Optimizing Bicycle Tire Pressure on Gravel - Part III (The Part where freshly-waxed chain break-in time is revealed.)

This is Part III describing studies to determine which tire pressure gives the least rolling resistance on gravel.  You should read Parts I & II first. Be aware that the results are unique to my weight, tire, and road surface conditions.

[For my genealogy research readers:   You are at the correct location.  As you know I get distracted at times.  More genealogy research results will come soon.  My most recent post on Y-DNA is here.]

In Part I, I found the surprising result (to me) that my drag at a tire pressure of 30 psi was less than my drag at 40 or 50 psi. I calculated the power savings to be about 10 watts. That is similar to the savings I obtained by getting into an aero position (all at about 15 mph).

In Part II, I made some improvements in the testing protocol, but I began to question the reproducibility of the data.

This study, Part III focused on reproducibilty of the protocol. The changes from Part II are:

1) Silca Pump used with advertised accuracy of 1%.
2) Wheel circumference was remeasured using new pump and a 10 revolution loaded-roll-out on smooth pavement. (Average of 5 roll-outs.  All pressures tested had measured circumferences with std dev of less than 0.03%.)
3) New test loop was in a traffic-free, wind-protected dip, with no paved sections.
4) 8 laps per test (vs 4 in Part II). Total distance per test was 2.21 miles vs. 2.09.
5) Freshly-waxed chain

The Test Variables

Only the tire pressure was varied.  All other variables were held constant.  Three pressures were tested: 20, 30, and 40.  The 30 psi test was repeated 4 times to test for reproducibility.

The results are shown in terms of Virtual Elevation (VE). (See Part I for short explanation of VE). The steeper the slope of a Test, the more power is being lost versus other Tests. [Note how much cleaner these data are than in Part I and II.  All of the laps are clearly visible. The VE of the dip is more consistent than Part II (due to lack of paved section?)]

Analysis of the data revealed that the break-in of the freshly-waxed chain overshadowed all of the other variables! The negative impact of using a freshly-waxed chain was about 14 watts during Test 1 versus Test 6. Even after 9 miles of riding there appears to be break-in losses. The abrupt change in slope in Test 3, when the pressure was lowered to 20 psi, suggests there could be some reductions in rolling resistance, making the break-in losses appear less severe.  When the pressure was increased to 40 psi in Test 4, the abrupt change in slope in the opposite direction indicates more loses versus the 20 psi Test.  The flattening of the slope at the end of Test 4 is a bit puzzling. The slope at the start of Test 5 could be viewed as a continuation of the curve begun back in Tests 1 and 2. The last 3.5 miles of riding appear consistent, indicating the break-in is complete.

Conclusion

The objective of showing reproducibility was not achieved in Part III. (Sorry...Part IV is coming.)

The freshly-waxed chain (SRAM 12 sp with SILCA Super Secret Hot Wax) was shown to take at least 10 miles or about 45 minutes of riding to reach a steady state. 

Also see 7:26 minutes into the Silca video: Chain Waxing? Avoid These 7 Common Failures! (youtube.com)

GCN beat me to it.  This Chain Waxing Mistake Makes You SLOW (youtube.com)

Optimizing Bicycle Tire Pressure on Gravel - Part II

This is Part II describing studies to determine which tire pressure gives the least rolling resistance on gravel.  You should read Part I first. Be aware that the results are unique to my weight, tire, and road surface conditions.

[For my genealogy research readers:   You are at the correct location.  As you know I get distracted at times.  More genealogy research results will come soon.  My most recent post on Y-DNA is here.]

In Part I, I found the surprising result (to me) that my drag at a tire pressure of 30 psi was less than my drag at 40 or 50 psi. I calculated the power savings to be about 10 watts. That is similar to the savings I obtained by getting into an aero position (all at about 15 mph).

There were some flaws with that study, which I tried to correct in Part II. The following changes were made to the study for Part II:

1) I used a two-sided power meter instead of one-sided.
2) I measured wheel circumferences at all tested pressures. (Lower pressures have smaller circumferences.  The circumference is a speed sensor setting.)
3) The new loop is at a dip in the road.  This allows for speed variations and easier identification of laps.
4) I started and ended tests at V=0 vs. in-motion starts and stops in Part I.
5) Test on a less windy day.

This study still had some flaws which I think I can address in Part III or IV. They are:

1) My pressure gauge is neither accurate nor precise.  (A new one is on order.)
2) There is a section of paved road at the bottom of the dip. This results in some noise in the data seen at the bottom of each data curve.
3) I only made 4 laps per test (vs. 6 in Part I), but the laps were twice as long as in Part I.

