FAQ for power-based training (version 12.04)

FAQ for power-based training (version 12.04)

Q: What is power?


A: To Henry Kissinger it was an aphrodisiac, but for our purposes, the definition comes from physics, and in particular the science of dynamics, which is a branch of mechanics.  Power is the rate of doing work or transferring energy, such that power = work/time, or P = Wt.  As relates to cycling, it is measured in international system (SI) units called Watts (W), rather than the familiar english unit of horsepower that is used as a measure of engine power (1 horsepower = 746 W).  Since work = force applied through a distance, or W = F × Δx, these two expressions can be combined and rearranged to express power as the product of force and speed, i.e., P = F × s, and this may be the best way to think of it: the speed you can maintain times the total force resisting your forward motion.  Similarly, power can be defined as pedal force (i.e., torque, which equals = [measured frequency – zero offset]/slope) × cadence, which means you can increase power by exerting more force on the pedals at a given cadence, by increasing cadence while exerting the same pedal force, or by increasing both force and cadence.

You should follow these best road bikes recommendations if you are planning to buy one road bike now. Source


Here are some examples that give an appreciation for units of power:


A 68 kilogram (150 lb) rider traveling on an 8.6 kg (19 lb) bike at 20 mph in on flat ground in with no wind requires about 177 W.


56.5 W are required to raise a 20 lb dumbbell 25 in. overhead in one second.



Q: Why is power-based training so important all of a sudden?


A: It’s no more so than it was previously, in fact, exercise physiologists have used calibrated ergometers for years to impose precise loads on study subjects.  Rather, the introduction of affordable on-bike power measurement systems (power/speed measuring device, handlebar-mounted computer, receiver/wiring and computer mounting bracket, download interface, software) have made it possible to use power in everyday training as well as racing, then analyze the resulting workout data.  This, and their widening use among both amateur and professional riders has generated considerable “buzz.”



Q: Why should I train by power?


A: (Eddie Monnier and Andrew Coggan)  Because it is the objective measure of exercise intensity, and as such directly determines physiological and perceptual responses to exercise, so training by power provides immediate and quantitative feedback on the intensity of effort.  300 Watts is 300 Watts, no matter how hot, windy, or hilly it is, or what your heart rate is – though it may “feel” easier or harder, depending on various conditions.


Three variables to control in any training program are intensity, duration, and frequency; of these, the latter two are easy to quantify objectively – duration is measured in hours, and frequency in sessions per week (the product of the two is volume).  Intensity, on the other hand, has traditionally been measured by perceived exertion (PE) and/or heart rate (HR).  HR is reliable enough at lower (i.e., aerobic-only) intensities, but for more race-specific (i.e., shorter but more intense) training, it becomes a much less effective proxy for intensity.  Besides being subject to numerous environmental and physiological variables (e.g., temperature, humidity, hydration status, altitude, overtraining, lack of sleep, nervousness, and upward “drift” as exercise progresses), HR responds slowly to workload demands, and thus is a lagging indicator of effort.  That is, it will be lower than power and during the early part of an effort, and higher afterward.  For example, if you bound up a few flights of stairs, your heart rate will take a while to reflect the effort, and will continue to beat at an elevated rate for a while even after you have stopped climbing steps.  The shorter the duration of an effort, the less useful HR is.



Q: So power-based training has made perceived exertion and heart rate obsolete?


A: Not quite, but they seem to have been relegated to a definite second and distant third, respectively!  Many still cling to HR an indicator of overtraining – though declining power for a given PE is the deciding (and often first) sign of that, too.  Nonetheless, there persists some popular, if not scientific controversy as to the role of HR, with some claiming that it indicates metabolic intensity, and therefore one should train by HR, while monitoring power.  In fact, just the reverse is true; particularly during outdoor cycling, metabolic load is more accurately reflected by power, integrated with PE, the latter being more reliable than HR and incorporating more physiological variables.  Power provides an objective standard by which effort can be quantified, thereby ‘calibrating’ PE, while PE serves to modulate power output.



Q: How do I measure power – I mean, what are some of the various power-measuring systems available?


A: Here are the four bicycle-based systems presently available, with links to each manufacturer’s site:


Ergomo Sport (a torque-measuring bottom bracket available in Campagnolo square-taper or Shimano OctaLink): and

Polar S-720i or S-710i (uses a chain vibration sensor that mounts on the right chainstay): or  User’s manuals are on-line at and


PowerTap (a torque-measuring hub that you build into a wheel):


SRM Powercrank (a torque-measuring crank that replaces your present model): and


Note: contrary to claims, Ciclosport models do not actually measure power, rather, they only give a rough estimate based on speed, total mass (rider/equipment), and road grade, which may be accurate on steeper grades, but is useless on flat terrain, particularly in group rides or if any wind is present.


Finally, you don’t need a high-tech gizmosystem to figure your power.  For instance, you can use a hill with a steady grade of ~7% or more by timing yourself over a measured portion of it, and then calculate power quite accurately (so long as air was sufficiently calm) using  You can even get a consistent estimate running up a constant grade or a flight of continuous steps, such as in a stadium:  Watts  =  (mass in kg ´ 9.81 ´ net elevation gain in meters)/time in seconds; kg  =  lbs ´ 0.4536.



Q: How do power-measuring devices work?


A: (Garth Rees and Charles Howe)  The various on-bike systems measure force the applied either at the crank (SRM), the rear hub (PowerTap), crank spindle (Ergomo), or chain (Polar).  (Note: a patent was granted to Shimano in November 2003 for a torque-measuring bottom bracket, U. S. Patent 6,644,135.)  The former two use strain gauges, which are fine polymer sheet, with ultra-fine wire or foil sandwiched in it, and the electrical conductivity of the metal changes as they are twisted or deformed when force is applied, due to the secure bonding to the material under test (the energy absorbed by the strain gauge is so close to nil that it can be neglected in any loss equations).  Strain gauges are fragile when not bonded, and typically no bigger than your small fingernail, often 2 × 4 mm or smaller, depending on application.  They may be in single, half-rosette (2 gauges, 90° offset), or full rosette (4 gauges, all at 90° offset, i.e., 2 opposed half-rosettes) configuration, with the last having the best accuracy of all, since it compensates best for the strains in the two major axes, resulting in good self-cancellation of any errors in the two devices.  The difference in accuracy from half to full rosette is not as great as is the implementation cost.  Here are pictures of the PowerTap hub mechanism (U. S. Patent 6,418,797) from and , showing the strain gauges in a full-rosette arrangement:


The strain gauges measure torque inside the hub, then this data is transmitted, along with wheel speed, to a seatstay-mounted receiver via digital radio frequency (RF) waves, and then by wire to a handlebar-mounted computer with a 16-bit microprocessor, where they are used to calculate instantaneous power, road speed, cadence, etc.


Similarly, the SRM senses torque exerted at the crankset, then multiplies it by crank rpm (cadence), measured with a crank magnet and sensor, to give power.


Both Polar models measure chain tension via a chainstay-mounted sensor that detects vibrational frequency; just like a guitar string, a chain vibrates faster as its tension goes up.  This is translated into an amount of force, which is then multiplied by chain speed, as measured by an optical sensor mounted on the rear derailleur, thereby giving power output: power (W) = chain tension (N) x chain velocity (m/s).


Finally, the Ergomo Sport uses a bottom bracket with a photointerruptor circuit actuated by two “combs,” or flat discs mounted on the bottom bracket spindle, each having numerous slots spaced evenly in a radial fashion.  Two optical sensors measure changes in the alignment of the slots, which is determined by how much the spindle twists, and hence how much torque is being exerted.  This value is then multiplied by crank rpm (cadence), which is measured by the bottom bracket unit, thereby yielding a value for power.



Q: Where can I buy a power-measuring system?


A: Check with your local bicycle or triathlete shop, or the manufacturers’ web sites for dealer listings.  You may also find the products in cycling catalogs and/or on the web, and many coaches are also dealers for the several systems.



Q: Which model is best?


