The idea of torsional compliance being good is strongly misguided. Sure you are providing a means of isolation for some side impacts on the tire, but...
The idea of torsional compliance being good is strongly misguided. Sure you are providing a means of isolation for some side impacts on the tire, but remember that these impacts are now imparting a directional change to the tire rolling axis. The result is that body inputs and extra steering inputs are needed to keep the bike upright and on-course. Its not actually a win-win. Furthermore, any intentional steering inputs will now be delayed. This creates something called a phase lag in the dynamic system, which degrades control-ability. I could dive into this in painful mathematical detail, but please just take my word for it.
Quantifying exactly how much torsional deflection occurs on trail is an engineering pitfall that you don't want to go down. Firstly, the task of defining tire loads for any vehicle is difficult because they are highly statistical in nature and very hard to calculate without lab data. Secondly, there is no standard for how much deflection is too much. We just end up with another comparison metric. Stiffness on the other hand, is a parameter within any dynamic system guaranteed to increase control-ability. That is why you hear engineers talk about stiffness all the time, because that is the parameter that can be easily measured, changed, and accounts for the lion's share of vehicle handling.
There may be a corner case at some point where, like in Motocross and MotoGP, we discover that a controlled amount of torsional and bending compliance actually benefits grip by smoothing out the loads on the tire contact patch. However, we are in a stiffness regime currently that sits far below this point. Hopefully discussions like this will lead to that eventuality where engineered fork stiffness is a thing, but first we need stiffer forks!
What I've been trying to get across in my recent posts to this thread is that we've been living within a paradigm dominated by strength-to-weight ratio, which has resulted in very low stiffness forks being offered for sale. That has created handling problems as speeds and loads increase. However, because this is what we are used to, no one realizes that fork stiffness is too low unless they spend a lot of time going back and forth between trail bikes, DH bikes, and motorcycles. If we shift our paradigm by changing the focus to stiffness-to-weight, then other solutions start to make more sense, like dual-crown forks for example.
I hope for a future where customers demand more of their forks so that dual-crown forks become the norm, with more USD options being available for DH applications.
I'm speaking from actual experience here. Not just in theory.
I have Dorados on my DH bike. They track better. They beat you up less.
There...
I'm speaking from actual experience here. Not just in theory.
I have Dorados on my DH bike. They track better. They beat you up less.
There is no issue with steering accuracy due to phase lag. The bike goes exactly where I aim it. It's not a wet noodle and it is in the sweet spot of torsional rigidity. For reference, I am not a lightweight either. I'm 95kg (about 210lb)
Yes, you need to adjust your expectation of how the bike feels and this might put a lot of people off.
I also disagree with the assertion that building compliance into the wheels is the way to go. When you take tension out of the wheel you lose strength, and they feel dead. It's fine for pros who have multiple sets of wheels to run weak low tension wheels to compensate for too much stiffness elsewhere but it's definitely not the optimum solution. Especially running super stiff carbon rims with low spoke tension. That is just dumb!
Did anyone say reducing spoke tension? That only weakens the wheel like you say, and I wouldn't suggest it at all
The entire construction of front...
Did anyone say reducing spoke tension? That only weakens the wheel like you say, and I wouldn't suggest it at all
The entire construction of front wheels is overbuilt already (Inherently stronger than the rear wheel which gets a harder time) so has plenty of room for different rim profiles, less spokes or narrower flanges
I'm not arguing that some wheels are overbuilt.
Lots of world cup pros are running reduced spoke tension because they are forced to run carbon wheels that are inherently too stiff.
Crank Bros carbon wheels are a better approach to this issue with the more vertically compliant rim profile and reduced spoke count in the front wheel. I'd still rather just run aluminium rims with 32 double butted j bend spokes
I personally hate 28 spoke wheels, and not sure I agree that front wheels are generally overbuilt.
Although this did end up in a twisted lower crown and bent axle so maybe a weaker front wheel may have saved other damage.
Rear wheels are somewhat compromised due to dish but with DH spacing you get equal spoke lengths and I have killed what more rear rims with dings and flat spots than bends.
I guess my point is that I would rather have torsional flex in my fork than dead, weak, floppy wheels.
Most people commenting on the flex in USD forks haven't ridden them and it isn't the issue people make it out to be.
I would also be interested to see if you could model the effect of the Manitou hex lock axle. That makes a big difference to the...
I would also be interested to see if you could model the effect of the Manitou hex lock axle. That makes a big difference to the torsional rigidity of the Dorado.
