Ask The Expert: Computrack's Greg McDonald On Why Marc Marquez Uses So Much Lean Angle

Introducing The Expert: Greg McDonald

Marc Márquez is a lean angle magician on a MotoGP bike. The Repsol Honda rider regularly uses 60° or more to wrangle his RC213V through corners. The rest of us – and his rivals – stand in awe, and wonder how he manages to get away with it.

Yet he has also said he would rather not use so much lean angle. There are easier ways around the corners, but with the 2019 Honda RC213V, using a lot of lean angle is the only way Márquez can maintain his competitiveness.

Why does he need so much lean angle? Keen to find an explanation for this I got in touch with Australian chassis expert Greg McDonald. McDonald worked as a GP and Superbike ‘insider’ for many years, but was previously reluctant to talk about the work he did for a string of factory teams, for commercial reasons.

These included Martin Adam’s Smokin’ Joes (American Honda) team in the USA, several projects with Team Roberts – including a little known 125GP machine - and a collaboration with the late Warren Willing to develop the Suzuki 500 on which Kenny Roberts Junior won the World 500 Championship in 2000.

These teams all bought Computrack machines, designed and built by Greg McDonald, to help make their bikes competitive.

More recently he collaborated with Willing on a 2013 test project with Ducati – prior to the arrival of Luigi Dall’Igna.

A quest for precision

McDonald is an ex-racer (dirt-track and road-racing) who ran a successful crankshaft re-building business for many years before branching out into straightening the frames of crash damaged motorcycles.

Frustrated by the lack of accurate equipment he decided to develop his own three-dimensional co-ordinate measuring machine in the 1980s. This was patented under the name ‘Computrack’ and won him The Australian (newspaper) Innovator of the Year Award in 1987.

Put simply, McDonald applied the same attention to detail from his crankshaft re-building days to chassis measurement. To him, a one millimetre tolerance is not to be countenanced; McDonald works to one hundredth of a millimetre.

It was not long before a few racers started showing up to have their crashed machines measured and straightened. From there McDonald started exploring the subject of chassis geometry with the aim of getting the best settings possible for good handling.

Hideo ‘Pops’ Yoshimura was his first overseas customer, the legendary engine tuner buying a Computrack as he saw it as “the way of the future.”

McDonald expanded his network to the USA and Europe and built up a huge database of chassis geometry from working on private customer’s bikes in addition to the factory-backed teams and OEM product development and quality audit.

Ask the expert

It is doubtful there is anyone else in the world with the breadth of chassis experience across a vast range of motorcycles.

Now here is his response to the question: why does Marc Marquez have to use such extreme bank angles with the Repsol Honda?

Michael Esdaile

On Friday at the Sachsenring, Marc Márquez told reporters he was having to use a lot of lean angle in order to make good lap times with the Repsol Honda. “If you check a little bit this year, we are using a lot of banking, too much, and why we are using banking is because the package is not turning. Then I use all this banking, not because it’s my riding style, not because I like it, but because I need to.”

Marquez’s comment was backed up by Cal Crutchlow: “Unfortunately, we need to lean this bike more to turn, but every time I lean the bike more, I slide off the bike, I crash," the LCR Honda rider said. "And that's just not my style."

This takes me back to the Honda RC30 days of chassis set-up with Aussie racer Shawn Giles’ bike that Tony Hatton and I collaborated on, a Moriwaki-powered RC30 he raced in Australia, and the Honda RC45’s I did with the American Honda, Commonwealth Racing bikes in the AMA Championships. The issue regarding what parameters are needed for the lean angle necessary for the racing line on a given radius corner is still not well known.

Checking the numbers

Using GMD Computrack and using a holistic approach – measuring all the numbers of a motorcycle’s geometry – I discovered there were eight parameters that needed to be correctly set to obtain the optimum grip to reduce the lean angle necessary for a given radius of turn.

The process is to change all the necessary geometry parameters in one step to get the chassis close to optimal, which gives the feeling of grip and very early warning to the rider. Then fine-tune from there (which is what Tony Hatton did with Giles’ RC30).

If a team does this one step at a time, they never seem to get it nailed. It is vital to look at the chassis geometry in totality and adjust everything that is necessary to get the complete package.

We learned a long time ago that when less trail is used, a rider needs to use more lean angle for a given rate of turn.

In many cases the pursuit of very light steering feeling does not mean better turning: but it does mean more lean angle is necessary.

The MotoGP teams work/develop in a very closed environment, with a very limited experience on only a few different bikes.

Seeking grip through geometry

Years ago, the Computrack 3D measuring machines were purchased by 12 OEM and many factory-backed GP and national race teams around the world. They were bought because they had proved their worth, with the data from each successful project being fed back into the database of bike set-up data, helping to refine the model.

