|
Price
The three systems evaluated range in price from
$300 for the power upgrade for pre-existing
Polar owners up to $4600 for the top of the line
SRM. For the purposes of this review, it
was assumed that one has no pre-existing
hardware on the bike (i.e, the consumer is
starting from scratch). Ratings will be
given on a scale of 1 to 4 (1 is poor, 2 is
fair, 3 is good, 4 is excellent). The SRM
that was evaluated was the Pro Model.
Performance
From doing a bit of research on the topic of
power meters, the question that seems to be on
the top of everyone’s mind is: which system is
most accurate? Unfortunately, since access
to an external dynamometer was not available,
this question will not be answered here.
It is, however, possible to make some
observations about how the data from each of the
systems compare to each other.
Furthermore, it should be noted that no system
will ever be 100% accurate, there will always be
some uncertainty associated with the
measurement. The best that can be done
here is to evaluate whether the data collected
satisfies the claims to accuracy by the
manufacturers of the systems (SRM claims +/- 2%,
PT +/-1.5%, and Polar claims +/-10%).
Additionally, it can even be argued that
absolute accuracy is not an issue, but rather,
consistency over time or measurement
repeatability is most important.
Regardless of which side is taken in this
debate, there are also other factors to consider
when evaluating the performance of power
measurement systems. The ones selected for
this review were: ease of installation, overall
weight, data quality, and software capabilities.
Ratings on a scale of 1 to 4 will be given in
each of these categories based on the same scale
as for pricing above.
Ease of Installation
Though it is an extremely small percentage of
the overall life of the product, the ease with
which the system is installed and adjusted can
immediately affect post purchase perception of
the product. A manufacturer survives in
the long term by receiving repeat business, and
anything that colors a consumer’s perception
negatively can have an impact on future
purchases.
An impact on accuracy of the system can also be
affected by the amount of consumer input during
the installation process. The best
installation is one that minimizes the
possibility of error by the consumer. An
analogy can be made to the computer world ��� the
best peripherals are the ones that are “Plug and
Play”. The end user simply plugs it in and
it works as desired. There will invariably
be some problems with some users if one has to
search for drivers, cables, etc.
Power Tap
The PT system is just about as close to “Plug
and Play” as one can get. The
straightforward steps of installing a cassette
and a tire on the wheel built with the PT hub
are nothing too difficult for your average bike
racer. Once these steps are accomplished,
one must simply drop the wheel in the dropouts
and install the receiver and CPU mount with the
supplied zip ties. The whole process
should take 30 minutes at most.
SRM
The SRM installation was only slightly more time
consuming. The cranks currently installed
need to be removed and the new ones (both left
and right) supplied by SRM need to be installed.
Prior to installing the SRM crank, though, a
chainstay sensor needs to be installed as well
as the front fork speed sensor. The rubber
band style of mounting the sensors makes it easy
to adjust or even move the entire system to
another bike. After the two sensors were
mounted, the process was completed with the
installation of the right crank.
Installation took approximately 45 minutes and
did require the ability to remove and install
cranks.
Polar
The Polar installation was a relatively
painful operation. Most of the anxiety
involved in the installation arose from plain
old uncertainty – was I doing it right?
The most tedious part in the installation was in
adjusting the height of the chain frequency
sensor. It needed to be raised, lowered,
and leveled multiple times while troubleshooting
some erroneous power readings. There is
also a chain speed sensor that needs to be
installed. This required removing the bolt
from one of the derailleur pulleys and
installing the one provided by Polar through the
attached chain speed sensor. The cabling
to this sensor must be adjusted to make sure it
is not damaged during normal operation of the
derailleur. The image below shows the
installation used by the Polar folks in their
booth at last year's Interbike show.
The speed sensor must also be placed on the
opposite chainstay using zip ties. This
was a straightforward operation – just be sure
to route it and tie down the cable so that it
doesn’t interfere with the tire.
The cadence sensor provided also did need some
special attention when used with the Shimano
Ultegra cranks I normally use on my bike.
The sensor needed to be rotated drastically
inward and often times rubbed the chain when in
the 53x12. It simply did not want to stay
in place, no matter how hard I cranked down on
the zip tie. The main problem with the
sensor was that the sharp angles of the crank
did not match well with the cadence sensor base.
A simple v shaped base might make this sensor
more adaptable to varying crank geometries.
However, when used with the SRM crank, the
cadence worked perfectly. Complete
installation took the better part of a Saturday
morning (3-4 hours) and subsequent tweaks to
troubleshoot some erroneous readings added
another hour or so.
The erroneous readings that were noticed while
riding were obvious. On the initial setup,
while riding along on a relatively level grade
and shifting through the entire gear range in
the big ring, power stayed constant. When
I shifted into the 53x21 (I run a 23 in the
back, so this was not the most extreme crossover
possible) I observed a 40 watt increase in
measured power. No matter how the chain
frequency sensor placement was adjusted,
(raising, lowering, leveling, rotating,
translating) the same phenomenon occurred.