That would be a no. As I've mentioned in the original post, all the joints, except the outer-to-stanchion, use a bonded setting, meaning it's almost like a single part. So none of the pressfits in the crown, screwed dual crowns, etc., are covered. Not ideal, but it's the same across all variants, so there is some stability there. The reason for this is that I do not have any engineering and manufacturing data to be able to determine the interference fits, which could influence on the stiffness too (with the fit opening up a bit under high loads for example).
When it comes to axles, the situation is even worse if anything. Quick release or most single crown axles are preloaded only axially, counting on the hub being squeezed together and the hub surfaces to provide enough stiffness. The axle to dropout fit is very loose, otherwise you couldn't insert the axle. Fox fixes that a bit with the Kabolt X, where you still use the pinch bolts.
Speaking of, dual crown or DH forks mostly use some pinchbolts (aaaaand look at the Boxxer!), which greatly increases stiffness of the interface, bringing it closer to the bonded joint I use. The hex lock axle achieves some more torsional stiffness with a QR axle, while it doesn't do that much when it comes to pinch bolts. I don't really see the point on the Dorado, except that you can run lower torque on the pinch bolts and still keep the torsional stiffness of the axle.
But overall, when it comes to real world quick release axles, the preload force will then have an effect on stiffness, the clearances, the stiffness of the hub axle, the dropout surface area (standard or torque caps), etc. etc. Overall, when it comes to the axle, single crown forks in this analysis are improved in anything compared to what we have in the real world. And sadly there have to be some 'shortcuts' taken to get meaningful results in a meaningful amount of time. Looking for that 1 % on the axle interface? That's job for the boys that work for the suspension manufacturers, that get paid to do that and that have a direct line with the manufacturing plant, where they can ask 'hey, what kind of tolerances can we achieve on this part with so and so production process?'. Or they have a database of those values. Sadly, I do not
@JamesR_2026: The purpose of this discussion is ultimately to bring experience and theory in line. One of my mentors in motorcycle racing defined this synthesis as "technology". What he meant was that experience not backed up by theory is useless. You always have to understand what you are doing and why you are doing it as a crew engineer or rider.
That said, I've no doubt that a flexible fork will "beat you up less", however as @TheSuspensionLabNZ wisely commented, fork flex is not a good way to manage rider comfort. A better solution would be to adopt a rigid fork chassis, and then add compliance to the handlebar or handlebar-to-fork interface (as in MX).
As to your second question, the effect of the Manitou hex axle is to create a mechanical interface that is a better approximation of the bonded contact condition that Primoz is using. The bonded condition is perfectly rigid, so the FEM is presenting a best-case scenario in that regard. There are more complicated methods (RBEs for example) for modeling contact, but I've not dared to ask Primoz to implement these!
@Primoz: I think you are discounting your own results, and the findings of an entire industry by saying that RSU forks will always win in terms of S/W. MotoGP would ditch their USD forks in a heartbeat if a RSU fork could be made to work at a lower weight.
There are many ways to model joints, with revolute joints you can for example add stiffnesses and dampings, friction, clearances, etc. But you often need some more data, measurements or experience to choose the correct parameters to model it. Ideally you'd have all of those, so you can prepare a model based on a real life state and then adapt it to a different product for example. Yet once again I don't have access to that data, sadly... Plus I'm not versed enough in FEA (yet) to be comfortable doing that kind of stuff. Anybody can make pretty colors, the question is if you trust that data.
I'm not discounting the results, it just makes me think the requirements are quite a bit different between the two applications. Even with a MotoGP bike you still steer with your arms, just like you are on your MTB. Yet the speeds and the loads lengthwise are much higher with any motorcycle than with any MTB. That's what makes me think that the bending vs. torsional stiffness ratio of RSU forks might not be wrong and the USD forks might have an issue for the MTB use case.
That's the same answer I've heard other places, referring to the fork lowers extending too far below the axles. I think that is totally valid, but not the entire answer since it only addresses motocross and not the rest of the motorcycle spectrum. We still have to wonder why street bikes (some Harley's excluded) with much shorter travel use USD forks.
I think the second part of the answer answer is that the fore-aft bending stiffness is massively more important for a heavy, high speed machine. To illustrate, the way you calculate spring rate for a street bike fork is you add up the machine and rider weight. You then divide that number by 90% of the fork travel. Divide that number by half since you have two springs and boom...spring rate. This is because under braking the entire weight of the motorcycle is being supported by the front tire. That front tire has so much grip that it can handle all the braking duties. If the forks were to flex too much under this load, the bushings would bind, causing a lack of grip, or a resonance could occur causing the front end to chatter. For mountain bikes this is not much of a problem, since our tires don't create enough grip to suffer from chatter. The binding problem still exists, but total deflections can be kept lower using less stiffness since the weight of the machine and rider is much lower.
