It’s been about 10 months since my last post about the powermeter development. Those who have emailed me have been stone walled about it, getting a response such as “I can’t talk about what I am working on in my new job”. The answer is that part of it was in fact a powermeter. So, to the loyal followers (and even the not so loyal) we are offering a discount promo code of Acuity for a 20% discount on all 4iiii’s products as a thank you for supporting me and following! Oh, here’s the “catch”. It’ll be the most accurate, lowest cost and user installed (no new crank needed for most riders!). And now I do what I love full time. So thank you!
UPDATE: Code was valid until midnight of Sept 12. The last day of interbike. Thank you all that participated.
Now, if you’re still reading but want to hear how this came to pass, I give you the long story
During my Masters degree I had used strain gauges. I needed a tiny load cell to test active suspension. It worked but wasn’t great. I decided to take these “student strain gauges” for a side project. My FSA vero crank, instrumented like almost all crank based meters, showed some results, but it was noisy, drifting, and unusable. After much experimentation I found that cheap off the shelf electronics were not cutting it. I shelved the idea and started working on some training software for use with the regressed power curve of my Kurt Kinetic. That was 2009 – 2011.
In the meantime a friend of mine in Waterloo was doing a Masters with someone at Babcock and Wilcox, the nuclear steam generator company, and passed off my resume. Five interviews, a security background check, and relocating my life from St. John’s, NL to Cambridge, Ontario I was now working in Nuclear Engineering. Most of my work was Finite Element Analysis, or FEA. Essentially I my intuitive understanding of how mechanics, stress, strain, deformation, etc work was being put to work for steam generators. I had been using FEA since the second year of my undergrad and at that point had used FEA on and off for almost six years. I even wrote a free book on using Ansys and FEA to design race car chassis' (link here).
It wasn’t long before a customer concern required experimental work measuring strain. Prior work was inconclusive. I heard the design manager going cube to cube in cube-ville (which was my home) and I ran after him (coffee in hand) and said I know a and used strain gauges, Eventually I disclosed my electronic hacking which had been featured on Hackaday and I was chosen as the ideal candidate.
However, I needed credentials. Everyone knows most engineers never touch things with their hands, but in order to do strain gauges I would. I’d add it to my machining and welding skills. So I was sent to a strain gauge course. After the courseof training later I knew exactly how to fix my vero crank powermeter! I was excited beyond belief. The problems I had seen were so minor!
For the next year I spent most of my time conducting experiments at B&W using strain gauges to great customer satisfaction. Between my FEA, electronics, and attention to detail combined with my knowledge of Design of Experiments and higher level statistics my work was deemed impeccable in execution. I was achieving 1% error with theoretical. A former Pratt and Whitney employee now at B&W stated his work was up to 15% error. I was so proud to be executing things at such an accurate level. In the mean time I had spent thousands of dollars and countless hours building up my own circuit boards and instrumenting cranks.
Vero, which was never self contained, was V1. Later a C-channel shaped SRAM Rival was V2. It was arduino pro micro based with an AP1 ANT+ stick from sparkfun and some analog components that weren’t suitable. V3 combined those onto a Rival OCT (hollow forged) crankset. However there was issues, clearances, etc. I had tried some different ideas and realized forces on a crank aren’t as simple as the industry thinks (or markets). However, I got my powermeter on the famed Hackaday and tens of thousands of visits to my little blog later I felt I was on to something.
So for V4, I moved to a lot of new components and a small custom board. I spent countless hours pouring over datasheets to chose the components. I built both a Rival OCT and Carbon S900 / Quarq unit. Both worked well. Very well. In fact I was on the verge of building up a bunch and starting to beta test. I had a listing of people who were interested from all over the world.
I got an email from the father of a lady who works at 4iiii’s. He was so interested in my work and has remained a strong believer of me. Eventually it seemed to fall to the wayside of the people there or was being evaluated at a slower pace than I wanted. However I liked this company. The founder had built Dynastream, the creator of ANT+. I was impressed. So I eventually sent him a message on Linked In.
A few super early morning calls later and I was using my barely touched vacation time to impromptu fly to Alberta to meet this guy and his team with my bike in tow. I had no idea what to expect. A short drive from Calgary to Cochrane later and I find a small building labeled the 4iiii Innovation Centre.
I had mucked around adjusting code on the carbon crank for more accuracy and messed up some math, so eventually I swapped in my Rival the night before. The battery kept disconnecting during the demo in his office but he was convinced.
Some interviewing and a discussion later on what he wanted to bring to market and I was confused, full of reservations but generally sold. He wanted user installable. He didn’t want into the crank buying, installing, selling market. He wanted to let people have choice. He wanted to drop the price and change the game.
I spent most of the next 6 months working on the basics. Can it be done? What material? What gauges? etc. What I found was that 2% bending error was not actually achievable on a crank arm with the generally accepted setup (a bending bridge). If you narrow your scope to a fine window it does but that didn’t sit well, but if you start introducing some cases that fall to the edge of the norm just a tiny bit (and encountered regularly in mountain biking) you’ll find that it’s closer to +/- 7% to +/-9% on torque accuracy depending on the setup. Let alone rotational accuracy from an accelerometer algorithm.
I’m 4 months in, looking at these numbers thinking I’m a not going to be able to pull it off - not with the accuracy everyone is claiming! I quietly start looking for a solution. How do I make this better. I’ve figured out the fundamental issues with user install but how do meet the accuracy. It doesn’t change from user install to pro installed. I tried angling the sensors, adjusting things, compensating for the non-symmetric cross section of the crank. Nothing!
Then I tried something else. What about adding a second setup! Something that can sense torque but that is very different from the first. Then, using my Design of experiments mythology and statistics I figured out that there is a super cool correlation. I was blown away! This was big, this was the answer to my hopes to get me that 2%. Not only that, but generally gave 0.5% error! This idea is so revolutionary that we put patent pending in place on this technology, among many other things.
We needed to package it, and admittedly the packaging isn’t perfect for the drive side and the last few years we’ve seen a mess in bottom brackets and now we’re seeing a mess on the chain ring bolt patterns causing incompatibility of some manufacturers cranks. This is sad. However I’m glad to say the last 3 generations of Shimano 105 (5600, 5700, 5800) work, Ultegra (6600, 6700, 6800 {this is actually epoxied together!]), and Dura-Ace (7800, 7900, 9000) all work. The latest Deore XT works for the Shimano MTB front. The 10 Speed SRAM Rival OCT works (my original workhorse on the prototype front) and the X9 looks good. FSA Energy is good too but their new 5 bolt asymmetric and SRAM’s 11 speed put bolts where we want the sensor. We are still searching for a solution. On the Carbon front the SRAM S900 works but we haven’t finalized our testing. We will be recommending using the calibration load cell that comes with the kit every year at least to check calibrations depending on crank arm.
Right now the product works, but in my bosses opinion is huge physically, so there will be a shrink in size and weight when we ship in a few months. This will also improve install-ability and compatibility as well as easing access to the battery.
However, one of the coolest things I’ve done which I hope to open to the public is my test data on the existing meters. It turns out it my original bending only design was better than existing solutions but was still +/-7% while the best off the shelf solution once we hacked into it’s sensor produced +/- 9% when we introduces pedal offsets.
Enjoy the promo code while it lasts. It’s our gift to you guys who believed in me on a personal level and going forward believing that we change the face of power for serious athletes at all financial levels.
*gage is generally considered the American spelling and gauge the British form.