As a last resort, I swapped out the chain for
a longer one and the problem vanished. I
can only speculate that the tighter chain
originally used was creating an effectively
shorter unsupported chain length (due to more
friction between chain and gear teeth) and,
thus, the higher vibration frequencies (measured
power) observed. The problem might
also have been resolved by tweaking the chain
speed sensor orientation, but that variable was
not manipulated at the time.
The main thing to take away from this is to
test out your installation before getting real
comfortable with the setup. Do this by
riding along at a relatively constant speed on
level ground and shift through the entire
spectrum of gears. Obvious problems should
be readily apparent. Be patient with this
system, one can get good results if due
diligence is done.
Weight
The scale doesn’t lie. It should be
remembered that power measuring systems are not
essential and any additional weight added to the
bike negatively affects overall performance –
though, at these levels the added weight has a
small affect on performance. This weight
penalty, of course, neglects any gains to be had
by training with power.
Data
Here is the good part. I was able to
acquire just over 8 hours of data using all
three systems and a few interesting things were
learned. First, though, let’s take a
closer look at an estimate of which system has
the potential to be the most accurate.
Uncertainty/Accuracy
A common practice in any experimental design
is to do an uncertainty analysis on the
instrumentation used. The results of this
analysis will show how good the results are, or
what claims can be made based on those results.
If the uncertainty is only plus or minus 20% and
one needs +/- 1% to answer a question, the
source of the error must be identified and
corrected.
Uncertainty of the instrument is a function
of all the parameters used to calculate the
parameter of interest. Essentially, the
fewer things that need to be measured and the
lower the uncertainty of the individual
instruments used to compute the quantity of
interest, the more accurate the measurement will
be. I will spare you the gory details.
It should be clear that the PT and SRM have
fewer fundamental sources of error than the
Polar system. It can also be speculated
that the largest source of error on the PT and
SRM will be in the calibration of the strain
gaged structure.
PT units can have their accuracy checked by
the user doing a static torque test. The
SRM goes one step further, and allows the
end-user to actually change the factory
calibration values in the event that the
original system calibration has drifted with
time. It is recommended that users do a
static check on their PT and SRM systems during
the initial installation and at least every 3
months to make sure that the system is stable.
Checking the systems statically requires a time
commitment, but it is helpful when analyzing
long-term power output trends – did you really
increase your power at lactate threshold by 10%
since last year, or did your power meter
linearity drift? Static calibration will
lend some insight into the answer to this
question.
Average Power
When using the data from all sets and
normalizing how the average power is calculated,
(dealing with how zeros are handled) it can be
seen that the values fall within a range of
approximately 5% of each other.
Peak power
Peak power measurement accuracy is primarily
driven by how long of a sample period the data
is averaged over. Therefore, the PT and
SRM are relatively similar, since the reporting
intervals are 1.26 seconds and 1 second
respectively. On the other hand, the Polar
unit only reports data every 5 seconds and
naturally has a much lower peak power reading
for similar efforts.
The above plot shows two five second sprints
done. The Polar under-reports the first
effort by 50% and completely misses the second
effort.
The longer the duration of the effort, in
this case 10 seconds, the more closely all three
systems will match each other.
Power by gear
It has been speculated that the Polar unit,
due to its tricky instrumentation, and widely
varying geometry of chain/stay configurations
may provide faulty power readings. For the
following data presentation, it should be noted
that the raw data was taken at different time
intervals (1 second for the SRM, 1.26 seconds
for the PT, and 5 seconds for the Polar).
In order to make the comparisons consistent, in
terms of total number of observations, the data
was pseudo-sampled digitally at 1 second
intervals. This may have a slight effect
on the results, so it is important to realize
that this is how the analysis was done.
In order to investigate and evaluate whether
bad data is acquired in a particular gear while
on the road, one must be able to calculate
instantaneous gear ratios. Several people
have presented this analysis technique before
(two sources that I know of include:
Robert Chung and
Tom
Compton). The fundamental math is
quite straightforward, though. The method
described on the sites linked above uses the
instantaneous velocity and cadence data to
calculate an instantaneous gear ratio.
It is then possible to determine which front
ring and which cassette cog one is in (assuming
one knows the cogs on the bike). For
example, during the 1400-1600 second range
above, it was determined that I was in the
39x23.
In the plot above, it should be observed that
the PT gear data is markedly different than
either the SRM or Polar systems – it does not
show a clumping of the data into discrete gear
ratios. This is due to the direct
measurement of cadence by the latter systems and
the indirect method employed by PT. PT
looks for torque pulses and assumes that each
one is alternating between left and right crank
signals. It then calculates a cadence
based on the time between pulses. Needless
to say, the data illustrates that the method
needs development. In fact, the new PT Pro
model ($899) is supposed to address this by
including a hardwired cadence sensor.
|