The Test Variables

Only the tire pressure was varied.  All other variables were held constant.  Three pressures were tested: 27, 32, and 37.  The 32 psi test was repeated at the beginning and end to test for reproducibility.

The data are reported as 'Virtual Elevation' (VE) versus distance. An increase in slope indicates more power was needed for that test condition versus the base case. Test 4 is the base case.

The data are a bit cleaner than the data from Part I, however the test is not reproducible from Test 1 to Test 4.  In my next study, I will do a duplicate at the beginning of the study (Test 1 and Test 2) and again at the end, to nail down the variability of this test.

If I use Test 4 as my baseline and choose CdA and Crr that will make Test 4 level, it can be compared to Test 2 and 3. In this case 37 psi is a 3 watt penalty and 27 psi is a 3 watt savings versus 32 psi.

I lack 100% confidence in these results.  However, I have changed my default riding pressure to 28 psi in front and 30 psi in rear, so I do have some confidence that what I am seeing is real.

Part III will be coming soon! 

Optimizing Gravel Tire Pressure using Chung Method - Part I

I am preparing for an upcoming gravel triathlon. What is the optimal tire pressure for me and my set-up on gravel? What if I find myself alone on the bike course with no one to draft behind? ...What is my optimal body position?

Generic "answers" to these questions can be found on the internet, but can I find the optimal conditions for me and my set-up?

I performed the seven tests below in under an hour.  Three of the tests were duplicates to verify that the tests were reproducible (1, 4, and 7).  The test conditions and results are shown below.  They are shown in terms of a "Virtual Elevation" (VE).  If the slope is upwards for a test, it means that extra power is needed for that test (like going uphill).  Conversely, if the slope is downwards for a test, it means less power is needed for those conditions.

In summary, it appears that a tire pressure as low as 30 psi is more efficient for me than 40 or 50 psi. As one would expect, getting into the drops is more aero (negative slope in VE) than being on the hoods.  Even greater savings is seen when I bring my hands in close on the bars (Test #6).

What amazed me was how quickly I was able to do these tests (under an hour) AND despite less than ideal conditions (it was windy, course was flat, and I did not alter speed very much) I had results that yielded valuable information.

Test Protocol

Each test was 6 laps on a quarter mile gravel loop.  The loop was flat.  (Better results are obtained on loops with a change in elevation.) I averaged about 14 mph for all laps. (Better results are obtained if the speeds are varied.) I used a calibrated Garmin Speed Sensor (more accurate than GPS).  I used a single-sided Quarq spindle power meter. (Better if two-sided.)  Tire pressure was measured using a Specialized floor pump with gradations every 5 psi....not that precise and probably not that accurate either. I captured T, RH, and P at the start and end of the test protocol using www.localconditions.com.  I weighed myself and loaded bike using a Tanita floor scale.  Each test was saved as a Garmin workout.  All seven workouts were exported from Garmin Connect as TCX files and imported into Golden Cheetah.  The seven workouts were combined and analyzed in the AeroLab Chart as shown above.

The Chung Method

Virtual Elevation is the output of the Chung Method.  The method solves the Power Balance equation for slope, using guesses for Crr and CdA. Slope multiplied by velocity gives the change in elevation which are strung together and plotted.  This elevation is referred to as "Virtual Elevation" as all of the unaccounted for "power" in the Power Balance, whether it is related to elevation gain or not, is converted to elevation. Chung has shown that this method is good at exposing even minor changes in Crr or CdA, even when the data is crappy...as mine are. You can read his paper here.  [Crr is the Coefficient of rolling resistance and CdA is the Coefficient of drag area.]

Data Analysis

Stringing all of the test data together allows a visual analysis.  I guessed a Crr and CdA which made the duplicate runs (1,4,7) as flat as possible. If the test is reproducible, 1,4, and 7 should all be the same. In fact, each of the 6 laps within a test should be the same.  If you look closely, you can count the 6 laps in each test. Test 1 had some anomalies on the first two laps. I did not do any practice laps. This was the absolute first time I had ever biked this loop. There was a conduit over the road which I rode over on the first 2 laps.  All subsequent laps, I biked around it.  It is interesting that this method is sensitive enough to show that. Test 4 had relatively good data. Test 7 appears OK for the first 3 laps, then trends up. The wind was gusting more, so that could be part of the cause. Considering how windy it was, I am surprised I got any meaningful data at all from these tests. With these differences in 1,4,7, I would want to do this test again on a calm day to confirm the results. 

