The zero wind tunnel option for serious cyclists

joseph.santaniello@gmail.com

On May 12, 8:29*pm, Andre Jute <fiult...@yahoo.com> wrote:
> On May 12, 6:23*pm, "Rik O'Shea" <rikos...@yahoo.com> wrote:
>
>
>
> > On 12 May, 18:14, "joseph.santanie...@gmail.com"
>
> > <joseph.santanie...@gmail.com> wrote:
> > > On May 12, 4:43 pm, Andre Jute <fiult...@yahoo.com> wrote:
>
> > > > I like the Compton idea of a large warehouse... Makes me wonder if one
> > > > could liberate some traffic cones and diversion signs and claim a
> > > > tunnel through a mountain or under a river for cycling tests for a few
> > > > hours early one morning before the authorities get their brains in
> > > > gear...
>
> > > I have considered this. But around here at least, there is often a
> > > stiff breeze in the tunnels. It may be constant however.
>
> > > This one ought to do:
>
> > >http://en.wikipedia.org/wiki/Image:TRS_070405_016.jpg
>
> > > It's 7.2km long under a narrow section of Oslo Fjord. The gradient is
> > > 7% on each side.
>
> > > There is so little traffic there, I'll bet you could dispense with the
> > > cones.
>
> > > Talk about a foolish endeavour. The tunnel cost an astronomical amount
> > > of money, and was built to alleviate some of the heavy traffic that
> > > goes through Oslo (and to service a new airport which never happened
> > > because they decided to build it someplace else!). But it is so steep
> > > that trucks use more fuel taking the tunnel shotrcut than taking the
> > > long route through Oslo (according to my father in law who is in the
> > > transport business here).
>
> > > Joseph
>
> > I used the Dublin port tunnel just before it was opened to traffic for
> > a very successful set of CdA field tests (it was just about to be used
> > for a 10 k running race).
> > Its perfectly flat, at sea level, over 4 km long and there is zero
> > wind in it!
>
> >http://en.wikipedia.org/wiki/ImagePT_inside.jpg
>
> Ha! Perhaps we should split this thread in three, naming the parts
> "Zero ]windtunnel[ (Andre)" to indicate that no windtunnel is
> required, "]Windy tunnel[ (Joseph)" to indicate that a tunnel is
> available but not ideal because wind blows in it, and "]Zero-
> wind[ tunnel (Rik)" to indicate that a tunnel is available and no wind
> blows in it.
>
> Choice rules!
>
> Andre Jutehttp://members.lycos.co.uk/fiultra/BICYCLE%20%26%20CYCLING.html

This one would be fun for an out-and-back time-trial!

http://en.wikipedia.org/wiki/Image:Laerdalstunnelen.jpg

Tunnels in Norway are fun. Once driving to Bergen with my wife in our
Lada, we got trapped (in a modest way) in a tunnel. Avalanche ahead,
jack-knifed tractor trailer behind. Don't forget the sleeping bags and
food!

Joseph

Robert Chung

On May 12, 11:29 am, Andre Jute <fiult...@yahoo.com> wrote:

> Ha! Perhaps we should split this thread in three, naming the parts
> "Zero ]windtunnel[ (Andre)" to indicate that no windtunnel is
> required, "]Windy tunnel[ (Joseph)" to indicate that a tunnel is
> available but not ideal because wind blows in it, and "]Zero-
> wind[ tunnel (Rik)" to indicate that a tunnel is available and no wind
> blows in it.

That's certainly one way to partition the posts. I'm sure there are
others.

Rik O'Shea

On 12 May, 18:56, Robert Chung <rech...@gmail.com> wrote:
> On May 12, 10:23 am, "Rik O'Shea" <rikos...@yahoo.com> wrote:
>
> > I used the Dublin port tunnel just before it was opened to traffic for
> > a very successful set of CdA field tests (it was just about to be used
> > for a 10 k running race).
> > Its perfectly flat, at sea level, over 4 km long and there is zero
> > wind in it!
>
> Could you describe your test protocol and how you could tell that they
> were very successful?


I used a power meter (Power Tap SL) and performed a number of steady
state speed trials in which I recorded power for various speeds.
I performed a linear regression of force versus speed*speed to deduce
CdA, Crr.
The tests were very successful as the results were statistically
significant with a very low standard error and deviation.

carlfogel@comcast.net

On Mon, 12 May 2008 10:56:30 -0700 (PDT), Robert Chung
<rechung@gmail.com> wrote:

>On May 12, 10:23 am, "Rik O'Shea" <rikos...@yahoo.com> wrote:
>
>> I used the Dublin port tunnel just before it was opened to traffic for
>> a very successful set of CdA field tests (it was just about to be used
>> for a 10 k running race).
>> Its perfectly flat, at sea level, over 4 km long and there is zero
>> wind in it!
>
>Could you describe your test protocol and how you could tell that they
>were very successful?

