Robotic beaters

The second task to work on was the development of the striking mechanism for the ceramic bells which would give a good sound. 

I knew something about how clay sounds when it is hit as I had been playing the sounds of pots since college days, when I stumbled across this after a bad firing. Frustrated at a kiln load of badly glazed pots, I was looking at them thinking about what to doand began to tap them with a wooden stick and was taken with the tonality and ring of the variety of shapes I had made. Soon after I went out and bought a pair of soft percussion mallets, normally used for a Marimba and began to play with the pots (now kept back for this purpose). At that time I never went much further with it, but have always hit the pots I make to listen to the tones and so, over the years, have built up a sort of empirical knowledge of the kinds of shapes which make good sounds. 

I started to make a range of bowl forms to experiment with and through a friend who works in the music instrument industry got in contact with companies that make percussion mallets. I settled on Chalkin mallets, beautifully made with a variety of hardnesses which give different tones.

At Kaleider testing initial ideas with the Informatics Lab team.

From the initial choice of materials, we began to explore how to make the beaters strike the pots and, as mentioned before in a previous post, I worked with the Fab Lab at Exeter Central Library and more specifically with Ian Woodbridge (of whom more will be said later).

The idea was to use a Raspberry Pi to act as the receiver of the information from the motion capture software and the device which would operate the servos which were to make the mallets strike the pots. We used 3 HAT’s from Adafruit which enabled us to add 16 servos per HAT to the Pi. However at this early stage it was really about learning how to operate the servos to enable the mallets to strike the pots in the way we wanted.

There was some code available to begin to make this happen but it took a while to get the servos to operate and hit the pots properly and eleimate initial issues. These were: understanding which angles of swing were best, how to stop the mallets dulling the sound of the bell by resting against the pot when struck and the speed of the mallet.

Initial tests at Kaleder with the Informatics Lab team.

In the initial tests the aim was to get the best sound ot of the mallet striking the bell as possible. We found that the mallet needed some momentum, so it had to swing at least 25 or 30 degrees in order to give a god sound. It had to have some speed, but not too much as the noise the servo makes at high speeds was too much and interfered with the sound of the bells.

Test at the Fab Lab in Exeter Central Library

Lots of different possibilities were experimented with and eventually Ian devised a way to make the mallet swing to hit the pot, pull back a little t allow the pot to resonate and then retract back to give the mallet momentum on the forward swing. This was the optimal swing pattern and one we have used since then.

The other important factor was how to position the mallets so that they would strike the right part of the bell to get the best resonance. This is normally around the top or outside edge of the bell and the servos needed something to support them and the mallets to hit the right place on the pot.

Initially I made boxes to support the servos and spent quite a time developing the shapes using a laser cutter to cut the panels of the boxes. However I found that the boxes amplified the sound of the servos to the extent that the bells could hardly be heard. I was able to rectify the problem using the offcuts of the laminated birch ply frame and discovered this by playing around in the carpenters workshop for an hour. Not only was the sound better, but the aesthetic and material came together in a way that I was much happier with.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.