Sunday, September 14, 2014

cycling in Basel, Switzerland: Gempen climb from Dornach

On Friday afternoon, I decided to try again for a paved climb up to Gempen. I wanted to climb the road I'd descended on Thursday, the one I'd seen so many riders climbing. My nominal goal was a Strava top 10. I thought this should be attainable from the VAMs I saw on the leaderboard if I was feeling good.

I decided to head out along the Rhein, crossing the river near where I was staying to the northern shore (German-side, but not Germany here), then heading east, crossing again near where the Birs river feeds into the Rhein from the south. Then I'd head southward to Dornach and begin the climb from there. I once again used Strava route-find, this time putting constraint points at the base of the climb and at the river crossing where I planned to recross the Rhein southward.

I was mostly able to follow the thus designed route, and it was again an okay route, but not as scenic as I'd hope. That bridge crossing wasn't great, with some slightly dicey navigation immediately following, but I made it fine into Dornach. Once there I clearly missed a turn because I found myself climbing a decently steep residential road. I viewed this as good warm-up for the main climb, and the power numbers I was seeing were fine for my perceived exertion, despite having done a middle-distance run after my ride the day before. At the top of this climb I stopped to check my position, and realized I'd taken an alternate, hillier route which got me with minimal added distance to the base of the main climb, which I could reach from this point via a short descent. So it worked out fine.

So there I was, at the outer edge of Dornach. The climb itself didn't begin immediately, the road starting generally flat before gradually transitioning into pure climbing mode. I wasn't sure where the segment began, so I needed to begin at a decent yet fully sustainable tempo before ramping up the effort only when I was sure I was on the segment. I expected the climb to take around 14 minutes.



Here's the profile, along with the grade. Comparing to Old La Honda, it's very similar in distance, but it gains close to 100 meters less. From where I hit the lap button (where I felt like the real climbing began) to where the grade finally drops toward zero is 5.2 km and 295 meters, an average grade of 5.7%. In contrast Old La Honda gains 394 meters in 5.4 km, an average grade of 7.3%. So the climb is clearly easier.

The day before there had been two weed-trimmers working the fringe of the road, and today they were still there. They presented no delay, however, as one passed me going the opposite way, and for the other I was waved pass in the opposite lane. Vehicle traffic was otherwise very light: just a few cars. The temperature was moderately warm. I was wearing my shorts, a wool undershirt I didn't need, my jersey, gloves, my Poc helmet, and a Poc cycling cap under that. Despite being a bit overdressed, I was fine, however. I'd rather be too warm than too cold.

But for some reason my power just wasn't where it's been this year on Old La Honda. I pushed and pushed but my legs just wouldn't let me keep it there. As I passed ten minutes, I tried to ramp it up a bit, not knowing when I'd reach the top. But as 13, then 14 minutes passed I still wasn't in sight of the buildings marking the outer fringes of Gempen, and the end of the segment. FInally I saw them, and the grade relented a bit as I approached. I'd been "ramping up" the effort for too long at this point, my lap timer now exceeding 16 minutes, and I was unable to avoid letting the power drop with the reduced grade. The penalty for doing so isn't as severe on gradual slopes, however, and I reached what I knew was the end.

The road continued to climb through the town, and on this portion I rode at strictly recovery mode, feeling fatigued from my effort. I knew from my time I was going to be disappointed with my Strava ranking. Why the bad legs? Loss of fitness from insufficient riding in the very busy weeks leading into my last week before leaving for Switzerland? Having run each of the preceding two days with insufficient running base? Residual jet lag after having arrived only the Monday 4 days prior?

I didn't think jet lag was a large factor at this point, but the running and fitness loss likely were. In any case, it was what it was.


Here's my power during the climb, comparing with what my power's been in recent Old La Hondas done on the Wednesday noon ride. You can see the rate of power loss during the climb is similar in all cases, it's just that the power on the Swiss climb is lower. The result was a 42nd place on one of the Strava segments for the climb, 51st for the slightly more extended one I'd been targeting (I lost rank on the reduced grade portion where my power had slackened). I was around a minute over my target 10th place. To reach 10th, I'd have needed to not only close the gap to my Old La Honda powers, but to go a bit higher still, as could be expected for a climb of shorter duration. With how my legs had felt on that climb that simply wasn't going to happen.

