Académique Documents
Professionnel Documents
Culture Documents
by
at
April 2017
Approved by___________________________________________________________________
Dr. Timothy Takahashi
Professor
Thesis Director
Approved by___________________________________________________________________
Dr. James Middleton
Professor
Second Committee Member
Accepted by___________________________________________________________________
Dean
Barrett, The Honors College
1
Planing and The Effect of Bicycle Frame Stiffness on Rider Performance
by
Graham Philip Sparks
Abstract
It is a common assumption in the bicycle industry that stiffer frames generally perform
better than flexible frames, because they transfer power more efficiently and absorb less energy
from the rider’s pedal stroke in the form of spring potential energy. However, in the last few years,
Jan Heine of Bicycle Quarterly has developed an alternative theory, which he calls “planing”,
whereby a flexible frame can improve rider performance by not resisting the leg muscles as much,
preventing premature muscle fatigue and allowing the rider to actually produce more consistent
power, an effect which overwhelms any difference in power transfer between the different stiffness
levels of frames. I performed several tests in which I measured the power input to the bicycle
through the crankset and power output through a power-measuring trainer in the place of the rear
hub. Heart rate data was collected along with most of these tests. Four bicycles were used with
three distinct levels of stiffness. After performing several ANOVA tests to determine the effect of
stiffness on the parameters of average power output during a sprint, maximum power output during
a sprint, maximum heart rate during a sprint, difference between power-in and power-out during
both sprints and longer efforts, and power quotient during a sprint, I found no effects of frame
stiffness on any of these factors except power quotient. The finding for power quotient suggests a
positive relationship between quotient and stiffness, which directly refutes the Planing Theory for
the test riders and levels of stiffness represented in this test. Also, no statistically significant effect
of stiffness on the difference between power-in and power-out was found, refuting the Power
Transfer Theory for the riders and levels of stiffness represented in this test.
2
Acknowledgements
I am extremely grateful to my thesis committee members, Dr. Timothy Takahashi and Dr.
James Middleton, for their help in facilitating this research. They both provided valuable feedback
on ideas ranging from theory to experimental design, as well as encouragement. I would also like
to thank Dave Skeeters, the manager at Giant Scottsdale, for allowing me to buy bikes at cost,
which allowed me to affordably equip my test frames with components, and borrow or use
equipment for my testing, such as the Wahoo Kickr trainer which allowed me to conduct my tests
on power transfer. I’m grateful for the help of Marty Ryerson as well, for allowing me to use a
demo Pioneer dual crankset power meter and head unit; this equipment was crucial for my data
collection and offered the highest quality data collection possible with a crankset-based power
meter. Last but certainly not least, I’d like to thank my test riders, Ben M., Ben Y., Tobie M., Alex
S., and Kevin W., for their very hard work riding the frames for data collection. I could not have
completed this research without the efforts of these individuals.
3
Contents
1 Introduction 7
1.1 Theories of frame performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7
1.1.1 Planing Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.1.2 Power Transfer Theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2 Experimental Design 9
2.1 Test Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10
2.2 Testing Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4 Conclusions 20
5 Further Research 21
6 Bibliography 22
7 Appendix A 23
8 Appendix B 26
4
List of Figures
2-1 Pioneer Dual-Arm meter with Wahoo KICKR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2-5 Left: Rider 2 during dynamic testing. Right: Bikes awaiting runs during dynamic
Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3-1 ANOVA Table for Effect of Stiffness on delta Power for Three-Minute Runs . . . . . . . .16
3-2 ANOVA Table for Effect of Stiffness on delta Power for Static Sprints . . . . . . . . . . . . .16
3-3 ANOVA Table for Effect of Stiffness on Average Power for Static Sprints . . . . . . . . . .17
3-4 ANOVA Table for Effect of Stiffness on Average Power for Dynamic Sprints . . . . . . . 17
3-5 ANOVA Table for Effect of Stiffness on Power Quotient for Static Sprints . . . . . . . . . .17
3-6 Interaction Plot for Effect of Stiffness on Power Quotient for Static Sprints . . . . . . . . . .18
3-7 ANOVA Table for Effect of Stiffness on Power Quotient for Dynamic Sprints . . . . . . .19
3-8 ANOVA Table for Effect of Stiffness on Maximum Heart Rate for Static Sprints . . . . .19
5
List of Tables
I Test Frame Tubing Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
6
Chapter 1
Introduction
Each school of thought approaches this issue of frame performance from opposite sides.
