Bicycle Quarterly Rolling Resistance Tests: No Surprises
 

I just finished going through this past fall's threads on the tire
performance testing reported in Bicycle Quarterly 5(1). I was amazed
at the collective amnesia displayed in this newsgroup (rbt).

On the correlation between Crrs as measured by drum tests of various
surfaces and road tests of various surfaces, see for example:

http://www.iea.org/Textbase/work/2005/EnerEffTyre/sandberg.pdf

This is a pilot study with no discussion section. For extensive
discussion on what the results mean, see this thre ad
http://tinyurl.com/2bsqr6
http://tinyurl.com/293kr7
which appeared just a few months before the BQ tire test threads, and
which people seem to have somehow forgotten about.

In my posts there, I noted that (a) in the historical development of
rolli ng resistance tests, which have been conducted by tire
manufacturers for a very long time, the drums were originally asphalt
or the like (b) there are drums currently available with all sorts of
surfaces (c) testers moved more to polished steel drums to r educe the
scatter, to make the results more replicable, and (d) now there is a
lot of interest in drum testing on rollers with various surfaces,
because now there are not just tire manufacturers wanting to reduce
the rolling resistance of the tires, but a lso people responsible for
the environment who want to reduce the energy losses particular to the
road surface. Since these tests are generally for car tires with
significant tread patterns, there is also an interaction between tread
pattern and road surf ace, so there is also a desire to optimize the
pairings. As the pilot study linked above shows, all this can also be
measured more cheaply on actual roads using instrumented trailers,
BUT... there is an awful lot more scatter with those methods,
especiall y, as with the study above, if the trailer is not encased
inside a fairing. Thus roller data, with curvature mathematically
corrected to a flat surface, is the preferred method, if you can
afford the equipment.

The amount of scatter observed testing with trailers- an inherently
far more controlled system than a bicycle+rider in a rolldown- should
be sufficient to make people think twice about the significance of the
particular rankings observed in the BQ tests, especially for nearby
results. This is even more the case when one considers a statistical
point which I bring up further below.

In fact, when one observes the scatter obtained even in the results
for a single drum surface at two different speeds, one should be even
more cautious about the validity of any resulting rankings, especially
for nearby results. As is well known, a tight correlation between two
methods does not ensure their interchangeability. Again, the
statistical problem I allude to above and which I will explain further
below makes the situation even more problematic.

Nevertheless, although the BQ tests were presented as overturning
tests done on polished drums, in fact they are yet another vindication
of them. This is best seen by the fact that a transformation based on
tire widt h and tread thickness normalized the results: all other
things being similar, these are exactly the two factors that
conventional rolling resistance theory, as explained in the threads
above, would point to. But not all other things were similar, and
further below I will single out another factor which was not taken
into account in the BQ tests.

In short, for the BQ tests, to quote a certain H. Inn, the best
surprise was no surprise. Consider the following:


-The fastest tire in drum tests was the same as the fastest tire in
the BQ tests: no surprise.

-Some extremely confusing statements about tire width:
----------------------------------------
#We expected wider tires to roll faster, and the wide tires we tested
#rolled slower than the narrow ones.

@However, while some are surprised that we found wider tires to roll
@faster, that is not inconsistent with previous results.

#I like 650B bikes, yet all currently available 650B tires scored
average
#or worse.

%One of our very fast tires, a Mitsubo shi 650 x 37 mm was very wide.
---------------------------------------

I take the average of these statements to mean that, all other things
being equal, wider tires roll faster. But this has been understood for
ages- again, no surprise.

-That a 24.5mm tire should be fastest should be no surprise: consider
what size Jobst has been riding all these years, and advocating as the
best compromise between rolling resistance losses and suspension
losses: none other than 25mm.

-It was claimed that the pressur e results of the BQ tests, which
showed that at higher pressures, not much difference in rolling
resistance results from raising the pressure further, were
inconsistent with the IRC drum tests. But that is not correct: Señor
Fogel already posted the resul ts below, using the rolling resistance
figures from those IRC tests:

--------------------------------------------
This calculator uses Jobst's rolling resistance data to predict
exactly
that kind of effect:

http://www.analyticcycling.com/ForcesTires_ Page.html
[...]
The range of difference in a 1-hour 40 km ride at about 25 mph is
less
than thirteen seconds for inflations from 120 to 100 psi.
---------------------------------------------

The BQ tests were done for a 25-30 sec, 184 metre ride. Sc aling the
above result down for the distance or time gives a difference of about
0.05 - 0.1 sec, about the same or in fact even less than what was
recorded in the BQ tests. Again, no surprises.


