Beruflich Dokumente
Kultur Dokumente
Abstract
Many scientists and coaches are interested in mechanical power produced during cycling, and use Schoberer Rad Metechnik (SRM) bicycle power cranks to obtain this data. However, it has been expensive and difficult to calibrate SRM cranks, causing much of the collected data to be unreliable. We present a static method, derived from first principles, for obtaining a calibration factor for SRM cranks. A known mass and lever arm (chainring of a known diameter) are used to apply a known torque load to the instrument in four positions, and the output frequencies are used to calculate the calibration factor in Hz/N m. The reproducibility of this method is 0.01 Hz/N m, which is acceptable for the application of the instrument, which is measurement of mechanical power application by cyclists at the crank. The method is reliable, inexpensive, and easy to set up, and will allow higher confidence in data collected using SRM power cranks. We recommend calibration of the power meter once every six months because of the measured drift of the calibration factor over time.
Keywords: cycling, mechanical power, ergometry, reproducibility
Introduction
A major advance in cycling physiology in the past decade has been made due to the invention and commercial availability of Schoberer Rad Metechnik (SRM, Jlich, Germany) power-measuring bicycle cranks. This system has allowed the direct measurement of power applied at the cranks during cycling Correspondence address:
Andrea Wooles Great Britain Cycling Team National Cycling Centre Stuart Street Manchester M11 4DQ UK Tel: +44 (0)161 274 2121 Fax: +44 (0)161 274 2095 E-mail: andrea.wooles@virgin.net
training, competition, and laboratory testing, and over 5000 units have been sold since the product came onto the market. The availability of this technology has allowed comprehensive monitoring of training in cycling, as well as a shift from training to heart rate zones to training to power zones. Professional and national cycling teams, physiologists, biomechanists, physiotherapists, and academics use the SRM system extensively to assess topics ranging from aerodynamics to race strategy to recovery from injury. The importance of this data was recognised within the Great Britain Cycling Team, and we have systematically employed SRM power cranks with the team with the aim of improving performance. This has required significant investment, and with such an investment comes the requirement that we understand and trust the data. Questions needed to be
2005 isea
137
Wooles et al.
addressed regarding comparability of data, and therefore about the accuracy of the power measurements. A calibration procedure was required that was cost-effective, quick, and could be approximated in the field as well as in a more controlled workshop setting. We hope that this procedure could also be adapted for use by single-set owners.
ship, and shifts depending on temperature, and torque on the chainring and crank arm bolts. It is a measurement of the frequency output of the system when no force is applied, and is subtracted from the measured frequency in the power calculation. This can be checked and set using the power control as many times as desired during a ride. It is important to set it at least once per ride, prior to commencing.
Calibration review
There is an apparent reliance by researchers on the reported manufacturers calibration of the SRM system prior to initial delivery, which is a moving system using an industrial lathe, a known lever arm, and a known mass. The reliability of this method is unreported. This is a large, complex, and expensive system, and is beyond the needs and capacity of most SRM users. A moving system for calibration is employed by SRM and also by the Australian Institute of Sport (Lawton et al., 1999; Woods et al., 1994), and this is often referred to as a dynamic calibration. A constant load is applied to a crank that is spun by a motor, which allows for measurement of the slope while the power meter is spinning, as it does when in use on a bicycle. However, this is not a dynamic calibration in the sense that the load is not continually changing, but is applied and left to stabilise as much as possible before a measurement is taken. This type of protocol requires large, expensive equipment, and has the disadvantage that the calibration rig itself
138
Wooles et al.
