RT50 wheels second generation: The story

The Velocite RT50 has been our flagship wheelset since we introduced it in late 2012. Our desire was to create a true road tubeless aero carbon wheelset and make it available for sale to riders that wanted to embrace this emerging technology, we believe we succeeded. The results and reviews speak for themselves. Our [internal] mantra here at Velocite is; We test, we research, then we do it again. We verify everything until we know it’s fast, light, strong and best in class. We do not make anything without doing this. We will not tell you fanciful stories just to sell you something. There are plenty of alternatives in the market who do that already.


We knew with the first generation of the RT50 we had created a benchmark not only for us, but tubeless wheels too (in the carbon aero category). Given our mantra we knew that we had to raise that benchmark. So we began testing, researching, again and again…

So where could we Improve the already great RT50? The first and most obvious area was, and is the holy grail of all racing products – aerodynamics. We wanted to make that wheel slip through air like silk more than ever before. Making a product truly aerodynamic is no easy task and requires substantial resources, when we first developed the RT50 we had huge resources at hand through various research partnerships we had developed. This was one of the cornerstones of our beliefs (remember that mantra?). This quest for real validation of aerodynamic qualities in our aero products was really built from that original research driven vision, and has helped us gain the knowledge, skill-set and equipment to be able to do original research for ourselves. We also continue to work on these partnerships and therefore have access to first class external knowledge and advice. This is also unprecedented for a company of our size.

So what is it? Do we have a wind tunnel? No. Wind tunnels are great but they can not help you with the initial design of aerodynamic products in the sense of telling you why something works, or not, you can only try to validate what you have. Trying to develop novel designs in the wind tunnel would not be useful without extensive and expensive instrumentation and visualization aids – We are not talking about the less then useful smoke trails, but water tunnels with fluorescent dies and such. This type of equipment is mainly found in the military sector.

What we have instead is custom written Computational Fluid Dynamics (CFD) software designed for one purpose: simulate airflow around bicycles and bicycle components, with or without moving, or rotating components (wheels, rider, whatever). This is not an off-the shelf software package, or something built into a generic CAD or 3D modelling program, but a novel, very high performance and high accuracy transient 3D flow CFD software package developed by reesearchers at the University of Geneva, Switzerland, for us.

As a result of this the smart money is on Computational Fluid Dynamics (CFD) which is sometimes referred to as a virtual wind tunnel when the simulations deal with external air flow. CFD allows you to visualize what is happening with the airflow as it deflects over each surface – it tells you what the air is doing, and to a skilled researcher it also tells why the air is behaving in a particular way. Knowing the answers to what? and why? allows us to make correct and educated design choices, without guesswork, or wishful thinking. In case you were wondering, how do we know that the answers that we are getting are real – we test and calibrate the software with other test subjects  [CFD software test cases, see Ahmed body, Onera M6, etc.] where the airflow patterns are well studied. When our software creates the same result we know it’s functioning and performing as it should be.

Before we talk about the RT50’s new shape and improvements, we must explain what the ‘virtual’ wheel went through to make sure that our new shape worked. Without the use of CFD we would never have been able to ascertain whether this new aerodynamic shape really worked in the way that we hoped for.

RT50 rim shown inside truncated simulation domain for visualization purpose. Actual simulation domain volume is 100% larger
RT50 rim shown inside truncated simulation domain for visualization purpose. Actual simulation domain volume is 100% larger

The similarities between wind tunnels and CFD exist as they both need the test subject to be held in an enclosed space [actually volume], a box if you like. In both the wind tunnels and CFD the boundaries of the box need to be set far enough away so that they do not cause interference with the airflow and unwanted interactions with the object that we want to test – technically in wind tunnels this is impossible to achieve, but in well run wind tunnels the contribution to the error is known and accounted for.

The wind tunnel test chamber is an actual physical “box”, however with CFD we need to create a volume to act as the test chamber, this is known as the domain (see above image). The size of the CFD domain [or box] for the RT50 was; length = 2.2m, height: 1.1m and width: 0.6m. We found this to be adequate for the wheel simulation at 0 degrees angle of attack. In the CFD simulation we also set up buffers at the edges of the domain to soak up any air flow coming off the test subject so that it wouldn’t bounce back and interfere with the results. Again, as in a wind tunnel we push air [virtual in CFD’s case] over the rotating wheel. For those that are curious the conditions at 0 degrees AoA were:

  • air inlet velocity was 13.4112 m/s
  • angular velocity of the wheel is 39.6898490677715 rad/s along the z axis,
  • boundary conditions are non-slip except for the ground which is set to constant velocity of 13.4112 m/s.