-The slowest tires in the BQ tests have not been publicly t ested on
drums, but were exactly those that would be expected to be slowest, on
the basis of conventional rolling resistance theory. The Rivendell
tires use the Tornado casing, otherwise abandoned long ago by
Panaracer, and for an obvious reason: since th e casing is Kevlar
reinforced bead to bead, it is a colossal energy sucker. (It should
though be more cut resistant than conventional sidewalls.) Again, no
surprises.

However, the presence of the Rivendell tires in the statistical
analyses therefore does pose a problem. Because of that bead to bead
Kevlar reinforcement, they are qualitatively different in
construction, and significantly outside the performance range of the
other tires. Thus they must function as outliers or leverage points
and ought to b e excluded from any regression or correlation analysis.
That done, the range of measured variation becomes smaller and the
relative size of the noise greater. One then has to start questioning
even more the significance of the rankings, especially for nearby
results.

-The BQ tests showed wearing gloves or not had a noticeable effect.
But this was already discussed in http://tinyurl.com/38xthb. In
particular, the wind tunnel testing done at MIT shows that gloves or
not has about the same aero effect as an aero front wheel or not.
Again: no surprises.


Now to some apparent anomalies that turn out not to be, and to some
erroneous comments:

-Some may wonder how it is that the Avocet Duro (700x28) scored worse
than the Avocet Cross (700x35). First of all, if it is really a Cross
and not a Cross II, then it is really a 700x32 tire, not a 35. The
Duro is a true 28. Look at this table:

Width Weight (g)
Cross Road
28 330 300 (Duro)
32 370 390 (Duro Plus)
35 395 ---

In other words, the Duro is a heavy duty tire, the Duro Plus even more
so, even more than the Cross. The two series use the same casings and
same rubber compound and same construction (save for t he tread
thickness, features, or lack thereof), so the weight differences are
all due to the tread thickness (tread coverage is also a possibility,
but these are similar in style). Thus if the Cross were very worn
down, as in fact it was reported to be, t hen we should indeed expect
it to be faster than the Duro. Again, no surprise.

I note that Jan Heine's remark that:

#Just because IRC or Michelin use drum tests doesn't mean it's a good
#way to test tires. In fact, both the Michelin Pro2 Race and the
Avocet
#Duro, which have been optimized on steel drums, scored much less
#well in our real-road tests, indicating that the current approach is
not
#the best. (They both were fine tires, but not as excellent as steel
drum
#tests would have us believe.)

is not apropos: there are no publicly available drum test data for the
Duro (or the Duro Plus). The famous graphs are only for the Time Trial
(20mm), Criterium (23mm), and Road (25mm) models. These are of a
different construction than the Duro, with thi nner rubber and
usually, 127tpi casings instead of 66 (old Road were 66).

-The Michelin Pro2Race and Continental Ultra Gatorskin in 23mm: the
Michelin was reported as better in the Tour tests, but the two scored
the same in the BQ tests. However, while the same models (presumably,
see further on) were tested, the same individual tires were not. There
is variation in production and we don't know how much. For example,
while the Tour test quoted had the Crr of the UG at 0.0058 and that of
the Pro2Race at .0042, the values given for these tires from some
other test that has been oft-quoted in this group (I think done by Le
Cycle, on steel rollers as well) are as follows:

Tire Prr Crr Speed*
Vittoria Open Corsa Evo CX 27.1 W .00334 23.22
Michelin Pro2 Race 29.2 W .00360 23.14
Michelin Carbon 34.7 W .00428 22.94
Panaracer Stradius Pro 35.4 W .00436 22.91
Hutchinson Fusion 39.6 W .00488 22.76
Continenta l Ultra GatorSkin 40.3 W .00497 22.73
Continental Grand Prix 3000 46.6 W .00575 22.50
*185lb rider + bike 250W rider output CdA = .32m^2 (racing crouch,
normal road bike) Transmission efficiency = 96%

Moreover, the UG was redesigned in 2 006 and it's not clear even if
exactly the same model was tested.