may have intrinsic errors. Because the crank is moving, the frequencies and powers reported to compare the calibration rig with the SRM are an instantaneous estimate, not an average for a given time period, or a truly stable figure. Researchers have used other dynamic systems to assess the systematic error of SRM systems. Systematic error has been found to be between 1% (Jones and Passfield, 1998) and 2.5% (Paton and Hopkins, 2001; Lawton et al., 1999), while some sets were found to have average errors of approximately 10% (Lawton et al., 1999). However, it is difficult to assess whether the measured error was attributable to the SRM power meter or to the calibration rig itself. The static calibration method addresses this issue by having minimal measurable error. A static system is affordable, easy to set up, and can be used to maintain confidence in the precision of the instrument. It can be used to assess both systematic and random errors of the SRM system. The need for this research has been previously identified (Paton and Hopkins, 2001). Power is a widely reported variable in the cycling literature, but with so many methods of measuring power without an accepted calibration procedure, comparability becomes questionable (Paton and Hopkins, 2001). While some researchers have stated that data was collected using SRM systems (Golich and Broker, 1996; Jeukendrup and Van Diemen, 1998; Balmer et al., 2000b; Balmer et al., 2000c; Bassett et al., 1999; Broker et al., 1999; Schumacher and Mueller, 2002; Craig and Norton, 2001), only a few have reported how or even if the systems were calibrated (Passfield and Doust, 2000; Paton and Hopkins, 2001; Stepto et al., 2001; Lee et al., 2002). Some researchers appear to have mistaken the setting of the zero offset frequency for calibration, and have reported that calibration was done using the manufacturers recommendations (Bentley et al., 2001; Balmer et al., 2000b; Balmer et al., 2000a; Millet et al., 2003; Balmer et al., 2000c). As the manufacturer does not provide recommendations for calibration, but does recommend frequent resetting of the zero offset frequency, this could be assumed to be a simple misunderstanding. However, the use of uncalibrated SRM systems casts uncertainty on the correlations drawn with other variables. Some researchers have used apparently uncalibrated SRM systems to assess the
validity or accuracy of other types of power-measuring devices, such as the Kingcycle (Balmer et al., 2000b), and Polar S710 (Millet et al., 2003). If routine calibration is to be expected of SRM systems as it is of all other laboratory equipment, an easy, affordable, and reliable method must be made available to all researchers using the systems.
2005 isea
139
Wooles et al.
fl3, fl4) in order to improve reproducibility. Therefore the expanded function becomes: ( f l 1 f o1) T1 + ( f l 2 f o2) T1 y = + 4
( f l 3 f o3) T1 + ( f l 4 f o4) T1 4
(4)
The calibration factor can then be checked by entering it into the data-logging system computer, and tripping the reed-switch cadence counter at a known frequency using a small, motorised plastic disk with an attached magnet. The power displayed by the system should equal the calculated power from the following relationship, where P is power (kg m2 s3 = W), is angular velocity (rad s1), and is velocity (rev min1): P = T =
(5)
Measurement protocol
Figure 2 Measurement system. The system used for determination of the calibration factor, pictured in the 0 unloaded position.
Functional relationship
Let the output variable y be the calibration factor, and let the input variables T, foffset and floaded be torque (N m), offset frequency (Hz), and loaded frequency (Hz), respectively. Torque (N m) can be defined as: T = mgr (1)
Each calibration is recorded on a separate sheet with a unique identifying number. SRM power meters are identified by the manufacturers serial number, as are SRM power control data-loggers. The serial numbers of the certificated masses used for the procedure are recorded, as is the environmental temperature. The date, time and name of the technician are also recorded.
where m is the mass (kg), g is the gravitational constant (m s2), and r is the radius (m) of the chainring used. In this experiment, torque (kg m2 s2 = N m) is constant at: T = 103.0 9.81 0.17985 = 181.73 The functional relationship can then be defined as: y = ( floaded foffset) T1 (3) (2)
e
b c
This relationship is true for one set of measurements taken in one position. However the mean of three offset frequencies and three loaded frequencies will be taken at each of four positions ( fo1, fo2, fo3, fo4 and fl1, fl2,
Figure 3 Principle of force measurement using SRM power cranks. Functional principle of the SRM power meter, where a is the force exerted on the chainring by the chain, b is the Wheatstone bridge circuit arrangement of four adhesive bonded metallic foil strain gauges, c is the crank arm, d is the direction of rotation, and e is the bicycle chain
140
Wooles et al.
Using a pre-checked SRM data-logging computer, the offset frequency is measured, including the mass of the bicycle chain and shackles. The weight stack is lowered using the trolley jack until the mass is fully suspended by the bicycle chain from the chainring. Once the mass becomes still, the loaded frequency is measured. This is repeated until three offset and three loaded frequencies are obtained. With the weights secured on the trolley jack in the raised position, the chain is removed from the chainring and the axle is loosened from the three-jaw chuck. The power meter unit is rotated clockwise through 90, and the axle is refastened in the chuck. The chain is repositioned on the chainring, and the measurements are taken for that position. This is repeated until four sets of measurements (0, 90, 180, and 270) are completed.
Frequency (Hz)
y = 20.149x + 493.94 R2 = 1
150
200
shown in Fig. 4. The y-intercept (494 Hz) of the regression equation represents foffset, and the slope of the equation (20.15 Hz/N m) represents the calibration factor. The R2 of 1 for this regression equation shows extremely good linearity for this power meter. Four other linearity checks were completed (on serial numbers 1934, 3499, 1797 and 1608), with R2 values of 0.9999, 1, 1 and 1, respectively. No difference was found in the calibration factor when weights were added incrementally to when they were removed incrementally, suggesting that there is no measurable hysteresis. Because SRM power meters are often used outside, it is important to know the effect of temperature changes on the calculated power. The strain gauges are mounted on aluminium, which expands and contracts as the temperature changes. This causes foffset to decrease as the temperature becomes lower, and to increase as the temperature becomes higher, leading to systematic error in measured power. From accumulated experience, foffset changes on average 7 Hz per degree Celsius. Calibrations are conducted in a workshop setting at a constant temperature, after the power meter has rested at room temperature for more than one hour.
