When our engineers were first contracted to build a dynamic calibration rig for a Sports Institute, they were very surprised to find that the so called “Gold Standard” of then current power meters had some serious flaws.
Most people at the time thought that the main problem with power meters at the time was the sheer expense of getting the information, but the dynamic calibration showed that this was only part of the issue.
At the same time, the average cyclist was moving to cheaper alternatives, mainly because the industry mantra (faithfully reported by the journalists and believed by many academics) was that all devices were accurate to +-2%. Over the last decade, this oft repeated accuracy idea became an accepted truth.
Now the market is being flooded by cheap power devices, each one more brilliant than the last apparently and of course much cheaper. All of them apparently accurate to +-2% except for two outliers. The ex-Gold Standard and the original usurper (Power Tap) both claim higher accuracy.
The problems with the +-2% or better assertion are many but in a nutshell revolve around some simple concepts;
- +-2% of what when and how often?
- If they are not measuring crank torque surely at best they are mathematically constrained estimators (see the embedded GIF where you can see the load path that needs to be measured for crank torque),
- All of them are measuring something that is NOT the tangential force – the force that drives the bike forward as a result of the energy (watts) employed by the cyclist.
InfoCrank started it’s life as a dynamic measurement tool and still is employed in that mode by some users including our own R&D teams. For example, some components companies use the InfoCrank to measure precisely the wattage lost or gained as they strive to produce new products. Some medical people use the InfoCrank to measure the success of knee replacement treatments and are producing the medical papers based on that research. A prosthetic manufacturer uses the InfoCrank to determine the energy lost with different prosthetic sleeves.
What is most interesting is that the average cyclist just assumes that +-2% is good enough for cutting edge performance. (That is, except for the top sports institutes and the leading coaches of the world). The reason is very simple. They all assume that +-2% means that the reading on the computer screen of the bike is within 2% at all times. Seen that way, 2% either side of true is not so bad and whilst precision has to be better, why pay for it particularly when you can get “close enough- good enough” for a third of the price?
The problem with this is that there is no evidence to back up the assertion believed by both riders and journalists that the cheap devices are in fact accurate to 2%.
When the InfoCrank is used in it’s former role as a dynamic measuring device and tests the cheaper pedal or hub based power devices, the results bear no resemblance to the marketing.
For example, in a recent test on the open road, a pedal based power device ($200 USD cheaper than the InfoCrank package) that claims to be accurate to +-1.5% was measured second by second against an InfoCrank. Both devices were zero’ed, even though the InfoCrank does not require it.
The test* was done on a climb (real road conditions) as steady as a trained cylclist could be. The target power was 250w and the achieved watts over the entire 45minute climb was 249watts, attained by watching the second by second watts on the InfoCrank.
The cadence and the HR were also measured and the separate computers recorded them. There was virtually no discrepancy in the numbers of these measures on the different computers, thereby proving that any differences had to be in the torque measurements.
What were the results?
- The pedal based device showed the correct number 6.5% of the time.
- It was within product specifications 41% of the time.
- It was therefore outside specification 59% of the time.
- The errors were a maximum of 36% and minimum of -23%.
- Basically the results were randomly strewn across the entire spectrum, meaning that even when the results were true, you would have no idea that they were true.
- The devices algorithm seems to control the error by about +-37% – a number that shows up over longer rides.
Coaches often advise cyclists to “smooth” these wildly oscillating numbers by using multi-second smoothing on their computer. The assumption behind this advice is that the numbers are actually measuring the correct thing and it is the cyclist whose power fluctuations is making it difficult to read.
Smoothing random numbers, even when constrained by an algorithm, still produces a random result. We followed the advice in a random segment to see the results.
In a 50 second segment, the 3 second smoothing meant that there were NO true results at all. (Compared to 10% in the non-smoothed results) The results were evenly dispersed from +-9% making it less wrong than the average of +-15%. Remember, that with random results, there is virtually no relationship between any one number and the true number.
Our experience in the short time that InfoCrank has been available is that more and more athletes and their coaches are getting fed up with numbers that bear no relationship to effort (energy) and that a sea change is about to happen in the world of bicycle power measurement, just as the biggest crank manufacturer brings another “me-too” +-2% power meter to market.
*The test was actually over nearly four hours, but included different elements, which did not favour the pedal based device. The 45minute climb was the most favourable segment for the pedal based device. This test produced similar results as all other tests on hub based and pedal based power devices.