-The article says that hard rubber, while faster on polished drums, is
not faster in the real world. But (at least for smooth tires) nobody
has claimed the former: it is universally agreed, and Jobst has said
it here many times, that harder, high-carbon rubber is a high loss
material, and for better rolling resistance, switch to something else
inferior in grip and durability. The question instead is, in the
search for a compromise between durability, rolling resistance, and
grip, which is better, hard thin high-carbon rubber, or softer thicker
rubber? The BQ tests do not and cannot address that issue, but most of
us know from experience that for durability and wet grip, you
certainly need hard, high-carbon rubber. Therefore, if you want good
rolling resistance nevertheless, you have to make it relatively thin.
The 1.5mm depth of hard, wide coverage, high-carbon rubber tread on an
Avocet Road (25mm, 127tpi, not tested by BQ) seems to be one of the
better compromises around. How many miles is a Deda Tre good for? And
what's the wet grip?
The only circumstance where one might expect hard rubber of a
given thickness to give better rolling resistance than soft, is where
there is a signi ficant tread pattern, e.g. knobs. The softer rubber
will squirm more and so this might overcome its inherent lower rolling
resistance. The testers wondered why the Grand Bois tires were only
middle of the pack; I don't know their TPI or tread depth, but i f
they were e.g. 66tpi or greater and not exceptionally thick, then
perhaps the problem is the file pattern tread combined with soft
rubber. Get them to make them with a smooth tread and hard carbon
rubber, i.e. to replace my Avocet Duro Pluses, and I mig ht even buy
some one day...

-There is a problem with the conclusion that width and tread depth are
correlated with performance, but weight is not, because weight ought
to be highly correlated with width and tread depth. This should be
explored further. P erhaps because there is a correlation, just not a
linear one?

-On that note, there are big differences in tire design not just in
tread depth and compound, but also in tread coverage. For example, the
Avocets have the tread rubber going much higher up th e sidewalls than
many other otherwise comparable tires.

-The article considered the various widths of the Michelin Pro2Race to
form a family identical in all but this one parameter. That does not
seem to be correct: if you plot the weights of these tires versus
their width- and compare with other tire model series (e.g. Pasela,
Continental Ultra 2000, Ultra Gatorskin, Michelin Dynamic, Avocet)- it
seems clear that the various widths of the Michelin Pro2Race are NOT
all just different widths of the same t ire. Look at this table:

Width Weight (g)
Pro2Race Avocet
20 210 215
23 220 235
25 265 255

We know the tread depths of these Avocet models are respectively 1mm,
1.25mm, and 1.5m m. Clearly the variation in tread depth between the
Pro2Race tires of different widths is even greater. Thus these are NOT
the same tire with just the widths differing.

-On the difference between rolling resistance and suspension losses,
and why one sho uld consider them separately:
http://tinyurl.com/2ltu8a (Note Jobst is being sarcastic when he
says "I suppose")
I add that since jiggling, deforming over obstacles and bouncing, are
not rolling, there is no reason why anyone should call the resulting l
osses "rolling resistance". They are suspension losses. As to the
matter of steel wheels, and why we should not consider it a problem
that they are not best on surfaces with texture, while they are on
polished drums: steel wheels have no pneumatic suspens ion, and so
every minor irregularity exerts what is essentially an unrecoverable
braking jolt, or even extra distance to travel, against the system. By
contrast pneumatic tires, even at pressures far beyond the ordinary
(200psi, as observed by Jobst), are still capable of swallowing normal-
sized road irregularities (visibly bulging upon rolling over an
obstacle). In an ideal tire, even if the pressure were infinite, a
conical object might still deform it to the height of the cone,
depending upon the assumptions. That's not the case for an ideally
hard metal. One has to be careful about arguing at infinity, it's a
strange place.


There should be very little discrepancy in suspension losses between
different tire models of qualitatively similar design at the same
width and the same pressure, since it is the air volume inside them
that does the bulk of the bouncing and the shock absorbing. Thus the
road surface was smooth enough that suspension losses are doubtful as
the explanation for whatever minor disc repancies from the drum tests
might have been observed. Perhaps a better one is the following:

There was at least one more factor not controlled for in the BQ tests.
Changing the diameter of the tire or the pressure changes the
("pneumatic") trail, which, regardless of whether the bicycle had a
lot of trail to begin with, changes the handling characteristics (I
leave aside any aerodynamic effects, but consider the difference
between gloves and no gloves...). This will tend not to worsen the
scatter withi n the data for a single tire, but will bias the results
between tire models. This may result in systematic errors that further
mess up the rankings.


Have I covered everything? Well, there is the oddity that tubular tire
performance got worse with higher pressure, but clincher tire
performance did not. Why?


Overall though: no surprises.

e