A: Would you believe the stadium steps?!  Accurate, reliable, and the least expensive!  Kidding aside, this is an area which can excite considerable controversy (!), so no recommendations are made here; each of the four models available can be a valuable training aid if used properly, and the final choice is largely a personal one.  Indeed, Wattage forum member Robert Chung has compiled several “Rosetta Stone” files comparing power data collected simultaneously, which showed close consistency between the models tested, as did Kraig Willett’s test at Kraig Willett’s test.  Here is a comparison chart:


  Ergomo Sport Polar S-720i/710i PowerTap Standard PowerTap Pro PowerTap Pro SL SRM Professional / Amateur
Measurement location Bottom bracket (Campagnolo or Shimano OctaLink) Chainstay and rear derailleur Rear hub (130 or 135 mm; 24, 28, and 32 hole drillings) Same as Standard Same as Standard Crank (Shimano OctaLink or Campagnolo; 167-182 mm lengths in 2.5 mm increments)
Method Photointerrupter circuit Chain speed and vibration frequency 4 strain gauges Same as Standard Same as Standard 4 strain gauges for Pro, 2 for Amateur
Claimed accuracy ± 2% ± 10% at any one instant, but 2-5% or less on average ± 1.5% Same as Standard Same as Standard ± 2.5% for Pro, ± 5% for Amateur
Recording interval Averaged values recorded every 5 sec. Current values recorded every 5, 15, or 60 sec. Current values recorded every 1.26 or 2.52 sec. Current values recorded 1.26, 2.52, 5.04, 10.08, or 30.24 sec. Same as Pro Averaged values recorded 0.01-30 sec.
Memory capacity 11 hr.

1 workout file

4:57-76:37 hr.

Up to 99 workout files

4 or 8 hr. depending on recording interval,

1 workout file

7.5-180 hr. depending on recording interval.

1 workout file

Same as Pro 0:45-225  hr. depending on recording interval.

Numerous workout files

Calibration By manufacturer only; accuracy can be checked via static ‘stomp test’ described below No; but accuracy can be checked on hill of known grade No; accuracy can be checked via static ‘stomp test’ described below Same as Standard Same as Standard Slope setting is user adjustable; manufacturer calibration now available in U.S.
Mass (grams) BB w/bolts & wires = 344 g

Computer  & mount =168 g

Sensors = 118* g

Computer = 53* g

Mount/wiring = 71* g

Hub = 579* g (w/o skewer)

Computer =39.5* g

Mount/wiring = 36* g

Same as Standard, plus slight added mass due to crank-mounted cadence sensor. Hub = 416 g (w/o skewer)

Remaining compo-nent masses are same as Pro

Pro = 560 g

Amateur = 640 g

Computer = 120 g

Mount bracket/wire = 30 g

Advantages 1. outstanding software (CyclingPeaks) with many useful analysis tools

2. third-generation design

3. fully hard-wired system is not affected by electronic or radio interference

4. easy installation

5. rechargeable computer battery lasts 5,000 hr., is good for ~30 hr., recharges in 2-3 hr.

6. almost no limit on component choice

1. least expensive of all options

2. feature-rich software, and extra hardware features like altitude

3. allows use of any wheel or crank that you want

4. large memory capacity, stores many workouts

5. incurs the smallest weight penalty

6. not affected by temperature

7. does not require calibration

1. easiest to move from one bike to another

2. affordable and accurate

3. compact, readable, easy-to-use display

4. most hub internals (axle, freehub, and drive side bearings) are all user-serviceable without disturbing strain gauges and electronics

5. easiest to install, andeasiest to remove for racing – just swap rear wheels

Same as Std., plus:

1. expanded memory (up to 180 hr.); can store only one file but can create unlimited number of intervals

2. display has time of day, and rolling average capability for power, speed, and cadence data; can be customized for these functions

3. can display “pedaling power” (excludes 0, i.e., coasting values)

4. can be used with fixed gear

5. measures actual cadence (more accurate than the Standard model’s “virtual” cadence)

6. easier operation of interval feature

7. mileage is programmable

8. faster downloading with Link software v. 1.04

Same as Pro, plus:

1. improved hub internals (4 sets of sealed cartridge bearings), but not user serviceable

2. hub is 162 g lighter than Pro or Std.

3. available in fixed gear and Cam-pagnolo freehub versions

4. improved software is also Java-based and Mac-compatible

1. excellent software

2. time-tested, reliable design

3. can display rolling average for current wattage

4. large memory capacity, can store multiple workouts

5. no limit on wheel choice


Drawbacks 1. large/heavy computer

2. bearings must be factory serviced ($300) every 15-20,000 mi.

3. not easily moved from bike to bike

4. cannot accept 2004 Dura-Ace cranks

5. averaged data can be accessed only by download (cannot be viewed during interval)

6. not useful on tandems

1. most difficult to set up properly

2. difficult to move from bike to bike (to the point that it will likely never happen)

3. small display is hard for some to navigate

4. the least “clean” installation (multiple sensors and cables)

5. averaged data cannot be viewed during intervals (or ‘laps’), only at the end of the ride

6. accuracy questionable on stationary trainers, possibly from harmonic vibrations effects

7. not practical on MTB, and cannot be used with fixed gear

1. mediocre software interface (also not Mac-compatible

2. limits wheel choice

3. wheel-based system (not hub itself) is more likely to be damaged in crash

4. no disc version; requires a cover (not USCF-legal after 1/1/07) to be used as a disc

5. no disc brake version for MTB

6. not available with Campagnolo freehub

7. drive-side bearings and cone are substandard quality

8. cannot be used with fixed gear

9. “virtual” cadence only (limited to 40-140 rpm)

10. reliability problems in wet weather with original version; Graber version has better seals and coated circuitry

11. limited memory (7.5 hr.), stores only one workout

12. no rolling average or “pedaling power” capability

Same as 1-7 for Std., plus:

1. hub requires modification to be used with fixed gear

Same as 1-5 for Std.


1. expensive

2. not made to be moved from one bike to another

3. some find display more difficult to read

4. daily calibration (takes ~30 sec.) recommended

5. user-serviceable, but factory service recommended every 1500 hr., and replacement interval for cranks (not including power measuring unit) is once yearly

6. crank is slightly more flexible than other models; Dura-Ace version available at significant extra cost

7. not useful on tandems

8. accuracy of Amateur deteriorates outside a ~100W range, and may drift significantly over the course of a season

Pedal analysis No Yes No No No Extra option
MSRP $1,200 720i, $575; add-on kit only, $315 $700 without rim and wheel build $900 without rim and wheel build;

$1,000 with built wheel

$840, hub only;

$1,200 without rim and wheel build;

$1,300 with built wheel

Pro, $2,650; Amateur, $1,770
Other Display and software give average power for pedaling time only. Display and software give only average power for with 0 (coasting) values.

Original grey case changed to yellow in 2004.

Data transmission is through carbon fiber “windows” in hub shell.  Electronics are completely contained inside hub; only batteries are accessible from cap. Average power display obtained only from pedaling time, but non-zero values (i.e., when coasting) included by SRM software.
*Actual mass; all others are manufacturer’s claims.

For comparison, the mass of a Shimano Dura-Ace FC-7700 rear hub is 312 g, while Campagnolo lists the Chorus at 260 g, and Record as 248 g (all masses without skewer).  A Dura-Ace FC-7700 bottom bracket is 201 g; an FC-7410 right crank and chainrings are 395 g; an Avocet 45 computer and mounting bracket are 20 g and 16 g; and a Polar Coach heart rate monitor (HRM) and mounting bracket are 40 g and 26 g, respectively.






Q: It looks like memory capacity varies considerably.  Why?


A: It depends on the chosen recording interval, and, for the Polar S-710, which features are turned “on” and “off.”  Scott Harvel has compiled this chart showing all the possibilities:


Altitude Speed Cadence Power 5 s 15 s 60 s
On On On On 4 h 57 min 14 h 53 min 59 h 34 min
On On On OFF 8 h 56 min 26 h 48 min 99 h 59 min
On On OFF On 5 h 35 min 16 h 45 min 67 h 01 min
On On OFF OFF 11 h 10 min 33 h 31 min 99 h 59 min
On OFF OFF OFF 14 h 53 min 44 h 41 min 99 h 59 min
OFF On On On 5 h 35 min 16 h 45 min 67 h 02 min
OFF On On OFF 11 h 10 min 33 h 31 min 99 h 59 min
OFF On OFF On 6 h 23 min 19 h 09 min 76 h 37 min
OFF On OFF OFF 14 h 53 min 44 h 41 min 99 h 59 min
OFF OFF OFF OFF 44 h 42 min 99 h 59 min 99 h 59 min
TOTAL ONE FILE 99 h 59 min
TOTAL ALL FILES 520 h 00 min



Q: If the data is sampled every 60 seconds, what good is that?  After all, it seems to change so quickly.


