The production of Virtual Belfry required a lot of careful planning and programming. But it also posed two significant challenges:

The Photography

This page describes the work involved in producing the twelve-bell version of Virtual Belfry. The original, eight-bell images also required a lot of work.

Camera Platform

For the photography to be successful, a strong, stable and accessible support for the camera was needed some five or six metres above the bells:


A lowerable beam was constructed, and a camera cradle made to slide along it into a central position. The cradle was specially designed to allow the camera to be pointed at the various bells in the tower in such a way that there were no problems with parallax shifts in the resulting images. In this way the overall belfry image as seen in the program appears to be made from a single, seamless photograph, when in reality it is a combination of many separate shots.

Video Photography

For the twelve-bell version (2.0 onwards) high-quality video photography of each bell being rung was taken. The resulting footage was then broken up into many thousands of separate frames. From these a suitable number for each bell (the tenor required 118) were saved and used to make the animation.

A lot of processing was needed to transform each original frame into the quality required for the program. These images show before-and-after versions of just one frame for one bell:


Technology

A Pentax K7 DSLR with a 35mm prime lens was used to capture the video. At the same time a smaller camera was capturing more video from a side-on perspective at bearing height. Comparing actual frames with their corresponding side-on frames allowed the angle of the bell to be determined for each image (information critical to the program to achieve a realistic animation).

The lighting was achieved with four 500 watt tungsten halogen floodlights pointing at the bell being photographed. These created a strong yellow cast in the images that had to be corrected in the finished result (see the shots above) but they were cheap to buy and easy to set up.

Image processing was done with Adobe Photoshop CS3, with bulk transformation and colour corrections achieved by the use of custom-written Actions.

Sound recording was courtesy of Michael Shelley who runs a local recording studio.

Virtual Bell Control

The mathematics required to satisfactorily animate a freely swinging bell are not hard (assuming there is no attempt to separately model the clapper). If you assume a bell moves in a straight line at a constant acceleration through a tiny arc of its circle (and at a different constant acceleration along a different straight line for the next tiny arc, and so on), then a combination of Newton's laws of motion combined with trigonometry allows you to plot a bell's movement around the whole circle.

But bells are not freely swinging - they are checked and pulled by the ringer to vary their rhythm according to the method being rung. Given that the clappers are visible in the photographs when they strike, it was essential to synchronise the sound with the pictures and to maintain an accurate rhythm for the whole performance. Modelling all the check and pull forces to vary the calculations in real time seemed to be the best way of doing this.

To control a bell properly, ringers need to think two blows ahead (if you have just struck in 5th place and you are hunting down, your handling to strike your next blow properly (in 4th place) will depend on what happens after that - whether you continue hunting down to 3rd place, make a place in 4ths, or dodge back into 5ths). Assembling all the possibilities, there are nine different two-blow "manoeuvres" that can be rung, as shown on the left. These nine must be doubled to eighteen if there is a handstroke gap (the handling to ring a certain manoeuvre starting with a handstroke will be different when ringing the same manoeuvre starting with a backstroke).

While not intending to go into the details of the algorithm used to calculate the forces for each manoeuvre, it is interesting to reveal a little blue line that was used for the purpose. This short stretch of ringing contains all eighteen manoeuvres with no repeats. A combination of forces that would get a bell successfully through this line, striking on time all the way and ending up travelling at the same speed as it started, would allow that bell to ring any method at all (jump changes excepted).

The calculations had to be repeated for every bell, for all numbers of bells being rung (from 3 up), and for every different peal speed setting that the program supports (2h30 up to 4 hours in 1 minute increments).

The program to calculate these forces took a while to design, several weeks to write and test, and many hours to do its job.