A post from the Brim Blog

The processing algorithm for the power meter is one of the critical parts of the system. The algorithm is the sequence of mathematical steps to calculate torque and power figures from the various sensor measurements. It’s relatively simple to put together a procedure to do this – as long as all the measurements are perfect. It’s not so simple to design an algorithm that will behave itself when the measurements are affected by noise (small unwanted variations) or calibration errors. If we’re not careful a small error in a measurement can cause a big error in calculated torque, so we’ve developed a computer model of the whole system that allows us to see what the effect of different types of errors would be under various conditions of torque, cadence and pedal angle. It took us a while to create the computer model, but it’s been worth it.

Over the last few months we’ve had a lot of fun using our computer model to examine and improve the algorithm’s performance in various realistic conditions. We’ve also been adjusting the algorithm to reduce the computer time needed to perform the calculations on the stream of measurements from the sensors. All of this has given us a decent version of the algorithm that we’ll use for our prototype tests, but we’ll come back and look at it again when we’ve done more testing of the whole system.

We’ve also been refining the construction of our force sensor, to make it smaller, more consistent in performance, and easier to produce. We got some new prototype sensor parts manufactured, and they seem to work as we hoped. We’ve built a sophisticated bench test system for the sensor, and developed simple and fast calibration and test procedures, so we can very quickly examine the effect of changing the sensor construction in any way. That allows us to try out possible improvements quickly.

Doing the first live tests of the sensors and algorithm a couple of months ago was very interesting, and not just because it confirmed that the technical development is on track. I was the test rider and the measurements from our system showed up the inefficiencies in my own pedalling style! It looks like I’m pushing the pedals at points where I shouldn’t, so I’m wasting energy. However, at this stage of my cycling career I think trying to change my style could cause more problems than it solves!

As I write this I’m on a family holiday in the Pyrénées in France, but the development work goes on. Every afternoon I work on writing and testing prototype code for the algorithm, with the day’s Tour de France stage on television in front of me (and the occasional glass of wine at my elbow). Very enjoyable. A week ago we went to the start of stage 9 of the Tour de France in St Gaudens, the one that went over the Tourmalet. Great to see Nicholas Roche in his recently won Irish champions jersey. Of course I took the opportunity to check out which riders were using power meters, and what ones they were! Some day they’ll be using ours, but meanwhile we’ve got some work to do.

This entry was posted on Sunday, July 19th, 2009 at 15:09 and is filed under Barry's blog. You can follow any responses to this entry through the RSS 2.0 feed. Responses are currently closed, but you can trackback from your own site.