This domain size (mentioned above) is then broken down into tiny cells (2.56mm actual resolution) for the entire volume edge-to-edge. This is important as some other common CFD systems would increase the cell size further from the object and leave it to the guesswork of algorithms to deal with the resulting errors rather than compute the actual results. The downside of our “brute strength” approach where we simulate the entire domain at the same high resolution is a much higher cost in terms of memory and processing requirement in order to hold all the information and compute the results. Our computational cluster computer has 136 Gb of RAM and 32 cores, and at times even this was inadequate.

So what are we testing, what data do we capture? Same as everyone else! The most common inlet velocities used for testing bicycles and wheels is at above 40 km/h. 40 km/h being where the effect of drag is very noticeable and where any gains from loss of drag give the rider potential ‘free’ power [watts] for his effort. Our tests were done at 48 km/h, which matches the de facto industry standard in the USA for such testing.


[embedplusvideo height=”456″ width=”640″ editlink=”http://bit.ly/1wg5ERg” standard=”http://www.youtube.com/v/6odtvaIzT2U?fs=1&vq=hd720″ vars=”ytid=6odtvaIzT2U&width=640&height=456&start=&stop=&rs=w&hd=1&autoplay=0&react=1&chapters=&notes=” id=”ep7630″ /]


What we are showing in the video animation (above, slowed down 300 times) is the real effect of what is happening as the wheel rotates through air while moving at 48 km/h, we have not cleaned up the simulation to create an artificial, idealized look at airflow, this is not a sanitised version to make our product look cool. It takes a very powerful computer and specialized software to do this (see above). Even so we only captured less than a second of information on what is happening to the airflow and to the wheel surface.

Specifically, we simulated 60000 time steps of airflow for each simulation run where each time step was just 0.0000074 seconds long! We took a snapshot of what’s happening every 400 time steps [each frame in the video represents 400 time steps], and ended up with a simulation whose  total duration is just 0.3 seconds, blink and you will miss it. This happened in a simulation domain with 132 million of those cells mentioned earlier! Each simulation of this intensity took four days to compute, each 400 time step snapshot is a file 3.1Gb in size that then needs to be post-processed, and there are 101 of those. The 0.3 seconds might sound like a minuscule amount of time, but this duration is long enough to create a quasi-steady state which allows us to really see what we need in the data. Alongside the visual data you see here are the numbers, without the numbers it’s very hard to tell if the simulation is valid so we plot the numbers to look at what is happening and if we need to revise the setup and redo the simulation.


Drag numbers for the Velocite RT50 rim.
Drag numbers for the Velocite RT50 rim. The chart displays aerodynamic drag, aerodynamic thrust (also known as lift) and the net drag which displays the characteristic decrease as the angle of attack increases and aerodynamic thrust exceeds aerodynamic drag.


As you can imagine, it took us a long time to process all of the data but it’s still more useful than a wind tunnel test which would just give us one data point – net drag – at every design iteration that we decided to test, without showing us why and telling us what was happening to the airflow. We must therefore stress that wind tunnels are really used for verification. No product could really be developed in a wind tunnel.

IN THE VORTICITY. Here we see the vorticity of the air on the RT50. This proves the aero ability of the wheels.
Here we see the vorticity of the air as it passes over the RT50 rim. The remarkable feature is almost complete lack of horizontal plane vorticity along the leading edge of the rim, indicating laminar flow over a stable vortex behind the front trailing edge.
Y-plane velocity slice and surface shear stress at 12 degrees angle of attack. The airflow remains remarkably stable both at the leading and the trailing sections of the rim. Shear stress on the surface of the rim indicates delamination of the airflow half way along the major chord which matches our target.
Z-plane velocity slice and surface shear stress at 12 degrees angle of attack. The airflow remains remarkably stable both at the leading and the trailing sections of the rim. Shear stress on the surface of the rim indicates delamination of the airflow half way along the major chord which matches our target.