As Jan Heine points out on his blog, Off the Beaten Path, the “critics [of the Planing Theory]
assume a constant power output” (Heine, 2013). Assuming that frame stiffness does not affect the
rider’s power output, the primary way that the frame affects performance is by absorbing energy
from the rider’s pedal stroke as spring potential energy. This phenomenon is considered a detractor
from performance, because for a constant force, a weaker spring can absorb more energy from the
pedal stroke. However, Heine and the staff at Bicycle Quarterly believe the opposite: “the frame
flexes as the rider puts out maximum power during the downstroke. This means that the bike can
accept peak power beyond what can be used to acceleration. The stored power then is transmitted
to the drivetrain when the frame springs back during the upstroke” (Heine, 2014).
There are several components to the Planing Theory as expounded in Bicycle Quarterly.
First, the flex in a less-stiff bicycle frame enables the rider to reach a higher aerobic threshold,
while a stiffer bicycle exhausts the rider’s muscles. The effect of the higher aerobic exertion is, if
not higher peak power, at least higher average power during an effort. Heine supports this
conclusion with data, published in the Summer 2009 issue of Bicycle Quarterly, which showed
that on short sprints with flexible frames, the average power was a higher percentage of the peak
power during that run. The “superlight” frame in that test was able to help the rider maintain an
7
average power output that was 79-86% of the peak power, while the “standard” frame only
experienced 70-76% of the peak power as an average (Heine, Vande Kamp, & Rossman, 2009).
Second, as referenced in the previous section, the flexible frame absorbs more energy during the
downstroke, and can productively return this energy during the dead part of the rider’s pedal stroke.
Third, since the flexible “superlight” frame allows a higher aerobic threshold, the rider may
experience a higher heart rate at the end of an effort, while a stiffer frame would promote muscle
fatigue and limit heart rate (Heine, Vande Kamp, & Rossman, 2009).
Since I do not have expertise in sports science or physiology, I cannot offer any insight or
opinion on the issue of muscle fatigue and aerobic exertion. However, my testing methods do
attempt to account for the claimed effect on heart rate, as described in Ch. 2. It is very difficult to
know if the spring potential energy of the frame during the pedal stroke is returned to the drivetrain
in any useful way, or if it is simply expended by returning the frame to its steady state condition
prior to the downstroke of the next crank rotation. We may be able to conclude that it is usefully
returned to the drivetrain if the power measured at the rear hub is not significantly less than the
power measured at the crankset for the flexible frames as it is for the stiffer frames. If the flexible
frames do allow the rider to maintain a greater level of exertion, as measured by the average
percentage of the maximum power output during an effort, then this phenomenon should be
observable as a statistically significant effect of frame stiffness on the Power Quotient, or the
average power of an effort or sprint divided by the maximum power in that effort.
8
The background of this theory relies on the model of a bike frame as a simple spring, for
which Hooke’s Law dictates that the displacement (x) of the spring is proportional to the force (F)
applied to it, holding the spring rate (k) constant:
F=k*x
E=.5*k*x^2
So, for a frame that is modeled as a simple spring in the direction parallel to the bottom bracket
spindle, with a spring rate of 100 kN/m, a force of 2000 N applied during the power stroke would
deflect the frame sideways by 0.02 m, and the potential energy of the spring at that point of
maximum deflection would be 0.02 kJ. Applying the same force to a frame with an effective spring
rate of 120 kN/m would result in a deflection of 0.0167 m, and a spring potential energy of 0.0167
kJ. This simple calculus leads many to believe that a stiffer frame, assuming the same pedaling
force, absorbs less energy from the rider’s pedal stroke, and this energy is usefully employed in
forward motion. However, this model neglects the biomechanical effects of stiffness. According
to advocates of the Planing Theory, higher stiffness interacts with the rider’s leg muscles,
promoting premature muscle fatigue and hampering the production of power. If the Power Transfer
Theory is correct, and we collect concurrent data from a rear hub-based power meter and a crankset
power meter, then the adjusted difference between the two will be significantly larger for the more
flexible frame than it will be for a stiffer frame.
Chapter 2
Experimental Design
With this experiment, I intend to test the previously outlined assertions of both opposing
theories of frame performance. To test the Power Transfer Theory, it is necessary to collect data
representing the power added to the frame-drivetrain system and the power leaving the system
through the rear wheel. To test the Planing Theory, it is necessary to collect data representing the
power produced by the rider, as well as the corresponding heart rate measurements. The original
tests performed in Bicycle Quarterly supported the Planing Theory with a double-blind test in
which frames of different down tube and top tube thicknesses were ridden on successive short hill
climbs by two riders, side-by-side. Power was measured by PowerTap rear hubs and the bikes
were exchanged after each trial. Heart rate data was apparently not collected. Gear ratio was also
an issue not rigorously addressed in the description of the test. Also, the tests were halted after 5
runs, due to apparent degradation of performance due to fatigue. These tests were repeated without
power data, using speed as a proxy for performance, for a bike with oversize tubing against the
9
stiffer bike from the previous test. Subjective impressions of each bike were collected before the
test was unblinded (Heine, Vande Kamp, & Rossman, 2009). With only five runs, it is difficult to
draw statistically significant conclusions about the performance of the frames. However, the riders
subjectively preferred the flexible frame in each test, and the data representing the Power Quotient
during the run seemed to favor the flexible frame in the first test.