Dear Robert,

I'm curious, too, since detecting the kind of differences that seem to
be involved should include weighing the rider and checking the tire
pressures before each run, as well as recording the temperature,
humidity, and barometric pressure.

As for the actual wind speed, many posters in this thread may be
unaware of comments like this from Andy Coggan:

"More recently, I've conducted three days of testing using Lim's
approach of multiple speed trials to separate CdA and Crr...despite
doing the testing very early in the morning on very flat roads,
precision was again only about 2%. One impressive finding from these
experiments was how little wind it takes to really skew the results -
for example, on a day so calm that a tiny piece of dry leaf dropped
from overhead falls to the ground less than 1 foot from vertical,
there are still significant differences in estimated CdA depending on
whether you are headed 'upwind' or 'downwind'. Similarly, trials
conducted less than a full minute after a vehicle passes down the road
at ~30 mph also produce abnormally high or low values, depending on
whether the trial was in the same direction or the opposite direction
in which the vehicle passed."

http://groups.google.com/group/rec....60f5e7d17060960

Cheers,

Carl Fogel

Rik O'Shea

On 12 May, 20:48, carlfo...@comcast.net wrote:
> On Mon, 12 May 2008 10:56:30 -0700 (PDT), Robert Chung
>
> <rech...@gmail.com> wrote:
> >On May 12, 10:23 am, "Rik O'Shea" <rikos...@yahoo.com> wrote:
>
> >> I used the Dublin port tunnel just before it was opened to traffic for
> >> a very successful set of CdA field tests (it was just about to be used
> >> for a 10 k running race).
> >> Its perfectly flat, at sea level, over 4 km long and there is zero
> >> wind in it!
>
> >Could you describe your test protocol and how you could tell that they
> >were very successful?
>
> Dear Robert,
>
> I'm curious, too, since detecting the kind of differences that seem to
> be involved should include weighing the rider and checking the tire
> pressures before each run, as well as recording the temperature,
> humidity, and barometric pressure.
>
> As for the actual wind speed, many posters in this thread may be
> unaware of comments like this from Andy Coggan:
>
> "More recently, I've conducted three days of testing using Lim's
> approach of multiple speed trials to separate CdA and Crr...despite
> doing the testing very early in the morning on very flat roads,
> precision was again only about 2%. One impressive finding from these
> experiments was how little wind it takes to really skew the results -
> for example, on a day so calm that a tiny piece of dry leaf dropped
> from overhead falls to the ground less than 1 foot from vertical,
> there are still significant differences in estimated CdA depending on
> whether you are headed 'upwind' or 'downwind'. Similarly, trials
> conducted less than a full minute after a vehicle passes down the road
> at ~30 mph also produce abnormally high or low values, depending on
> whether the trial was in the same direction or the opposite direction
> in which the vehicle passed."
>
> http://groups.google.com/group/rec....60f5e7d17060960
>
> Cheers,
>
> Carl Fogel

Here's the Required Test Conditions

(1) Select an appropriate test course; this can be a “speed trap” i.e.
a section of road between 200m and 1000m long or a velodrome. The
speed trap dictates that you should have reached your desired steady
state speed before you enter the start of the trap. This speed is
maintained while in the trap (i.e. there should be little or no
acceleration). As you exit the trap and end your test run your speed
should be the same as when you entered the trap. The advantage of a
longer speed trap is that it may allow slight fluctuations in speed,
power and wind to be averaged out allowing a more accurate average
power and speed to be recorded.
(2) The course should be flat with no corners. The road or track
surface should have minimum variation in texture; smooth or even
surfaces provide the best results.
(3) Calm wind conditions should exist. A 30 cm length of single-ply
tissue paper held hanging straight down reveals whether or not calm
wind conditions are prevalent.
(4) The course must be free from interference from road traffic. You
should wait at least one minute after a vehicle passes before
performing a test trial.

All these conditions were met in full - in particular (3) which is
difficult in an open air course.


Here the exact protocol

(1) Record the total mass of rider + bike
(2) Record the air pressure, air temperature and humidity - used to
calculate air density.
(3) Perform the desired number of test trials recording the average
speed and power (the complete series of test trials took less than 30
minutes).
- Use a fixed speed rather than power for each test trial.
- To start choose speeds with roughly equal intervals. At higher
speeds the interval can be progressively decreased.
- The order of test trials involving various speeds should be
randomized. This is optional.
- Before you enter your speed trap or start your recording interval
you should have achieved the desired steady state speed. You should
strive to maintain a steady speed for the duration of the test trial,
i.e. no acceleration or deceleration.
- Maintain the exact position on the bicycle for each trial.
(4) When you have completed the field test use your power meter
software to extract the data for each test run i.e. to select the
starting and finishing points of each interval so that the start and
finish speeds are equal. Then obtain the average speed and power for
each test run.

