This article was first published in AutoSpeed.
Earlier in this series we’ve covered the ideas behind the new semi-leading
arm suspension design.
So what advantages did that approach have?
It’s a mouthful but we’ve described the way in which the design dynamically
adds camber and castor in bump and (to the outside wheel) in roll; how it has
the potential to be lighter than many other systems because of having only
single arms, and only two pivots for each of those arms; how it leaves plenty of clear space
forwards for pedalling legs; and how it allows the use of a low motion ratio
That’s all well and good – but how to make it?
Semi Leading Arms
Let’s take a look at the stresses and strains each front suspension arm will
be subjected to.
1. Vertical forces trying to bend the arm primarily upwards. These forces are
developed by the weight of the trike and in vertical bumps. However, because the
spring is out near the wheel, this component is quite small.
2. Horizontal forces trying to bend it backwards. These forces are developed as
the wheel runs into a bump, but, much more importantly and in the other
direction, when the brakes are applied. Braking develops the greatest forces
these arms need to withstand – and braking forces can also be applied very
suddenly. Braking also causes the leading arms to twist – torsion is applied.
3. Buckling and bending forces. These forces are developed when the vehicle
corners and a sideways load is developed.
Purists can pick holes in these points – most times, the arms are subjected
to multiple variations of these forces. But the only reason I went for
relatively thick 1.2mm wall thickness (a decision that caused me physical pain
with the thought of the extra weight!) is to withstand braking forces. There’s
nothing like actually being able to see the suspension in action to make you
realise how high braking forces are...
So the arms, each about 380mm long, were made from 35mm OD, 1.2mm wall chrome
moly steel tube.
On my first trike I used for suspension pivots the pictured hardened, chrome-plated steel
rods (ex car shock absorber shafts) that rotated in greased polyurethane bushes
(off-the-shelf suspension bushes from an obscure car). In terms of strength and
durability, this approach is a wonderful way to go. The assemblies absorb road
vibration and are very strong and durable.
But (1) they’re heavy. And (2) they allow deflection.
In fact, even when using the widely spaced double wishbones on my first
design, I found that in hard braking, the upper bushes noticeably deflected –
and so I replaced them with nylon. With this new HPV design, using a front
suspension design with just a single semi-leading arm (and so closely spaced
mounting points at the chassis end of the arm), it was very important that the
pivots have effectively no deflection at all.
And they needed to be light....light....light.
I looked long and hard at all sorts of pivot systems, including:
1. steel shafts running in greased bronze bushes (the system used in the
Greenspeed trikes for the kingpins and steering arm pivot)
2. plastic bearings made from acetal rod
3. plastic bearings injected as hot plastic (there are car ball-joint repairers
who have the equipment to do this)
4. rod-end bearings (ie Heim or rose joints)
5 ball bearings
But when I looked at the available load data, the cost and the weight, plain
old sealed ball bearings were way out in front.
I have an old NSK roller bearing book and it lists, for example, the stats on
a sealed roller bearing with a 30mm OD and a 10mm ID as 520 kg force dynamic and
229 kg force static - and yet each bearing weighs only 32 grams. That seems to
me to beat a plastic bearing with completely unknown specs and durability...
(Incidentally, note that mountain bike suspension systems have available very
roller bearings, but they’re rather expensive.)
The new design uses a suspension arm in the shape of a T (arrowed). The length of the
long arm of the T is 380mm and the cross-piece is 80mm long. The ball bearings
mounted in each end of the top of the T. The 30mm by 10mm bearings are a loose
fit in the 35mm OD tube, but if a 32mm x 0.9mm tube is slipped inside the 35mm
tube, the bearings are a nice fit inside the paired tubes. By cutting off short
sections of the 32mm tube, placing them inside the 35mm tube and then brazing
them into place, a very lightweight bearing mount could be made. Bearing
adhesive was used in the final assembly to absolutely secure the bearings.
Passing through the bearings are high tensile steel, counter-sunk head bolts,
one entering from each end of the assembly. To reduce weight, the two bolts
screw into high tensile nuts brazed to each end of a 19mm x 0.9mm wall chrome
moly steel tube positioned between the bearings. Weight is reduced
because a single through-bolt is not used, and this internal spacer also holds the
bearings correctly spaced apart.
The swing arms pivot on brackets folded from 1.6mm chrome moly steel sheet.