Anecdotally, lots of motocross riders recall the transition from RSU to USD coinciding with a massive increase in front end stiffness. Some people hated this because the bikes were now more physically strenuous to ride. However, the limit of performance was now higher, if you were strong/good enough to reach it. We've mentioned before how motocross RSU forks didn't (and still don't) have bridges, so the transition to USD definitely caused a relative increase torsional stiffness.
But to add to this, I've also been thinking about the relative difference in axle-to-crown lengths between MTB and motoX. On a motoX bike with RSU forks, the fork tubes make up a significantly longer portion of the a-c distance. Replacing all but the last 12" of fork with large diameter stanchions therefore made a lot of sense. Mountain bikes just don't have very long tubes, so the proportion of tube to a-c distance is lower.
My 2018 Boxxer lowers are 13" long, and the distance from the seals to the top crown (exposed bendy bit) is about 14". Switching to an inverted fork would mean the stiffer stanchions would now be 19" long, and the exposed tubes 8". For a motoX bike you are looking at about 14" long lowers, and 23" long exposed tubes. An inverted design is going to have 25" long stanchions and 12" long exposed fork tubes. So for the MTB the bendy bit is about 51% the length of the assembly, while for the motoX its 62% (if both are RSU). Therefore, we'd expect the motoX fork to see a greater increase in bending stiffness by going inverted than the MTB. Looking at the less-bendy bits, the MTB stanchion length increases by 36% by going to USD, while the motoX increases by 78%! I'll bet that these proportions may re-balance the findings from your FEA investigation, with the USD fork looking relatively acceptable in terms of torsional stiffness even when compared to a bridged RSU.
There are many things in engineering that require a threshold of size or complexity in order to become advantageous. Concrete is a fantastic material if you don't want your building to be super tall. Above a certain threshold you switch to steel because its strength-weight ratio is better and the weight of the completed structure won't crush the lower levels. Titanium is a fantastic material, but you tend to use it only where aluminum can't take the heat. As for forks, it seems that travel and a-c length require a USD design above a certain point where RSU designs become too flexible overall. Below that threshold, bridged RSU forks offer superior torsional stiffness at a lower weight, making them the best choice.
Regarding street bikes, keeping the same topology (stick to what you know) could be a reason. And assembling it as well, as street forks don't have any axle offset, the offset makes it much easier to assemble the fork using bolts in the centre of the stanchions. There might be a few other factors too, besides the ones you mentioned.
I had a pair of really hard nose-cases this winter. Part of the hazards of trail building means that you get the be the one to test things out...sometimes the bike pays the price. I had one so bad that I could actually feel the bushings bind and grind to a halt. It felt like catching a medicine ball and of course I took the mandatory inverted journey through the air as the fork unloaded. Of course all this abuse led up to a nice CSU creak developing and so the fork is currently with RockShox under warranty return. The fun bit is that I noticed the lower head set bearing seal had become dislodged on one side. I take this as evidence that the steerer tube bent so much as to cause the seal to pop out! This then made me wonder back to Primoz's original single crown analysis and the constraints used on the steerer tube. In that analysis the outer surface of the steerer tube was fixed. However, my experience demonstrates that a real steerer tube experiences much more deflection than can be ignored.
A dual crown fork should have to react the same loads in that region, however the stresses can be distributed between three structural entities, greatly reducing the strains. My guess is that including a more realistic constraint set for the steerer tube will show a much greater benefit in terms of fore-aft stiffness for the dual crown fork.
The dual crown analysis holds the steerer in two bands only, simulating the headset bearings and the like. The original (single crown only) analysis fixed the whole steerer tube, that wouldn't work with dual crown forks, so I had to change it here.
The band is a fixed constraint, which is far from true for a normal headset bearing. And it's butted up to the crown as well, so yeah, not ideal. But that's the issue with simulations, it can never be ideal, as there are too many factors playing into it. To make it realistic to a headset, make it elastic? With what kind of stiffness and damping factors? How stiff is the steerer tube then? Etc.
That's the issue and it's easier to simplify some things to at least know what you're doing. And when doing comparisons, judging if it will still make for a good comparison. Offsetting the ring would make dual crown forks seem a bit more stiff, yes, but it wouldn't be a game changer. Not compared to upside down forks. The material simply isn't there, as a lot of bending also happens right under the crown as well, in the stanchions.
So yeah, not ideal, with many assumptions, as I've said in the original post. Would love to compare it with real world numbers though.