Since it is difficult for a rider to exactly control the level of exertion during each run in a
quantifiable and rigorous way, Bicycle Quarterly asked the riders to sprint as hard as they could,
using the side-by-side test format to ensure that the effort remained consistent between runs, since
each rider would try to outrace the other by maximizing power (Heine, Vande Kamp, & Rossman,
2009). I also tested with maximum-output sprints, stressing to each rider the importance of
sprinting with maximum effort in each run. Bicycle Quarterly also did not consider runs during
which the average power was less than 85% of that rider’s maximum average power. I did not
adopt this policy, instead using all data sets and assuming that differences in performance between
frames could account for a difference of 15% in power output. Also, I had riders perform more
than five runs, believing that more data would strengthen the conclusions of the experiment.
Scheduling constraints and the lack of availability of a pair of riders of similar size and ability
prevented me from using the side-by-side testing method employed by Bicycle Quarterly. I used
four frames, with three different levels of stiffness. Cost constraints prevented the use of the
“superlight” tubing specification (0.7-0.4-0.7). Instead, I used frames with top and down tubes
with 0.8 mm ends and a 0.5 mm middle section (0.8-0.5-0.8) in the lightest and most flexible
category. The middle level of stiffness had 0.9-0.6-0.9 top and down tubes, and the stiffest frame
had 0.8-0.5-0.8, with oversize tubes (top and down tube were 0.125 inches larger in diameter). The
entire tubing specification is detailed in Table I. There are two frames in the lightest, most flexible
category. Each other category had one frame. The frames had the top and down tubes covered in
foam to conceal the oversize frame. The test was then randomized by a worker at Giant Scottsdale,
who labeled the frames 1-4.
Unlike Bicycle Quarterly, I collected data from five test riders, not two. Using more riders
gives a better picture of how stiffness interacts with different riding styles. These riders varied in
height, but all were fit riders and experienced cyclists. To account for gear ratio, the riders were
instructed before testing that they could only change gear ratio in between sets of four runs, in
which each frame was ridden. This would ensure that for each gear ratio used by each rider, every
frame would have the same number of runs. The testing was divided into two categories: Dynamic
and Static. Static testing occurred on a trainer that measures power, enabling the analysis of the
difference between power in and power out among the different levels of stiffness. Dynamic testing
occurred on a short, steep hill on Lake View Drive in Tempe that rises from the Tempe Town Lake
to Curry Road. The climb gains 18 meters of elevation and required between 25 and 35 seconds to
ride, depending on the rider. Static tests occurred at Giant Scottsdale. Three of the riders performed
a constant three-minute effort at a constant 180-200 W on each frame, providing a relatively noise-
free, but small amplitude sample to perform comparison tests. Dynamic data was only collected
from two riders, due to concerns about heat exhaustion and scheduling issues.
10
2.1 Test Equipment
To measure power at the crankset, a Pioneer FC9000 Dual-Arm power meter with
accompanying SGX-CA500 computer was used. A Garmin chest strap heart rate monitor was also
paired with this computer using the ANT+ wireless frequency protocol. A Wahoo ELEMNT
computer was paired with a Wahoo KICKR trainer for the static test data collection. Both devices
recorded data at an interval of 1 second.
Figure 2-2: Pioneer SGX-CA500 head unit on left, Wahoo ELEMNT on right
11
The four test bikes were made from 4130 steel with identical geometries, fork blades, and
rear triangles. Frame 1 was made from oversized (31.7 mm down tube and 28.6 mm top tube) 0.8-
0.5-0.8 tubing in the top and down tubes. This frame formed the stiffest level in the analysis. Frame
2 had standard diameter (25.4 mm top tube and 28.6 mm down tube) 0.9-0.6-0.9 tubing. Frames 3
and 4 had standard diameter 0.8-0.5-0.8 tubing, and composed the most flexible level of frames.
The numbers of these frames refer to the numbers they were assigned during randomization.