The progressive staff wanted to know what the chassis tolerances were, and what the defined overriding parameters were from that database. The answer was simple, replace the words ‘weight’ and ‘flex’ with the word ‘grip’, and then the chassis geometry set-up parameters that needed adjusting became clear.

Superior motorcycle handling is about controlled grip, or traction. Much is written about weight bias and chassis flex etc. but if a motorcycle’s suspension and geometry is not working in harmony to capture the opposing forces and convert them to grip, then chassis flex etc. will not give the desired result.

Stiction vs flex

We never hear how the impact of the chassis flexing affects stiction (sliding friction – the resistance of the fork stanchion from sliding in the outer tubes) - the most common problem we find. Riders often misinterpret excess stiction as tyre grip, but even small binding in suspension movement takes away the feel of early warning from the front tyre. Even a small twisting of the forks from a crash, or in practice sessions or the race (often due to less than optimal fork clamps) can cause stiction.

With the very fine tolerances in the bushings of the beautifully engineered modern forks, smooth motion is vital. In addition to optimal chassis geometry, our focus is on the first one to two millimetres of suspension motion in any change of direction with low inertia, such as the release of the front brake and the initial throttle opening. Smooth, non-binding fork action at this moment is vital for maintaining front tyre edge grip and providing critical early-warning feedback to the rider.

Without that, there is a vastly increased risk of front-tyre ‘tuck under’ – or ‘closing the front’ as the Europeans put it.

It is hard to see how increasing chassis flex with the attendant risk of suspension binding, which alters the feedback to the rider, always increases grip.

Stored energy

It has been stated in many forums that motorcycle suspension does not work when the motorcycle is at extreme lean angles. This certainly is a challenge, and it is at these extreme lean angles where it is critical to have the suspension working smoothly, for the reasons already outlined.

Stiction and ‘chatter’ (sometimes called ‘patter’) are the result of the delayed release of stored energy, mostly caused by suspension ‘binding’.

The focus of the GMD Computrack user network is to identify these real chassis problems, optimise the parameters and make them work for the rider.

The aim is to identify and rectify. Or put another way, optimise without compromise.

I will go on record as saying that from our findings, so much that is written on this topic today is "running wide and off-line".

There is a huge difference between the data obtained around the world over 30 plus years by those using Computrack to measure bikes and set their geometry numbers, as compared to those who obtain expertise from a very small sample.

Leaning versus turning

The lean/turn issue we discovered in 1988. We learned that optimising the entire chassis geometry was the way forward. The optimal steering rake must be matched with the correct amount of trail, swingarm pivot height and the other five parameters, working to very fine tolerances.

That's why I say that Marc Márquez is having to use so much lean angle. The Repsol Honda is using geometry that gives very light steering, but the excess lean angle is a side-effect of that. To change the geometry would make the steering seem heaver, but not change the rate of turn, it’s a matter of the riders feeling preference, and would need a few track sessions to assess. This is exactly the sort of problem the GMD Computrack is designed to solve.

I hope to explain a bit more about the whole process in a series of articles here, to help people understand how the geometry of motorcycles affects how they behave on track.

Greg McDonald

People who strive for excellence is chassis technology may contact Greg McDonald via The purpose of these articles is to share knowledge and reach out to the next generation of riders and teams, to pass on what I know to all interested, intelligent people. is the perfect forum for that.

People who strive for excellence in chassis technology can contact Greg McDonald via The purpose of these articles is to share knowledge and reach out to the next generation of riders and teams, to pass on what I know to all interested, intelligent people. is the perfect forum for that.

If you enjoyed this article, please consider supporting You can help by either taking out a subscription, by making a donation, or by contributing via our GoFundMe page. You can find out more about subscribing to here.


Back to top


I get what you're saying, there was basically no actual info in there, just a lot of self promotion. But it would appear to be the first of a series so - as I said about Simon Crafar's potential-filled-but-semi-disasterous performance on the mike earlier this year - perhaps we should allow the benefit of the doubt at this stage.

GMD have certainly done well for him/themselves.  I actually remember the ads in REVS motorcycle news and the like - way back in the 80's.  It's very easy for a great idea like that to be taken away and used by others, so GMD are obviously not only technically clever but also savvy in business.  Hopefully in deciding to write he's accepting that he'll have to actually share a slice of his knowledge to make the articles worthwhile.  I look forward to the forthcoming articles - and Simon in pit lane for the last time in 2019 this weekend.

Have you watched his tech talks? So good! While he doesn't seem able to breathe and relax, I have no problem enjoying Crafar! Some delivery awkwardness as he tries to push the ocean through a straw doesn't deter me from hearing how spot on he is so often, especially describing what riders are experiencing in an array of situations. Yet again, to each their own?