I'll have another chance, at least, before I leave Switzerland. Next week I'm working out of my company's Zurich office. Hopefully I can work in some decent runs that week despite my relatively long Basel - Zurich - Basel commute. This isn't a "riding" or even "training" trip, but hopefully I can find a way to at least retain what limited fitness I had. Indeed since my 50 km race in April, this year has been a massive downer on the training and fitness front due to work and moving. The goal on this trip is to get myself pointed in the right direction.

Saturday, September 13, 2014

cycling in Basel, Switzerland: Strava route-find and a Gempen dirt climb

As I write this, I'm on a TGV waiting to leave the Paris Gare de Lyon station for Zurich. I get off in Basel. The total trip is little more than 3 hours. Really remarkable. It's considerably more distance than the train between Santa Barbara and "San Francisco" (really Emeryville), the Coastal Starlight which is a 1-a-day affair in each direction. Delays of several hours are common. The TGV, on the other hand, is of course precisly on time. Indeed the entire timing board in Gare de Lyon was "á l'heure".

Prior to the weekend in Paris, I was in Basel last week. A search on Strava Segment Explorer turned up one obvious "local" source of climbs: the road to Gempen, a nearby hilltop town, small and rural, remarkably isolated from the Basel semi-sprawl.

There's two main paved climbs from Basel to Gempen. The most heavily trafficked, by far, is the 2-lane road from Dornach to the west. Dornach is somewhat inland from the Rhein, and reaching it from central Basel isn't too bad, given the extensive bike infrastructure including bike paths, lanes, and bike-specific signals. But navigating off my iPhone without a corresponding mount for my Ritchey Breakaway meant I had to frequently stop, remove my phone, unlock it, and check my position on the Strava app. It was slow going, but if I'd known the way it would have been better. Certainly it was no worse than riding my bike across San Francisco, which on routes I know well I find quite enjoyable.

A shorter way is a climb which begins further to the north, in Muttenz, gains a bit more vertical on the climb proper, but is on a narrow, bidirectional one-lane road which is certainly scenic but which may demand a bit more caution than is optimal for producing an optimal effort. I didn't know this, however, when I requested from Strava route finder a way from where I'm staying in Basel and Gempen.

A nice feature I had thought route-finder had was, given one or more segments, to find routes which include the requested segments in the order given. This feature could certainly be provided as long as segments could be mapped to the roads database used by the route finder. This is a non-trivial if. Segment definition and segment matching know nothing about roads. They're just a list of longitude and latitude coordinate pairs to which Strava's segment-matching algorithm compares the longitude and latitude points measured for the activity. This is important because bikes aren't constrained, generally, to pre-defined roads. They can ride pretty much anywhere. However, the option of constraining segments to roads would be a very useful feature, as long as the road data are sufficiently reliable. It could improve the integrity of segment definitions for road rides. But this is a major digression.

So I'd done a route expecting Strava to take me on either of these two paved climbs. Since the resulting route clearly wasn't going anywhere close to Dornach, I assumed it was the northern route, not really looking at it so closely.

As I set off early Thursday afternoon from Basel, I had two options. One was the route loaded into my Edge 500. The other was to track the route on the Strava app. Despite past frustrations with the Edge navigation, the most infamous where it wasn't able to navigate me south on Highway 1 without multiple "off course: turn back" warnings, I was going to give that another try. If it works, it's by far the best option, since it sits on my stem and I can navigate without stopping. Unfortunately it never has really worked, but there's always the possibility of a first time.

But not this time: as soon as I turned on the Garmin I got a "battery low" warning. So off the bike it came. When I plugged it into the charger it read "2%". At some point I'd inadvertantly left it on and it had drained down before finally shutting itself off. So it was the Strava app to save the day.

The Strava app, on the other hand, doesn't seem to do any real navigation: it just draws your point history along with the route itself on the same map. However, this is fine in many instances. While it won't tell you to "turn left", "go straight", "approaching a climb", or whatever, the map clearly shows the route to be taken and the route taken so far.