Frame 1 Top Frame 1 Down Frame 2 Frame 2 Down Frame 3-4 Top Frame 3-4
Tube Tube Top Tube Tube Tube Down Tube
Diame 28.6 31.7 25.4 28.6 25.4 28.6
ter
Wall 0.8-0.5-0.8 0.8-0.5-0.8 0.9-0.6-0.9 0.9-0.6-0.9 0.8-0.5-0.8 0.8-0.5-0.8
Thick
ness
Part NOV_CRDT_ NOV_CXDT_ NOV_CRT NOV_CRDT_ NOV_CRTT_8 NOV_CRDT_
Numb 858_650 858_650 T_969 969_650 58_600 858_650
er
Table I: All Dimensions are in mm
12
The test bikes were equipped with components borrowed from Giant Contend 3 road bikes,
omitting the front derailleur and rear brake for the sake of expediency. The longer chainstays on
the test bikes limited the gear range to the five smallest cogs on the rear when the large (52 tooth)
chainring was in use. Rider 2, being shorter than other testers, used an 80 mm stem; the other
testers used the stock 100 mm stem.
A testing session begins by fitting the bike to the rider. The riders are mostly of similar
build, so this usually involved moving the seat to the correct height, as measured from the bottom
of the bottom bracket shell to the top of the seat with a tape measure. I install the rider’s pedals of
choice onto the crankset. The computers are initialized, and I ensure that they are receiving data
from their respective power meters and heart rate monitor. The rider performs a test run, allowing
them to find a suitable gear ratio to begin testing. The computers are set to begin logging, and the
rider rides the first frame in a randomized list of frame numbers. The rider spins up to about 60
RPM, then tells me when they are beginning the sprint. I time them, and tell them when 20 seconds
has elapsed. The rider immediately stops pedaling, and I use this zero power value to index the
end of the sprint in the data analysis and processing. I switch the Pioneer crankset to the next frame
in the sequence, then repeat this procedure for data collection. The switch takes two to three
minutes, allowing the rider a short period to rest. We stop data collection when the rider reports
exhaustion or the three-hour testing period comes to an end. Differences between dynamic and
static testing procedures are several. In static testing, I switch the crankset between bikes and place
the bike with the crankset on the trainer, clamping it in place of the rear wheel, whereas in dynamic
testing, I merely change the crankset. Also, rather than timing the rider in dynamic runs, they begin
the sprint at a mark on the road and push as hard as possible in their selected gear, without shifting,
until they reach the mark at the top of the hill, where they stop.
13
Figure 2-4: Rider 4, mid-sprint during static testing. Note foam tube covers.
Figure 2-5: Top: Rider 2 during dynamic testing. Bottom: Bikes awaiting runs during dynamic
testing.
14
Chapter 3
15
3.1 Effect of Stiffness on Power Transfer
Figure 3-1: ANOVA Table for Effect of Stiffness on delta Power for Three-Minute Runs
Riders 2, 3, and 4 performed three-minute long runs on each frame, attempting to ride at a
constant 200 W effort. The riders were permitted to see the computer that recorded the power at
the trainer, in order to control their power input. The Wahoo KICKR recorded data for these runs
along with the Pioneer power meter. The KICKR did record values higher than the Pioneer, when
it should have recorded lower values due to power loss from the chain, frame flexure, and other
factors. Therefore, the difference was adjusted by 35 W. According to the ANOVA test for these
data sets, the p value of 0.5099 indicates a nonsignificant effect of stiffness on the difference
between the average power recorded at the trainer and the crankset. The Power Transfer Theory
predicts that the stiffer frame will record a smaller average difference than the flexible frames.
This effect was not observed. However, given the relatively small magnitude of the 200 W effort,
there may be a larger difference recorded for a sprint. The next test addresses this concern.
Figure 3-2: ANOVA Table for Effect of Stiffness on delta Power for Static Sprints
Using the static sprints, which logged much high average power values than 200 W, there
is still no significant effect on the difference between the crankset and the trainer, with a p value
of 0.6403. The adjustment for the difference in power was higher at 75 W.
16
3.2 Effect of Stiffness on Average Power
Figure 3-3: ANOVA Table for Effect of Stiffness on Average Power for Static Sprints
The effect of the stiffness on the average power output for the static runs was
nonsignificant. However, one can see the p value for the Rider factor is significant at near zero,
which indicates the effectiveness of the blocking of Rider. There is an equally insignificant
interaction effect detected by the ANOVA between rider and stiffness.
Figure 3-4: ANOVA Table for Effect of Stiffness on Average Power for Dynamic Sprints
The effect is more significant for dynamic sprints with a p value of 0.1591, but still not
near the critical 0.05 value. Blocking on rider for dynamic sprints is still very effective.