Perhaps it's because I forgot it's his second year and the awkward bit was early 2018 not 2019.  And reading it back the last bit could sound like I'm looking forward to it being the last time I hear Crafar for 2019.  But you guys have totally misinterpreted my comment. 

Here's what I wrote after the 2018 Qatar race roundup when someone was bashing Crafar's debut performance:

but yeah, cut Crafar some slack. It was indeed clumsy compared to last year but I thought he didn't do too badly given it was his first go at it.  He was clearly uncomfortable, so once it becomes 'normal' for him and he develops relationships with all the guys he is interviewing he will come good.  One thing was that he made a comment quite early (QP2?) that Zarco seemed uncomfortable and stressed, not his normal self.  Sure enough, after qualifying came comments from Zarco's side  that he had been stressed and yada yada...

Give him a chance.

So.... to clarify... SC's first couple of meetings were indeed "semi-disasterous" but the potential was there and I was waving the Anzac flag defending the guy because I could see he would be brilliant in time.  And I was right, it transpired exactly as I wrote above and by mid year he was relaxing into the position and is now a total natural and one of the best parts of the coverage.

And it's the parallel I was drawing against the first poster who slagged this series off before it had even had a chance to prove itself.  Like Crafar, I get the feeling Greg McDonald has a lot to offer but the first installment only offered a hint of it.



Take a look at yourself man.  And you wonder why some people aren't willing to pay for the experience of reading comments like this?

Sand in my crack?  You betcha.  Cash for comment, no matter how ignorant, vs constructive input that is screened out?  It's a very strange echo chamber indeed, and I don't need to be heard that badly.

Not that it'll ever get exposed to the light of an OLED screen, but back on topic: Greg's comments re-accurate measurements and data echo those of Mike Sinclair (book "King Kenny's Spannerman"). Basically if you don't/can't measure it you are just guessing.  

The other point that jumped out at me was his comment about making a step change rather than incremental improvements.  Mat Mladin said something similar: basically don't bother with tiny adjustments, "give me something I can feel" so he would know straight away if they were going in the right direction.  In contrast, how often have we heard from Maverick or Dovi lately that some change or other didn't really make any difference?



At last, words from the now retired world leading expert, McDonald, on motorcycle set-up for maximum grip. To date this knowledge has been limited to those with access to the GMD Computrac database. Maybe not anymore.

This is the sort of information that racing enthusiast like myself who also like to understand the nitty gritty of why one bike does a lap time so differently to the next, wants to know. Great to hear from someone who has worked on this topic at the coal face.

Thanks David for including this on your site.

I don't think it was... there was a need to increase power (from 18) and it seems that many things had to be changed, one major item was the airbox  (probably made larger or a different shape) and that changed many things on the chassis, the steering damper for one (gone from a more conventional linear to a rotary one...inside the "tank"), and that lead to the new chassis that only Marc can ride successfully. He even said in preseason testing that he was having problems with it, so no, i don't think it was designed for him but it's only him who can get the best out of the chassis, and there's the difference.

From the last interview with Takeo, Takeo admitted that they were designing the MY2019 by prioritizing Marquez's needs. They knew that to achieved "faster bike on the straight" they had to compromise RCV's handling, and they knew that only Marquez can wrangle it.

If Marquez said that he had a problem riding it, at pre-season, well it's no secret that he still recovering from his shoulder injury -at that moment. Or, maybe he just lied, like he always does.......

An assumption that Marc likes the 2019 Honda's handling seems ill conceived. He asked for motor and got it. Now he doesn't have to overdo it on the brakes so much. I haven't seen even a blip indicating he wanted/likes the front end porridge feel.

Looking forward to the next articles fleshing out the point re geometry synergy and grip. Cool point to extrapolate from. Same for relevance/impact of stiction. Ears peeled. We will hear some re why Marc leans so far...but not how. That is where the big question is!

Perhaps we will get a peek at the Ducati geometry as well. It needs a rear wheel steering blaster onboard again. It is a better bike than the Honda recently. 2018 a solid argument could be/was made that it was the best bike on the grid. On the whole this yr w how Yamaha began the season, it may still have been. Marc flatters that waggling vague rocket turd. We can't quite see the full potential of this Suzuki with only two bikes and so little resources, but it was the best railing Yamaha on the grid until Summer break and remains a conventional handling masterpiece. The rate of improvement of KTM has been incredible, and can be expected to continue, and while it is more Honda-like than anything else out there, it is really it's own sort of wobbly beast. As is the pumping Ducati.

