Wind Tunnel Methodology

Ground Control to Major Tom.The Rail prototype did extremely well in the wind tunnel.  Our hope going in was that it would be aerodynamically close enough to the 404FC that with the wider inner brake track width, lower profile, and lighter weight, the Rail could be able to make a compelling case for itself - economics aside - as one of the best all around wheels on the market.  I'm in a bit of a state of shock after learning that that's exactly what the test results show. 

We put a fair amount of pressure on ourselves by making a big deal about our trip to the wind tunnel before we went.  Hypotheses precede testing, claims follow it.  As I recently told Mike, "until you know, you don't know, and if you don't know, you can't claim anything."  While we hadn't made any claims before the trip, it might have been a bit smarter to go there in stealth mode in case the results stunk, but where's the fun in that anyway?  Failing's no fun unless you get to look like an idiot in public while doing it.

The real risk we had in front of us was what we've come to regard as "supplier risk."  Supplier risk means simply the risks to which we were exposed in the event that the test went badly.  Yes, we'd look a little stupid, and perhaps our credibility would have taken a knock.  More importantly, we have a significant amount of time and money into this project, which would more or less go down the drain with a bad result.  I didn't have any "better design" or alternate take up my sleeve; what we sent down there was my final answer.  A bad test would at the very least have meant a big delay in bringing this thing to the market, and losing time on the market in a performance driven segment is never a good thing. 

Of course we could have sidestepped that element of "supplier risk" by not testing, assuming but not confirming that our hypotheses would be borne out in the tunnel, and gone ahead on that basis.  That would have introduced what we have come to think of as "consumer risk": that you would buy a product based on assumptions that haven't been objectively proven, with the risk of it not performing up to claims borne by the people who bought it.  Prototyping and wind tunnel testing are expensive and time consuming, but in this day and age they are accessable enough that you must do them.  Until you know, you don't know, and if you don't know, you can't claim anything.  You must know. There is still development and testing going on, we've got to ensure (and test, and therefore prove) that braking and heat management are good, and that the wheels are stiff.  Those all still reside firmly in the venue of "supplier risk" and we will not transfer them to "consumer risk."

But you're all here now because we were in the wind tunnel yesterday, so let me go through that a bit. The reason why we sent a 404FC to be tested among this batch is simply that transferability between tests is problematic at best.  We had planned to send each wheel with a tire already installed, and then we were advised that even different tires of the same exact make and model can skew results - the same exact tire should be used on each wheel.  The benchmark needed to be as exact as possible. We now know how the Rail compares aerodynamically to the 404, and each of our other wheels, under conditions that could not be replicated more closely. 

If the difference between two of the exact same tires could skew the results, you can imagine how introducing different variables like different types and widths of tires could throw things out of whack.  We knew we were sending a worst case down there on behalf of the Rail, as what we sent was an unfinished prototype with extra spokes compared to what the production version would have.  While more spokes is never aerodynamically better than fewer, the magnitude of the effect is unknown.  On one wheel, taking out four or six spokes might have a huge impact while on another it might have none.  We sent two 58s down there, one with 20 spokes and one with 24, to act as something of a learning point on what going from 20 to 24 spokes might do, but we had no plan to rely on the over-spoked 58 as a proxy for the over-spoked Rail.  While it might be tempting to do so, we aren't going to imply that the increased drag from the extra spokes on the 58s is directly transferable to the Rail.  That would constitute consumer risk.  Instead, we will re-test the Rail with the production spoke count (and we will re-test the 404FC at the same time to maintain that control baseline as accurately as possible) and then and only then will we make any claims about how the final Rail compares to the 24 hole prototype.

At this point, I've looked at so many aero graphs that I can somewhat reliably pick a 404FC out of a lineup based on the shape of its curve. Sometimes it is higher up on the y-axis, sometimes it's lower, but the shape is generally very recognizable.  The 404FC is the lingua franca aero baseline wheel, which is why we chose to send it with our test.  If the shape it produced in our test was markedly different than how it usually looks, even though we used a 23mm tire that we think more accurately represents what people actually want to use instead of the 21mm for which it's optimized, that would have caused us some doubt about our baseline's validity.  As we plotted out yesterday's results, I was very happy to see that familiar pattern start to appear.  The absolute values along the y-axis may or may not be exactly what they are in other comparisons, but we know that the shape is good and therefore we have a good, if necessarily imperfect, idea about how we stack up among other wheels.  As I said at the outset, we're just shy of over the moon with where that is.  