A: (Alan and Jean-Joseph Coté, Jason Yanota, Andrew Coggan, Hunter Allen, Tom Compton, and Dave Wendt)  On longer rides, it is likely sufficient to record data with relatively infrequency, whereas on the track, every second may not be often enough.  Still, 60 seconds seems too long to be of use, at least for Polar units, which record current values at the end of each interval, rather than an average over its duration.  They do so since they are designed as multisport devices, and the longer recording intervals are more appropriate to other data they collect (HR, temperature, altitude), which does not change as rapidly as power.


There are three different measures of the ‘data stream’ which must be distinguished:


Signal rate the number of times torque is measured in a given period.  The strain gauges in the PowerTap hub do this 60 times per second (Hertz), while the sampling rate for the SRM is 200 Hz.  The Ergomo measures torque 72-144 times second, i.e., every few degrees, which is why it or a future model has the potential to provide data on variations in torque within a single revolution.


Display (or refresh) interval the length of time between each update of the readout (display).


Recording interval the length of time between each record of elapsed time, distance, speed, power, and cadence that is stored in memory for downloading.


The Polar S-720i and S-710i group power data by crank revolution, and consider each crank revolution to be indivisible with respect to the actual power reading.  Values saved in display memory are an average over the last few crank revolutions.  The number of crank revolutions isn’t always the same, i.e., when you pedal really slowly, it doesn’t average over a really long time, but over the recently completed revolutions (in a specific time window) where the instantaneous samples are grouped by revolution.  At higher cadences, more revolutions are included in the calculation.  Since the update is faster than 5 seconds, the instantaneous values during a given revolution may be included in more than one successive reading.  It isn’t a straight average, i.e., the total work over the interval divided by its duration, but rather an average that is weighted toward the more recent revolutions.  The details of this calculation are proprietary, but the bottom line is that viewing it as a 5-second average is a pretty good approximation.


The values stored in memory that is downloaded are not averages of some interval, but rather, the current numbers as displayed on the monitor which are captured and stored once every 5, 15, or 60 seconds.  Since the update interval is faster than the recording interval, a high value may be displayed (and then saved as the maximum value on the S-720 “FILE”), even though it doesn’t occur at the point when the value gets stored for downloading, so the true maximum value might not appear in the downloaded data.  Again, the 5 second interval is really the only useful one for recording wattage.


Other than the Polar units, power-measuring devices display current power as an average over some short time period, which leads to a problem known as the “precession effect.”  That is, unless you are pedaling at a rate where one or several revolutions are exactly completed in each averaging interval, an extra quarter-revolution can occur, and that partial turn of the crank may be either a power stroke or a dead-center (and perhaps the opposite for the next sampling period), which will produce a less consistent reading, especially for short intervals; the maximum power value captured in the PowerTap’s display memory, for example, is significantly affected, since it is the highest average value achieved over just 1.26 seconds.  Thus, averaging over one (or just a few) crank revolutions would reduce variability in the current power display, track power more nearly as a rider senses it, and result in more accurate maximum values for instantaneous power.  Recorded power values could, and perhaps should, still be based on time.


For current power, the PowerTap Standard displays only the power calculated every 1.26 seconds, and when set to record every 2.52 seconds, discards values calculated at 1.26 seconds, i.e., it records every other value without averaging.  The Pro model, on the other hand, can display average over the last 1.26, 2.52, 5.04, 10.08, or 30.24 seconds for the current power value, but like the Standard, it records the instantaneous value at the selected recording interval, so for instance, when at the 10.08 second recording interval, every 8th value is stored, and the other 7 are discarded.  Some have noted that displayed memory is often a couple Watts higher than what is downloaded.  In fact, the “raw,” recorded data represents is the most accurate and unaltered information, coming directly from the hub.  The reason the display is slightly off is that it uses lower-precision arithmetic, rounds improperly, or computes running averages using a method that is prone to accumulated errors or truncation.  These corners are cut because memory and CPU computing power are at a premium.


The SRM averages torque during each pedal revolution, then multiplies the result by the average angular velocity (cadence) during the revolution, then makes calculations according to the specified interval:


0.1 second – all completed revolutions are averaged, if a revolution hasn’t been completed then the previous data is sampled again.


1 second – all completed revolutions in the previous second are averaged.  One revolution will be sampled in the first sample, two revs will be used in the second sample, etc.


10 second – all completed revolutions are averaged; at 90 rpm this would mean the average of the previous 15 pedal revolutions.


Instantaneous power is estimated using the torque analysis function, which samples torque at 200 Hz, and in this way, SRM claims there are no artifacts in its power calculations, however, this is only an estimation of instantaneous power, because we don’t know instantaneous crank speed, and speed variations, though slight, do occur while pedaling.  The crank torque and angular velocity that are combined to calculate power aren’t necessarily time-aligned properly, which can be an issue if cadence is changing rapidly.


As previously mentioned, the Ergomo takes 72-144 measurements per second (depending on cadence), averages them each second, and temporarily saves the result.  Each 1-second average in turn is averaged every 5 seconds, and then this number is recorded for download.  For example, 300 W, 300 W, 300 W, 305 W, and 310 W will be averaged by the computer and recorded as 303 W.  The Ergomo display is updated every second from a rolling average of the last 8 rpm, so a new number appears in the display each second as the rolling average is kicking out the last number.


Although both the Ergomo and SRM measure the torque, or twisting force, generated by the rider’s leg(s), the Ergomo measures it at the bottom bracket, whereas with the SRM, it is measured between the chainrings and the right crank arm.  When you push down with the left pedal, that torque is transmitted through the bottom bracket spindle, to the spider, and then to the chainrings.  When you push down with your right leg, however, the torque is transmitted only through the spider to the chainrings – none is transferred to the bottom bracket spindle, hence, the Ergomo measures the power output of the left leg only (and then doubles it), whereas the SRM measures the power output of both legs.  (If you use your right leg to help lift your left back to the top of the stroke – and many of us do – then there is some torque applied to the bottom bracket spindle.  This is, however, considerably smaller than the torque generated during the downstroke with either leg, furthermore, it is in the opposite direction.)  Some therefore claim that an imbalance between left and right leg strength (due perhaps to an injury or even just occurring normally) renders the Ergomo inaccurate, even as it may give consistent and repeatable values, however, this has yet to be demonstrated.



Q: It sounds like each ride produces a lot of data.  How to make sense of it all?


A: Software is included with each system, and there are also several aftermarket packages with enhanced capabilities, including,, and (comes in versions for both Mac and PC).  Lastly, has some useful analysis features.



Q:  I just had my first ride with my new power meter.  Is it normal for current power to fluctuate so much?


A: First-time power meter users are almost invariably surprised at how “jumpy” the current power display is, and question the readout’s reliability.  This is a result of having become accustomed to the heart rate monitor (HRM) as a gauge of intensity, and being fooled into thinking that the energy requirements of outdoor cycling are relatively steady by the delayed response heart rate to changes in intensity, an effect that is accentuated by the smoothing algorithm incorporated in the HRM’s firmware.  Although some of the variability in power is due to instrument artifact (the “precession effect,” as just discussed), the energy demands of road cycling do indeed fluctuate very rapidly and widely (sometimes referred to as the “stochastic” nature of on-road power output), something that can easily be verified by comparing power data collected outdoors against that obtained from most any indoor trainer.  Using your power meter’s interval feature, if it has one, or setting it to smooth (average) readout data over a period of time such as 30 seconds can help to ‘settle’ the display.



Q: My friend says he can average 275 Watts for 30 minutes.  Is that any good?


A: It all depends.  Power is somewhat ‘personal,’ such that three riders traveling the same flat section of road together, at the same speed, side-by-side (not drafting each other), might each have considerably different power outputs, so absolute Watts do not necessarily provide a valid basis of comparison.  This is even more outstandingly true going uphill, where the force you must overcome is determined largely (75%+ for an 80 kg bike/rider putting out 300 W on grades over 5%) by weight.  For instance, if Dan’s mass is 70 kg and he averages 315 W on a particular climb, while Felicia is 49 kg and maintains “only” 270 W on the same hill, she will drop him easily, since she is putting out 5.5 W/kg, while he generates only 4.5 W/kg.  On flat terrain, by contrast, the main resistance (80%+) is from air, so speed is a question of Watts per square meter of effective frontal area (CDA), which determines air drag.