Rather than the bearing through-bolts passing through just the 1.6mm wall
thickness of the brackets, steel sleeves are used to spread the load.
Furthermore, the steel sleeves are flared so that the countersunk through-bolts
self-align when they are tightened. The flared sleeves were made by using an oxy
torch to heat the sleeves red-hot and then forcing them down over a sacrificial inverted
countersunk bolt. The resulting sleeves are very light, fit the countersunk
bolts perfectly, and were vastly easier made in this way than by a lathe.
A very important advantage of the semi-leading arm design is the reduction in
the number of pivot points.
As described above, the chassis suspension pivots of my first trike were made
from hardened steel rods working in polyurethane bushes. Taking the front double
wishbone suspension, there were four pivots per side, one for each base of the
upper and lower A-arms. That makes in total 8 chassis suspension bushes. Then,
on the wheel upright, there were two more pivots – small (but relatively heavy)
ball-joints. That makes another four pivots – so we’re up to 12.
In comparison, the semi-leading arms design has only four pivots in total.
Two are the chassis pivots (formed by the two pairs of ball bearings) and two
are the steering swivel bronze bush kingpins.
Whatever way those pivots are formed, a reduction from 12 to 4 is a very
major weight saving.
As any good suspension textbook will show you, swing arm suspensions (ie
where the suspension pivots are parallel with the longitudinal axis of the car)
have a roll centre that’s very high.
On the other hand, pure trailing arms have a roll centre that’s at ground
So what of semi-leading arms? This diagram, taken from 30m3performance.com, shows how the roll centre position is calculated for semi-trailing (or
semi-leading) arms. Basically, depending on the angle of the arms (and as you’d
expect from the above) it’s somewhere between ground level and very high!
So of what significance is the position of the roll centre? Well, the roll
centre height (lateral position isn’t an issue - it should be on the
centreline of the vehicle) will influence
the amount of roll that occurs in cornering. The roll centre is almost always
lower than the centre of gravity (COG), and the closer it is to the height of
the COG, the less roll that will occur. So having a high roll centre sounds a
pretty good thing, no?
The trouble is, the higher the roll centre, the more that the lifting force
being developed at the tyre tries to corner the vehicle. This diagram shows the
situation - there’s an upward component in the force vector. Now if the vehicle
rises on its springs during cornering, the rising pivots cause positive camber on both wheels, resulting in a loss
of grip. This development of positive camber is most likely if the vehicle
passes over a mid-corner hump. The result is sudden sliding: the trait for which
early Volkswagen Beetles and the Chev Corvair were known with their swing-axle
rear suspension designs.
But of how much importance is this to my HPV? The short answer is that until
I ride the thing hard, I honestly don’t know.
That’s not to say that I haven’t considered it in great detail - but it’s
just all too hard to try to work out on paper. For example, say that there’s
60kg vertical load on the front suspension. I corner hard enough that I lift the
inner wheel: now there’s 60 kg vertical load on just the outer wheel. If I am
cornering at 0.5g, that translates to 30kg lateral force. With a 30 degree angle
between the tyre contact patch and the roll centre, that 30kg lateral force
translates to (...thanks for the maths, Dad...) a 15kg vertical force. So the
front suspension gets lifted as if the weight has dropped by 25 percent.
But – so what?
There’s still the same downwards force on the tyre – the trike hasn’t got any
lighter. And even with a 15kg reduction in front ‘weight’, the camber will still
be negative, or at worst, zero. In short, yes the virtual swing arm is very
short, but that’s what I want to give me the major camber variation on bump. As
a result of this major camber variation and short virtual swing-arm, the roll
centre is very high – but again, does it matter?
You won’t find the answer in suspension design textbooks, and nor will you
find it in on-line resources devoted to suspension and handling of normal race
cars. However, I did find a gold mine of pertinent information on a discussion
group devoted to those who build their own off-road racers – people not afraid
to push the boundaries in terms of dynamic castor, huge camber variations and
yes – short virtual swing-arms. One clue came from the point: "We usually don’t
need much anti-roll bar, so our roll centres are pretty high" – the more you
think about it, the more interesting that line is....
But really, for this HPV, it’s very much a wait and see exercise.
Off-the-shelf Greenspeed kingpin assemblies are used. These set the scrub
radius at zero (that is, for 20 inch wheels, the steering axis intersects the
tread at the road half way across the tyre’s width) and they’re cheap and well
made. Using them was a case of using a known entity – in terms of strength,
contribution to total mass and scrub radius, I knew what I was getting. The
kingpins use bronze bush greased bearings.