Figure 3-5: ANOVA Table for Effect of Stiffness on Power Quotient for Static Sprints
17
Here we see the first discernable, significant effect. Stiffness has a clear relationship here
with the static power quotient. The static sprint data set is the largest, with four riders and 48
sprints. The ANOVA table tells us that there is once again not a significant interaction effect. An
interaction plot is used to determine if the effect works as predicted by the Planing Theory:
Figure 3-6: Interaction Plot for Effect of Stiffness on Power Quotient for Static Sprints
The Planing Theory predicts that the flexible frames (S3) will have a higher power quotient
than the stiffest frame (S1), due to the more flexible frames preventing muscle fatigue throughout
the run. Here, the opposite effect is observed. The S3 level with the flexible standard diameter 0.8-
0.5-0.8 frames averages a power quotient of slightly more than 0.9, while the stiffer frame averages
almost 0.93, a significant difference. A regression analysis was not performed, because, at this
time, I have no rigid, reliable fixture with which I could test the frames’ exact drivetrain flexibility
and assign numerical values to the S1, S2, and S3 stiffness levels.
18
Figure 3-7: ANOVA Table for Effect of Stiffness on Power Quotient for Dynamic Sprints
The effect of stiffness on power quotient is not significant for the dynamic runs, although
the blocking of the rider factor is still effective, with a p value of 0.0243.
Figure 3-8: ANOVA Table for Effect of Stiffness on Maximum Heart Rate for Static Sprints
There is no significant effect of stiffness on heart rate either. The Planing Theory predicts
that because the flexible frames allow the rider to reach a higher cardiovascular level of exertion,
the heart rate at the end of the run (maximum) will be higher. This relationship was not observed;
it seems the maximum heart rate is only strongly correlated with the rider, not the frame. This is
partially expected. The maximum heart rate is a factor of each rider’s individual fitness. Heart rate
was apparently not a factor tested by Bicycle Quarterly, only alluded to anecdotally. This missing
piece of the Planing Theory does not support the assertions of the Bicycle Quarterly staff.
Below is a sample chart showing data from Rider 5, the strongest rider, including average
power, maximum power, and power quotient:
19
Frame Number Average Power Maximum Power Power Quotient
1 683.9 750 0.9118
2 717.4 753 0.9527
3 759.7 816 0.9310
4 685.8 713 0.9618
2 685.9 721 0.9513
4 706.3 738 0.9570
3 690.1 734 0.9401
1 717.2 758 0.9461
4 726.5 795 0.9138
2 660.8 681 0.9703
1 662.2 683 0.9695
3 698.1 737 0.9472
1 666.7 707 0.9429
2 641.5 677 0.9475
4 656 700 0.9371
3 619.6 659 0.9402
Table III: Power Data for Rider 5 Static Runs
One reason for the lower power quotient on Frames 3 and 4 could be that the greater
flexibility of the frame somehow enabled a higher peak power output. I offer the data from Rider
5 as a counterpoint to the results of the ANOVA on static power quotient. Rider 5, at 6’0” tall and
180 lbs, most closely fits the physical profile of the Bicycle Quarterly testers. Looking at the
maximum power column, the two highest values come from Frames 3 and 4, in the least stiff S3
category. In the average power column, three of the top five values, including the highest value,
come from Frames 3 and 4, while only one comes from Frame 1 in the S1 category. Looking at
just the first four runs, the highest maximum and average power values come from Frame 3, which
is in the flexible S3 category, while the lowest average power and power quotient come from
Frame 1, which is the stiffest. So, although the ANOVA was not able to detect a statistically
significant effect, it does seem at first glance that the best run in most categories came from the
flexible frames, which matches the experience of the Bicycle Quarterly testers. Perhaps if they had
performed 16 runs instead of 5, their results might have differed significantly.
20
Chapter 4
Conclusions
These results directly contradict many of the main assertions for both the Planing Theory
and the Power Transfer Theory. There was no significant effect of frame stiffness on difference in
power between the crankset and rear hub, for either high-power sprints or for low-power three-
minute runs, leading us to conclude that, in the range of frame stiffness and sensitivities of power
measurement represented in this experiment, there is no measurable effect on the transfer of power
from the crankset to the rear hub. Additionally, there was no significant effect of stiffness on
average power and heart rate for either static or dynamic tests. There was a significant effect of
stiffness on the power quotient, but it was the opposite effect predicted by the Planing Theory; the
stiffest frame consistently allowed the rider to produce an average power that was a higher
percentage of the peak power during that effort than that of the most flexible pair of frames. What
do these results mean? First, one should recognize that the range of stiffness represented in this
experiment is very small. A frame made from a standard diameter 0.8-0.5-0.8 tubeset is not much
more flexible than one made from a standard 0.9-0.6-0.9 tubeset. If this test was repeated with
tubes of much greater differences in diameter or thickness, the results of the test might better
support the Power Transfer Theory or the Planing Theory. Additionally, using more sensitive hub
based power measurement equipment than the Wahoo trainer might yield a different result. The
Pioneer Dual-Arm meter is state-of-the-art. There is no publicly available crankset-based power
meter that might provide data with better resolution or accuracy. I suspect that, given the very
small differences in bottom bracket stiffness of the frames in this test, one would need extremely
accurate power measurement equipment at both ends to detect what is likely a vanishingly small
amount of power “absorbed” by the more flexible frames. Additionally, the amount of data
collected per rider in my experiment, although much larger than that collected in the Bicycle
Quarterly test, is still very small and increases the effects of random confounding factors.