The amount of good data on power-generating capabilities for cyclists across different skill levels, disciplines, and gender is limited, and of course, statistics don’t determine the outcome of a race; if they did, we could just set up trainers at the starting line, run a few tests, and proclaim the winner!  Still, it may be useful to gauge your own power against others, or those with whom you hope to compete, and if so, be sure to “normalize” (divide by) body mass or frontal area.  For timed events, such as a 40 km time trial, you can get a reasonable idea of the power needed to achieve a certain time for yourself by using an online model like those available at,, or


For his World Hour Record in 1996, Chris Boardman averaged an estimated 442 W, while Miguel Indurain needed about 510 W when he broke the same standard in 1994 (both about 6.5 W/kg), and an analysis of Lance Armstong’s time of 38 minutes 1 second in climbing L’Alpe d’Huez during the 2001 Tour de France produced an estimate of 6.5 W/kg, which came at the end of a 209 km long stage featuring two prior “hors categorie” (beyond category) climbs.  In setting a new women’s record of 54 minutes 2 seconds at the 2002 Mt. Washington (NH) Hillclimb, Geneviève Jeanson averaged an estimated 278 W (5.56 W/kg).  At the other end of the power-duration relationship, the best male match sprinters have hit 23 W/kg, females ~20 W/kg, however, comparable and even higher values than these have (somewhat surprisingly) been recorded by non-cyclists, such as weightlifters, hockey players, etc.



Q: What in the heck is this ‘CDA’ you refer to?


A: It’s the product of your aerodynamic drag coefficient, CD, and frontal area A.  Let’s start with A, which is simply the area of the profile a rider presents to the air they move through.  Stand directly in front of a rider, look at his or her outline, and you’ll see that a smaller rider has a smaller frontal area, while larger rider has more area to push through the air, and therefore must put out more power for a particular speed.


CD is a measure of how ‘streamlined’ you are, i.e., how smoothly air flows around your body/bike without swirling behind you.  Imagine two riders of exactly the same size and position, where one is using a Cervélo, an aero helmet, shoe covers, etc., while the other has a standard round-tubed bicycle, a Pneumo helmet, and no shoe covers (both wheelsets have 42 spokes, but the first has 58 mm deep-section rims, while the second has standard box-section rims).  Although both riders present the same frontal area, the former will have a lower CD, encounter less aerodynamic drag, and go faster at a given power output.


The product of these two components is CDA, or effective frontal area, and it is most accurately determined in a wind tunnel, but it may be possible to measure it with a power meter, on a flat course in calm air.  As a rough rule of thumb, an 0.005 m2 reduction in CDA = 0.5 seconds/kilometer = 0.1 lbs difference in drag at 30 mph = 5 W.



Q: How can I test my progress when training by power?


A: (Eddie Monnier)  Keep in mind that testing does not necessarily guarantee racing results, rather, it allows for periodic evaluation of your condition and training program.  If you do this often enough, you will have a bad test from time to time, so it’s important not to get too hung up on any one result.


There are several test protocols to determine what might be termed ‘functional threshold power,’ including the critical power and maximal aerobic power (MAP) tests (the latter is usually administered indoors).  The method proposed in the training guide referenced further on, however, is not a lab test, but a functional test of average power for a 40 to 60 minute time trial which is used for determining training levels since it integrates the underlying physiological mechanisms of endurance exercise: maximal VO2, lactate threshold, and efficiency, thereby giving a sort of “bottom line” measure of fitness.


You may also want to test your power-generating capacity at various durations, depending on race objectives and personal development needs; for instance, a criterium specialist will be more interested in maximal and average power over a 200-meter sprint than a climber, who will tend to focus more on average power on a particular climb.  In this regard, something like Power Profiling can be helpful.



Q: What are “normalized” power, intensity factor, and training stress score?


A: Created by Andrew Coggan, Ph.D., a noted exercise physiologist, this is obtained via an algorithm that adjusts for variations in power with reference to lactate threshold and other physiological responses.


Suppose you race a 1 hour criterium, where you are frequently sprinting out of corners, covering breaks, etc., and you race to your limit, such that there are very few if any slack periods.  Average power with coasting time will nonetheless be considerably lower than a flat 1 hour time trial where you paced steadily and had nothing left at the end, yet you feel just as stressed physically.  The normalizing algorithm adjusts for variations in power, such that the resulting normalized power value will be very close to what you would have achieved in a TT of equivalent duration.  In short, it is meant to more accurately reflect the actual metabolic strain that the body incurs, rather than the average stress load imposed.


Here’s how it is calculated: first, a rolling 30-second average (mean of the last 30 seconds) is applied to the wattage values from the downloaded workout file, since the body does not respond instantaneously to rapid changes in exercise intensity, rather, most physiological responses follow a predictable time course with a half-life of approximately 30 seconds.  Next, each of the values obtained from this is raised to the 4th power, just as blood lactate response has been shown to closely fit the plot of y = x4, where y = blood lactate and x = power output; indeed, many critical physiological responses (e.g., glycogen utilization, lactate production, stress hormone levels) are similarly related to exercise intensity in a curvilinear, rather than linear relationship.  Finally, all these values are averaged, and the 4th root is taken.


If that all seems a bit complex, just think of adjusted power as the equivalent power you would have achieved if the race course had been perfectly flat and the pace perfectly steady, with no variations.


Two other metrics of workout performance, intensity factor (IF) and training stress score (TSS), are derived from normalized power, though space considerations preclude further discussion; the previously-mentioned CyclingPeaks Software includes all three of these features and has a nice explanatory article as well, at, while there is a free on-line calculator at



Q: How does altitude affect power output?


A: The effects of altitude on VO2 uptake (and hence aerobic power) are highly individual, so it is difficult to predict to what extent any one person will be affected, although as a general rule it has been shown that elite athletes, as compared to normal individuals, have a greater decline in VO2max under conditions of reduced ambient pO2 (partial oxygen pressure).  This is caused by their higher cardiac output, which results in a decreased mean transit time for the erythrocytes (red blood cells) within the pulmonary capillary, and thus less time for equilibration between alveolar air and blood in the pulmonary capillary.


These equations from Bassett et al.1 were generated from 4 groups of highly trained or elite runners, so they are population-specific to that group, but can be used to estimate aerobic power at a given altitude as a percentage y of what is normally available at sea level, where x = elevation above sea level in km:


acclimatized athletes (several weeks at altitude):  y = -1.12x2 – 1.90x + 99.9 (R2 = 0.973)


non-acclimatized athletes (1-7 days at altitude):  y = 0.178x3 – 1.43x2 – 4.07x + 100 (R2 = 0.974)


whereas Peronnet et al.2 found y = -0.003x3 + 0.0081x2 – 0.0381x + 1.  Here is a table derived from these equations:


(feet above sea level) Bassett et al.1 Peronnet et al.2
acclimatized non-acclimatized  
0 99.9% 100.0% 99.9%
1,000 99.2% 98.6% 98.8%
2,000 98.3% 97.0% 97.8%
3,000 97.2% 95.2% 96.8%
4,000 95.9% 93.2% 95.8%
5,000 94.4% 91.1% 94.7%
6,000 92.7% 88.9% 93.5%
7,000 90.7% 86.5% 92.2%
8,000 88.6% 84.2% 90.7%
9,000 86.3% 81.7% 88.9%
10,000 83.7% 79.3% 86.7%
11,000 80.9% 77.0% 84.3%
12,000 78.0% 74.7% 81.4%
13,000 74.8% 72.5% 78.0%
14,000 71.4% 70.4% 74.2%


1Bassett, D.R. Jr., C.R. Kyle, L. Passfield, J.P. Broker, and E.R. Burke.  Comparing cycling world hour records, 1967-1996: modeling with empirical data.  Medicine and Science in Sports and Exercise 31:1665-76, 1999.

2Peronnet, F., P. Bouissou, H. Perrault, and J Ricci.  A comparison of cyclists’ time records according to altitude and materials used.  Canadian Journal of Sport Science 14(2):93-8, June 1989.


Thanks to David Bassett, Jr., Ph.D., for his contributions to this section.



Q: Where can I get more information on training by power?


A: Wattage forum member Andrew Coggan has created a ‘schema’ of Power-based training levels (backup sites #1 and #2), which Charles Howe has included in Part 1 of The Road Cyclist’s Guide to Training by Power, at  Another article by Coggan is Racing and Training With a Power Meter (backup sites #1  and #2), while a number of power-related articles can be found at the  Power Tap (archived), SRM (plus more at The Bike Age), and Cycling Peak Software web sites.  Additionally, Robert Chung and Amit Ghosh have personal web sites with many good articles, as do and


Finally, the Wattage Forum ( can provide much help and advice from that list’s many members; supplemental list archives are at as a StuffIt file, while the StuffIt Expander utility needed to open this file is at



Q: What about power-based training indoors?