In a way the term ‘spring support’ is misleading: it may be supporting the
top of the spring but in fact its most important function is to take the weight
of the HPV acting through that wheel.
But it’s an interesting type of weight – a very interesting weight
Let’s assume that the spring support takes only the forces acting through the
spring – in other words, the bump-stop is mounted somewhere else. This means
that the weight is cushioned and absorbed by the spring: the very high peaks
that would be experienced without the springs are gone. Hit a big bump and
instead of the impact being immediately fed into the frame, it’s ameliorated by
the compression of the spring. In short, the instantaneous loads that need to be
carried by the spring support are much lower than are carried by the frame of a
machine that doesn’t use springs.
Furthermore, and this is not obvious until you think about it, the
maximum weight acting through the spring support is only present when
there’s weight on the machine!
Huh? What does that mean?
Well, like a bicycle, the majority of weight of a loaded recumbent trike
comes from the presence of the rider. The trike itself might weigh only 20kg
unloaded – but 100kg when loaded. To put this another way, 80 per cent of the
weight of the loaded trike is made up of the person riding it. So, assuming
one-third of the weight acts through each wheel, an unloaded trike will have a
maximum force acting upwards on a spring support of about (20/3=) 7kg. That’s
very little. However, with the trike loaded, that jumps to about 33kg, with
dynamic increases to 66kg on 1g bumps.
Now, remember that all the rider’s weight is pushing down on the seat. This
in turn makes the loaded seat able to cope with pretty high vertical force
inputs – in fact, you need to push up with a combined force of more than
80kg before the seat even tries to lift from its mounts. So where is this
discussion going? In short, the front spring vertical loads can be fed into the
seat. You wouldn’t want to apply full loads to the frame without someone being
in the seat, but then why would that ever occur?
In fact, I initially decided to use simple extensions of the front seat
supports as the upper spring mounts. This meant the front spring supports were
made from small tube – just 19mm diameter x 0.9mm wall. The lower spring
supports, which mounted the base of the spring on the suspension arm, were also
made from the same size tube.
However, testing of the frame showed that the upper spring supports had major
deflection – as much as 15mm! This occurred for two reasons – (1) the 19mm tube
was not strong enough, and (2) my design placed part of the upper spring support
in torsion – and in torsion, the 19mm tube was especially weak. So despite the
increase in weight, there was nothing for it but to remake these supports in
much larger 32mm x 0.9mm tube (arrowed). The new design places this tube only in bending
(not torsion) and still incorporates the seat supports.
The 19mm tubular lower spring supports were retained.
(More on spring supports in the coming Part 11 of this series – which is on
the design and construction of the frame.)
One thing that immediately struck me when the suspension was first built was
the amount of track change that occurs as the arms move through their arcs. From
full bump to full rebound, the track alters by something like 130mm. (Since that
suspension movement should almost never occurs, in more normal use the lateral
movement is more likely to be about 20mm). In other words, the tyres are forced
to scrub sideways across the road as the suspension moves up and down.
The disadvantages of this include increased tyre wear and greater
But there is also a potential major advantage: the suspension movement is
damped by this scrubbing. In other words, the sideways movement of the tyres
creates resistance to suspension movement.
On my first HPV, where double wishbone front suspension was used, the track
changed through full suspension movement by about 20mm. Together with friction
in the suspension pivots, the damping caused by this track change was sufficient
that external dampers were not needed.
To put it another way: If the use of external dampers can be avoided, but
tyre wear is greater, it’s a good trade-off in terms of reducing HPV weight. But
of course, the degree of trade-off depends on the amount of tyre wear!
(Another thought: if the tyres’ sideways movement is providing through
damping, the lower the friction of the surface, the less damping that will
occur. This implies that on wet roads, the suspension will be less damped –
something that’s normally desirable!
Full deflection bump stops are normally considered to be intrinsic to good
suspension design: they’re the rising rate springs that take over to absorb and
cushion the last movements of the suspension. And especially when dealing with
the miniature Firestone airbags, which don’t have internal bump rubbers, the
presence of bump rubbers would seem to be critical.
However, after some thought I decided not to use front bump rubbers. Why?