We learned a lot from this wind tunnel experience, and we owe a big thanks to Dave Salazar at A2 for his help.  I know a lot of people are chewing on the bit for us to post results and pretty graphs, which we will do soon enough (days not weeks, and well before we offer it for sale), but before we do that we need to make it as clear as we can what, and how much, the results of this round mean. 

9 Wheels and No Cameras

As of Tuesday of next week, we will be able to claim that all of our wheels are "wind tunnel tested." Does that mean we can raise our prices?

Our wheels are not at the wind tunnel for marketing purposes though. Believe it or not, this trip is motivated purely by R&D. We've got 9 wheels in boxes that just arrived at A2 in North Carolina, with nary a camera in sight.* To let you know what precisely we are testing, I'll start with the lineup of wheels we are sending (all of these are front wheels):

1. Canary Thunder, our yellow solid plastic prototype of the Rail, built with 24 radial laced CX-Ray spokes (more on spoke count in a bit) 
2. RFSC 85 w/ 20 CX-Ray spokes
3. RFSC 58 w/ 24 CX-Ray spokes
4. RFSC 58 w/ 20 CX-Ray spokes
5. RFSC 50 w/ 20 CX-Ray spokes
6. RFSC 38 w/ 20 CX-Ray spokes
7. RFSC 38 w/ 20 Laser spokes
8. FSW 23 2/ 20 CX-Ray spokes
9. Zipp 404 Firecrest Carbon Clincher w/ 16 bladed spokes

Every wheel will be tested with the same 23mm Vittoria Corsa Evo CX tire rotating in the same direction. A single tire will be transferred from wheel to wheel to eliminate any variability that might result from variations in the tires, even if they are the same make and model. 

Curiously, Canary Thunder has 24 spokes. No, we do not intend to go to market with that many, but have you ever tried building a wheel with a 20 pound solid plastic rim? We get one bite at this apple and the extra spokes help ensure the wheel will remain true at the tire pressure and rotational speed needed for the testing. We realize it will skew our data to the unfavorable side, but we're fine with that as we're looking to get smarter in the tunnel, not proclaim ourselves the victor when we emerge.

You will see also that we have a pair of RFSC 58s, one with 20 and the other 24 spokes. Testing both of these will give us some insight into how much drag is attritutable to 4 extra spokes. Everyone assumes fewer spokes are less drag - won't it be nice to know by how much? We think so. Testing 20H vs 24H will also help us normalize the testing of Canary Thunder by extrapolating the data to suggest how CT would perform with 20 spokes (and start to build a hypothesis about what 18 would look like).

We're testing every depth of our carbon rim for two reasons: 1) we want to be able to give quantifiable response when someone asks what is the difference between this depth and that. We can say that the shallower rim sheds a certain number of grams, and by providing the aerodynamic delta as well we think our customers will be better able to calculate their personal trade-off between speed and light weight; 2) we are not fully convinced a range of 4 (or even 3) different wheel depths is warranted. As we move from open mold to proprietary designs, we will likely remove offerings that offer variety with no real additional value. 

Note that we also have a pair of 38s in there - one with CX-Rays and the other with the less expensive but same weight Lasers. CX-Rays are provided standard on a lot of wheels because of their purported aerodynamic benefit. How much is that benefit? Does it justify the additional expense? Let's find out.

Our FSW in our new 20H front skinny build also made the trip. We sent the 20H so we can compare it to comparably built carbons in different depths. It made the cut in part for baseline purposes. Most of the testing you see uses an alloy wheel as a baseline. But that wheel is usually a 32H Mavic Open Pro (front, no less!) that is about as suitable a benchmark for a performance wheel as Toyota Sienna is for a sports car. Our FSW is an outstanding wheel, but is not what anybody would consider aerodynamically remarkable. So let's see how all the carbons stack up against a racing quality wheel that isn't even trying to be fast. (And in the process, let's see how fast the FSW is in its own right.)

Finally, we've sent a Zipp 404 Firecrest carbon clincher. There are a host of reasons for this choice alone, most of which have to do with benchmarking - our current wheels against a Zipp with the same tire (and not the 21mm Zipp uses in their own testing but nobody uses on the road), our Zipp data against everyone else's Zipp data, and our new proprietary design against Zipp's recognized shape. We have no idea what to expect, but also nothing to fear. Does anybody expect our wheel to beat Zipp's? Spoiler alert - with 24 spokes instead of Zipp's 16 and a rough plastic surface that is not nearly as slippery as smooth carbon, our prototype will not be faster than their production wheelset. But this is a training race for us, not one where we expect to win. We just want to see how much work we have to do, and gauge our design's potential.

