in front on wrists (as per the Novakordo) and attached to the main box via a duct. Chambers in the main box will have ports so the effective volume (and hence) pitch of the box can vary – bongos and tables rolled into one!

On the NovaKordo front, there are 3 main areas that I am working on. First, is an upgrade to the differential wrist. Whilst version 1, surprisingly worked ok, I felt that the belts (known as T2 and found widely in 3D printers – hence being cheap) would be too flimsy for playing live. With too much pressure the belts could be made to slip over the teeth of the gears.

There is a more robust belt that is common in robotics (T5), and I have considered a rebuild with these, but I think there is an easier way. With the belt system, it is possible to rotate the top part 360 degrees or more. In practice, the keyboard assembly will only turn through about 120 degrees max. This means that instead of using belts, we can use ropes. Very small ropes. And I have identified a fishing line that is incredibly strong (intended for sea monsters). Because this can be attached to fixing points on the wrist, this will eliminate any chance of slippage. Another advantage is that the parts can be thinner, rather than fatter if we used the T5 belts

Here is a prototype of the new wrist. Incidentally, I have moved from Tinkercad to Fusion 360 for the more complex designs. I still use the former for a quick and easy print as Fusion is still somewhat baffling.

You can see that cords can be attached from the horizontal wheel (on the right) to the vertical wheels (behind). These, in turn, are connected to yokes on the stepper motors (not yet attached).

Because one can create 3D changes in sound via hand movements (remember that the object of the wrists is to allow the keyboard assemblies to follow the movement of my hands via “9 dof” sensors and thus get variable timbres, modulation, and volume changes, I have been using the 3D joystick from the earlier version of the Novakordo, to play around with potential sound setups. Particularly interesting are:

my Hammond organ experiments where drawbars can be moved dynamically together with the speed of a Lesley cabinet. Using the UVI Acoustic Samples B5
a mix between upright and electric basses. Playing with various Ample Sound basses
a mix between Bossendorfer, Rhodes and Wurlitzer pianos. Here I am a longtime fan of the Moddart Pianoteq plugins
movement between the soprano, alto, tenor and bass parts of a choir – in this case the Spitfire Audio Epic Choir
and similar to above with various sax sounds (and enhanced with a breath controller) using Audio Modelling SWAM plugin

More on this with demos in a further blog.

Finally, I want to start playing with the lights in the frame again soon. So far, I have just used a simple demo program but I need to model the lights to the shape of the frame, try out various effects, and then have a means of controlling them easily for different songs, and probably have the lights triggered by the sounds themselves. I’ve just discovered a new set of tutorials here by Scott Marley and here is a link to a set of tutorials from Dave’s Garage.

Where has the year gone?

News

I made a major breakthrough in the early part of the year when I decided that we would become even more un-accordion like – that is a keyboard and a buttonboard (Launchpad) for each hand. In addition, and possible a world first, I proposed a scheme whereby the keyboards would track the movement of the hands. They would do this by monitoring the hand positions (using sensors attached to the palms). Robotic “wrists” would perform the actual movements to maintain the alignment.

In the previous post, I showed a simple proof-of-concept to test out various sensors and ways to be able to move the keyboards in the appropriate planes. And what I have been doing since, and no small feat, is to learn robotics, and test out a system that would be able to move the keyboards (and at nearly 2kg for the keyboard/Launchpad combo have sufficient oomph). Here are some views of my first prototype – they work but I am not convinced that they are robust enough so I am currently working on version 2.

The movement in 2 planes is achieved by a differential mechanism (see here and below) using 2 stepper motors. Whilst the system works fairly well, I don’t think that the belt drives I chose are strong enough. Because I don’t need 360 degree movement in either plane, I am working on a modified system where the belts are replaced with strong cords and fixed in place on the pulleys. All will become clear soon.