A: (Charles Howe, Andrew Coggan, and Bruce Sargent)  Stationary trainers can offer an important form of supplementary training, not merely over the winter months or when the weather is foul, but even during good weather when a controlled, structured workout is desired.  Stand-alone trainers such as the SRM and Velotron appear to be well-calibrated, as is most any good lab ergometer, such as those manufactured by Lode.  In evaluating the wattage readout accuracy claims of bike-stand models, however, it is important to realize that just because a device has a digital display (such as the Computrainer, Cateye CS-1000, and various Tacx models) does not mean it accurately or reliably reports power output – and precision in administering the exercise load is one of the most important benefits of power-based training.  Indeed, of the various models that claim to be calibrated for power, only the Velodyne ( appears to be consistently accurate and precise (see, while realistically replicating the actual demands of cycling on the road.  It achieves this with a feedback control system that measures and adjusts the resistance of an electrically-controlled brake as well as a 10 kg flywheel that simulates the “inertial” forces encountered during actual riding.  When properly calibrated, it faithfully reproduces the experimentally validated speed-to-power equation (by software design), although it assumes a set value for frontal area which may or may not correspond to that of the individual riding it.  Additionally, it will simulate drafting (though the magnitude of the simulated reduction in air resistance is unknown), but not headwinds or tailwinds.  In ergometer mode, it will hold wattage constant down to ~5 mph, i.e., power will not vary as your speed changes at levels over 100 Watts.  The ability to maintain constant power in “erg” mode is a bit speed-dependent in the other direction too, i.e., rolling resistance accounts a large enough fraction of the total power demand that it is hard to get the power down when wheel speed is too high.


Overall, an on-bike power meter in combination with most any bike-stand load generator is the most affordable, flexible, and accurate arrangement for precision power-based training indoors; those with a flywheel adequately heavy to simulate inertial forces (designated with an “*” below) are recommended.  Except for the Velodyne, an accurate power meter is the only way to know the energy requirement of bike-stand models, which renders the load generator’s readout superfluous.  (Note: the accuracy of the Polar S-720i/710i on indoor trainers has frequently been observed to be highly unreliable.)


Stand-alone trainers:

Cardgirus – (See also Andy Birko’s review)

Cyclesimulator –

Kettler Ergo Racer – or

Lode ergometers –

PowerTap –                    SRM –

Velotron –


Bike stands:

1up USA CPR A-2000 –

*Bike Technologies Advanced Training System –  and

Blackburn Fluid and Mag Trainers –

Cateye CS-1000 –

*Chaindriver –                Cycleops –

Kinetic – (*custom version:

Kreitler –                             Tacx –


Power requirements of numerous bike-stand models have been charted at, while Robert Wells has posted a careful evaluation of the Tacx Flow at



Q: What does accuracy matter anyway, so long as a unit’s readout is consistent?


A: (Andrew Coggan)  Accuracy is important if:


  1. you wish to compare your performance over the long term with different devices;
  2. you wish to compare your performance to others;
  3. you want to use a self-assessment tool such as use my “Power Profiling” tables, or
  4. you wish to do any modeling, e.g., to predict your time on a new course.


It is true, however, that precision (reproducibility) is probably more important than accuracy – but that includes across various power outputs, pedaling speeds, etc., as well as across brands of powermeter.  Care must be exercised when comparing data collected using different power meters, even if they have all been carefully calibrated.  In my own case, for example, the slight improvement in power I’ve seen at various longer durations recently, as compared to previous years, can potentially be accounted for entirely by my switch to SRM from PowerTap, since the former measures power “upstream” of the chain, whereas the PowerTap measures it “downstream,” or after the chain.  On the other hand, my power at 5 seconds is down significantly, as are peak values at 10, 15, and 30 seconds, but 1 minute isn’t, which makes me believe that it is real, and not an artifact of the change in systems.  Without careful assessment of the data collected with each powermeter, however, an incorrect conclusion about whether a certain type of training is or isn’t working might be reached.



Q: Can a power meter be used as an aid in dieting?


A: Since they accurately measure energy output, power meters can be used to estimate metabolic energy expenditure in kilocalories (simply “calories” in common usage.)  Most models give the total work for a ride in kilojoules (kJ), but if not, average power output for the ride can be converted to kJ when multiplied by ride duration in hours (decimal form) and a factor of 3.6.  For instance, if you averaged 142 Watts for 1 hour 22 minutes, that’s 142 W ´ 1.37 hr ´ 3.6 = 699 kJ. Since the body is ~20-25% thermodynamically efficient, this roughly cancels out the unit conversion factor (4.184 kJ = 1 kcal), and the work accomplished in kJ during a ride is pretty near equal to kcal burned by the body.  Unfortunately, efficiency varies during a ride, increasing directly with intensity, and it must be determined in a lab, but here are factors for converting kJ to kcal over a range of values for efficiency:


If you are 25% efficient, kJ × 0.96 = kcal, and 87.1 W are produced by each liter of oxygen uptake

for 24% efficiency,  kJ × 1.00 = kcal, and 83.6 W are produced by each L of O2 uptake

for 23% efficiency,  kJ × 1.04 = kcal, and 80.1 W are produced by each L of O2 uptake

for 22% efficiency,  kJ × 1.09 = kcal, and 76.6 W are produced by each L of O2 uptake

for 21% efficiency,  kJ × 1.14 = kcal, and 73.2 W are produced by each L of O2 uptake

for 20% efficiency,  kJ × 1.20 = kcal, and 69.7 W are produced by each L of O2 uptake


The OwnCal feature of Polar HRMs only estimates calories expended, based on averages derived from large samples, and thus can vary widely by individual, as the manufacturer itself admits.



Q: I’ve heard that temperature really affects the accuracy of the SRM and PowerTap.  True?


A: (Andrew Coggan, Chris Cleeland, Jesse Bartholomew, and Andy Birko)  A recent study found that both read higher in colder air than warmer (8° C, or 36 F) as compared to the lab (70° F), but this was because the investigators tested them without re-zeroing.  In other words, they deliberately disregarded the manufacturer’s recommendations for use, and the error should therefore be viewed as a worst-case scenario due to improper operation.  If you reset the zero at the same temperature at which you collect data, then accuracy will be unaffected.


The PowerTap autozeros when coasting (i.e., whenever there is zero torque applied), however, if there is an offset of more than ±8 in-lbs, the unit will require the user to re-zero.  This usually occurs due to a large temperature change, so to obtain the most accurate data, you shouldn’t just look for non-zero watt values while coasting, you should look for non-zero torque values.


This page from a software company shows that proper engineering can detect strain in the presence of thermal-induced stress.



Q: How do I calibrate my power meter?

A: (Andrew Coggan)  Neither the Polar S-710 nor the PowerTap require calibration after initial set-up.  Calibration of the SRM via slope adjustment can be performed by the user, as described in the Owner’s Manual at (click on “Troubleshooting,” then “Calibration check”), and a more complete calibration procedure is now available un the U. S. as well.

Technically, the PowerTap cannot be user-calibrated, but its accuracy can be checked using a simple test that is similar to the SRM calibration check.  First, check that the transmission icon is on, and if not, give the rear wheel a spin.  Then, enter the torque mode by holding the “Select” button down for 2 seconds or longer (the “WATTS” designation will disappear from the top line.)  Apply the rear brake sufficiently to lock up the rear wheel.  Now, measure torque as follows: with the cranks exactly horizontal (right crank at 3 o’clock), hang a known weight of at least 50 lbs from the right crank, or simply stand on it – hence the name ‘stomp test’!  Measured torque  =  (weight in lbs)  ×  (crank length in mm)  ×  (1 in/25.4 mm)  ×  (cog teeth/chainring teeth).  For a 159 lb rider standing on a 175 mm crank, with the chain on the 39 tooth ring and the 23 tooth cog, 159 lbs  ×  175 mm  ×  1 in/25.4 mm  ×  23/39  =  646 in-lbs.  Compare this to the displayed value by calculating % error as (measured torque – displayed torque)/measured torque.



Q: Can I race with a power meter?