David, can you please ask these micrometer magician-geeks to compare and contrast/critique geometry and flex for all the manus? Perceived weakness particulars and context? Inline 4 vs V4 chassis dynamics? How MotoGP-scale sheer power management affects handling? Particularly the Ducati. I have assumptions that their stability is needed for the V4 grunt, and that Yamaha gets some of it's rails via softer power delivery. Intuition only goes so far.

Thanks mate!

My only assumption is Honda built the bike around Marc, not that Marc loves the 2019's handling. Marc loves faster bike, and Honda gives him a faster bike -with an ill handling. As Marc said, "I love faster bike, and I don't mind if I had to work around it". And there, Honda gives him that kind of bike, sacrificing the others, especially Jorge Lorenzo who love the bike did its part without the need to wrangle it first.

I'm sure Marc still longing that front end feeling from the 2018 Model-Year and hoping that the 2020 bike still retains some front end feel to make him work less every session (while still being faster), and so did Honda. However, if Honda can't find the compromise from the chassis balance, and Marc still needs the faster bike, I'm sure that what Honda will give to him, no matter what result it will direct to the other riders riding the same bike.

IMO, this year looks like Honda is in the 90's situation again. Honda’s to Marquez looks like Honda to Doohan. To win, Honda needs to build a faster bike and their main rider like it because he can win with it. But, did he love it? Finally Doohan able to make his bike rideable to him. Once the other starts catching up with the same bike, Doohan asked for the “faster yet harder to ride” screamer. The problem is, in the 97’s the factory team permitted to have a different type of machine, while current regulation restricted it. Hence, it painful to be Lorenzo right now, since he can’t ride a different bike compared to Marquez’s once the season starts.

Since I didn't get much from the article I'll posit this: could fork tubes could be made oval with the long axis front-to-back? Seems to me this would provide for and aft rigidity with side-to-side flex.

it wouldn't be cheap (but Ohlins forks aren't cheap anyway)

Two thoughts: if it were better surely someone would have at least tried it by now.

Why telescopic anyway? There are plenty of other options.

Engineering conservatism: that's why there's not much difference between MM93's Honda, and Les Graham's AJS - they are both strengthened bicycle frames with an engine bolted where the pedals should be.

As the Motogp engineers are still trying to find ways to achieve and tune chassis flex at lean angles I would expect that they haven't tried everything. Oval pistons have been tried, why not oval fork tubes? Thinking "someone would have thought of it already" is limiting when trying to find solutions to problems such as chatter at max lean.


I'm nobody too, just a guy who's been following racing for 40 years

Hitting the wayback machine, MotoCzysz was working on such forks for the C1 and looks like Czysz continued to develop them for the his E-bikes. Sealing a non-round tube (reference Honda oval pistons) is not simple. Czysz moved fluid-containing parts to outside the stansions themselves likely to not have to solve that problem.

I was aware of Motoczysz but not the oval fork. Interesting articles...of course Ibis was correct: someone had thought of it.


This is the sort of information that racing enthusiast like myself who also like to understand the nitty gritty of why one bike does a lap time so differently to the next, wants to know. Great to hear from someone who has worked on this topic at the coal face.

Thanks David for including this on your site.

A little unpolished but gives great insight when discussing a riders mindset - I think he's an asset to the paddock broadcasts.


Do they really work to a tolerance of 0.01mm? To put that in context, the paper in a writing pad is usually around 0.15mm thick, so 0.01mm is 15 times thinner than that. I used to be a precision sheet metal worker fabricating all sorts of high spec aero components and achieving 0.1mm was pretty good going. I doubt we could have even measured 0.01mm let alone do that consistently. I'd have thought ambient temperature changes alone would make that, at the very least, challenging, especially on something like a frame.

Anyway, Im intrigued to read more when it comes, as this is an area that has puzzled me for all sorts of reasons.

A good CNC machine with built in tool measuring every tool change and time limits on tools etc. Will run under 0.02 all day, everyday. If you can keep it running non-stop, pallets, robot arms etc. Maybe even a little less. Now for real accuracy, you may want a hard turning machine or something similar and 0.01 is definitely on (easily)these generally measure after a first cut and compensate accordingly. Of course, temp controlled lube oil and cutting fluid and all the rest aint cheap. But they don't turn up with a bad attitude or a huge hangover either.

I could have put this reply on any of the sub-threads but the bad attitude bit made me laugh. Why is for another day, but yeah, I know what you mean. To my shame I was out by a decimal place, we of course worked to 0.015mm or 5 thou in old money. Even though it's thirty years since I wielded a vernier in earnest you'd think I'd at least remember what that converts to, even if it does now seem an incredibly small thickness.

