Grinding a Bicycle Crank for Power Analysis

For [Mark] and [Brian]’s final project for [Bruce Land]’s ECE class at Cornell, they decided to replicate a commercial product. It’s a dashboard for a bicycle that displays distance, cadence, speed, and the power being generated by the cyclist. Computing distance, cadence and speed is pretty easy, but calculating power is another matter entirely.

The guys are using an ATMega1284 to drive an LCD, listen in on some Hall Effect sensors, and do a few calculations. That takes care of measuring everything except power. A quick search of relevant intellectual property gave then the idea of measuring torque at the pedal crank. For that, [Mark] and [Brian] are using a strain gauge on a pedal crank, carefully modified to be stiff enough to work, but flexible enough to measure.

A custom board was constructed for the pedal crank that measures a strain gauge and sends the measurements through a wireless connection to the rest of the bicycle dashboard. It works, and the measurements in the classroom show [Brian] is generating about 450 W when pedaling at 33 mph.

Video below.

35 thoughts on “Grinding a Bicycle Crank for Power Analysis

  1. The usual way of measuring torque is by measuring how much the axle twists. You have two optical or capacitive stripe patterns going around the axle at one end near the cog, and another at the other end near the pedal. The harder you pedal the more the phase of these two repeating patterns shift relative to one another because the axle twists slightly.

    This is how you don’t need to fix electronics to a spinning piece of metal. A couple discs with stripes and optical gates will do, or even a single disc if you know how fast the crank is turning at all times.

    Torque measurement is a requirement in e-bikes in the EU because they can’t turn on unless the user is pedaling, so a common way is to measure torque from the crank and add power to the back wheel in proportion.

    These things are sold on the market:

    Another version is to measure the chain tension with a displacing idler wheel, which gets pulled perpendicular to the chain the harder the rider pedals. A bit of math then reveals how much torque you have.

    1. It seems that the Chinese pedelec sensors attach the front sprocket to the crank axle with a bit of springy steel, and use a hall-sensor to magnetically measure how much the spring deflects when pedaling.

      Effectively, they’re measuring how much the sprocket is twisting into a spiral as a result of torque being applied.

    2. “Torque measurement is a requirement in e-bikes in the EU because they can’t turn on unless the user is pedaling”

      That’s not true; the requirement is that the user is pedaling, but there is no requirement for the user to actually provide any torque. Most cheap pedelec bicycles only measure if the user is pedaling, by counting pulses from a series of magnets passing by a hall sensor. The dutycycle of these pulses is different in the forward direction vs. reverse, so they can determine the speed and direction with a single sensor, but they don’t actually measure the torque.

      Many higher quality pedelecs, though, do measure the torque (or something related to the torque, like the chain tension), and use this to determine the torque of the electric motor. This results in a much smoother and more natural ride, but is definitely not required by law.

      1. In practice it is a requirement, because simple cadence measurement makes for a jerky ride. The speed slows down with more torque requirement uphill or in headwind, which reduces cadence and sends a signal to reduce motor power, and vice versa. If the bike goes downhill, the speed increases and the motor increases power. It works the opposite way it should.

        With high power motors that isn’t a problem because the user only has to keep the pedals moving along and the motor is essentially working on/off, but EU bikes are limited to 250 Watts, so a significant portion of the motive power has to come from the rider and the power assist isn’t working correctly for that.

    3. If it were as simple as you say, no doubt any of half-a-dozen-plus companies making bicycle power meters would’ve taken the approaches you describe. Yet…none do.

      SRM, Powertap, power2max, ROTOR, Stages, Garmin, and Crank Brothers…all measure force using strain gauges. The bottom of the market is occupied by force measurement inside the rear wheel hub, and a slightly more expensive meter measures bending of the left-side crank arm. The rest of the market is dominated by torque sensing at the crank spider, with one or two exceptions – Garmin and Crank Brothers measure at the pedal, but Garmin’s system is overly difficult to calibrate and not very reliable. Crank Brothers has yet to bring their system to market.