A: Sure!  The most obvious application is time trialing, where it is invaluable for pacing, particularly in the initial stages of the race, and even for pursuit events on the track, as well as short (<10 minute), prologue-type events on flat terrain.  Although criteriums allow fewer situations where power data can be conveniently accessed during the race, it can be used in road races to judge effort when off the front, in a breakaway, or bridging up, and when seeking the “sweet spot” in a paceline or echelon.  Even if not useful during the race, a power meter can be used as a “black box” (ride data recorder), allowing informa-tion to be analyzed afterward to quantify the demands of the race, and training programs to be tailored accordingly.  Still, some who train using a power meter choose to race without it for psychological reasons, and ultimately, its use in competition is a matter of personal preference, like an HRM.



Q: How does the PowerTap calculate cadence without a sensor? 


A: (Andrew Coggan)  This “virtual cadence” feature estimates crank rpm based on the time from one peak in torque to the next as your legs pump up and down.  Such peaks occur very frequently (e.g., every 333.3 milliseconds at 90 rpm) and have to be identified “on the fly,” so any slight variation in either when pushed down the hardest or when the computer thinks you pushed down the hardest will therefore have a significant effect.  Depending on how/how fast you pedal, the cadence values can therefore be quite erratic, even though the power measurements are still accurate.



Q: Can I use the PowerTap just as a computer, without the hub?


A: (Andrew Coggan)  An undocumented function of the original (grey case) PowerTap Standard, this is now explained in the owner’s manual for the PowerTap Pro.  Anyway, in the normal (not interval) mode, scroll to current cadence (it should be all dashes if the cranks are not turning).  Hold down the “select” (right-hand) button for about 3 seconds until it says “OFF” on the top line where Watts are normally displayed.  Speed, heart rate, and distance will now be shown, but not cadence or Watts.  Return to normal mode by reversing the process (you will have to cycle back to current cadence, since as soon as you let up on the right button during the above procedure, the computer jumps to average cadence).

You need to mount a magnet on the rear wheel and ensure it passes very close to the sensor (5 mm or so).  According to a PowerTap, the receiver on the bike may be more “particular” about magnet strength and location than your average cycling computer, which may be why this function was left undocumented.



Q: The drive-side bearing cone in both my PowerTap Pro and Standard hubs wore out after less than 3,000 miles – it makes a crunching sound and does not turn very smoothly.  Any suggestions?


A: This is a clear deficiency in materials that should be corrected by the manufacturer.  Until they do so, try using item SH-3AO9803, a right cone for Shimano Dura-Ace rear freehub FH-7700, available from Loose Screws for $12.20 each.  Grind or machine the narrow end of the cone down a few millimeters since it is too long, and file the inboardmost ‘step’ off the aluminum spacer that comes with the PT hub, but once you do, it all works fine, and since there is a rubber lip seal on the cone, the hub will be double-sealed.  There are other Shimano cones that may fit better, but I don’t know which ones for sure, and this one is the best quality.


Note that the non-drive side bearings are sealed, and must be serviced by the manufacturer.



Q: What can I do to improve the waterproofness of my PowerTap hub?


A: (Chris Cleeland and Lindsay Edwards)  Get some tune-up grease, also known as dielectric grease (or heat sink grease in the electronics world, although that tends to have thermal conductivity properties as well as being dielectric) from the nearest auto parts store (or “auto spares” as they say in the U. K.)  This is the stuff that’s made for the inside of spark plug wire boots to ensure that they can be removed, but won’t conduct electricity.  Squeeze a liberal portion of this on to your finger, then smear it all over the leaf contacts both on the cradle and the nubs on the back of the CPU.  This will keep water and moisture out of the contacts, but maintain the connection.


The other issue is water in the hub itself, which happens to me less often, perhaps 1/3 of the time I ride in rain (though I have yet to be caught in an all-out downpour).  It also happens in heavy fog occasionally.  A simple overnight period where you take the cover off is enough to dry it out and get things working again.  I’d suggest using tune-up grease here, too.  It’s a little thicker than Pedro’s syngrease, doesn’t break down in heat as much, and if it does, it won’t affect electrical connections.  Here’s a link describing Permatex’s product and to a place selling it online:


Finally, apply some silicone sealant around all the joints of the receiver, paying special attention the point where the cable enters the body of the receiver.



Q: I want to build custom wheel from a PowerTap hub, but I’m not sure how to spec it.


A: Critical dimensions for both old and new versions are given below, and you also need the effective rim diameter (ERD).  Armed with these parameters, you can determine spoke length using one of the on-line calculators at  The following table provides specifications for selected rims (more ERDs can be found at the above link as well).


OLD PowerTap Hub (130 mm over-locknut distance)

Mfd. before 2001 by Tune; painted matte silver finish.

Center-to-flange width, left side: 36.2 mm
Center-to-flange width, right side: 16.7 mm

Diameter through spoke holes, left side: 78.0 mm
Diameter through spoke holes, right side: 66.0 mm
Spoke hole diameter: 2.4 mm

NEW PowerTap Hub (130 mm O.L.D.)

Introduced late 2001 by Graber; shiny polished silver finish.

Center-to-flange width, left side: 32.7 mm

Center-to-flange width, right side: 16.7 mm

Diameter through spoke holes, left side: 78.0 mm

Diameter through spoke holes, right side: 66.0 mm

Spoke hole diameter: 2.4 mm

PowerTap SL Hub (130 mm O.L.D.)

Introduced late 2004 by Saris/CycleOps; carbon fiber center section.

Center-to-flange width, left side: XX.X mm

Center-to-flange width, right side: XX.X mm

Diameter through spoke holes, left side: XX.X mm

Diameter through spoke holes, right side: XX.X mm

Spoke hole diameter: 2.4 mm








CAMPAGNOLO Sydney (32° only) C 30 553 581
MAVIC Open Pro CD C 18 439* 602
SUN ME14A C 20 421 601
SUN Venus C 25 440 592
VELOCITY Aerohead C 21 405 598
VELOCITY Deep V C 30 520 582
VELOCITY Pro Elite T 30 500 582
ZIPP 415 C 38 415 567
ZIPP 280 T 38 280 569
ZIPP 505 C 58 568* 529
ZIPP 360 T 58 360 530
*Actual mass; all others are manufacturer’s claims.

ERD – effective rim diameter    C – clincher    T – tubular.

For 28 spokes, use 2-cross, for 32, 3-cross (all rims available in both drillings unless noted).  As a general rule, round the calculated spoke length down if using brass nipples, up for alloy nipples.


The CH Aero wheel cover, available from Excel Sports, essentially converts the PowerTap wheel to a disc, although the hub opening on the left-side cover must be cut larger to ~73 mm diameter; a carbon fiber version is available by calling the manufacturer at (800) 227-6751.  To make a home-made disc/wheel cover, see these instructions from Warren Beauchamp and Bob Schwartz, as well as an additional note from Ken Lehner.  Note: wheel covers will become illegal under when the U. S. Cycling Federation adopts UCI bicycle regulations on January 1, 2007.



Q: What about a PowerTap hub with 135 mm over-locknut distance?


A: (Rick Moll and Jeese Bartholomew)  The hubs are the same with except for axle length and spacing; a 5 mm spacer is added to the left (non-drive) side, so the right side flange is shifted 2.5 mm away from the hub center, while the left side flange is shifted 2.5 mm towards it, therefore:


Center-to-flange width, left side = 32.7 mm – 2.5 mm = 30.2 mm

Center-to-flange width, right side = 16.7 mm + 2.5 mm = 19.2 mm



Q: How can build up a Power Tap hub as a fixed-gear wheel?


A: Check out this article:



Q: How can I mount the PowerTap harness on my stem?


A: (Chris Mayhew)  You can do this by crossing the zip ties so that they exit one side of the mount but cross over and enter the other side.  A cleaner method, however, is to tightly wrap electrical tape around the stem and the lower part of the mount, behind the ‘ears.’  Be careful how much tape you use; too much will cause a poor fit between the harness and PT.  With both methods it’s best to put a very small piece of pipe insulation under the harness to fill in the gaps.

An off-the-shelf mount can be purchased at

Q: My bike has a Campagnolo derailleur, but the PT has a Shimano 9-speed freehub.  What to do?

A: (Brian Smith and Eddie Monnier)  Quoting Sheldon Brown, “For reasons that are not quite clear, 9-speed hubs/cassettes seem to work pretty well with the opposite brand of 9-speed derailleur/shifter,” the operative words being “pretty well,” so results may vary, but many report doing so without any problem (avoid using a Campagnolo chain on Shimano cogs, however).