Well, with 5 inches of suspension travel (yes – 130mm!) at the front and an
appropriate spring rate, the bump rubbers should be impacted hard only
extremely rarely – if ever. So with full bump rubbers you’re carrying around
with you the extra weight that may never even be used. Because the airbags are
clearly visible to the rider, it will also be fairly obvious when full bump is
being reached, and, even when that does occur, the frame and suspension is built
strongly enough to take it. I may later fit some very light expanded foam
bumpers, but at this stage I decided to leave them out of the front
However, full droop stops are absolutely required. This is because
whenever there’s no rider on the machine, these stops are brought into action –
get off the trike and the airbags immediately expand to their full constrained
length. The force the airbags apply in this situation is quite large, so the
full droop stops are subjected to high and frequent forces. After looking at
lots of options (lots – including stainless steel cable loops,
chrome-moly-tube-and-rubber-stops located at different positions, simple ropes –
and more!), to keep weight down to a minimum and provide high strength, very
heavy duty plastic ties were used.
As described earlier in this series, an anti-roll bar is a prerequisite if
linked front airbag springs are to be used. However, the anti-roll bar caused me
quite a lot of concern – it was very hard indeed to make it light.
Not knowing how stiff an anti-roll bar would be needed (that depends on
maximum cornering lateral acceleration, roll centre height, dynamic spring
stiffness and tolerable roll angle, sway bar geometry and linking) I went on the
experience of my previous suspension HPV design. In short, I thought that a
bloody stiff anti-roll bar would be needed! (Note that on a three wheeler, the
rear wheel contributes zero to roll stiffness....) I used 19mm x 0.9mm wall chrome
moly tube and linked to the leading arms about half way along their length.
The chassis mounts are fabricated steel D-shackles, mounted on the same
brackets that support both the front and rear suspension pivots, while the
bushes are single piece high density polypropylene, a self-lubricating plastic.
I initially intended to use rubber bushes each side of mounting plates to
cope with the angular change between the sway bar links and the suspension arms.
However, after spending nearly a full day making different prototypes (none of
which were light, strong and didn’t put side loadings on the links during full
suspension movement), I decided to use rose joints (ie rod-ends). These are 6mm
designs (ie the threaded ends and ‘eye’ holes are 6mm in diameter).
Note that the mounts do not place the through-bolts in double shear, which is
normal best engineering practice. This was done deliberately: double shear
mounts fail catastrophically, while single shear through-bolts will first bend
if the loads are too great. Without knowing the magnitude of the loads, I wanted
an early indication if the links weren’t up to the task. (I think a sway bar
failure at high downhill cornering speed would likely throw you off the
So the front suspension is built but how will it perform? With its short
virtual swing arms length and high roll centre, will it have problems in
‘jacking’? Will the huge variation in camber cause steering problems? Will the
change in castor (something really radical!) cause steering wander or massive
changes in steering weight? Will the changing track wear out the tyres in just a
Watch this space!
I had intended to tack-weld the trike together using an arc welder, and then
take the frame to a welder to have it TIG welded. However, tacking the very thin
wall tube with an arc welder proved impossible, even when using low hydrogen
rods designed for high tensile steel.
Instead, I changed approach dramatically and invested in an oxy acetylene
set-up. This allowed me to braze the frame together using high strength
flux-coated nickel-bronze rods. The huge advantage was that welding could be
done on-site and different design approaches could be tried and tested (eg
changing the upper front spring supports – see above). This speeded-up the
construction process immensely.
So how strong is the brazing? That’s a hard question to answer – but I don’t
think as strong as good TIG welding. I brazed together some sample tubes and
then took to them with a hammer. I couldn’t break the join. However, during
construction I broke several short tacks that I’d made. These tacks failed much
more easily than similar TIG’d tacks.
To retain maximum strength, I decided not to grind back any of the braze
welds. I also brazed fillets (ie built up the weld to quite great thickness)
where I thought additional strength was required.
The flux-coated rods use a flux that’s very hard to remove – a normal
hand-powered wire brush just runs over the top of it. The brazing material
itself is also very hard. As a result, during construction I decided to give
each join just a cursory wire-brushing. The finished frame will be sandblasted
before being powder coated; I’d expect the blasting to remove the flux properly.
(Black paint has been used here just to stop rusting during the testing
A final note: the flux-coated nickel-bronze rods are very expensive –as high
as AUD$7 each – and can be hard to obtain.