The problem is I don't have an iPhone mount. Tour Magazine did a review of these in the February 2014 issue, which I coincidentally downloaded for another article. LifeProof was the top scorer, followed by a 3-way tie between Armor-X, Biologic, and Wahoo. I should really get one of these 4 to make app navigation more workable. Without it, I found myself frequently stopping, occasionally backtracking to reach turns I'd missed. My only big question about the case is whether I can avoid the phone requesting a security code when the phone shuts off. I want to be able to put the display back on just with a touch, and the iPhone 5S fingerprint matching is confused with sweaty fingers. But I can look more into that. (added later: Armor-X looks good)

route through Basel

One of these times was a turn I'd seem but rejected. Riding along a paved bike path, I was looking for a left turn which I knew to be forthcoming. I saw a dirt road to the left, and I rejected it. I went further, seeing a roughly paved road which immediately set off at a ferociously steep grade. Was this it? Ahead the path appeared to T-intersect into a busy street.

I stopped, pulled out my phone, entered my code after (as typical when riding) the fingerprint reader rejected me, then brought up the Strava app. Sure enough, it had intended for me to take the dirt road. The paved road extended for only a short distance before giving way to a second dirt road which intersected the first. Indeed the dirt went on for a considerable distance. Strava had routed me this way, a third option in addition to the two paved options I'd expected.



This wasn't my first similar experience with Strava route finder, but in the past it's been with roads and paths I've known, so it was immediately obvious upon first glance what it had done, and I was able to constrain the path to stick to pavement. I don't blame it -- I asked it to connect points A and B with cycling routes, judging the quality of the path based on distance and popularity, and as long as a sufficient number of mountain bikers take the dirt option, Strava considers that a viable candidate.

I like the dirt, so off I went. The riding was a gradual, long climb, with a somewhat rocky but non-rutted surface which was fine on the road bike. The quality of the experience was somewhat attenuated by the need to too-frequently check the phone at junctions.

One such junction was where the wide path extended ahead to a series of piles of cut logs. To the right, a narrower, rockier, steeper route climbed onward. I was intended to take this latter option.

I got into my 46/26 and climbed up a steep initial portion before the grade leveled out. But then I heard the unmistakable sound of descending mountain bikers. I pulled to the side of the road as first one then a second downhiller flew by. This didn't bother me: I wasn't in a rush.

But I couldn't get enough traction to restart. My rear wheel would slip on the rocks. I was glad I'd brought mountain pedals and shoes on this trip: my maiden voyage with BeBop pedals, which I really like. They're staying on the Breakaway -- with a low stack (requiring some Dremel surgury on the shoe pads to allow axle clearance part of the standard installation procedure), impressively low mass, and a float similar to Speedplays, the pedals are a great solution where some walking will be required or dirt clearance may be an issue. I'll stick to Speedplay for the racing bike.

So I walked a bit, which offended my "death before dismount" sensibilities, but was functional. And then I was to the paved road.

The Strava route had me turn right on the paved road, take that a bit, then take another dirt path which was to act as a short-cut for a less direct, but no doubt less steep, path climbed by the pavement. I decided to stick to the pavement.

The climbing was surprisingly steep, despite the relatively meandering route. But after more of this than I expected, I was in Gempen.

With all of the stopping to check directions, all of the tentative riding on dirt roads whose character I didn't know, and finally with one to two hundred meters of outright walking, I knew I wasn't going to set any ranking times on Strava this day. Yet as I finally rolled into Gempen, the small agricultural town, I decided I was done climbing for the ride. I descended the wide, open road to Dornach, a steady stream of riders coming the opposite direction, along with two large vehicles with weed-trimmers working the road fringe. This, I decided, I would climb the next day, with a real effort to get a decent Strava ranking.

I semi-improvised my route home, deciding instead of the relatively heavily trafficked route I'd taken out to ride toward the Rhein, then head back along that. This was partially successful, but also involved some decidedly less nice riding. I made it back fine, however.

Friday, September 12, 2014

Thursday, September 11, 2014

Fairwheel Bikes brake test: mechanical ratios

Fairwheel Bikes, the most technically engaged bike shop I know and one which has made an incredible impact for its size at numerous bike shows including Interbike and NAHBS, earlier this year published the results of a brake test on the WeightWeenies forum. They did a number of tests, including stiffness and braking force. Here I'll look at the brake deflection versus cable pull.