The RT50 simulations were ran at multiple angles of attack (0, 6, 12 [see above] and 22 degrees) to establish not only the drag characteristics, but for the real world even more importantly, its handling characteristics in crosswinds. Handling characteristics of a front wheel are not determined by the side forces (the traditional measure) but by the torque that is generated (or pitching moment when talking about airplane wings) and transmitted through the fork and handlebars. The symmetrical airfoil of the new RT50 performed as expected and produced exactly zero torque at all angles of attack. This means that in crosswind conditions the new RT50 wheels will not cause you to veer off course, or stand up from the drops or extensions and consequently lose your aero position and potentially the race. We wouldn’t want anything else would we?

Another variation (call it improvement) to standard bicycle wheel aerodynamics simulation and test protocols is that the air inlet velocity was adjusted for the angle of attack in order to maintain the constant moving ground and rotating wheel conditions at 48 km/h. This is not done in wind tunnels, or with other published bicycle CFD data where the inlet velocity is kept constant regardless of the angle of attack of the test subject, leading to incorrect variation in test results at any angle of attack greater than 0 degrees.

So as you can see it takes a lot just to capture a split second of aerodynamics data for a wheel, and almost as much effort and resources to process the results. We should thus add that the technology, hardware and software that we own and operate in-house is often put to work on massive supercomputers to calculate global weather patterns and other huge, often classified, tasks. In fact two of our compute nodes are identical to those used by the NASA Pleiades supercomputer. What we are trying to say is; our CFD capabilities are very advanced and we are the first company to apply this exact form of CFD analysis to analyze the performance of bicycle components. Having this capability is really allowing us to build some of the most truly aerodynamic wheels (and we are working on frames in the same manner – more information to come) that exist.

AIRFLOW STREAMERS. Here we see the air flow over the wheel, as you can see the airflow hitting the front of the wheel is remaining nice and stable for a considerable way beyond the physical boundaries of the rim
Here we see the air flow over the wheel in the form of streamlines. As you can see the airflow hitting the front of the wheel is remaining nice and stable for a considerable way beyond the physical boundaries of the rim

So what did these awesome CFD resources create? A monster! They created the fastest wheel that we have ever made. However, let us get more specific. Firstly the rim width, it’s increased to 25mm, up from the 23mm of first generation RT50. The reason for this is of course aerodynamics. We tested the increased rim width with a 23mm tire and our findings are based on the end user also using a 23mm tire. The extra width allows the 23mm tire to form a lower profile over the rim, and complete the airfoil shape. Thus it’s the casing of the tire that is actually helping to create the nice symmetrical profile shape. The airflow image (above) clearly shows how nicely uniform the air is as it passes the front of the wheel.

Other advantages to this lower tire profile means the casing will deform less which will offer more confident cornering, whilst this is not an aero advantage it has the potential to help you, the rider, be that little bit faster again. There is a lot of hype over 25mm wide rims but we feel the advantages when using a 23mm tyre on our 25mm rim are not hype. Obviously if you use wider tires for comfort you will lose some of this advantage. So we recommend 23mm tires for optimum performance. Given that the RT50 is a road tubeless wheelset, allowing you to pick your own tire pressure, you should find it a comfortable wheelset to use while going very fast.

The other change to the rim, besides the new more aerodynamic and more neutral handling profile is that we now have a trackless braking surface which allows the air to follow a smoother path and hold it’s shape that little bit longer until the inevitable happens and the airflow is whipped up and dispersed behind the wheel. Again from the images and video animation the RT50 is incredibly good at handling that airflow. The trackless braking surface also creates a much nicer aesthetic on a 50mm aero rim. As with the first generation, the decals are easily removed if you want that stealth look or have sponsor obligations to [appear to] use their wheels. The layup of this rim has obviously changed but the resin has remained the same high quality as last time which has been tested with a resin glass transition temperature (Tg) in excess of 190° C. Meaning safe braking in most circumstances. We continue to use the excellent Gram hubs and Pillar 1422 aero spokes which are laced 20/24 with 2:1 rear to improve the wheel stability and potentially improve torque transfer.

Currently these wheels are in the final stage of production and are being hand built at this very moment. Expected weight will be the same as last time at around the 1630g per set but we will verify that once we have a set fully built. Price is expected to remain at what is a very reasonable cost of $1,789.00 per set, given these are unlike any wheelset out there, it’s a bargain.

So, get ready for the second generation Velocite RT50 wheels to spin you that bit closer to victory, or a personal best you never imagined you could achieve. These monsters will roar you there. All the while knowing we crunched terabytes of data to help you.