A note on Test 3: On the fifth lap, a camper pulled out and I had to go onto the shoulder to go around him. On the sixth lap, he was still there. All of that is visible in the virtual elevation plot. Now I know to redo the test if that happens again.

Crr on Gravel

My testing qualitatively showed that my set-up at 30 psi had less rolling resistance than 40 psi or 50 psi.  My tires are tubeless Panaracer GravelKing SS TLC 40 (40-622) (700x38c) tires. I use Silca Sealant. My total weight (biker + bike) is 71.7 kg.

BicycleRollingResistance.com has tested these tires.  They show that the rolling resistance decreases as psi increases....the opposite of what I found in real life!! Of course, their testing is done on a drum that mimics a paved road at 18 mph and 94 lbs load and 70-73 °F. Their Crr are 0.00498 at 54 psi, 0.00528 @ 45 psi, 0.00576 @ 36 psi, and 0.00674 @ 27 psi. You would expect the Crrs to be much higher on a rough surface like gravel.

The Silca Tire Pressure Calculator suggests (for my weight and tire) about 43 psi for Category 1 gravel (well-packed), 40 psi for Category 2 Gravel (not packed), 35 psi for Category 3 gravel (very rough) and about 31 psi for Category 4 gravel (off-road). This is getting closer to the pressures suggested by my testing on what I would call Category 2 Gravel.

My data are not really good enough to quantify a value for Crr for my conditions, but we know it is greater than 0.005.

CdA on a Gravel Bike

The range of CdAs for cyclists vary from 0.2 m² (time-trialers) to over 0.7 m² (riding upright). 

The Chung Method may not always be able to quantify the value of CdA, but one can visually see a change in CdA from one set-up to another.  In Tests 4,5,6, all conditions remained the same except for body position as seen below.

I have also used AeroTune to measure CdA on a timetrial bike and find the Chung method not only easier to perform, but much easier to get a visual feel for the accuracy of the results.

Test 4 - On Hoods

Test 5 - In the drops

Test 6 - Tuck with hands on bar

Part II addresses some of the flaws in this study.

Andrew Campbell b1747 - YDNA

I have a special fascination with an "Andrew Campbell" who served in the Continental Army for all 8 years of the Revolutionary War.

I wrote an article about him for the Journal of the Clan Campbell Society of North America. 

I blogged about his military discharge at Snake Hill in New Windsor, New York.  

I even started an autosomal DNA project to find my improbable 7th cousins. 

I searched for Andrew's descendants and (more importantly) I searched for people doing the same.

It has been over 10 years since the CCSNA article. It has been a long slow journey, a journey that continues. This week marked a huge step in finding out ... Who is "Andrew Campbell?"

As recently as 2020, I believed that Andrew had no paternal descendants. However, another family researcher, Debi L., continued to uncover Andrew's children and their descendants. A few years ago, she found a living paternal descendant.  It took me another few years to make the contact and make arrangements for a YDNA test with Family Tree DNA.  The test is now "in the mail."  Results are expected in a few months.

Why is this so interesting to me and to other descendants of Joel and Andrew? Currently the parentage of Andrew is unknown. His geographical proximity to Joel in 1775 makes a relationship more than a remote possibility.  Joel was in the same local militia company in Hanover Precinct, Ulster County, NY. 

Andrew was in the minuteman company of Captain Peter Hill of Hanover. Joel is enumerated on the same page as Captain Peter Hill in the 1790 census. [The Campbells and the Hills had lived in the area since the 1760s.]  Peter was the son of Nathaniel Hill referred to in a prior blog.  His home still stands.

Andrew also served in the Continental company of Captain William Jackson of Hanover Precinct (as he testified for his pension application in 1818.) Nathan, Samuel, and Robert Campbell were also in a militia company commanded by Jackson the same year, prior to Jackson raising a company in the Continentals. Nathan and Samuel were brothers of Joel. 

Was Andrew a nephew? (son of an older sibling of Joel?) Or perhaps a 2nd cousin (a grandson of a brother of Joel's father.) Or possibly related to the Alexander Campbell family living in that area in the 1760s who were part of the Lachlan Campbell immigration party of 1739...and unrelated. YDNA will answer most of those questions.

Here is the genealogy from the father of the Tester to Andrew:

Arthur L Campbell 1911-1989
Richard James Campbell 1886-1946
Albert S Campbell 1842-1918
Richard Campbell 1798-1869
Andrew Campbell 1747-1833