This protocol was followed exactly and the results had a very high
statistical significance.

Robert Chung

On May 12, 2:16 pm, "Rik O'Shea" <rikos...@yahoo.com> wrote:

> This protocol was followed exactly and the results had a very high
> statistical significance.

Forgive me for persisting. When you write "a very high statistical
significance," what was the hypothesis you were testing for which the
significance was so high?

FWIW, we've been discussing this, and another, method over on the
Wattage List.

Tim McNamara

In article
<d0d36acc-40fb-4298-ab39-239c2c40e63a@d45g2000hsc.googlegroups.com>,
"Rik O'Shea" <rikoshea@yahoo.com> wrote:

> This protocol was followed exactly and the results had a very high
> statistical significance.

Meaning?

Rik O'Shea

On 12 May, 22:36, Robert Chung <rech...@gmail.com> wrote:
> On May 12, 2:16 pm, "Rik O'Shea" <rikos...@yahoo.com> wrote:
>
> > This protocol was followed exactly and the results had a very high
> > statistical significance.
>
> Forgive me for persisting. When you write "a very high statistical
> significance," what was the hypothesis you were testing for which the
> significance was so high?
>
> FWIW, we've been discussing this, and another, method over on the
> Wattage List.

Let me rephrase a little - the standard error and standard deviation
associated with the derived CdA value was quite low. I am aware of
some of the "other" methods. One problem I see with them is that there
is no easy way of determining a confidence interval with the CdA
result plus the reliance on "guessing" the Crr value.

Andre Jute

On May 12, 10:36*pm, Robert Chung <rech...@gmail.com> wrote:
> On May 12, 2:16 pm, "Rik O'Shea" <rikos...@yahoo.com> wrote:
>
> > This protocol was followed exactly and the results had a very high
> > statistical significance.
>
> Forgive me for persisting. When you write "a very high statistical
> significance," what was the hypothesis you were testing for which the
> significance was so high?
>
> FWIW, we've been discussing this, and another, method over on the
> Wattage List.

Perhaps it will be useful if you were to describe the two methods, and
give us a summary of the consensus, if any, and remaining
difficulties, perceived or real.

Andre Jute
http://members.lycos.co.uk/fiultra/...%20CYCLING.html

Robert Chung

On May 13, 2:06 am, "Rik O'Shea" <rikos...@yahoo.com> wrote:

> Let me rephrase a little - the standard error and standard deviation
> associated with the derived CdA value was quite low.

Ah, thanks. I understand that was a one-off test but, with those data,
how small of a difference in CdA do you think you could have detected
reliably?

> I am aware of
> some of the "other" methods. One problem I see with them is that there
> is no easy way of determining a confidence interval with the CdA
> result plus the reliance on "guessing" the Crr value.

That depends on what you're interested in. If what you want to do is
determine which of two (or more) different aero setups has smaller CdA
then Crr isn't an issue. That's the classic approach of coast downs,
where you don't really care what Crr is since you're coasting down the
same surface (presumably with the same tires and other equipment) and
altering only your position. Although the precision of coast downs is
poor, it's not because of reliance on "guessing" the Crr.

As for the CI, that's largely an issue of sample size and some methods
are more parsimonious with the data than others.

Robert Chung

On May 13, 5:55 am, Andre Jute <fiult...@yahoo.com> wrote:

> > FWIW, we've been discussing this, and another, method over on the
> > Wattage List.
>
> Perhaps it will be useful if you were to describe the two methods, and
> give us a summary of the consensus, if any, and remaining
> difficulties, perceived or real.

We've actually been discussing more than two field methods, though
each of the ones we've been discussing require that power and speed be
known using an on-bike power meter.

OTOH, I'm still fascinated by your method which requires nothing more
than a bike and the road on which to ride it. Perhaps it will be
useful if you were to describe what you meant by "a very high degree
of accuracy" and give us a summary of the consensus, if any, on your
method? Can you provide an example from data you've collected using
your method?

Rik O'Shea

On 13 May, 14:55, Robert Chung <rech...@gmail.com> wrote:
> On May 13, 2:06 am, "Rik O'Shea" <rikos...@yahoo.com> wrote:
>
> > Let me rephrase a little - the standard error and standard deviation
> > associated with the derived CdA value was quite low.
>
> Ah, thanks. I understand that was a one-off test but, with those data,
> how small of a difference in CdA do you think you could have detected
> reliably?
>
> > I am aware of
> > some of the "other" methods. One problem I see with them is that there
> > is no easy way of determining a confidence interval with the CdA
> > result plus the reliance on "guessing" the Crr value.
>
> That depends on what you're interested in. If what you want to do is
> determine which of two (or more) different aero setups has smaller CdA
> then Crr isn't an issue. That's the classic approach of coast downs,
> where you don't really care what Crr is since you're coasting down the
> same surface (presumably with the same tires and other equipment) and
> altering only your position. Although the precision of coast downs is
> poor, it's not because of reliance on "guessing" the Crr.
>
> As for the CI, that's largely an issue of sample size and some methods
> are more parsimonious with the data than others.