I'm still slightly sceptical that you can even make a large-ish, complex-shaped welded frame and especially from alloy to that kind of tolerance but maybe things have moved on. Or they jig it to the high heavens to prevent movement while welding, which I suppose you would do if you're investing six figure sums or more to get what you want. And I'm even more impressed if they can straighten a wrecked frame back to that tolerance after a crash. I've sometimes wondered just how good those bikes are after they've  cartwheeled through the gravel traps a few times. I'd have thought there must be a fairly small scope for re-usability beyond which you just can't still make adjustments using reliable references. Because one thing I do remember all too well is that it really doesn't take much of a knock to a lightweight structure to have you chasing your tail to get it back to tolerance.

Forgot to mention all my machined dimensions were in mm so 0.02 is 8 tenths of a thou in imperial and 0.01 is 4 tenths. As mentioned above no real problem on a decent CNC. You are entirely right for the welded up stuff, they must madly jig things up. Just going to mention quickly, cos maybe it'll strike a chord. Place I worked about 15 years ago started making Apache helos dashboards out of a single billet. Boeing loved it. Machined 3 times, with settle times in between. Flex and chatter were the only real problem the machine could monster the dimensions. 

Think those crashed frames just became Cola cans, remember when upside down forks came in? All the bikes were killing the Al Alloy beam frames in a crash cos the forks were stronger than the frame. 

I would hazard a guess that the fames are made of components that are roughed in relatively close on things like steerer head bearing bores, swing arm pivot points, etc., then welded up (yes, good fixturing would be important here), and heat treated, before moving on to final machining. Choosing one feature as a reference point (for example, the swingarm pivot), the frame can be fixtured and moved around so that all the other important features are machined relative to the initial one chosen.

One thing about this kind of fabrication: there are many ways to get from raw material to finished product; not all work as well as one might hope. By that I mean you can paint yourself into a corner by using the wrong approach, or sequence. Since welding and machining tend to add or move stresses in the material around, I think that trying to weld up finished pieces would be an invitation to chase yourself around and never get everything where you want it.

Yes, I can easily picture how I'd go about it in a state-of-the-art factory and imagine I could probably get all the crucial gubbins to stay where they should be, less so the longer bits of section that join them all together. Mind you, a blessing of metals is that they behave consistently so once you know what's going to happen and by how much when you apply force or heat, you can take that into account especially if you use robots to apply the treatment itself consistently. Though whenever I see the welding on close-ups of the frames it looks like it's done manually and I'd assumed that, because the number of units made each season is probably very small - two or three dozen? - much of the fabrication is indeed hand made. 

They can measure using the latest digital 3d methods whether it's laser or a mechanical system. There are lots of variants on the market that would do the job just as well. 10 microns in Aluminium! yes measure but not  tweak. Welding up a chassis in Aluminium gives it a lot of residual stresses that changes the shape on cooling down, this has to be either allowed for or adjusted after and then heat treated until stable and the correct hardness/ductility is achieved. Temperature fluctuations will cause more movement than 10 microns during a day or race situation.

It is interesting to read that less trail dictates bigger lean angles for the same corner speed. It would be great to understand WHY this is the case. Perhaps this is disclosed in the following article(s)?

I remeber working on the Honda powered CRT design some years ago and the technical staff in the Gresini team being impressed that the FTR chassis with all adjustments on zero, both bikes measured the same while the Hondas of Bautusta could vary by up to 3mm on wheel base with all zero settings.

Goes some way to also explain why some rifers have a #1 bike which they prefer.

There was a video doing the rounds back when the KR Proton project was underway. I think it was John Barnard that was involved, and driving towards F1 levels of accuracy in the frame construction. The video showed the bearing seat of a headstock being clocked with a dial gauge all held in a pretty heavy duty jig. That thinking is what had led us to the machined from solid frame sections that all the Al chassis now use. I had the chance to have a close look at a Kalex frame a couple of years ago, and they are a work of engineering porn! Its easy to understand FTR working in the same way, the F1 industry is mostly based out of England after all.

Lilyvani (and anyone else with fifteen minutes of your life you will never need to get back),

I can't speak for Computrack (I am sure they have the capability that allows them to measure conformance to specifications, which is what matters) but I do have a a few insghts on the subject from my professional experience as a subject matter expert (SME) for dimensioning and tolerancing specifications.

There is a big difference between resolution and accuracy. Resolution is how many digits your measurement equipment displays (typically that may be a CMM, or Coordinate Measuring Machine, but there are a lot of other methods commonly used), while accuracy is how many of those digits are actually valid. Having a CMM display read to 0,0001mm is not uncommon. But those same CMM's may only be accurate to 0,002-0,004mm, and that accuracy is very much dependent on the distance from the measurand (a completely geeky word meaning the characteristic being measured) to the Datum Reference Frame (or DRF, which is the origin of measurements for physical features when Geometric Tolerances are specified with Datums).