I think you'd need to start using some more advanced constraints like RBE's. You'd create a fixed node in space and then connect nodes on the steerer tube to the fixed reference node via the RBE's (I think RBE2 would be what you want) If you split the cylindrical face of the steerer tube that'll create an edge along which you can apply constraints to the member nodes.
Simpler yet would be to apply a translational fixed constraint (tx, ty, tz) to just one node at the bottom of the steerer near the crown. Since the upper bearing is able to slide on the steerer tube, I would chose a node at the top of the steerer to provide the ty, tz constraint (assuming your load is applied in the x direction, I don't remember the orientation of your coordinate system). This would give you a simply supported steerer tube, but you would also see some pretty gnarly stress concentrations around the constrained nodes. Ignoring those, I think the deformation of the tube would be more accurate.
All this analysis aside, I think its time to start pulling on some things and seeing what happens. I would like to rig something up in my garage. Currently I have a Boxxer, a Pike (on its way back from warranty service), and a Trust Message that I could test. My thought is to machine a simple headset tube, and weld said tube to a plate that can be bolted to the concrete floor. This should be pretty damn rigid. The axle will also be very close to the floor so I should be able to use a set of calipers to measure displacement. Any ideas for loading schemes are welcome. It'd be nice to do something that would avoid having to purchase a load cell. Buckets of water or lead shot come to my mind.
Mount it higher and use lifting weights? And just add a plank next to it to measure deflection?
As for RBEs and the stuff, I would need to read into it, this was more of an intro to FEA for me. It's not an ideal analysis, I have no problem saying that (I said it in the original post too). Regarding rotation of the bearings and the like, some stiffness values would need to be considered as well. And yes, the steerer can move, but the bearings and the steerer are also axially preloaded by the preload bolt. The current analysis uses two revolute joints in roughly the areas of the bearings, which means more or less no out of plane rotation, so it is likely overconstrained. But like I mentioned, things like these are rabbit holes and I'd prefer to (knowingly) stay on the meadow and compare daisies on there than to chase the rabbit and losing myself in the tunnels
Without being able to measure actually simulated products after the fact, it's hard to tune the model properly. And it is quite possible the graphs in the first post are not correct in some way, shape or form. But I would be surprised if the general scale of results changes drastically in real life though.
If you were to make the steel tube fit a headset, you could also check headset preload (torque on the star nut bolt) vs deflection, if it has any effect.
I got around to welding up a quick test fixture today and finally got some numbers on the single crown fork. The fixture is just a piece of L-angle welded on top of two pieces of steel strap and a base to create a nice box structure. The fork steerer sits in the L-angle and then gets clamped on one side so that it won't fall out. I cut the L-angle to 4.5in so that its about the length of most head-tubes. This way the steerer under deflection will get supported at the bottom by the angle, and at the top by the clamp, simulating the two-points of support provided by the upper and lower bearings in the headset.
The load is provided by a small jack sitting on top of a bathroom scale. Initially I was worried that the scale might not be beefy enough for the loads I'd need to apply, but I was surprised upon testing at just how flexible the fork is. It only took about 50 lbs to get noticeable movement of the axle. The axle was threaded in, but not tightened so as to keep from flexing tubes inward. The lack of any pre-load shouldn't matter in the fore-aft bending axis being tested with this setup.
Our volunteer test subject is a brand new 140mm travel RockShox Pike (27.5 in). I would expect this to be the least stiff "real" MTB fork out there, maybe equivalent to a Fox 34. The procedure was to take a measurement from the axle to the scale with the caliper. Then I zeroed the scale and applied 75 lbf of load to the axle. I measured the new axle to scale distance and recorded the value.
I repeated my measurements 5 times and got pretty consistent results between 125 - 130 lbf/in for fore-aft stiffness. All displacement measurements were taken with a caliper, so should be fairly accurate. My main concern is with the digital scale, which is not the most accurate, but hey its paid for!
I will now put my bike back together and repeat this with the 2017 Boxxer on my DH bike after lunch. How does this measurement compare with the FEA results for a SC RSU fork? I for one am a bit amused by this number as it shows just how flexible the fork is. Consider that reducing fork offset from 51mm to 44mm changes handling noticeably. To get that same amount of change via deflections requires just 35 lbf of fore-aft load on the fork!
It was 1,5 mm at 100 N of force, but at 25 % of travel for a 160 mm 29" fork. So roughly 66 N/mm. Yours is 21 N/mm. Which makes sense considering your real world fork has some play in the axle and the bushings at least. The simulation of course dealt with a perfect fit.
