- Home
- Custom & OEM
- Custom & OEM Motion Platforms
Custom & OEM · The mechanism, and what changing it costs
Custom motion platforms, sized from your load case rather than a catalogue row.
Most enquiries arrive as one changed number — heavier, taller, wider, faster. On a parallel motion platform no number changes on its own: raise the payload and the actuator grows, the legs get heavier, the first structural mode falls, and the bandwidth somebody promised falls with it. This page is about which numbers are ours to move, what each one takes from the others, and where the honest answer is a standard machine.
Only have a mass figure so far? That is one of three numbers, and it is the least decisive of them — the payload section shows what centre-of-gravity height and inertia do to the same 500 kg.
6DOF motion platform with electric linear actuators supporting a 1,000 kg test load.
The thing that makes custom different from configurable
Nothing on a motion platform moves one number at a time
A configurator lets you pick a table size, a stroke and a payload as if they were independent. On a real machine they are three views of the same structure, and moving one of them spends something you were relying on somewhere else.
Take the most common request there is: the same platform, but for twice the payload. The actuators have to supply roughly twice the static share and twice the inertial force, so they move up a frame size. A larger cylinder has a larger tube, a heavier rod and a heavier motor, so the moving assembly gains mass that has nothing to do with your payload. That extra mass lowers the first structural mode of the assembly, and the first mode is the ceiling on usable bandwidth. The machine that was quoted as a faster one arrives slower.
None of that is a reason to avoid custom work. It is the reason a custom platform has to be sized as a system rather than assembled from a list, and it is why the useful early conversation is about the four or five quantities below; a model number comes out of it, not into it.
The propagation grid on the right of this page is not a sales device. It is the order in which our own sizing runs, and it is the reason a serious quotation takes longer than an afternoon: each row has to be closed before the next one can be evaluated, and closing the last row sometimes sends you back to the first.
A custom platform is not a list of options you add. It is a budget you spend.
Been given a number by a supplier that looks too good for the payload? Send the specification and we will say which row it skipped →
Raising stiffness to put the mode back where it was is a structural change, not a bigger motor. That is the pass most quotations skip.
Custom motion platform diagram showing how increased payload leads to larger actuators, greater moving mass, lower natural frequency and reduced usable bandwidth.
Read your own request in the left column
The 6-Line Change Propagation Grid
Six requests cover almost everything anyone asks a platform builder to change. Each one has a predictable second effect, a cost that lands somewhere the enquiry was not looking, and one input that has to arrive before it can be answered at all.
The 6-Line Change Propagation Grid What you ask for, what moves with it, and what we need to answer
| You ask to change | What moves with it | What it costs, and where | What has to arrive first |
|---|---|---|---|
| More payload | Actuator frame size, leg section, moving mass, base footprint, drive and supply rating | Bandwidth, through the first structural mode. Floor loading, through the base reaction | Mass, centre-of-gravity height and inertia — all three |
| More stroke | Leg length, tube diameter, joint size, closed height, guarding envelope | Buckling margin, which is recovered by making the leg fatter and therefore heavier | Which axis actually needs it, and at what payload |
| Bigger table | Joint circle radius, table stiffness and mass, tilt available from the same stroke | Angular travel. A wider table gives less tilt from the same leg stroke | The real mounting pattern, not the bounding box |
| Higher acceleration | Motor and drive sizing, thermal duty, structural loads, anchoring | Continuous rating. Peak is cheap; the duty cycle behind it is not | A duty profile — amplitude against time, not a peak figure |
| Lower height | Leg inclination, joint layout, force resolution into the actuators | Efficiency. A flatter leg angle needs more actuator force for the same vertical load | The hard ceiling and the pit depth, dimensioned |
| Different interface | Command layer, coordinate origin, fault reporting, safety terminations | Usually nothing structural — this is the cheapest column on the page | What your host already outputs, in its own units |
Line 06 is worth noticing precisely because it is cheap. Interface work, coordinate conventions, command sets and fault behaviour are software, and software is where a machine built in this company can be bent without a structural consequence. The details of that are on the controller and API page; everything else on this page is mechanical, and mechanical changes are the ones with a bill attached.
Lines 01 to 05 all share one property: none of them can be answered from a single number. That is not a stalling tactic. A supplier who quotes line 01 from a payload figure alone has either sized for the worst case you did not describe, in which case you are paying for a machine you did not need, or sized for the best case, in which case you find out during acceptance.