Psychological factors are very important insofar as they affect the riders’ effort. Bicycle Quarterly
accounted for the necessity for effort by having riders of similar ability ride side-by-side. It is
impossible to know if psychological effects confounded the performance of the riders run-to-run
in my test. Future tests should devise a better, more repeatable method to control for psychological
factors.
The results cast doubt on the Planing Theory’s assertion that flexible frames help riders
avoid muscle fatigue and “bogging down” during hard efforts. The riders in my experiment
performed better on the stiffer frame in this regard, logging significantly better power quotients in
the static tests. The purpose of this experiment was partially to replicate the results of the Bicycle
Quarterly test to verify the theory, and some was to expand both theories by performing tests
which address heretofore untested factors, such as the heart rate assertion of the Planing Theory.
The results supported neither theory. My claim is not that stiffness does not matter in pedaling
performance, but that for my test group and levels of frame stiffness, it did not appear to make a
21
discernable difference. Bicycle Quarterly also suggested that planing may be an effect specific to
the rider; in their subjective double-blind tests, their third rider was not able to discern a difference
in performance between the flexible and stiff frames. Perhaps my test group was like this third
rider, for whom the flexibility of the frame has little effect. Perhaps they even experienced the
opposite effect, as observed in the results for the static power quotient, through some factor of
riding style and testing methodology. The results of this test offer limited insight into the
performance of frames with flexibilities significantly greater or less than that of 0.8-0.5-0.8
standard and oversize. However, given the results detailed in this report, it does not appear that
Planing Theory applies to the riders in this test. There was also no evidence of the power loss
between crank and rear hub that was predicted by the Power Transfer Theory.
Chapter 5
Further Research
I believe there is much more to be done in this area of bicycle performance. Clearly, it is
strongly felt among industry members that frame stiffness is a crucial aspect of performance, but
the results of this experiment clearly demonstrate how difficult it is to rigorously determine what
exactly the effect of various levels of stiffness is. In my opinion, the way forward is to test pairs
of frame stiffness levels of significant difference, for example 0.7-0.4-0.7 standard diameter vs.
0.8-0.5-0.8 oversize. Also, it is obvious that the chainstay stiffness plays a significant role in the
drivetrain stiffness of a frame, but the effect of chainstay stiffness is likewise not well understood.
Having identical frames, but changing the chainstay stiffness and performing the same tests would
be informative of any existing effects. Also, it has often been suggested by Bicycle Quarterly and
others that the effects of stiffness on the rider are more dramatic at the end of long rides or efforts
(Heine, Vande Kamp, & Rossman, 2009). Devising a new testing methodology to collect data with
maximally consistent conditions and longer data sets would give some insights into the validity of
these claims. Future test could look at the effect of gear ratio as well, because some have suggested
a bimodal model in which stiffer frames are superior for “mashers” who pedal high gears at a low
cadence, while flexible frames are better for “spinners” who pedal low gears at a higher cadence
(Heine, 2011). Also, future tests should better account for psychological factors, as described in
Chapter 4.
There may be companies in the bicycle industry that have performed similar research on
the effects of frame stiffness on performance, but these results have apparently not been published.
Generally, debate centers around how stiffness is measured, but the advantage of a stiff frame is
taken for granted. I urge bicycle companies to perform some rigorous research in this field to
support the claims made by their marketing literature. If frame stiffness really matters very little,
then perhaps cyclists would be better served by prioritizing other qualities in bicycle frame designs.
22
Bibliography
[1] V. (2013, July 8). About frame stiffness – what is stiff? Retrieved April 10, 2017, from
http://blog.velocite-bikes.com/2013/07/about-frame-stiffness-what-is-
stiff/#sthash.4AZdYfZS.b9dBcFoh.dpbs
[2] Buchanan, I. (2016, February 05). Stiffness & Compliance. Retrieved April 10, 2017,
from http://fitwerx.com/stiffness-compliance/
[3] Heine, J., Vande Kamp, M., Wetmore, A., & Rossman, H. (2008). Frame Stiffness
Matters: Double-Blind Test Confirms the Planing is Real. Bicycle Quarterly, 6(4), 42-48.