An excellent cassette to convert the PowerTap for use with Campy 10 is available from Wheels Manufacturing but the one from American Classic is not  recommended (see “Important Notes” at – “The following wheels and hubs are incompatible: Shimano pre-built paired spoke wheelsets, and ALL Powertap hubs.”)  It seems that the spacers on this model, and on the cassette from Miche as well, are fixed so that the smallest cog (e.g., the 11) just barely seats onto the PowerTap freehub body, making it vulnerable to “spinning” on the freehub body.  With the Wheels cassette you will get a few spacers that are wafer thin, so you can fine tune how much the smallest sinks into the body.  This gives a more positive fit so that you shouldn’t have any problems.



Q: Which of the two types of PowerTap pickups should I use?


A: (Jesse Bartholomew)  All PowerTap hubs made by Tune (matte silver-grey hub body) and some CycleOps PowerTap hubs are designed to be used with a receiver that mounts 7″ from the hub for optimum signal transmission; these have a serial number of 27383 or less.  In a successful attempt to limit data drops, we “tuned” the hub and receiver, resulting in a new receiver that needs to be mounted closer (3-5″) to the hub; these have a sticker indicating how to mount them on one side, a CycleOps brand sticker the other side, and serial number greater than 27383.


Part of the tuning was to desensitize the receiver a bit, and because the new PowerTap SL hub transmits through carbon “windows” in the hub shell, the signal is weaker, and the receiver won’t pick it up consistently, so we’ve gone back to the original 7″ style receiver, but we recommend mounting it no more than 3″ from the hub to maximize consistent data transmission.  I know that’s terribly confusing, but the short version is that the only combination that won’t work together is the SL with the 3-5″, current receiver.  So if you do upgrade to the SL you’ll need to use the SL model receiver with whatever other hub you are using for training.



Q: I have a Mac G3 Powerbook without serial ports and want to run the Polar software with IR interface.  How can I do this?


A: (Bill Pence)  Use the Keyspan High Speed Serial to USB adapter and the serial version of the IR receiver from Polar.  You need to be running Virtual PC 5.0 and Windows 2000.  The Mac OS needs to be 9.2, as 9.1 does not seem to work too well with VPC, nor does OSX (OS 10).  I’ve run the Polar PPP under VPC 3.0, which worked fine, but the IR adapter didn’t work.  I found that I needed to upgrade my device driver to Keyspan V 1.9.


Finally, I needed to plug the Keyspan Serial Port adapter into a USB port on the back of my G3 – for some reason it was not happy plugging into the spare USB port on the keyboard.  I have not been able to make the USB IR interface work with this setup, but serial port works just fine the way I have it set up now with my S-710.


After that it is a matter of setting all of the dialog boxes correctly.  With the IR adapter plugged into Port 2 of the adapter, the Keyspan Control panel it will advise you of the devices attached.  Click Advanced Settings for more detail.  Pull down the menu in the dialog box to Port #2.  It should read something like P#2USA28X02.  This is specific to the Keyspan device, and identifies port number 2 (where you plugged the adapter).  If funny things have been happening, you may reset the port here.  Mine is set to receive FIFO of 16 and a buffer of 64, both default values.  Also, make sure interrupt endpoints is set.  I do not know what effect it has, and am not anxious to find out.


Leaving Keyspan and launching VPC, once Windows has booted up, look in the menu bar on the Mac side of the house (hold down the Apple key and the Mac menu bar appears) and pull down “Edit Windows 2000 Settings’” which brings up the settings list.  Click on COM 1.  A dialog box will appear to the right of the window with various radio buttons.  Click “Mac Serial Port.”  P#2USA28X02 should appear below it; this is your Keyspan Port with the IR adapter on it.  Check “Non-Modem device” on the next line below, and the COM 1 port is now mapped to the Keyspan port that the IR Adapter connects to.

One more step.  Launch the Polar Software (new versions are best).  Pull down the “Options-Preferences” menu.  A dialog box will appear labeled “Software Preference,” and click on the “Hardware” tab.  In the top of the box is a section devoted to the Polar S-series HR monitor.  Set the pull down menu to COM 1.  Click the “Options” button.  I have mine set to USB Autocheck and Keep HR in Connect Mode.  I don’t think it matters unless you check “Use Windows Internal IR” port, which would be very bad.

Set the 710 in front of the IR receiver and click the connect button.  It should connect and work just like running on a Windows PC, and that should be all there is to it.  I do not believe the USB IR adapter will work with Virtual PC.  I tried.  A lot.  VPC is not happy sharing USB devices with the Mac OS.

Q: (Brian McLaughlin)  Hold on there!  I have been using OSX with VPC 6, using the USB Keyspan converter model USA-19QW.  It all works fine with PowerTap software and CyclingPeaks software.  The driver CD that I have is 1.2 for OS X, 1.9 for 8.6-9.x, and it works fine as long as I make sure it is recognized in the set-up of VPC before I try to download data.  I use COM 2.

A: Thanks!

Q: When starting VPC 6 on my iBook (OSX), it takes forever – so much so I have never actually opened it.  My iBook has 128 MB – and I have a suspicion that VPC 6 needs 256 MB.  Does anyone know?

A: (Anne Grofvert, Chris Bartholomew, and Jeff Lawson)  192 MB of RAM is what Microsoft specs for 6.1, but to run VPC you really need to allocate much more RAM than recommended.  I have 756 MB and have allocated almost 400 MB to VPC to get it running smoothly.  When you exit VPC be sure to “save all and quit” to preserve your settings, so that VPC it doesn’t take forever to load the next time.

The beauty of the new Panther operating system is that it allocates memory to the programs being used, so VPC does not affect the functioning of the machine when you are using other applications.  VPC does not seem to run on a G4 either, again, due to lack of memory capacity.  It just sits there and does not load.

Q: My Polar S-710 has been difficult to install.  Any suggestions beyond what’s in the owner’s manual?

A: Wattage Forum member Robert Chung has devoted a page to this at his web site:, and there is a video at the Polar web site

(Tom Anhalt)  The angle of the chain across the sensor and whether or not the sensor module is parallel to the chain do not matter; all that counts is to position the module so the chain is no farther than 30 mm, in all usable gear combinations, from an approximately 1” square area centered on the “middle” mark on the module.  If this requirement is met, and if the cadence sensor is properly positioned (which depends on the particular magnet you use), you’ll get consistent readings, otherwise, the chain vibration signal will be weak and the signal processor will tend to “lock on” to signal noise, causing erroneous readings.

Some comments on the Polar installation video:

  1. why mark the center of the chainstay? This is the first thing shown, but it’s not used for anything. The location of the module on the chainstay is driven solely by the placement of the magnet on the crank, and then placing the module so that the cadence sensor lines up.
  2. the routing of the speed sensor wire just begs for it to get snagged and ripped out. There are much better techniques for routing and securing this wire to the derailleur that will minimize this threat.
  3. it is a mistake to make the vertical spacing measurement 5-10 mm in the small-small combo. It’s the wrong end of the range from which to make this critical measurement, since the chain will be much farther away than 30 mm even in the small chainring-large cog combo. I run a pretty “normal” gear setup (53/39, 12-25 cogs), and if I try to run the 39 x13 (which I don’t because of the cross-chaining), the chain actually rubs on the sensor.
  4. there is no mention of making sure the chain passes over the sensor in all gear combinations, a significant omission.

Finally, to protect the sensor module, I first tried some mylar, but that didn’t last long.  The best thing to do is to grab a couple of black zip ties and wrap them around the module right over the top of the magnetic frequency sensor (that’s where the chain will be pulled down).  This way, the chain will rub on the “sacrificial” zip ties instead of the top of the module.

Q: I’ve had problems with the cadence magnet that came with my Polar power kit.  It just doesn’t want to stay in place on the crank arm.

A: (David Bilenkey and Tom Anhalt)  The magnets supplied with the kit work poorly, if at all – don’t bother with them.  Instead, get a 1/2 or 3/8” diameter “rare earth” magnet, such as from Radio Shack, Lee Valley Tools, or National Imports.

These are small disks, 1/8” – 1/16” thick, and should cost less than $2.  If your pedal spindles have any ferrous content (“stainless steel” may or may not), just drop one on the backside of the spindle.  Align the magnet in the best location to make the little cadence light blink.  No tape required; it’s strong enough not to fall off, but not strong enough to pull the chain over against the crank.