Recognise your project in one of these six lines? Send that line with your numbers against it →
The most common gap between an enquiry and a quotation
Payload is three numbers, and most enquiries send one
A mass figure describes what the platform has to hold up. It says almost nothing about what the platform has to hold still. Two payloads of identical mass can differ by a factor of four in the moment they demand, and the difference is entirely in how they are mounted.
What the height of the centre of gravity does
Take 500 kg on the table. Accelerate it horizontally at 0.5 g and the payload pushes back with about 2.45 kN. That force acts at the centre of gravity, so it also applies a moment about the table equal to the force times the height of that centre of gravity above it.
With the mass mounted low — a centre of gravity 0.3 m above the table — the moment is roughly 0.74 kN·m. Raise the same 500 kg so its centre of gravity sits 1.2 m up, and the moment becomes about 2.94 kN·m. Tilt the platform 15° and gravity adds another 1.52 kN·m at the high mounting against 0.38 kN·m at the low one.
Reacted across joints on a circle of 0.8 m radius, that difference is about 2.1 kN of extra force on one leg and the same amount taken off the leg opposite — a first-order figure that the actual joint geometry moves, but not by enough to change the conclusion. The same mass, mounted four times higher, asks for four times the moment.
What inertia does, and why it is worse than it looks
Angular acceleration needs torque, and torque needs the payload’s mass moment of inertia about the axis the platform rotates around — not about the payload’s own centre. The parallel-axis term is the one that surprises people.
A 500 kg payload with an inertia of 150 kg·m² about its own centre, mounted with that centre 1.2 m above the motion reference point, contributes an additional 500 × 1.2² = 720 kg·m². The mounting height has added nearly five times the payload’s own inertia. Size the pitch and roll axes from the 150 figure and the machine will be short of torque by a factor that no amount of tuning recovers.
This is why the enquiry form on this page asks for a height. It is a single dimension, it is usually available from a general-arrangement drawing, and it changes the answer more than any other number in the brief.
Where you mount it usually matters more than what it weighs.
Comparison infographic of two 6DOF motion platforms showing a 500 kg payload at low and high mounting heights, highlighting center-of-gravity height and increased overturning moment.
The 3-Number Payload Statement What each number decides, and what happens when it is missing
| Number | Usual source | What it decides | If it is left blank |
|---|---|---|---|
|
Mass kg, moving assembly | Weighed, or summed from the bill of materials | Static leg force, drive continuous rating, base reaction, floor loading | Nothing can start. This is the only one nobody forgets |
|
CG height m above the table | A general-arrangement drawing, or a two-scale weighing | Overturning moment under acceleration and under tilt; peak leg force; joint sizing | We assume a conservative height, and you pay for actuators you may not need |
|
Inertia kg·m² about each axis | The CAD model of the payload, or an estimate from its geometry | Angular acceleration achievable, torque margin, servo tuning limits | Estimated from mass and envelope, which is the roughest figure in the whole sizing |
Where a payload does not exist yet — a cabin still being designed, a test article not yet built — send the target and the tolerance you expect to hold it to. A stated range is a design input. A number that quietly grows by 30 % between order and delivery is a different machine.
One more distinction worth keeping straight, because two suppliers can quote the same machine and mean different things: gross moving load is everything above the actuators including the table and any fixture, while payload is your article alone. Every figure on this site is gross moving load unless it says otherwise.
Have the mass but not the height or the inertia? Send the general-arrangement drawing and we will take the other two off it →
Why the same request costs different amounts on different machines
A longer stroke arrives as a fatter, heavier leg
Stroke looks like the cheapest thing to change on a parallel platform. It is not, and the reason is a piece of nineteenth-century structural mechanics that has not become any less true: a strut in compression fails by buckling long before it fails by crushing, and its resistance to buckling falls with the square of its length.
Euler’s result gives the critical compressive load of a slender strut as proportional to its bending stiffness divided by the square of its effective length. Double the length and the critical load drops to a quarter of what it was — the same tube, the same steel, one-quarter of the margin.
Recovering that quarter means raising the second moment of area of the section by a factor of four, and for a tube the second moment of area rises with the fourth power of the diameter. So the diameter has to grow by roughly the fourth root of four, about 41 %. That larger tube is also about four times heavier, because it is both fatter and longer.
Which lands back at the top of the page. A stroke request that looks like a small change to a drawing turns into a leg that weighs several times what the original weighed, six of which are part of the moving assembly, which lowers the first mode, which lowers the bandwidth. Long-stroke machines are not slower because of some vague engineering compromise. They are slower for a reason you can compute.