[4] Heine, J., Vande Kamp, M., & Rossman, H. (2009). Frame Stiffness and Pedal
Stroke. Bicycle Quarterly, 7(4), 10-13.
[5] Heine, J., Vande Kamp, M., & Rossman, H. (2009). Planing and Oversize
Tubing. Bicycle Quarterly, 7(4), 14-15.
[6] Heine, J. (2011, October 03). Science and Bicycles: Frame Stiffness. Retrieved April 10,
2017, from https://janheine.wordpress.com/2011/10/03/science-and-bicycles-frame-
stiffness/
[7] Heine, J. (2011, February 27). A Journey of Discovery, Part 5: Frame Stiffness. Retrieved
April 10, 2017, from https://janheine.wordpress.com/2011/02/27/a-journey-of-discovery-
part-5-frame-stiffness/
[8] Heine, J. (2013, February 19). Laws of Physics. Retrieved April 10, 2017, from
https://janheine.wordpress.com/2013/02/19/laws-of-physics/
[9] Heine, J. (2014, December 31). The Biomechanics of Planing. Retrieved April 10, 2017,
from https://janheine.wordpress.com/2014/12/31/the-biomechanics-of-planing/
[10] Heine, J. (2014, November 23). What is Planing? Retrieved April 10, 2017, from
https://janheine.wordpress.com/2014/11/23/what-is-planing/
[11] Manning, J. (2016, June 10). The Cyclist guide to frame stiffness. Retrieved April 10,
2017, from http://www.cyclist.co.uk/in-depth/1310/the-cyclist-guide-to-frame-stiffness
23
Appendix A
MATLAB script used to process data and perform ANOVA:
24
StatPowerPioneer2=[mean(R1SP173(:,1)) mean(R1SP182(:,1)) mean(R1SP191(:,1))
mean(R1SP204(:,1)) mean(R1SP211(:,1)) mean(R1SP223(:,1)) mean(R1SP232(:,1))
mean(R1SP244(:,1)) mean(R1SP253(:,1)) mean(R1SP261(:,1)) mean(R1SP272(:,1))
mean(R1SP284(:,1)) mean(R3SP052(:,1)) mean(R3SP064(:,1)) mean(R3SP073(:,1))
mean(R3SP081(:,1)) mean(R3SP094(:,1)) mean(R3SP102(:,1)) mean(R3SP111(:,1))
mean(R3SP123(:,1)) mean(R4SP054(:,1)) mean(R4SP063(:,1)) mean(R4SP071(:,1))
mean(R4SP084(:,1)) mean(R4SP092(:,1)) mean(R4SP101(:,1)) mean(R4SP113(:,1))
mean(R4SP121(:,1)) mean(R4SP132(:,1)) mean(R4SP144(:,1)) mean(R4SP153(:,1))
mean(R4SP162(:,1)) mean(R5SP011(:,1)) mean(R5SP022(:,1)) mean(R5SP033(:,1))
mean(R5SP044(:,1)) mean(R5SP052(:,1)) mean(R5SP064(:,1)) mean(R5SP073(:,1))
mean(R5SP081(:,1)) mean(R5SP094(:,1)) mean(R5SP102(:,1)) mean(R5SP111(:,1))
mean(R5SP123(:,1)) mean(R5SP131(:,1)) mean(R5SP142(:,1)) mean(R5SP154(:,1))
mean(R5SP163(:,1))];
StatStiffness2={'S3';'S2';'S1';'S3';'S1';'S3';'S2';'S3';'S3';'S1';'S2';'S3';'
S2';'S3';'S3';'S1';'S3';'S2';'S1';'S3';'S3';'S3';'S1';'S3';'S2';'S1';'S3';'S1
';'S2';'S3';'S3';'S2';'S1';'S2';'S3';'S3';'S2';'S3';'S3';'S1';'S3';'S2';'S1';
'S3';'S1';'S2';'S3';'S3'};
StatRider2={'R1';'R1';'R1';'R1';'R1';'R1';'R1';'R1';'R1';'R1';'R1';'R1';'R3';
'R3';'R3';'R3';'R3';'R3';'R3';'R3';'R4';'R4';'R4';'R4';'R4';'R4';'R4';'R4';'R
4';'R4';'R4';'R4';'R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5'
;'R5';'R5';'R5';'R5'};