But first, here's the brakes in the test, with mass for the pair measured without brake pads or holders, and a price. No price is listed for the Campagnolo single-pivot, perhaps because it is typically sold paired with a dual-pivot brake (single for rear, dual for front):

KCNC C7111 grams$330
THM Fibula116 grams$1429
KCNC CB3126 grams$335
EE Cycleworks139 grams$610
KCNC CB4150 grams$200
KCNC C6161 grams$180
Far and Near168 grams$290
Sram Red190 grams$350
Campag Single pivot212 grams 
Shimano 7900216 grams$400
Shimano 9000218 grams$400
Campag Dual Pivot232 grams$355

Brake pads and holders would add around 50 grams.

They attached the brakes, optionally with Edge grey pads, to an apparatus which pulled on the cable by a specific amount. By not using brake levers, they avoided conflunding the results of the brake with the design of a specific lever. For brakes designed to work with a specific lever (for example, Campagnolo, Shimano, and SRAM are presumably all designed to work best with their specific levers), this may provide misleading comparisons. But other, 3rd-party brakes don't have matching levers, so this allows comparison of different brake options. Since Fairwheel is a shop which specializes is less common, typically extra-light parts, this was the best approach. But the results need to be interpreted in the context that the lever will also make a difference, and different levers have different mechanical characteristics.

Jan Heine, in Bicycling Quarterly, has done an incredible review of the history of brakes. One conclusion from that work was that some relatively recent models which have been considered novel were instead based on designs which existed long ago. We sometimes think the history of brakes began with the single-pivot calipers which dominated in the 1980's and into the early 1990's, but this is incorrect. More recently, direct-mount brakes have become increasingly popular, but direct-mount brakes have a long, long history, pre-dating the single-pivot design.

In any case, Fairwheel tested a variety of brakes of different design. Campagnolo has two designs, a dual pivot and a lighter single-pivot. Shimano, which started the long run of dual pivot dominance (at least on road racing bikes) with its 105 brake in the early 1990's, is dual pivot, as is Far & Near and KCNC. SRAM Red is a modified single-pivot design. EE Cycleworks is relatively complex multi-pivot design: Jan did a specific review of that brake in Bicycle Quarterly which I highly recommend, but I don't have access to that article to reference right now (on train from Basel to Zurich: my BQs are sorted on my bookshelf in San Francisco). I wish they'd included the Camillo Zero-Gravity brake, which I use, but that's considered generally inferior: it uses a cam-enhanced single-pivot design similar to the SRAM Red which followed it.

As Jan described in his Bicycle Quarterly article, brake design is a balance between range of motion (allowing the brakes to clear the rims with sufficient margin) and mechanical advantage. If I pull the lever a particular amount the force exerted on the pad is proportional to the ratio of the cable pull to the pad deflection. This is why disc brakes are considered powerful: they sit extremely close to the rotor and so relatively little deflection is required, providing a high ratio of cable pull to pad deflection.

A compromise is to use a variable rate of deflection. As the cable is initially pulled, the brake pad moves relatively rapidly, closing most of the gap to the rim. But as the pad gets close to contact, the rate of deflection with respect to cable motion decreases, providing greater mechanical advantage. This provides the best of both: plenty of clearance for out-of-true rims or removing wheels with tires whose profile extends beyond the rim width, while allowing for plenty of braking force once the brakes contact the rim.

A key here is that the mechanical advantage felt by the hands on the brake lever come from three sources. One is the conversion of lever motion to cable pull at the brake lever. The second is the conversion of pull at the lever to pull at the brake. This is ideally 1:1 unless the cable stretches. The third is the conversion of cable pull at the brake to brake pad deflection. This third stage consists of two parts: one is the unloaded conversion ratio, and the other is a reduction in this ratio due to bending of the brake with increasing load. Stiffer brakes will maintain closer to the unloaded ratio. The important thing is the product of these stages. So I can tune the characteristics of the mechanical advantage either at the brake or at the brake lever.