The other thing stroke buys, and the geometry that takes it away
Angular travel on a Stewart platform comes from a difference in leg lengths across the joint circle, so the same stroke gives less tilt on a bigger table. To tilt the top platform by an angle, two opposed joints have to separate vertically by about twice the joint radius times the sine of that angle.
With 400 mm of differential stroke available, a joint circle of 0.8 m radius gives roughly 14.5°. Widen the table so the joints sit at 1.6 m and the same 400 mm gives roughly 7.2°. Nothing about the actuators changed. The customer who asked for a bigger table and kept the same tilt specification has asked for a longer-stroke machine without knowing it.
These are first-order figures. The real numbers come out of the kinematic model with the actual joint positions, leg inclinations and hard stops in it, and that model is run before a quotation goes out, not after it.
Stroke is bought with diameter, and diameter is paid for in bandwidth.
Capacity falls with the square of length, so the diameter has to rise by the fourth root of the shortfall to bring it back.
Side-by-side comparison of two servo electric cylinders with different stroke lengths, showing that the longer-stroke cylinder has a larger tube diameter and heavier body.
The Stroke-to-Structure Ladder What a stroke increase actually asks for, rung by rung
| Rung | What changes on the drawing | What it does to the machine | Where it shows up for you |
|---|---|---|---|
| R1 Leg length | Longer tube and rod, longer screw or belt run | Buckling capacity falls with the square of length | Nothing yet — this is the rung that looks free |
| R2 Section | Tube diameter and wall increased to restore the margin | Second moment of area recovered; mass rises steeply | Unit price, and the crate dimensions |
| R3 Joints | Larger spherical and universal joints to carry the higher force | Joint mass and clearance both rise; preload becomes more critical | Backlash specification, and the calibration effort |
| R4 Closed height | Minimum retracted length grows with the stroke | The machine gets taller even at rest | Ceiling height, pit depth, door and crane clearance on your site |
| R5 Moving mass | Six heavier legs, a stiffer table to match | First structural mode falls; usable bandwidth falls with it | The frequency content you can actually reproduce |
| R6 Drives | Larger motors and drives, higher supply rating | Continuous torque restored at the new inertia | Electrical supply, cabinet size, heat into your room |
The ladder is worth reading in both directions. Read downwards it explains why a stroke increase is expensive. Read upwards it explains something more useful: a project that can accept a shorter stroke gets a lighter, faster, cheaper machine that fits under a lower ceiling. On roughly a third of the enquiries we see, the stroke asked for is inherited from an earlier specification rather than derived from the motion the job actually needs.
Carrying a stroke figure from an older specification? Send the motion you need to reproduce and we will derive the stroke from it →
The specification most often quoted and least often qualified
Mass moves the frequency before it moves the force
A platform is a mass on a set of springs, and every mass on a set of springs has a first natural frequency. That frequency is the ceiling on what the machine can be asked to reproduce, and it is decided by the structure long before anyone tunes a controller.
Natural frequency rises with the square root of stiffness and falls with the square root of mass. The practical consequence is blunt: adding payload lowers the frequency whether or not the actuators can still carry the load.
Suppose a moving assembly weighs 420 kg and its first mode with a 500 kg article on it measures 18 Hz. Add another 500 kg and the total moving mass goes from 920 kg to 1,420 kg. The frequency falls by the square root of that ratio, to about 14.5 Hz. To put it back where it was, the structure needs 54 % more stiffness — a new table, a different rib pattern, sometimes a different joint layout. It is not something a larger motor fixes.
Usable command bandwidth then sits well below that first mode, because approaching a resonance means the machine amplifies rather than follows. Where the job is faithful reproduction and not a felt cue, a separation factor of three to five between the highest commanded frequency and the first mode is the range these machines are designed into.
What the manual’s own figures say about the top of the range
Two figures from the 2024 product manual are worth quoting because they bound the conversation. The vibration-class machines listed there are specified to a system response of 20 Hz or better with a frequency deviation within 0.1 Hz, and the micro-vibration configuration is specified to a maximum working frequency of 25 Hz. Those are specific machine classes, not a blanket claim — a 30-tonne motion base does not do 25 Hz and never will.
Which is the useful point. Frequency capability on a parallel platform is a property of one design at one payload, so it belongs in a project specification with the payload and the amplitude written next to it. A frequency quoted on its own tells you nothing about the machine that would arrive.