      Your suggestion for measuring chain tension is only practical for determining torque at a very gross level. The industry standard in terms of accuracy is 2%, and it’s extremely unlikely you could match that. The pulley system you describe would interfere with, or be difficult to design to accommodate, front derailleur shifting systems. It’d also add a spring-like action to the drivetrain, which would be extremely undesirable; cycling drivetrains are designed specifically to avoid twist, give, etc.

      The optical systems you describe simple are not capable of detecting the very small deflections in a bicycle crank.

  2. There is another way to do that. Measuring the resonant frequency of the chain with a magnetic sensor, and knowing its specific mass and length, with a bit of math you can calculate the chain tension. In this way you can install the system on an unmodified bike.
    With more maths you can calculate the power developed with each leg, synchronizing the data capture with the crank rotation.
    In this way work some commercial devices.

      1. I’ve done 55mph uphill(in a 35 mph zone…it was fun passing cars) and I’ve made it to 70 mph on a mostly level road. That was in 1987 on a Fuji customized for racing(but I think I was using standard steel rims at the time due to the terrain).

        1. World record for the flying 200 works out about 47.8 mph. Tour sprints are roughly in the 40’s (mph) and coming down a long mountain decent pro’s will hit 60/70 mph. So 70mph peak on the flat is quite something….

          1. For perspective: Eddy Merckx who had the “hour record” between 1972 and 1984 averaged 49.4 km/h (~30.7 mph). In ’87 it was Francesco Moser with 51 km/h (~31.7 mph). Both with steel tubing frame bikes. Sebastiaan Bowier has the current top speed record for a pedaled vehicle on a flat surface, unpaced. 133.78 km/h (83.13 mph) in 2013 with a recumbent carbon bike.

  3. This is very similar to my project that was posted here before. I’ve often thought about creating a kit so people can strain gage more easily to the arduino or other microcontrollers. I have so many libraries written for different ADC’s, know the suppliers, the glues. I just unsure if accurate force sensing is something the hacker community needs.

    1. There’s always somebody who’ll need it, sooner or later and often sooner. Who knows how many people have been unable to measure weight etc, and have either given up or used some inferior method? You might not get famous but you’ll be loved by the people you help.

      Experience is a valuable thing and can’t be bought. Put your knowledge together right and you’ll be saving others a lot of trial and error. I’d just put the info and advice up, wouldn’t invest money in making a business for something that’s a bit of a niche.

      If you really want to make a kit, maybe design one for one of the hobby component suppliers? Let them do the business side. Stick to your expertise.

  4. I cycle regularly (also use a power meter with a turbo on occasion) and their power figures seem a bit out of whack. 450W for 33mph with no wind or turbo resistance set seems a bit high. They also show about 180W @ 18mph for 0 resistance which seems pretty unlikely (too high).

    The methodology is good, similar to SRM powertap etc but they need some calibration I’d guess.

    Of course I could be totally wrong as the resistance of that horribly rusted chain must be worth 50w on its own.

      1. Those curves are with resistance set to a specific level on the turbo, similar to what trainer road do for virtual power. You can hear @4:40 that they don’t have resistance set.

  5. It’s a cool project, but it’s inexpensive direct copy of something that that exists on the market from quite a few companies (they say that in the video). For example ‘Stages’ has an identical system where they bond strain gauges to one crank and transmit it using ANT+ (a lightweight 2.4GHz sports equipment protocol) to any ANT enabled head unit. Once again there are a few head units, the most popular one is made by Garmin. Other places to bond strain gauges are, the pedal spline (Garmin Vector) and the Spider (SRM, SRAM/quarq). Polar used the chain resonance system for a while but it was a total failure so they went to a pedal spline strain gauge.
    The head units display all sorts of metrics, including as described in the video, instant, 30s and 1 minute moving average of power. The also log the data for upload to your PC for analytics or strava.
    Interestingly, the cadence (the crank speed) does not usually require a magnet for crank based meters. They can work it out based on the forces and the other sensors. I’m guessing they also have an accelerometer.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s