Maybe try it out by preloading the fork to say 20 lbs, measure it, then load it to 120 lbs or so and measure again, to see how it behaves once it's preloaded. Theoretically it should be stiffer in this case if I understood your original procedure correctly.
Plus maybe compare it with a hub clamped in the fork, if you have the option? Maybe even just a correctly sized tube going over the axle?
TLDR;
-The rig isn't perfect, which accounts for some of why the measured values are lower than the FEA values
-The measured relative stiffness difference SC to DC RSU is ~20%
-A perfectly rigid test rig would see this value increase
-DC forks are better at preserving steering geometry, primarily fork offsett
If you are doing the FEA the way I think you are, then the program is doing a tet-mesh. Those elements add approximately 10 - 15% stiffness to structures. Tet meshes are great for speed because its easy to write algorithms that can auto-mesh just about any geometry. They are noted however for adding artificial stiffness to things, and ( quite interestingly) artificial viscosity to CFD simulations. Quad meshes for 2D and Hex meshes for 3D are the most accurate, but take the most time to generate due to the largely manual creation process they usually require. At least I've never gotten to play around with a mesher smart enough to replace me yet.
I also have no doubt that there is some flex in the base-plate of my little test rig. I can watch it move ever so slightly, and I made some attempts to stiffen it up, so lets say that's also a source of bendy-ness. Bushing and axle play are on the order of thousandths on the other hand, so I don't think it would be within my tolerance of repeatability to consider those real effects.
As I've mentioned before, the flex in the headtube is REAL. I tried to make a rig that would allow that to happen. I wish I could show you the amount of deformation that occurred in my lower bearing to pop the seal out this winter, but I just can't get a camera angle that would do it justice. Your FEA essentially neglected this effect, so that is another source of simulation stiffness over real world.
All these effects together, and I'm glad we are at least within the same order of magnitude, and the FEA is predicting stiffer. Both of these match my expectations! I did measure deflections at both 50 and 75 lbs and used the two values to make sure my stiffness was coming out nice and linear. Therefore, I'm sure the spirit of what you propose above has already been captured.
I just finished with the Boxxer and it came out to 150 lbf/in. So lets say 20% stiffer than the Pike...which is a nice round number to hold in your back pocket, and quite a bit more than what your analysis predicted. Once again, I think this comes down to your treatment of the headtube flexibility. Keep in mind if my test rig is super flexible (and I admit its not perfect), then my measured difference would be smaller than actual, so I don't really see a point in pursuing increased fixture stiffness to try and bring this difference down to the FEA value. Rather, a stiffer fixture should bring the relative deflection difference further apart.
My conclusion is that the feeling of superior control I get on a DC bike is not just an illusion, but a real result stemming from the superior preservation of steering geometry, primarily fork offset (and by extension trail) due to superior fore-aft bending stiffness. The SC fork should die a fiery death as it is an artifact of a bike industry lazer focused on weight as the primary measure of performance.
Hey justin, thanks for your input, cool to have some real world measurement! Question here, were you measuring the boxxer at full 200mm travel? I assume so, if that was the case, I can only imagine it would be significantly stiffer still when lowered to 150mm (which should give roughly the same AC as 140mm pike), it would be pretty easy to just slide the stanchion tubes 5cm further up in the crowns to measure to simulate that.Just for an idea for the future maybe.
Lots of world cup pros are running reduced spoke tension because they are forced to run carbon wheels that are inherently too stiff.
Crank Bros carbon wheels are a better approach to this issue with the more vertically compliant rim profile and reduced spoke count in the front wheel. I'd still rather just run aluminium rims with 32 double butted j bend spokes
I personally hate 28 spoke wheels, and not sure I agree that front wheels are generally overbuilt.
Although this did end up in a twisted lower crown and bent axle so maybe a weaker front wheel may have saved other damage.
Rear wheels are somewhat compromised due to dish but with DH spacing you get equal spoke lengths and I have killed what more rear rims with dings and flat spots than bends.
I guess my point is that I would rather have torsional flex in my fork than dead, weak, floppy wheels.
Most people commenting on the flex in USD forks haven't ridden them and it isn't the issue people make it out to be.
When it comes to axles, the situation is even worse if anything. Quick release or most single crown axles are preloaded only axially, counting on the hub being squeezed together and the hub surfaces to provide enough stiffness. The axle to dropout fit is very loose, otherwise you couldn't insert the axle. Fox fixes that a bit with the Kabolt X, where you still use the pinch bolts.