So while Canary Thunder is the impetus for this trip, we hope to learn a lot about how much a difference additional spokes, rim depth and spoke shape make, completely independent of whether we are on target with our prototype. We'll share what we learn here.

*Not entirely true. We asked the guy who runs the bicycle business at A2 and who will be conducting our tests to snap a couple of pictures of "Canary Thunder" with his phone.

Proprietary Rims and Wind Tunnel Objectives

Road trip!This morning, we're packing up a whole lot of wheels to spend some time at wind tunnel camp, at the A2 Wind Tunnel in NC.  We're sending every one of the clincher wheels we currently offer, as well as "canary thunder," the pretty yellow prototype of our proprietary rim design.  The wind tunnel is a significant step for us, in many ways.

We've questioned the validity of how other companies use the wind tunnel, specifically in Mike's post of about a year ago.  You see glamor shots of finished products in the wind tunnel, and you think "why are they sending finished product to the wind tunnel?  If they learn it stinks are they going to scrap the mold?"  We've got our current product line going down there because we want to learn more about it, and we want to be able to benchmark all of it.  We're also sending the wheel that's the de facto aero standard, which will by extension allow us to compare our wheels against pretty much any others out there.  If we find out we've got a clear winner (see below for what that means) on our hands, hooray - we'll have the "finished products in the wind tunnel" photos soon enough.  If we've got a lemon, not just in color but performance, you'll see Tweety Bird mk 2 flapping into that tunnel before long.  This isn't just some exercise in window dressing.  It's important for both you and us to know what kind of performance we're delivering.  If we were going to just pull some fancy performance claims out of our butt, we'd save ourselves the significant cost of this exercise. 

One way in which our wind tunnel aims differ from a lot of other companies is that if other companies produce less than the provable best, walking out without that graph that puts their wheel's line bottom most of all that are out there, they've failed.  It's kind of amazing how many have actually walked out with that golden graph line, but we'll leave that be for now.  In any case, we don't go there needing to stake a claim to that bottom-most line, for a couple of reasons. 

First, aerodynamic competitiveness is a leg of our triangle, but it isn't our triangle.  We also want to avoid going super deep for reasons of weight and practicality.  There's a lot of value in having one really great wheel set, and as we talked about before, we feel that the 50mm neighborhood is a very effective place to be in that regard.  If we're in the aerodynamics class of good wheels that are 10mm deeper, we'll be quite happy with that.  We are also married to the idea of our internal width, as we're absolutely convinced of the benefit, so if the magical aerodynamics fairy was to say "you could show a 3 watt improvement at angles of attack between 5 and 10 degrees if you made the whole thing 6mm skinnier" we'd take a pass.  We need to have good aerodynamics, and if the wind tunnel shows that our design does not have very good aerodynamics, it's back to the drawing board.

Second, it's sort of a crazy game to play in getting to the point where your wheels are demonstrably the aerodynamic best.  There's a cost to benefit thing that goes on, where it becomes more and more likely to geometrically increase your investment to incrementally increase the aerodynamic performance.  How much is a watt worth?  In terms of time, it's worth about 3 seconds over a 40k TT in constant conditions.  In terms of drag, it's worth about 10 grams the way it's figured in the bike world.  In terms of money, for some people it's invaluable, priceless even.  It trumps everything.  For a lot of other people, once you get to where something is quantitatively not giving much, if anything, away from best in aerodynamic class wheels, and taking some back in terms of handling, practicality, weight and build quality, that's a big fat win. 

Developing your own proprietary rim is a challenging and resource intensive proposition. It's taken a lot to get to where we are with this, but at this point the proctor has called "pencils down." We're passing our blue book forward, confident that we've earned a good grade, but we won't know until we know. And if that sounded hopelessly anachronistic to everyone who graduated college after the turn of the millenium, I apologize. 

A Blank Sheet Of Paper

Mike described a huge part of the design brief for the Rail in the last entry, and the part that he described really can't be minimized.  When I started racing my mountain bike more, I quickly realized that my biggest gains weren't to be made in going faster, they were to be made in not slowing down.  Guys were killing me in the turns, which was a double bonus to them - they could gap me in the turn, and then since they were holding more speed out of the turn they'd extend the gap after the turn, while I burnt matches to accelerate to their speed.  Handling counts for a lot, on every surface. 