Payload capacity and bandwidth are the same budget seen from two ends.
The outer bar is the first structural mode; the shaded block inside it is the band a command can actually use, at a separation factor of three. Both are anchored on a measured mode rather than a modelled one — modal testing on the finished machine is what puts the anchor there.
The outer bar is the first structural mode; the shaded block inside it is the band a command can actually use, at a separation factor of three. Both are anchored on a measured mode rather than a modelled one — modal testing on the finished machine is what puts the anchor there.
The 5-Layer Bandwidth Ceiling Which layer is holding you back, and whether money can move it
| Layer | What sets the ceiling | Can it be raised? | What raising it costs |
|---|---|---|---|
| L1 Structure | First natural frequency of the moving assembly with your payload on it | Only by redesign — stiffer table, different rib pattern, different joint layout | A new design cycle. This is the hard ceiling and it is set before anything else |
| L2 Actuator | Peak force and peak velocity at the required amplitude | Yes, by moving up a cylinder frame size | Mass, which pushes L1 down. The two fight each other |
| L3 Drive | Current loop and continuous thermal rating over the duty cycle | Yes, by a larger drive and motor, or better cooling | Cabinet size, supply rating, heat into your room |
| L4 Controller | Interpolation cycle — 500 µs on the dual-core EtherCAT master | Rarely the limiting layer on a mechanical platform | Nothing, in most projects. Detail on the controller page |
| L5 Your host | The rate your own software produces a new pose, and the link it uses | Yours to raise, and often the cheapest place to look first | Your development time, not a machine change |
The reason this table exists is that bandwidth complaints usually arrive pointed at the wrong layer. A platform that will not follow a 12 Hz command may be limited by its structure, by its actuators at that amplitude, or by a host that is only producing 50 new poses a second. Those are three different bills, and only one of them is a machine problem.
Have a frequency requirement but no amplitude next to it? Send the signal you need reproduced and we will tell you which layer decides it →
Scope, stated before a drawing exists
What we change, what stays catalogue, and what we will not do
Custom work is only useful if its edges are visible. Three lists follow: what is genuinely ours to move on a machine, what is bought in on purpose and stays that way, and what we decline. The third list is the one worth reading.
Ours to move
- Stroke and travel envelope on every axis, within the structural consequences above
- Table size, mounting pattern, material and stiffness; motion reference point placed where you need it
- Joint layout — six-strut parallel, three degree-of-freedom, and the two- and seven-axis variants
- Actuator frame size, lead, gearing and the resulting force and velocity envelope
- Base footprint, anchor pattern, closed height and pit or plinth arrangement
- Cable routing, slip rings, guarding, service access and lifting points
- Command layer, coordinate origin, units, sign convention and fault reporting
- Documentation set, language, marking, spare-parts list and acceptance procedure
Deliberately catalogue
- Motors, drives and encoders, in standard commercial frame sizes
- Bearings, gearboxes and screws from mainstream suppliers
- Fasteners, cable, connectors and pneumatic fittings in metric standards
- Fieldbus physical layers — EtherCAT and CANopen as published, not a private variant
We will not
- Supply a hydraulic or pneumatic motion platform. The range is servo electric and stays that way
- Supply a certified functional safety system as a product, or sign a safety case for an installation we do not own
- Certify a passenger-carrying amusement device to EN 13814 or ASTM F2291 — we build to those inputs, we do not issue the certificate
- Build for explosive atmospheres or ATEX-classified zones
- Fit our controller to someone else’s mechanism without an engineering review first
- Quote a custom platform from a payload figure alone
- Take on sub-millimetre precision-positioning work below roughly 20 kg — that is a different instrument class and a different supplier
This list is a design decision, not a cost saving. A maintenance team in another country can source every item on it without contacting us, which is the point. A machine whose spares only exist in one building in Nanjing is a machine with a hidden liability in it.