This test looks only at the brake. Here I plot the inverse of unloaded mechanical advantage, which is the ratio of the brake pad deflection to the cable pull. A high value means the pads are closing rapidly, but the braking force is proportionally low. A low value means the pads aren't moving much, but the ratio of braking force to lever force is higher. I plot some of the brakes tested by Fairwheel, omitting only the KCNC models:


One thing to note about this plot: Fairwheel measured the brakes at 3 mm, 6 mm, 9 mm, and 12 mm of cable pull. There is also an implied measurement at 0 mm cable pull when the brake is not in contact with a rim. To measure the marginal mechanical ratio, I'd take the slope between measurement points, for example between 0 and 3 mm, or between 3 mm and 6 mm. This results in a slope estimated at the average pull of the two points. So using 0 mm and 3 mm cable pulls, I get the slope estimated at 1.5 mm cable pull. So from their measurements I get ratios at 1.5 mm, 4.5 mm, 7.5 mm, and 10.5 mm.

Then in this plot I show the KCNC brakes. I separated them because there are four of them and including them would have cluttered the preceding plot:


You can see that most of the brakes have a relatively uniform rate of cable pull to brake pad deflection, while a few have more variable relationships. The deflection axis here is logarithmic, so a given distance on the axis corresponds to a different ratio, as opposed to a different absolute difference, in deflection ratio.

The more variable relationships come from the Campagnolo single-pivot an the Shimano 9000 brakes. The Shimano 9000 has a relatively broad range of relatively uniform ratio between 3 and 9 mm of pull. The Campagnolo brake is more variable, however, with no real uniform range.

It may seem like a good idea to engineer the variable deflection into the brake. However, a disadvantage of doing this is that if the brakes are set up so the sweet spot of optimal mechanical advantage comes with the pads just touching the rim, when the brake pads wear the brakes will need to be deflected more to initiate braking. Either the "sweet spot" needs to be extended to accommodate a range of rim widths and brake pad thicknesses, which reduces the available clearance, or the braking quality will vary from optimal.

If instead you put the mechanical tuning in the brake lever, then when brake pads wear or a wheel with a different rim width (measured at the braking surface) is installed then all you need to do is rotate the barrel adjuster and you've restored the lever range of motion. The mechanical characteristics are thus in tune again.

So a uniform mechanical response is the most resiliant approach.

To analyze this, I took ratios as plotted (incremental ratios). The mean mechanical advantage is considered by looking at the average such ratio. The variability in the mechanical advantage I evaluated using the statistical standard deviation of the natural logarithm of the slopes. I took the natural logarithm because I'm interested in the fractional change of the mechanical ratio, not the absolute change.

Here's the sorted result:

brakevariabilityavg. ratio
Thm Fibula0.03100.5800
Sram Red0.03600.6075
Shimano 79000.04990.5267
Far and Near0.05360.6550
KCNC C60.06590.7117
KCNC CB40.06720.7900
KCNC CB30.06950.7900
Campag Super Record Dual0.07200.7050
KCNC C70.08260.6333
Campag Super Record Single0.10550.9150
Shimano 90000.13010.6167

So the Fibula, which is also the 2nd lightest of the brakes tested (116 grams compared to 111 grams for the KCNC C7), also has the most uniform mechanical ratio by a fairly good margin. Only Sram Red comes within 20%. The Shimano 9000 and the Campagnolo Super Record single pivot are the most variable, although as I noted the Shimano at least has a relatively uniform range mid-deflection, while the Campagnolo brake is variable through its range to the resolution of the test.

The EE brake is an interesting case. It scores only mid-range in ratio variability. But notice that the variability is due to the first 3 mm of cable pull. The brake has excellent linearity from 3 mm to 12 mm, as seen in the plot. This provides additional clearance when the brake is open relative to, for example, the THM Fibula or Shimano 9000 brakes. So the brake is fine, arguably better with this design.

The average ratio is is mathematically the same as the ratio of deflection with 12 mm of cable pull to that 12 mm cable pull. I show the same data ranked by this:

brakevariabilityavg. ratio
Shimano 79000.04990.5267
Thm Fibula0.03100.5800
Sram Red0.03600.6075
Shimano 90000.13010.6167
KCNC C70.08260.6333
Far and Near0.05360.6550
Campag Super Record Dual0.07200.7050
KCNC C60.06590.7117
KCNC CB30.06950.7900
KCNC CB40.06720.7900
Campag Super Record Single0.10550.9150

The Shimano 7900 has the least pad motion followed by the THM Fibula, Sram Red, EE, and Shimano 9000 all fairly tightly bunched behind. All of these brakes have an excellent reputation for stopping power, so it's clear the mechanical ratio is significant here. The Campagnolo Super Record single-pivot is a solid last place in this ranking, with KCNC brakes taking the 3 spots immediately behind (the CB4, CB3, and C6).