We also thought it was critical to optimize around a 23mm tire.  A lot of wheels out there (Zipp 404, for example) are optimized around 21mm tires, and have a narrower inside and outside dimension at the tire hook.  All of their, and seemingly everyone's, aero data is presented based on testing with a 21mm tire.  Everyone we know rides on 23mm tires (many even ride 25s, which becomes somewhat superfluous when you've got a wide enough inside dimension), 23mm tires are most widely available (both variety of offerings and availability for purchase), and several studies have shown that wider tires (up to a point) have lower rolling resistance.  We knew how well the 18mm inside dimension with a 23mm tire handled, and a whole lot of other people agreed with us, so optimizing around the 18mm inside width and 23mm tire combination made absolute sense to us.  Of course our baseline raw drag data from the wind tunnel will probably appear a bit higher than other manufacturers show, because they all test with 21s.  To help you (and us) make more accurate comparisons, we're testing a benchmark wheel with 23s instead of the 21s around which they're optimized, but which no one uses. 

Once we knew that we wanted to wind up with 18mm for the inside width, the next parameter to nail down was depth.  We chose 52mm as a start point for a couple of reasons.  First, there is a significant aerodynamic threshold right around there.  Ever thought why the UCI is such sticklers about shapes that are more than 3:1 deep?  Well, given an 18mm inside width, we knew we'd wind up at 25mm for the outside width.  25 times 3 is 75.  A 23mm tire plus a 52mm rim is 75.   Makes sense?

If 52 gets you right to that threshold, why didn't we go to 60 or 70?  First, weight.  Added depth costs weight, and we didn't want to wind up with good aerodynamics at the expense of practicality on different courses.  Many of the new rims are showing that there are only very small gains to be made with deeper rims, enough so that rims much beyond the 50mm range are really being pushed into the "special use" category (pretty much tt only).  Second, even though the shapes we were investigating promised a lot of gains in crosswind stability (more on this in a bit), at some point surface area is surface area, and that turns into a liability in windy conditions.  So we figured 52 would get us to that magical nexus of three inflection points. 

Now we had width and depth, but what about shape?  Fortunately, I'd worked with a lot of foil shapes before, and there's a ton of transferability between sailing shapes and rim shapes.  Step 1 was to tap into the (extrememly public) database of NACA shapes.  So much of innovation seems to be finding new ways to apply old knowledge, and NACA is nothing if not old knowledge - the agency opened its doors in 1915.  It is also, in fact, rocket science.  I skipped straight to the 4 digit, 00XX series sections, as I knew that the 4 digit series sections would put the maximum thickness in about the right place along the chord, and that I wanted a symmetrical foil (side to side) which is what the 00XX sections have.  From there, it was a game of figuring out ratios.  Since both the tire and the rim are both leading and trailing edges, I wanted to try and get them roughly symmetrical.  That was pretty easy to do with the 0024 section.  Oddly enough, just days after we'd really committed to that route, Mavic came out with their new 80mm rim, which also relied on the 0024 section.  A classic "good news, bad news" situation - clearly our choice had a lot of validity, but we'd also be accused of being serious copycats.  Fortunately (at least we hope), we slice and diced it such that our setup would replicate an 0024 with both tire and rim, where their tire side goes 0024 but their rim side goes 0011.  Their rim is less blunt than ours - more on this in a bit. 

NACA 0024 foil plot

Speaking of rim bluntness, I don't think it's necessarily possible to accurately model steering response as a function of rim shape, but again I was fortunate to bring some sailing experience to bear.  I've trimmed A LOT of sails, and part of what makes the good guys good at it is the ability to match the conditions with the needed shape. The illustration below, looking down from above at cross sections of sails, helps make the point.  When conditions are more turbulent or challenging (for example, when the waves are steep), or when the person driving the boat maybe isn't quite world class, you give a deeper, more forgiving entry shape.  On the other hand, in more ideal conditions, you set up a narrower entry angle. 

image courtesy UK Sails

The sharper entry angle will be faster in ideal conditions, but it is far more prone to stall.  In the case of a rim with a narrow entry, the flow can switch from side to side, which would be a jarring input.  As I explored in a post about TdF TT equipment this summer, a narrow rim like the HED 3 spoke does very well in low apparent wind angles.  Mavic choosing the 0011 shape for their 80mm rim makes sense, as it's more of a dedicated TT wheel, which will operate most often at the low yaw angles where sharper, narrower rims are going to be relatively strongest.  With a sail, you have the basic shape and then a bunch of controls that help you modify and manipulate that shape to suit the conditions.  Obviously you can't do that with a rim, so a forgiving all purpose shape works well for an all purpose wheel. 