The required accuracy of the measurement depends on the Specified Tolerance (generally the required accuracy will be 1/5th to 1/10th of the total tolerance, with the trend towards the 1/5th value these days). If your specified tolerance is +/-0,02mm (or 0,04mm total), then your measurement must be accurate to no more than 1/5th that total value, or 0,008mm. If your CMM set-up is cerrtified accurate to that value or less, you are good to go. The previously cited example of a CMM system accurate to 0,002-0,004 will be perfect for our needs.

Additionally, all measurement includes uncertainty, and that uncertainty is always reported along with the actual value measured. This can be confusing for people that are staring at far more desplayed resolution than can be used. For example; let's assume we have a bore for a nominal 48mm bearing, and we want a cold press fit at assembly, so the bore tolerances might be for an H7 Fit. The bore feature has a specified tolerance of 48 +0,025mm/-0mm. We measure the bore and record 48,024mm with measuring equipment that has a cetrtified accuracy of 0,002mm...does the feature conform to specifcation? Answer? We don't know. And that is because the measured value plus the uncertainty of the processing equipment might exceed the specified tolerance limit (the as-produced feature could actually be as small as 48,022mm, or as large as 48,026mm after we factor in the measurement accuracy). Since the uncertainty of our measured value does not allow us to verify conformance (or reject for non-conformance) everyone starts swearing and then calls engineering to see if they will analyze the discrepency and buy off on it...but they probably won't so they call metrology and see if anything more accurate is available to measure with. Metrology comes through and you get the same repeatable value, but now with processing equipment with an accuracy of 0,001mm. And you are golden because the maximum actual bore size would now be the measured 48, the new uncertainty of 0, you a worst case of 48,025mm, which is conforming. 

At first look tolerances themselves are pretty straight forward: Tolerances are specified so the form, fit, and function of the detail component or assembly shall meet the functional requirements of the end item. For instance, the bore size in the previous example was selected from a range of industry standards to provide a good location fit, such as might be required for a pair of swingarm bearings. It needs to be tight enough to securely locate the bearing, but also allow reasonably cost effective assembly via a cold press fit (or even dedicated drift and a soft mallet).

It gets a bit trickier though where a single feature may have multiple functional requirements, and as-such will require not one but several tolerances to control all the feature characteristics (and this is more common than not). In the previous example we just have a Size Tolerance (that ensures the bearing can be correctly installed in the bore). But that bore can't just be anywhere, so it will also require a Location Tolerance. Ah, but the bearing works with a twin on the other side of our swingarm, and if we locate them individually, even though they might separately meet the Location Tolerance, because of tolerance stacking they may wind up slightly offset from each other. If so, the swingarm (or its pivot) will have to distort to be assembled, and will probably bind in use. Not a very happy state of affairs and certainly not meeting the functional requirements of the end item. So we treat the two bearing bores not individually, but as a Pattern by specifying an additional Coaxiallity Tolerance to control any mis-alignment to an acceptable value. Then there is the depth of the bore, the surface finish requirements, and so forth...and you can see where all this is going; we wind up with a whole stack of Specified Tolerances (to support all of the Functional Requirements), with most of them having different values and which as-such may also require different measurement accuracies to verify. And by now we are thinking if we measure all that stuff our massive factory will be producing three motorcycles a year.

So we don't measure all that stuff. First, becasue in a series production environment measurement of the as-produced features is a mostly a non-value added activity. This statement disturbs a lot of people, but consider; it doesn't matter how many ways I measure the bearing bore, it is exactly the same feature when I finish as when I started. I have not improved the Size, Location, Coaxiallity, or Durabillity of the feature one damned bit just by taking measurements. If I want a better feature, I need to have a better manufacturing process...period. And so the whole manufacturing world avoids this issue of measuring your way into poverty by using statistical controls, which are more commonly known as Process Capabillity. This is a whole different subject that I will not go into...too much...but the short version is we ensure that the manufacturing equipment is of sufficient accuracy and reliabillity that we already know, to a high degree of certainty, that the manufactured feature will conform to the specified tolerances. For example if those bores are made with a high precision machining center or robotic controls, we will first run controlled tests on the manufacturing process to establish that processes's dimensional standard deviation, and then that can be analyzed against the tolerances required to meet the functional requirements of the end item. From that we get our Cpk values and Sigma numbers. The Sigma is just the number of Standard Devaitions (SD) from the Upper or Lower Specified Limits (USL, LSL) to the Nominal Value of a centered process. Example: If we have a feature specified as 25+/-0,01mm, and an SD of 0,002, we have a Five Sigma process. And that should ensure that we have no greater than 233 defects per miillion oppurtunities, which is pretty sporty. Our Cpk value would be 1.66. Cpk's are different from Sigma's in that Cpk's can account for process drift, which matters because everything, including our high precision menufacturing processes, eventually wears out and gets a bit wobbly. And, OK, we still measure a bit as a failsafe (and to comply with statistical process control standards), but no more than required (it very much depends on our confidence level in the manufacturing process. Lower confidence = more measurement). For a robust process, if we had a production run of 600 swingarms we would probably sample the first one (to make sure evereything was tooled correctly), the last one (to make sure we had no significant process drift during the run), and if we are the cautious types we might sample another one maybe every 200 items. Almost all the measurement today (in series production) is done in validating the Process Capabillity of the manufacturing equipment, because that is where you can make improvements. If your process is making oval bores, measuring them after the fact will never make them round...fixing the manufacturing side will. As a result of all this, nobody spent any significant time at the factory measuring the dimensional accuracy of our production motorcycles. They validated the machines that made them...and then pushed the "start" button.

