Recently, Polar released its new S710, a heart rate monitor with cyclecomputer features. One of the long-awaited features of the S710 is an optional add-on (at about $350US) to measure the wattage generated while cycling. That power sensor is now generally available. Because the S710's power sensor is quite different from the other two on-bike power-measuring systems, the Graber (née E-tune) Power Tap and the SRM, a great deal of interest has been focused on the accuracy and precision of the S710.
Alan Coté, one of the developers of the power-measuring technology used in the Polar S710, has kindly agreed to share some data collected from a rider who simultaneously used both a Power Tap and a Polar S710. I have split these data into two comma-delimited ASCII files. Some spreadsheets, like the nearly ubiquitous Microsoft Excel, can directly import CSV files without further ado. I have combined the files here (comparison.zip, 208KB total).
I have plotted the power reported by the two devices for the 5-hour ride, and you can see them below. Each graph shows one hour's worth of raw and unsmoothed wattage comparisons, with Power Tap wattage in black and Polar S710 wattage in green. Here they are:
I think you will agree that they appear to match well, certainly well enough for training purposes. An additional Power Tap-Polar comparison file can be found at the bottom of this page, and a three-way comparison of the Power Tap, Polar S710, and SRM can be found here.
The data appear to have been recorded during a single ride of approximately 120km (74 miles) over the course of 5 hours during the Fall of 2001. The rider is a male in his 50's of about 65kg. He is well-known for his cycling, but not for mass-start races, and this ride does not appear to have been a race. The ride appears to have included a climb from a minimum altitude of about 250 feet up to about 2300 feet, and the ensuing descent. I do not know what kind of bicycle was used, nor if the Power Tap or S710 were self-installed and properly calibrated. These latter conditions may be critical, as can be seen here.
Note that the Power Tap data are recorded at an interval of 2.52 seconds, while the S710 data are recorded at 5 seconds. In fact, if one takes the S710 interval of 5 seconds as given, you get slightly better fit if you presume the Power Tap data were collected at an interval of 2.53 seconds, but only anal people would care. Hi there. Note also that the Power Tap data are metric (distance is in kilometers, speed is in kilometers per hour, torque is in Newton-meters), while the S710 reported speed in miles per hour and altitude in feet. The S710 does not record distance; however, one can integrate speed over time to do this. If I do this in a quick-and-dirty fashion, I get 74.01 miles from the S710, compared to a recorded 119.16 Km (74.05 miles) from the Power Tap.
I haven't bothered to compare the speed or HR measurements, though the cadence measurements for the Polar appear much more precise. See the section below for why that is so.
Both the Polar and the Power Tap report cadence, though since the two devices use different ways to measure cadence and cadence appears to have less focus than wattage, it is not surprising that there are discrepancies between what is measured by the two. The key is that, unlike the Polar, the SRM, and most cadence-measuring cyclecomputers on the market, the Power Tap doesn't actually measure cadence directly by placing a magnet on the crank and counting the time interval as that magnet goes past a switch--rather, the Power Tap calculates cadence by analyzing the forces measured in the hub. These forces typically have high and low spots within each crank revolution (with relatively lower forces as the rider's foot reaches 12 o'clock and 6 o'clock, and relatively higher forces at 3 o'clock and 9 o'clock). The Power Tap counts the time between peaks to calculate cadence.
There are advantages and disadvantages to the Power Tap approach, but one of the disadvantages is that the cadence measurement appears less accurate and precise than with other approaches.
Alan has also provided two other files of comparison rides he himself has made. These files are in MS-Excel format (i.e., they are .xls files, not comma-separated ASCII as are the files above), and include some graphs and calculated averages. Those files are available as a single zipped file of about 240KB here.