Experimental results
From April 2001 to November 2002 only two positions, with the crank arm vertically upwards and downwards, were measured as part of the static method. The method evolved over this period, and as such a number of different masses (73.18 103.00 kg) and chainring radii (0.0928 0.17985 m, measured with vernier callipers) were used. The positional variance found in these 332 calibrations was minimal
2005 isea
141
Wooles et al.
(mean = 0.15 0.03 SE, 95% confidence interval 0.10 0.21), but could be improved upon. Based on the observed positional differences, we decided to implement the four-position method to improve the precision of the calibration factor. We also began performing a series of maintenance checks prior to initiation of the calibration method, as the outliers and extreme values were largely attributable to mechanical faults. Positional variance measured in 343 calibrations (mean = 0.11 0.02 SE, 95% confidence interval 0.07 0.15) using the four-position method with 181.73 N m of torque is presented in Fig. 5. The mean change in the calibration factor between measurements was 0.4%, with a standard deviation of 2.74, using the two-position method (188 measurements). The measurements are normally distributed, with a small number of outliers and extreme values, representing power meters with mechanical faults. Using the four-position method, 185 repeat measurements have been completed to date (mean percentage drift = 0.22, standard deviation = 2.01). However, if three outliers are removed from this sample, because they identified mechanical problems with the power meters that were subsequently resolved, a more representative description can be obtained (mean percentage drift = 0.15, standard deviation = 1.51). A graph is presented of percentage change for the four-point method, excluding three outliers, in Fig. 6. An example of calibration drift for one power meter, serial number 2085, is presented in
Figs. 7 and 8. It is expected that the calibration factor will vary around the unknown true value over time, as this is consistent with the behaviour of strain gauges and electronic components.
142
Wooles et al.
Calibration date Figure 7 Absolute calibration factor drift over time for SRM power meter serial number 1553
Calibration date Figure 8 Calibration factor drift in percent change over time for SRM power meter serial number 1553
serial number 2097 on 22nd May 2003. The first foffset measurements were incorrect, due to unbalanced placement of the chain on the chainring. This resulted in a calibration factor of 18.31, which was in error. Therefore the data from the first calibration were excluded from the uncertainty estimation calculation. The mean calibration factor for the remaining nine calibrations was 18.65 Hz/N m, with a standard uncertainty ux = 0.005. The expanded uncertainty for k = 2, n = 9 is ux 1.19 = 0.006. This provides a 95% confidence interval of 18.6418.65 Hz/N m for this power meter. This variation is largely due to changes in the local environment, placement of the power meter and bicycle chain, electrical and mechanical characteristics of the SRM, and imperceptible swinging of the hanging masses. Other sources of uncertainty include values assigned to the masses applied and the radius of the chainring. These are minimised by using only traceable masses, and Vernier callipers for length measurements. We checked for changes in the performance of the SRM data-logging computer by applying a known torque at a known cadence, repeatedly tripping the reed switch using a motorised spinning plastic disk with an attached magnet. The logged file was then downloaded and inspected to ensure that the power and cadence readings were as expected. The battery in the datalogging computer is changed annually to prevent malfunctions due to low power. The calibration method itself does not incorporate any approximations
until the final step, as calculated values are not rounded until the final calibration factor is obtained.
Conclusion
SRM power meters can be calibrated with a reproducibility of 0.01 Hz/N m using this static method, showing the remarkably small random error inherent to this instrument. Once this calibration factor is entered into the power control to be used for calculation of power, this amounts to less than 1 W systematic error for 1000 W as calculated by the SRM system. If this calibration method was used by researchers wishing to publish SRM data, and they clearly stated the calibration method and zero offset frequency setting procedure, much greater confidence could be placed in their findings. Valid comparisons could be made between research findings, and other power measurement equipment could be tested against the SRM system. Calibration of power meters once every six months is recommended to minimise systematic errors due to calibration drift. It is advisable to perform a comprehensive maintenance check on all components of the system prior to commencing the calibration procedure, although the method itself can be useful in diagnosing instrument malfunctions. This method is easily reproduced, relatively inexpensive to set up, and could be used by anyone wishing to improve the precision of SRM power cranks.
2005 isea
143
Wooles et al.