One note: these are unloaded deflection tests, and this translates to force only if the brakes are perfectly stiff, which they are not. Fairwheel also did several force-based tests but since I don't fully understand the results I won't comment too much on them here. But basically the actual brake performance under loaded braking may be different.

So what do I conclude from this? It seems that if you want a lot of stopping power, the Shimano 7900 is the top choice, but without much margin to clear the rim. The THM Fibula, EE, and Sram Red have a bit less mechanical advantage but more clearance. The EE gains a bit extra clearance by having an increased pad-to-cable motion ratio during the initial stage of motion. Since my rims aren't always perfectly true I like that, and so if I had to pick one of these curves, I'd probably go with the EE. But the super-low mass combined with a very uniform ratio through its range of motion make the Fibula also an attractive choice.

Tuesday, September 2, 2014

Tour magazine frame mass tests: carbon vs titanium vs steel

Here's a comparison of Tour magazine tested mass data for medium-to-large frames taken from three articles:

  1. steel bikes, Aug 2013, on-line summary
  2. "sub-800 gram" carbon bikes, Dec 2013, on-line summary
  3. titanium bikes, Sept 2014, purchased today in print version


This sample is biased in favor of the carbon since the carbon frames selected are the lightest, most expensive available, while in steel, for example, a custom builder like Rob English, emphasizing weight, can get lower than most stock steel frames, which have less emphasis on weight. But I think the Crema represents this quite well: only 1512 g for the frame.

Forks are heavier on the steel and Ti bikes, in general, but I am focusing on frames, since forks can be selected, especially with custom steel, which is still a lot cheaper than those carbon bikes.

Result: best carbon is 704, best Ti is 1350, and best steel is 1512. So the cost of Ti is around 646 grams, and the cost of steel around 808 grams, relative to carbon.

808 grams: that's not bad at all. Still within typical daily body mass fluctuations.

Monday, September 1, 2014

Fit change from Cervelo R to 2015 Cervelo S5 geometry

Cervelo's new geometry for the S5 is a move toward the industry standard for stack and reach, with shortened head tubes reducing the stack and increasing the reach of each size.

I decided to see what this new geometry did for my fit on the bikes. Would I end up with a hideous tower of spacers?


Here's a chart I posted before, with Cervelo S5 2015 replacing the Cannondale Evo. The Cervelo S5 and Evo geometries are fairly close, and so I removed the Evo curve to reduce color. I also show the Trek H1, an example of an aggressive geometry, and the Fuji SL1, which is my bike.

My bike is small, with an 11 cm +6 degree stem establishing the position set for me by 3D Bike Fit in San Francisco. Then from those coordinates (to the right of the plot), I trace back to the frame using a variety of stem options, using the head tube angles of the Cervelo R-series bikes.

With the R-series (also S2-3, and also the 2014 S5), I could ride a 48 cm frame with a 13 cm +6 degree stem, a 51 cm frame with a -6 degree 11.5 cm frame (adding 1 cm of spacers), or a 54 cm frame with a -17 degree, 10 cm stem (with around 5 mm of added spacers). The 51 cm is the most likely candidate here.

To the S5, shown with dotted lines: now the 51 cm would require an 11 cm, +6 degree stem, removing around 5 mm of my present 1 cm stack height. But with the 54 cm frame, I'd be very close to my present fit using a 10 cm -6 degree stem. It would be even closer if I'd adjust for the fact that the 54 cm has a 0.9 deg steeper head tube and the displayed stem trajectory is just slightly too steep (upward).

The result: with a -6 degree stem, I can use the 51 cm R-series with an 11.5 cm stem, or the 54 cm S-5 with a 10 cm stem. Personally I'd prefer the 10 cm stem.

So really no spacer difference is involved: to the contrary, in fact. With the S5 I end up with fewer spacers than I would have for the R-series. And the reason is with the S5 I can upsize to the 54 cm frame with a -6 degree stem.