So how many samples do you need using the other method(s) to obtain a
CdA with 95% CI of +/- 0.01 m*m ?

Andre Jute

On May 13, 3:05*pm, Robert Chung <rech...@gmail.com> wrote:
> On May 13, 5:55 am, Andre Jute <fiult...@yahoo.com> wrote:
>
> > > FWIW, we've been discussing this, and another, method over on the
> > > Wattage List.
>
> > Perhaps it will be useful if you were to describe the two methods, and
> > give us a summary of the consensus, if any, and remaining
> > difficulties, perceived or real.
>
> We've actually been discussing more than two field methods, though
> each of the ones we've been discussing require that power and speed be
> known using an on-bike power meter.

That's the necessity my method is intended to circumvent. The common
coastdown method takes a spot reading on some maximum speed and,
generally speaking, derives both power and Cd from there, a procedure
with considerable potential for error.

> OTOH, I'm still fascinated by your method which requires nothing more
> than a bike and the road on which to ride it. Perhaps it will be
> useful if you were to describe what you meant by "a very high degree
> of accuracy"

I certainly don't mean 2% repeatable. I mean a very high degree of
accuracy relative to the spot method, all other conditions holding. My
purpose is a useful reading to a particular cyclist on his own roads,
under his own conditions, without presuming expensive instrumentation.
I deduce that it will be more accurate than the spot reading on
coastdown method because of the internal structure of my method, which
calculates the cyclist's power by arguing that he outputs maximum
power not only at a maximum speed but on acceleration, and then takes
repeated acceleration readings and averages them, thus reducing
measurement error progressively, and also averaging uncontrolled
factors over time. Note that my formulation has an as yet undeclared
internal check by which total tractive effort must equal total
tractive resistance; this is illustrated in the graphic on p162 of my
book from which I adapted these methods (Andre Jute: Designing and
Building Special Cars, published by B T Batsford, London, 1985,
published in the States by Robert Bentley of Boston).

>and give us a summary of the consensus, if any, on your
> method?

It's a newly formulated method for cycling; by the nature of people's
attitudes to the novel, I would expect zero consensus from the stick
in the muds and growing consensus over time from those with enquiring
minds and the engineers who either have an automobile (or possibly
railway) background or are the possessors of better brains so that
they can follow the logic which, I admit, I haven't laid out as fully
as in my book. I was in fact pleasantly surprised to hear people
instantly agree that the formulae are correct and state that their
reservations are about applicability to cycling. However, in
automobiles, the consensus is 100%; these are methods proven by time,
taught to first-year mechanical engineering students, generic to every
comprehensive textbook. Special car builders write to me to express
their delight at how easy it is once the method is understood, and
expressing surprise too at how consistent the results are with
expectations and the experience of others "in the club".

>Can you provide an example from data you've collected using
> your method?

It's a new formulation for cyclists, so I have no data on cyclists
yet, and my data on automobiles by which I tested my formulations for
(uninstrumented) special cars builders was published in my book cited
above, where you can look it up.

HTH.

If you're a young fit, preferably slender, cyclist with a friend or
two to help in these tests, perhaps you should start collecting data
to see how well my method works. It is described here:

http://members.lycos.co.uk/fiultra/...parameters.html

and aspects of methodology re recording are discussed in this thread,
and Rik O'Shea also in this thread has provided a more rigorous test
protocol than I laid down -- I'm happy with a pretty casual approach
as long as casual doesn't become slack, and as long as the test
iterations mount up inexhorably.

Andre Jute
http://members.lycos.co.uk/fiultra/...%20CYCLING.html

Robert Chung

On May 13, 7:18 am, "Rik O'Shea" <rikos...@yahoo.com> wrote:

> > As for the CI, that's largely an issue of sample size and some methods
> > are more parsimonious with the data than others.
>
> So how many samples do you need using the other method(s) to obtain a
> CdA with 95% CI of +/- 0.01 m*m ?

Depending on the method, .01 m^2 could be pretty big. If reasonable
protocols are followed with the right method, a few to several minutes
of samples. With A-B-A testing, one guy detected the change due to an
aero water bottle of less than that size. Another guy estimated the
change in CdA due to a 5cm x 5cm reference object (i.e., .0025 m^2)
as .003 m^2, once again validated with A-B-A test runs. That said,
I've never been able to come close to that kind of precision with my
runs but I'm a lousy experimentalist.