Speaking of, dual crown or DH forks mostly use some pinchbolts (aaaaand look at the Boxxer!), which greatly increases stiffness of the interface, bringing it closer to the bonded joint I use. The hex lock axle achieves some more torsional stiffness with a QR axle, while it doesn't do that much when it comes to pinch bolts. I don't really see the point on the Dorado, except that you can run lower torque on the pinch bolts and still keep the torsional stiffness of the axle.
But overall, when it comes to real world quick release axles, the preload force will then have an effect on stiffness, the clearances, the stiffness of the hub axle, the dropout surface area (standard or torque caps), etc. etc. Overall, when it comes to the axle, single crown forks in this analysis are improved in anything compared to what we have in the real world. And sadly there have to be some 'shortcuts' taken to get meaningful results in a meaningful amount of time. Looking for that 1 % on the axle interface? That's job for the boys that work for the suspension manufacturers, that get paid to do that and that have a direct line with the manufacturing plant, where they can ask 'hey, what kind of tolerances can we achieve on this part with so and so production process?'. Or they have a database of those values. Sadly, I do not
That said, I've no doubt that a flexible fork will "beat you up less", however as @TheSuspensionLabNZ wisely commented, fork flex is not a good way to manage rider comfort. A better solution would be to adopt a rigid fork chassis, and then add compliance to the handlebar or handlebar-to-fork interface (as in MX).
As to your second question, the effect of the Manitou hex axle is to create a mechanical interface that is a better approximation of the bonded contact condition that Primoz is using. The bonded condition is perfectly rigid, so the FEM is presenting a best-case scenario in that regard. There are more complicated methods (RBEs for example) for modeling contact, but I've not dared to ask Primoz to implement these!
@Primoz: I think you are discounting your own results, and the findings of an entire industry by saying that RSU forks will always win in terms of S/W. MotoGP would ditch their USD forks in a heartbeat if a RSU fork could be made to work at a lower weight.
I'm not discounting the results, it just makes me think the requirements are quite a bit different between the two applications. Even with a MotoGP bike you still steer with your arms, just like you are on your MTB. Yet the speeds and the loads lengthwise are much higher with any motorcycle than with any MTB. That's what makes me think that the bending vs. torsional stiffness ratio of RSU forks might not be wrong and the USD forks might have an issue for the MTB use case.
14:07, if the timestamp doesn't work. Jordi explains why moto forks are upside down.
I think the second part of the answer answer is that the fore-aft bending stiffness is massively more important for a heavy, high speed machine. To illustrate, the way you calculate spring rate for a street bike fork is you add up the machine and rider weight. You then divide that number by 90% of the fork travel. Divide that number by half since you have two springs and boom...spring rate. This is because under braking the entire weight of the motorcycle is being supported by the front tire. That front tire has so much grip that it can handle all the braking duties. If the forks were to flex too much under this load, the bushings would bind, causing a lack of grip, or a resonance could occur causing the front end to chatter. For mountain bikes this is not much of a problem, since our tires don't create enough grip to suffer from chatter. The binding problem still exists, but total deflections can be kept lower using less stiffness since the weight of the machine and rider is much lower.
Anecdotally, lots of motocross riders recall the transition from RSU to USD coinciding with a massive increase in front end stiffness. Some people hated this because the bikes were now more physically strenuous to ride. However, the limit of performance was now higher, if you were strong/good enough to reach it. We've mentioned before how motocross RSU forks didn't (and still don't) have bridges, so the transition to USD definitely caused a relative increase torsional stiffness.
But to add to this, I've also been thinking about the relative difference in axle-to-crown lengths between MTB and motoX. On a motoX bike with RSU forks, the fork tubes make up a significantly longer portion of the a-c distance. Replacing all but the last 12" of fork with large diameter stanchions therefore made a lot of sense. Mountain bikes just don't have very long tubes, so the proportion of tube to a-c distance is lower.
My 2018 Boxxer lowers are 13" long, and the distance from the seals to the top crown (exposed bendy bit) is about 14". Switching to an inverted fork would mean the stiffer stanchions would now be 19" long, and the exposed tubes 8". For a motoX bike you are looking at about 14" long lowers, and 23" long exposed tubes. An inverted design is going to have 25" long stanchions and 12" long exposed fork tubes. So for the MTB the bendy bit is about 51% the length of the assembly, while for the motoX its 62% (if both are RSU). Therefore, we'd expect the motoX fork to see a greater increase in bending stiffness by going inverted than the MTB. Looking at the less-bendy bits, the MTB stanchion length increases by 36% by going to USD, while the motoX increases by 78%! I'll bet that these proportions may re-balance the findings from your FEA investigation, with the USD fork looking relatively acceptable in terms of torsional stiffness even when compared to a bridged RSU.