This is a seriously condensed (and yet challengingly detailed - thanks for following along this far) version of the design process, but it gives a good sense of how setting an initial parameter made the rest fall in line.  Now we have a yellow plastic manifestation of that figuring, which we are about to take to North Carolina to have wind blasted at it so that we can know how fast it is or isn't.  We'll write all about that experience in a couple of weeks when we've got the results and have had time to digest them. 

First Look at the Rail, Our Proprietary Design Wide Carbon Clincher

There is lots to digest from that title alone. Let's start with the name of our new wide carbon clincher - the Rail. The phrase "corners like it was on rails" is as ubiquitous to high end bike reviews today as "laterally stiff, vertically compliant" was a few years ago. The difference is that it actually means something to racers. Cornering matters. If you do it with confidence and aplomb, it's free speed. If you're unsure or cautious, gaps open that you have to burn matches to close. We're calling our new wheel the Rail because it is designed not just to excel in the wind tunnel, but in the real world as well. Racing is accelerating, holding speed on the flats, climbing, bombing descents and railing corners. When we thought about what attributes we wanted to design into a new racing wheelset, we kept all of this in mind. The Rail is (we think) the first wheel optimized for every part of the race course. 

That leads us to the next part of the title of this blog I want to focus on - Wide. What is a wide rim? The most common answer is 23mm. But if you think about the attributes of a wide rim - that it results in less casing deformation and broadens the tire's contact patch on the road, increasing road feel and traction - it becomes clear that the width that matters is not the brake track width, but the inside width. The inside width of the Rail is 18mm. By comparison (and not coincidence) the inside width of the Velocity A23 rims we use in our alloy FSWs is also 18mm. We positively love the road feel of good rubber on a set of our FSWs, and deliberately modeled the Rail to replicate that feel with the same inside width.

But alloy brake tracks are thinner than those on carbon clinchers. In order to achieve an 18mm inside width, we needed to expand the brake track to 25mm wide. So technically the Rail is 25mm wide at the brake track, but when we say "Wide" we're actually referring to the 18mm wide on the inside. That's the wide that matters.

By way of reference, here are the brake track and inside widths of some other popular clincher rims:

RIM: Brake Track / Inside
November Rail: 25.0mm / 18.0mm*
ENVE SES: 26.0mm front, 24.0mm rear / 18.0mm front, 15.5mm rear*
November FSW: 22.7mm / 17.8mm
HED C2: 23.0mm / 17.2mm
Stan's NoTubes Alpha: 20.2mm / 17.1mm
Zipp 404 Firecrest: 23.5mm / 16.25mm*#
Mavic Open Pro: 19.7mm / 14.5mm
November RFSC: 21mm / 14.0mm
Mavic Ksyrium: 20mm / 14.0mm
Kinlin XR-300: 18.5mm / 13.7mm

*carbon rim
# Brake track width measured at the tire bead. Zipp FC rims widen to 25.2mm at the bottom of the brake track. 

You'll see that the delta between inside and brake track width is higher on carbons, which means that a rim that the brand describes as 23mm or even 25mm wide may actually have a narrower inside - where it matters - than a 20mm wide alloy. The width at the brake track likely does matter in the wind tunnel, which is easier to measure and market than "road feel" (something I expect we'll learn all too well in the near future), but our philosophy is that inside width matters more for wheels designed to race fast instead of test fast.

This solid plastic prototype of the Rail goes to the wind tunnel for preliminary testing next week.And the final word from the blog title I want to point at here is Proprietary. We've made no secret of our practice of using open mold rims and framesets. Our belief is that performance is a meaningful differentiator, but simply being different does not always mean improved performance. We really like our original open mold RFSCs for their light weight, stiffness and durability. When we elected to switch to a wide carbon clincher however, we couldn't find anything in an open mold that offered the performance we wanted. We found rims at 23mm - 24mm wide, but the inside width wasn't the 18mm or so that we feel strongly is critical for all-around race performance. We also see the trend of going as wide as possible in the bulge for aerodymamics, but that adds a lot of weight that racers jumping out of 80 corners every Saturday start to pick up on. When we couldn't find what met our needs in the market, we set out to build it ourselves. Starting from scratch was the only way we could offer what we think is the ideal racing wheelset, combining road feel, aerodymanics and low rim weight (our target for the Rail is between 475g and 495g). 

Does this mean we're abandoning open molds altogether? Not necessarily. What it means is that we will take whatever path necessary to pursue a product strategy that keeps November at the intersection of performance and value, even as the bar is raised on both. 

Also, start training, seriously. It's January for chrissake.