But as-noted, that is for series production. What Computrack does is very different, so in their case everything has to be measured (so I am glad they are serious about it). What they are doing is improving an existing detail component or assembly that already exists, either by changing (or correcting) the production tolerance values (which are specified at the OEM level to a cost), adjusting assemblies to account for manufacturing tolerance stacks, or changing the Nominal OEM target values (which may not be optimized for maximum performance because they were specified for an average population instead of a specific one, or were simply wrong to begin with (it happens)). They have evidently developed a series of very specific dimensional target values that work, but to arrive at those values you also need to know where you are starting from, and then to verify that you actually hit the targeted value. And that is where Computrack's measurement capabillity would be required, because they will need to do far more of it than the OEM manufacturer would, and to a greater degree of accuracy.

Anyways, sorry about the length of this screed, I hope some of it was of value to you. And at least now you know better than to ever get me going about Tolerances, because over the years I developed a real talent for taking someone's simple request to find out what time it is...and turning that into an extended lecture on The Complete History of Watchmaking. And if you made sheet metal parts to a 0,1mm tolerance, you have my complete respect. Cheers.

PS - You are spot-on about the thermal issues. The tightest tolerances I ever had to verify were on the order of 0,0013mm (and the feature I was measuring was over a meter in diameter), so it all took place in an environmentally controlled room...but only after the component had a 36 hour heat-soak (in that same room to remove any temperature gradients in the component). Heck, you couldn't even touch it with your bare hands after it was thermally stabilized because yoiur body heat might impact the results. And components we might normally think of as rigid bodies become decidedly less so when the tolerances shrink that radically. Unless you are orbiting on the ISS everything is non-rigid if you look close enough (damned gravity). We ran a really spifffy Fourier Analysis program the Navy had MIT develop for us to wash out the sag values which, though quite small, were still a significant percentage of the specified tolerance. Cheers.


Maybe you were working on nuke sub prop shafts. Just a guess. My family business was precision sheetmetal fabrication. I personally made parts that went to the moon. We routinely worked down to +/- .0015" or .04mm.

That was a hell of a good stab at it, mate. Scary good actually. Main propulsion shaft seals for Ohio class boomers and Los Angeles class fast attack boats to be precise (also the prototype testing for the Seawolf seals, and maybe the last set ever for the NR-1)...all while at EB. I loved it there...for reasons that now elude me...but then one day some airplane wranglers showed up and doubled my pay to join them at the Lazy-B Ranch. Yee-Ha. Cheers. 

That's brought me up to speed though you're a bit iffy on target time, you might need to recalibrate that watch. And, sadly, I actually enjoyed the read. There's something vaguely disturbing about the number of comments around this topic, I picture a lot of old geezers now feeling an urge to head to the garden shed, to caress and gaze fondly on measuring instruments. I'm impressed with myself too now I've remembered we worked to 0.015mm, especially as a lot of it was prototype small batches. I bet some of the work involved in building race machines is nothing less that fine tolerance art.

Thanks for the teaser David & Greg McDonald, more please.

Yes I remember GMD Computrack from last century, thought it was interesting then. Still interested.

Jinx I am really enjoying your stuff, thanks. Would you like to explain the difference between Precision & Accuracy ?

Anyone have a good explaination of the physics of how & why a  motorcycle turns?

This will get me (ex-fitter/machinist) through the long off season.