There are many things in engineering that require a threshold of size or complexity in order to become advantageous. Concrete is a fantastic material if you don't want your building to be super tall. Above a certain threshold you switch to steel because its strength-weight ratio is better and the weight of the completed structure won't crush the lower levels. Titanium is a fantastic material, but you tend to use it only where aluminum can't take the heat. As for forks, it seems that travel and a-c length require a USD design above a certain point where RSU designs become too flexible overall. Below that threshold, bridged RSU forks offer superior torsional stiffness at a lower weight, making them the best choice.
I had a pair of really hard nose-cases this winter. Part of the hazards of trail building means that you get the be the one to test things out...sometimes the bike pays the price. I had one so bad that I could actually feel the bushings bind and grind to a halt. It felt like catching a medicine ball and of course I took the mandatory inverted journey through the air as the fork unloaded. Of course all this abuse led up to a nice CSU creak developing and so the fork is currently with RockShox under warranty return. The fun bit is that I noticed the lower head set bearing seal had become dislodged on one side. I take this as evidence that the steerer tube bent so much as to cause the seal to pop out! This then made me wonder back to Primoz's original single crown analysis and the constraints used on the steerer tube. In that analysis the outer surface of the steerer tube was fixed. However, my experience demonstrates that a real steerer tube experiences much more deflection than can be ignored.
A dual crown fork should have to react the same loads in that region, however the stresses can be distributed between three structural entities, greatly reducing the strains. My guess is that including a more realistic constraint set for the steerer tube will show a much greater benefit in terms of fore-aft stiffness for the dual crown fork.
The band is a fixed constraint, which is far from true for a normal headset bearing. And it's butted up to the crown as well, so yeah, not ideal. But that's the issue with simulations, it can never be ideal, as there are too many factors playing into it. To make it realistic to a headset, make it elastic? With what kind of stiffness and damping factors? How stiff is the steerer tube then? Etc.
That's the issue and it's easier to simplify some things to at least know what you're doing. And when doing comparisons, judging if it will still make for a good comparison. Offsetting the ring would make dual crown forks seem a bit more stiff, yes, but it wouldn't be a game changer. Not compared to upside down forks. The material simply isn't there, as a lot of bending also happens right under the crown as well, in the stanchions.
So yeah, not ideal, with many assumptions, as I've said in the original post. Would love to compare it with real world numbers though.
Simpler yet would be to apply a translational fixed constraint (tx, ty, tz) to just one node at the bottom of the steerer near the crown. Since the upper bearing is able to slide on the steerer tube, I would chose a node at the top of the steerer to provide the ty, tz constraint (assuming your load is applied in the x direction, I don't remember the orientation of your coordinate system). This would give you a simply supported steerer tube, but you would also see some pretty gnarly stress concentrations around the constrained nodes. Ignoring those, I think the deformation of the tube would be more accurate.
All this analysis aside, I think its time to start pulling on some things and seeing what happens. I would like to rig something up in my garage. Currently I have a Boxxer, a Pike (on its way back from warranty service), and a Trust Message that I could test. My thought is to machine a simple headset tube, and weld said tube to a plate that can be bolted to the concrete floor. This should be pretty damn rigid. The axle will also be very close to the floor so I should be able to use a set of calipers to measure displacement. Any ideas for loading schemes are welcome. It'd be nice to do something that would avoid having to purchase a load cell. Buckets of water or lead shot come to my mind.
As for RBEs and the stuff, I would need to read into it, this was more of an intro to FEA for me. It's not an ideal analysis, I have no problem saying that (I said it in the original post too). Regarding rotation of the bearings and the like, some stiffness values would need to be considered as well. And yes, the steerer can move, but the bearings and the steerer are also axially preloaded by the preload bolt. The current analysis uses two revolute joints in roughly the areas of the bearings, which means more or less no out of plane rotation, so it is likely overconstrained. But like I mentioned, things like these are rabbit holes and I'd prefer to (knowingly) stay on the meadow and compare daisies on there than to chase the rabbit and losing myself in the tunnels
Without being able to measure actually simulated products after the fact, it's hard to tune the model properly. And it is quite possible the graphs in the first post are not correct in some way, shape or form. But I would be surprised if the general scale of results changes drastically in real life though.
I got around to welding up a quick test fixture today and finally got some numbers on the single crown fork. The fixture is just a piece of L-angle welded on top of two pieces of steel strap and a base to create a nice box structure. The fork steerer sits in the L-angle and then gets clamped on one side so that it won't fall out. I cut the L-angle to 4.5in so that its about the length of most head-tubes. This way the steerer under deflection will get supported at the bottom by the angle, and at the top by the clamp, simulating the two-points of support provided by the upper and lower bearings in the headset.