We could improve precision further by using more than four positions, but time limitations make this impractical. The precision given by the four-position method is acceptable for the application of the instrument. The sensitivity of the instrument to zero offset changes makes it necessary to ensure that the bicycle chain is replaced evenly for each set of measurements, as is shown by the error incurred during the uncertainty estimation trial. It also reinforces the potential for and magnitude of user error if the zero offset is not checked and set prior to instrument use in the lab or the field. In the field, it must be set before each ride, and again after the power meter has reached environmental temperature. In research, it should be set before each test, and checked afterwards, and both readings should be reported along with the data. Further development of this static method is in progress to increase the applied torque load to 125% of the instruments maximum functional range. From data collected in our lab, we estimate the maximum torque load to be approximately 350 N m. To achieve this, we will need to increase the mass and the radius of the chainring. However, we are currently limited by the ability of the three-jaw chuck to hold the axle without slippage. We have identified a stronger method of mounting the axle, and will continue with development.
Acknowledgements
We should like to thank the Great Britain Cycling Team for supporting this project, and Dr. Cory Sutela for his valuable suggestions during preparation of the manuscript.
References
Balmer, J., Davison, R.C.R. & Bird, S.R. (2000a) Peak power predicts performance power during an outdoor 16.1 km cycling time trial. Medicine & Science in Sports & Exercise, 32, 14851490. Balmer, J., Davison, R.C.R. & Bird, S.R. (2000b) Reliability of an air-braked ergometer to record peak power during a maximal cycling test. Medicine & Science in Sports & Exercise, 32, 17901793. Balmer, J., Davison, R.C.R., Coleman, D.A. & Bird, S.R. (2000c) The validity of power output recorded during exercise performance tests using a Kingcycle air-braked cycle ergometer when compared with an SRM powermeter. International Journal of Sports Medicine, 21, 195199.
Bassett, D.R., Kyle, C.R.J., Passfield, L., Broker, J.P. & Burke, E.R. (1999) Comparing cycling world hour records, 19671996: modelling with empirical data. Medicine & Science in Sports & Exercise, 31, 16651676. Bentley, D.J., McNaughton, L.R., Thompson, D., Vleck, V.E. & Batterham, A.M. (2001) Peak power output, the lactate threshold, and time trial performance in cyclists. Medicine & Science in Sports & Exercise, 33, 20772081. Broker, J.P., Kyle, C.R. & Burke, E.R. (1999) Racing cyclist power requirements in the 4000 m individual and team pursuits. Medicine & Science in Sports & Exercise, 31, 16771685. Craig, N.P. & Norton, K.I. (2001) Characteristics of track cycling. Sports Medicine, 31, 457468. Golich, D. & Broker, J. (1996) SRM bicycle instrumentation and the power output of elite male cyclists during the 1994 Tour of Dupont. Performance Conditioning for Cycling, 2, 2. Jeukendrup, A. & Van Diemen, A. (1998) Heart-rate monitoring during training and competition in cyclists. Journal of Sports Sciences, 16, S91S99. Jones, S.M. & Passfield, L. (1998) The dynamic calibration of bicycle power measuring cranks, Blackwell Science, Oxford. Lawton, E.W., Martin, D.T. & Lee, H. (1999) In Fifth IOC World Congress, Vol. 1999 International Olympic Committee, Sydney, pp. 199. Lee, H., Martin, D.T., Anson, J.M., Grundy, D. & Hahn, A.G. (2002) Physiological characteristics of successful mountain bikers and professional road cyclists. Journal of Sports Sciences, 20, 10011008. Millet, G.P., Tronche, C., Fuster, N., Bentley, D.J. & Candau, R. (2003) Validity and reliability of the Polar S710 mobile cycling powermeter. International Journal of Sports Medicine, 24, 156161. Passfield, L. & Doust, J.H. (2000) Changes in cycling efficiency and performance after endurance exercise. Medicine & Science in Sports & Exercise, 32, 19351941. Paton, C.D. & Hopkins, W.G. (2001) Tests of Cycling Performance. Sports Medicine, 31, 489496. Schoberer, U. (1998) Operating Instructions: SRM Training System, SRM, Julich. Schumacher, Y.O. & Mueller, P. (2002) The 4000 m team pursuit cycling world record: theoretical and practical aspects. Medicine & Science in Sports & Exercise, 34, 10291036. Stepto, N.K., Martin, D.T., Fallon, K.E. & Hawley, J.A. (2001) Metabolic demands of intense aerobic interval training in competitive cyclists. Medicine & Science in Sports & Exercise, 33, 303310. Woods, G.F., Day, L., Withers, R.T., Ilsley, A.H. & Maxwell, B.F. (1994) The dynamic calibration of cycle ergometers. International Journal of Sports Medicine, 15, 168171.
144