So one final post. This is my simple understanding which more knowledgeable folk can no doubt correct if wrong. Turning is about pivoting around an invisible point. Imagine a horizontal line at 90 degrees to the centre line (axle) of each wheel. As you turn the handlebars those lines intersect, and where they intersect, that's the pivot point. The more you turn the front wheel the smaller the pivot radius. Leaning shifts the front contact point further forward in relation to the axle so the more you lean into the turn, again the smaller the pivot radius. Similarly, the shorter the wheelbase the less turn or lean needed to make a given turn radius and vice versa. Ideally you wouldn't lean at all (but where's the fun in that) because you sacrifice grip but you need the fulcrum effect of shifting weight inwards to counter centrifugal force. So leaning has two effects.  Don't ask me about centrifuges, gravity and all that stuff, you need a proper engineer or physicist for that. But the same is true with turning (changing the angle of the front wheel), you sacrifice grip. And I imagine in racing at this level you are constantly trying to find the sweet spot between turn and lean that just gets you round the bend without losing grip to the point of crashing, including allowing the wheels to slip/slide to alter the pivot point.



...and then a rider shows up on your bike and does something qualitatively different. Stoner (what if we rear-wheel pivot and massage the throttle?) was a favorite example prior to our current Alien's Alien (what if we drift-skate slide the front AND rear?).

Anyone going to bite on V4 vs I4 chassis dynamics/geometry? We've got all Winter in here in 5 short days...

P.S. pondering chassis, riding styles, and appreciating Michelin, great tires out there now!

No worries, mate.

Precision is the reproducibility of a reported measurement. I.e., the degree to which reported dimensional values of the same physical characteristic(s) are in agreement with each other. Precision by itself does not ensure Accuracy.

Accuracy is the uncertainty of a reported measurement with respect to the actual dimensional values of the measurand. I.e., the documented maximum variation between the dimensional values being reported and the actual dimensional values of the physical characteristic(s) being measured. Accuracy by itself does not ensure Precision. The accuracy uncertainty value is always stated as part of the reported measurement.

Resolution is the degree of detail to which dimensional values can be reported by the measurement system. I.e., a resolution of 0,0001mm can theoretically report dimensional variations of a smaller magnitude than a resolution of 0,01mm. But resolution does not by itself ensure either Precision or Accuracy. A measurement system with a lower resolution can be more accurate or more precise (or both) than a measurement system with a higher resolution. A measurement system with an artificially increased resolution may lack the sensitivity to reliably detect the smaller magnitudes of dimensional variation required to support the (perceived) higher resolution. This is a case where resolution exceeds sensitivity (which may lead to measurement reporting errors...a situation all too common when vendors use the displayed resolution of a measurement system primarily as a sales tool).

Sensitivity is the magnitude of the dimensional variations that can be reliably detected. A measurement system may be able to detect dimensional variations of a smaller magnitude than it can report, because its sensitivity is greater than its resolution. This is the preferred relationship between sensitivity and resolution.

Measurand is the sum of the dimensional characteristic of the feature(s) we are measuring. I.e., the physical characteristic(s) defining Size, Form, Orientation, Location, or Profile.

Below is a commonly used illustration to show the differences between Accuracy and Precision. Cheers.


to operate to these tolerances when making a frame, then suspend it off three shop-bought springs soaked in oil, on top of two rubbery bags full of air. 

@jinx, could you possibly crack on and write a full article for this esteemed website? Preferably one about motorcycles although I suspect you could probably write an interesting and informative article about the nocturnal habits of the Komodo dragon. Thanks in advance from me + all visitors to the motormatters website.

Jinx is correct.  Do not assume anything when dealing with a machine that demands precision.

Are both front springs the same length and strength?

Is the steering stem aligned to the centre line of the motorcycle.  

Are the fork clamps true?

Do the forks stroke through their travel with no binding - when the front wheel is installed?

Is the swing-arm pivot mounted perfectly horizontal?

Is the swing-arm pivot mounted so that the right and left-hand ends are the same distance from the front axle ends?

Is the rear suspension unit mounted so that it does not bind as it goes through its stroke?

Are the rear suspension linkages free from any 'slop'?

Are the front and wheels all exactly the same?



They make reading the not so interesting post worthwhile !!!!

But remember it's summer down under - we have plenty to do so will find it hard keeping up with your writings :-)
Actually I'm in Sydney, so we don't even really have "Winter" any more - that's for places like Phillip Island (any time of the year).

"We never hear how the impact of the chassis flexing affects stiction (sliding friction – the resistance of the fork stanchion from sliding in the outer tubes) - the most common problem we find. Riders often misinterpret excess stiction as tyre grip, but even small binding in suspension movement takes away the feel of early warning from the front tyre. Even a small twisting of the forks from a crash, or in practice sessions or the race (often due to less than optimal fork clamps) can cause stiction."

Many times have we seen a rider release the front brake lever after braking for a corner, THEN have the front tyre tuck under, sending bike and rider sliding down the track.  It would be interesting to learn what the reason is for this.  One would expect most front-end instability when the front suspension is compressed under heavy braking, thus reducing trail.  So why the crashes following the release of the brakes when the front suspension is extending and trail is increasing?  Cannot wait for the next article.