The load is provided by a small jack sitting on top of a bathroom scale. Initially I was worried that the scale might not be beefy enough for the loads I'd need to apply, but I was surprised upon testing at just how flexible the fork is. It only took about 50 lbs to get noticeable movement of the axle. The axle was threaded in, but not tightened so as to keep from flexing tubes inward. The lack of any pre-load shouldn't matter in the fore-aft bending axis being tested with this setup.
Our volunteer test subject is a brand new 140mm travel RockShox Pike (27.5 in). I would expect this to be the least stiff "real" MTB fork out there, maybe equivalent to a Fox 34. The procedure was to take a measurement from the axle to the scale with the caliper. Then I zeroed the scale and applied 75 lbf of load to the axle. I measured the new axle to scale distance and recorded the value.
I repeated my measurements 5 times and got pretty consistent results between 125 - 130 lbf/in for fore-aft stiffness. All displacement measurements were taken with a caliper, so should be fairly accurate. My main concern is with the digital scale, which is not the most accurate, but hey its paid for!
I will now put my bike back together and repeat this with the 2017 Boxxer on my DH bike after lunch. How does this measurement compare with the FEA results for a SC RSU fork? I for one am a bit amused by this number as it shows just how flexible the fork is. Consider that reducing fork offset from 51mm to 44mm changes handling noticeably. To get that same amount of change via deflections requires just 35 lbf of fore-aft load on the fork!
Maybe try it out by preloading the fork to say 20 lbs, measure it, then load it to 120 lbs or so and measure again, to see how it behaves once it's preloaded. Theoretically it should be stiffer in this case if I understood your original procedure correctly.
Plus maybe compare it with a hub clamped in the fork, if you have the option? Maybe even just a correctly sized tube going over the axle?
-The rig isn't perfect, which accounts for some of why the measured values are lower than the FEA values
-The measured relative stiffness difference SC to DC RSU is ~20%
-A perfectly rigid test rig would see this value increase
-DC forks are better at preserving steering geometry, primarily fork offsett
If you are doing the FEA the way I think you are, then the program is doing a tet-mesh. Those elements add approximately 10 - 15% stiffness to structures. Tet meshes are great for speed because its easy to write algorithms that can auto-mesh just about any geometry. They are noted however for adding artificial stiffness to things, and ( quite interestingly) artificial viscosity to CFD simulations. Quad meshes for 2D and Hex meshes for 3D are the most accurate, but take the most time to generate due to the largely manual creation process they usually require. At least I've never gotten to play around with a mesher smart enough to replace me yet.
I also have no doubt that there is some flex in the base-plate of my little test rig. I can watch it move ever so slightly, and I made some attempts to stiffen it up, so lets say that's also a source of bendy-ness. Bushing and axle play are on the order of thousandths on the other hand, so I don't think it would be within my tolerance of repeatability to consider those real effects.
As I've mentioned before, the flex in the headtube is REAL. I tried to make a rig that would allow that to happen. I wish I could show you the amount of deformation that occurred in my lower bearing to pop the seal out this winter, but I just can't get a camera angle that would do it justice. Your FEA essentially neglected this effect, so that is another source of simulation stiffness over real world.
All these effects together, and I'm glad we are at least within the same order of magnitude, and the FEA is predicting stiffer. Both of these match my expectations! I did measure deflections at both 50 and 75 lbs and used the two values to make sure my stiffness was coming out nice and linear. Therefore, I'm sure the spirit of what you propose above has already been captured.
I just finished with the Boxxer and it came out to 150 lbf/in. So lets say 20% stiffer than the Pike...which is a nice round number to hold in your back pocket, and quite a bit more than what your analysis predicted. Once again, I think this comes down to your treatment of the headtube flexibility. Keep in mind if my test rig is super flexible (and I admit its not perfect), then my measured difference would be smaller than actual, so I don't really see a point in pursuing increased fixture stiffness to try and bring this difference down to the FEA value. Rather, a stiffer fixture should bring the relative deflection difference further apart.
My conclusion is that the feeling of superior control I get on a DC bike is not just an illusion, but a real result stemming from the superior preservation of steering geometry, primarily fork offset (and by extension trail) due to superior fore-aft bending stiffness. The SC fork should die a fiery death as it is an artifact of a bike industry lazer focused on weight as the primary measure of performance.
Post a reply to: How stiff are dual crown forks?