StatMaxPowerPioneer2=[max(R1SP173(:,1)) max(R1SP182(:,1)) max(R1SP191(:,1))
max(R1SP204(:,1)) max(R1SP211(:,1)) max(R1SP223(:,1)) max(R1SP232(:,1))
max(R1SP244(:,1)) max(R1SP253(:,1)) max(R1SP261(:,1)) max(R1SP272(:,1))
max(R1SP284(:,1)) max(R3SP052(:,1)) max(R3SP064(:,1)) max(R3SP073(:,1))
max(R3SP081(:,1)) max(R3SP094(:,1)) max(R3SP102(:,1)) max(R3SP111(:,1))
max(R3SP123(:,1)) max(R4SP054(:,1)) max(R4SP063(:,1)) max(R4SP071(:,1))
max(R4SP084(:,1)) max(R4SP092(:,1)) max(R4SP101(:,1)) max(R4SP113(:,1))
max(R4SP121(:,1)) max(R4SP132(:,1)) max(R4SP144(:,1)) max(R4SP153(:,1))
max(R4SP162(:,1)) max(R5SP011(:,1)) max(R5SP022(:,1)) max(R5SP033(:,1))
max(R5SP044(:,1)) max(R5SP052(:,1)) max(R5SP064(:,1)) max(R5SP073(:,1))
max(R5SP081(:,1)) max(R5SP094(:,1)) max(R5SP102(:,1)) max(R5SP111(:,1))
max(R5SP123(:,1)) max(R5SP131(:,1)) max(R5SP142(:,1)) max(R5SP154(:,1))
max(R5SP163(:,1))];
StatPowerQuotient=StatPowerPioneer2./StatMaxPowerPioneer2;
%Static Heart Rate
StatHR=[max(R1SP173(:,2)) max(R1SP182(:,2)) max(R1SP191(:,2))
max(R1SP204(:,2)) max(R1SP211(:,2)) max(R1SP223(:,2)) max(R1SP232(:,2))
max(R1SP244(:,2)) max(R1SP253(:,2)) max(R1SP261(:,2)) max(R1SP272(:,2))
max(R1SP284(:,2)) max(R3SP052(:,2)) max(R3SP064(:,2)) max(R3SP073(:,2))
max(R3SP081(:,2)) max(R4SP054(:,2)) max(R4SP063(:,2)) max(R4SP071(:,2))
max(R4SP084(:,2)) max(R4SP092(:,2)) max(R4SP101(:,2)) max(R4SP113(:,2))
max(R4SP121(:,2)) max(R4SP132(:,2)) max(R4SP144(:,2)) max(R4SP153(:,2))
max(R4SP162(:,2)) max(R5SP011(:,2)) max(R5SP022(:,2)) max(R5SP033(:,2))
max(R5SP044(:,2)) max(R5SP052(:,2)) max(R5SP064(:,2)) max(R5SP073(:,2))
max(R5SP081(:,2)) max(R5SP094(:,2)) max(R5SP102(:,2)) max(R5SP111(:,2))
max(R5SP123(:,2)) max(R5SP131(:,2)) max(R5SP142(:,2)) max(R5SP154(:,2))
max(R5SP163(:,2))];
StatRiderHR={'R1';'R1';'R1';'R1';'R1';'R1';'R1';'R1';'R1';'R1';'R1';'R1';'R3'
;'R3';'R3';'R3';'R3';'R3';'R3';'R3';'R4';'R4';'R4';'R4';'R4';'R4';'R4';'R4';'
R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5';'R5
'};
StatStiffnessHR={'S3';'S2';'S1';'S3';'S1';'S3';'S2';'S3';'S3';'S1';'S2';'S3';
'S2';'S3';'S3';'S1';'S3';'S2';'S1';'S3';'S3';'S3';'S1';'S3';'S2';'S1';'S3';'S
1';'S1';'S2';'S3';'S3';'S2';'S3';'S3';'S1';'S3';'S2';'S1';'S3';'S1';'S2';'S3'
;'S3'};
25
%ANOVA Testing Effect of Stiffness on Wahoo/Pioneer Power Difference
%3 minute 180-200 Watt Sessions
[p,table,stats] =
anovan(longDeltaPower,{longStiffness},'model','interaction','varnames',{'Stif
fness'});
%Sprints
[p,table,stats] =
anovan(StatDeltaPower,{StatStiffness1},'model','interaction','varnames',{'Sti
ffness'});
%Results show insignificant difference in Power Transfer for DeltaPower
%caused by stiffness
26
Appendix B
Detailed chart of test runs for each rider (first four static runs for Rider 2, 3, and 4 are long 3-
minute runs):
27