To remove the magnet, remove the pedal and slide the magnet sideways to get a grip on it and peel it off.  If your pedals have titanium spindles, or non-ferrous stainless steel, simply place a piece of electrical tape across over the magnets so they won’t fall off during bumps.

You might also try gluing the rare earth magnet to the center of one side of a ½” diameter ceramic magnet (making sure you match a south pole to north, or vice versa) and then glue this “stack” to the backside of your pedal spindle.  This should eliminate any problems with chain interference.

Q: As an aside, how does the Polar power module ‘know’ the free length of the chain?

A: (Jean-Joseph Cote)  Since cadence, wheel speed, and chain speed are measured, there’s enough information to calculate the number of chainring and sprocket teeth, and from there, the diagonal length of the vibrating section of the chain can be obtained (this is in the patent).  Polar chose not to display the gear numbers, presumably due to limitations of the display size.

Q: Is it true that downloading drains the battery in the PowerTap Standard computer?

A: It will if you leave it in the download cradle.  Remove it after downloading, replace it in the handlebar mount, and let it “fall asleep.”  Then, remove it and do a “clr all.”  It is now in its most efficient mode.  When the battery starts to get low, HR function seems be the first thing to go, becoming unreliable, with many “data drops.”

Q: I’m running XP Pro on a Pentium 4 CPU, and when trying to install the PowerTap Link software (vers. 1.02), I get the same crash.

In the “Quick Access” Screen, whenever I click on rider management, I get the following run-time error 

Runtime Error


Method ‘~’ of object ‘~’ failed

A: (Rick Sladkey)  It sounds like your data access components did not get properly updated by the link installer.  You might try manually installing the latest Microsoft Data Access Components (MDAC) 2.8:

Q: I need a converter so I can download the PowerTap to a USB port, rather than a serial port for which it was designed.

A: Try item the adapter Keyspan High Speed USB-Serial Adapter USA-19HS (recommended by PowerTap; formerly USA-19QW), at, item GUC 232A from iogear at, or Belkin item USB-A/DB9M at  There are also and

Q: Help!  My PowerTap Standard stops downloading after only 250 records!

A: (David Easter)  First, some background: during a download, the firmware in the PowerTap CPU transmits data in blocks of 256 records.  Each record contains data from a single sample (once every 1.26 seconds or longer, depending on the recording interval).  Each block is terminated with a calculated “checksum,” i.e., a consistency value designed to detect any errors that might be introduced on the serial link between the PowerTap and the PC.  As the PC receives each block of data, it calculates the check value from the received data and compares it with the value transmitted by the PowerTap.  If it matches, the download continues, if not, you get the download error.

For reasons nobody seems to understand, the PowerTap sometimes generates bad check sequences.  That’s why the download goes to ~250 every time and quits.  The data isn’t being corrupted on the serial line but the PC thinks it has, and bails out after the first block.  I’ve had this problem most recently after fighting with what turned out to be a bad receiver.  Once I got a good one, did a “clr all,” went for a ride, and then got a “250” download error.  Curiously, it got to 250 records, thought it had succeeded, but generated a file containing only the header record.  The next day, I removed and reinserted the battery in the PowerTap CPU, and all has been well since.  I’ve also seen variants of the problem with a CPU that stopped receiving power data and also started showing screwy data while in magnet mode.

One theory is that the CPU doesn’t do proper range checking on data coming from the receiver, and allows noisy data from interference, or a bad receiver, to corrupt the internal memory in such a way as to break the download protocol, however, data in the CPU still seems to be good.  The trick is to get it out.  A download program that simply ignores the check values might work in some cases, though I haven’t tried it, but in other cases, some of the data has obviously been corrupted and the CPU probably can’t generate the normal download stream anyway, regardless of what the PC would do with the check value.

I have created a web site at to rescue data trapped by this problem.  Note: this recovery process works only with a (grey) PowerTap Standard CPU; the newer PowerTap Pro CPU (yellow) is different, and this process won’t work with it.

Q: I’m having problems importing a PowerTap file that was e-mailed to me.  I’m getting a message, “error with .csv file.”

A: (Brian McLaughlin, Craig Upton, & Robert Chung)   The workaround is to have the sender compress the files.  Have the sender WinZip or StuffIt the file.  These compression softwares will “protect” the file, so that when the receiver’s MS Outlook handles it, nothing is done to the file that is enclosed in its compression shell.  When you receive it, download the attachment, put it on your desktop, un-Zip or un-StuffIt.  It should work fine.  This also happens with Entourage e-mail software, as well as Hotmail, which apparently uses some compression of its own to alter the file.

Comma separated value (.csv) files are actually plain text files where the fields are separated by commas.  For reasons unknown the PowerTap software requires not only that the values be separated by commas but also that the spacing be exact, which sort of defeats the philosophy of CSV.

For odd historical reasons, e-mail programs were often allowed to treat text mail differently than other kinds of data streams.  The foregoing CSV problem is akin to the CR/LF annoyances that used to occur when sending text mail to and from Unix systems.  In some sense, the blame is equally shared: e-mail programs for modifying files, and the PT software, for marrying the inefficiency of text files with the inflexibility of binary files.

Q: I forgot to make sure I had zeroed out the torque on my PowerTap before my last ride, and so all the values for power are off.  How do I now go back through the file and correct the error?

A: (Rick Sladkey and Rick Moll)   You need to subtract the unzeroed torque value recorded when you were coasting from each value of measured torque, and then recalculate power from torque and speed.  Find a section where you know you were coasting and actual torque was zero (i.e., when power and cadence were zero) and pick a representative torque value, then compute actual torque:

(1)  torqueactual  =  torquemeasured – torquecoasting, then re-compute power as

(2)  poweractual (Watts)  =  1746 × torqueactual (N-m) × speed (km/h)/wheel circumference (mm)

where wheel circumference is the same as in setup (about 2093 for a 700C × 23 mm tire).  Chris Mayhew has posted a spreadsheet at to perform these calculations.  Alternatively, equation 2 can be divided by

(3)  powermeasured =  1746 × torquemeasured × speed/wheel circumference

and both sides multiplied by powermeasured to give

(4)  poweractual =  powermeasured × (torqueactual / torquemeasured)

which, if used in a spreadsheet, must be protected against division by zero.

Since torqueactual / torquemeasured is not a constant, neither is poweractual / powermeasured, however, equation 4 demonstrates that power is proportional to torque.

Q: What should smoothing percentage be set at? 

A: (Andrew Coggan and Rick Sladkey)  It depends on what you’re looking for.  There are times when you might wish to apply really gross smoothing, e.g., to better detect any overall downward trend in power during a very long ride.  On the other hand, there are times when you don’t want to smooth the data at all, such as when trying to capture the details of a 500 meter race on the track.  The same logic applies to data recording frequency . . . during very long rides it is probably sufficient to record data relatively infrequently, whereas on the track, even 1 second intervals may not be frequent enough.  Properly designed hardware and software should give the user maximum flexibility with regards to these issues.

To calculate the equivalent rolling average for a given smoothing level, multiply the duration of the ride or interval in seconds by the selected smoothing percentage.  For instance, 1% smoothing for a 1 hour ride would be calculated as 3600 seconds × 0.01 = 36 seconds, however, if the ride (or an interval within a ride) was 30 minutes (0.5 hours) long, then smoothing to 1% would be equivalent only to an 18 second rolling average, whereas 2 hours 1% smoothed would be the equivalent of a 72 second average.

It should be noted that the smoothing which the PowerTap Link software (versions 1.04 and lower) gives is nothing like a true rolling average, rather, it smoothes a curve by taking fewer points and then simply connecting them with sinusoidal curves.  This is so poor both mathematically and visually that 1% and 2% are the only useful settings.

Q: My PowerTap keeps ‘cutting out’ – both power and heart rate will plummet, then come back up.  This shows up as a lot of zeroes in the downloaded file.  I have checked the location and orientation of the receiver on the chainstay, and both the hub batteries as well as the CPU batteries are fine.  What next?


A: This is commonly known as a case of the “data drops,” and you have taken the first steps to correct it, but if they don’t work, perhaps the firmware in the PowerTap CPU needs upgrading.  You can download the latest from, or contact


Jesse Bartholomew, PowerTap Product Manager

[email protected]; 1-800-783-7257, ext. 159



Q: How can I open a PowerTap database file in Microsoft Access?


A: You need to know the password, which is “link.”