The Custom-Change Boundary Crosswalk Six requests we get regularly, and the honest answer to each
| The request | Answer | Why | What usually happens instead |
|---|---|---|---|
| “Same machine, twice the payload” | Yes | A different machine wearing the same name. Every line of the propagation grid applies | We size it properly and quote both, so you can see what the second one costs |
| “Match this hydraulic rig’s force in the same envelope” | Usually no | Hydraulic actuation carries more force per unit volume. Electric wins on control, maintenance and running cost, not on force density | We say so early and, where force density really is the requirement, that we are not the supplier |
| “A bigger table, same tilt angle” | Yes, at a price | Tilt comes from differential leg length across the joint circle, so a wider table needs more stroke | We quote both the wider table and the shorter stroke, because one of them is often acceptable |
| “Put your controller on our existing platform” | Review first | The mechanical baseline, the drives and the existing safety arrangement decide whether it is worth doing | A written review that sometimes concludes the machine is not worth the retrofit |
| “Something like your standard model, with three changes” | Yes | This is the most common and the cheapest kind of custom work | A configured machine off an existing platform, with a much shorter lead time than a new design |
| “One unit now, twenty a year later” | Yes | Repeat supply is a different engineering problem than a one-off, and it has to be planned from the first unit | A first article built under configuration control — see the repeat supply section |
Row two is on this page on purpose. Steering a buyer to the right machine, including when it is not one of ours, is how a company that builds non-standard equipment earns the second project. A hydraulic requirement that we pretend to meet electrically becomes a failed acceptance test with our name on it.
What is in the supply — scope of delivery, documents, installation, training, spares — is set out on the Custom & OEM overview. This page is about the machine itself.
Not sure which of the three lists your requirement falls into? Describe it in one paragraph and we will place it →
Before anything is cut
The machine is run in software before it is run on the floor
A one-off has no prototype to learn from, so the learning has to happen somewhere else. Four models are built and exercised before the first plate is machined, and each one answers a question that would otherwise be answered by an expensive surprise.
Kinematic model
Forward and inverse solutions for the actual joint positions, with the workspace swept for singularities, leg-length limits, joint-angle limits and self-collision.
Dynamic model
Actuator forces across the duty cycle, built from the Jacobian and a Lagrangian formulation of the moving assembly with your payload, centre of gravity and inertia in it.
Structural model
Finite element analysis of the table, base and legs under peak load, and a modal analysis that locates the first natural frequency of the assembly.
Multibody model
The whole mechanism run as a system so that control, structure and payload interact — the model that catches problems each of the other three would individually pass.
3D rendering of an enclosed simulator cabin on a 6DOF motion platform with six yellow linear actuators and a steel base frame.
Workspace Analysis of the Stewart Parallel Robot
Motion platform roll and pitch response plots showing time-domain signals, amplitude spectra and phase responses, including 0–2 Hz amplitude details.
The tooling is conventional, and naming it says more than any adjective would: finite element and modal work in ANSYS, multibody dynamics in ADAMS, kinematics and control in our own code alongside MATLAB. The structural design rule the models are checked against is the one in the 2024 manual — a minimum safety factor above 1.5× on structural members at peak load.
What the models are blind to
This is the more useful half of the section, because a simulation result presented without its blind spots is a marketing asset rather than an engineering one. Four things do not come out of these models with any authority:
- Damping in bolted joints. Modal frequencies from FEA are reliable; modal damping is not. It is measured on the finished machine, not predicted.
- Foundation compliance. The model assumes the base is held. A real floor, plinth or pit has its own stiffness, and on large machines it moves the first mode measurably.
- Cable and hose stiffness. Small, ignored in modelling, and occasionally the reason a lightly loaded axis has a stiction problem at low amplitude.
- Wear over service life. Joint clearance grows. The model describes the machine as delivered, which is why the acceptance record has to describe the same machine measured, not simulated.
Need the analysis before you commit to the order? Ask for the sizing study as a priced first step →
Manufacturing when the quantity is one
A batch of one is built differently, and it shows in five stations
Series production earns its accuracy from fixtures, jigs and a pilot run that finds the mistakes. A custom platform has none of those: the first article is the delivered article. So accuracy has to come from measurement at each station instead, and the parts are cut to a datum that was measured, not assumed.
ST 01
Cylinder build and run-in
Screw, nut, bearings and motor assembled, then each cylinder run against its own encoder through the full stroke and reversals until steady-state error and velocity accuracy settle.
ST 02
Joint preload
Spherical and universal joints set to a preload torque and checked for backlash under reversal, one joint at a time, before anything is welded to anything.
ST 03
Machine to a measured datum
Base and table faces machined, then the actual joint positions measured and recorded — the numbers that go into the kinematic model are the ones from this machine, not the drawing.
ST 04
Assembly and unloaded calibration
Full assembly, wiring, cabinet and controller. Home position, zero and mid positions set, then the kinematic model calibrated against measured pose instead of nominal geometry.
ST 05
Loaded dynamic test and FAT
Ballast built to your mass, centre-of-gravity height and inertia. Travel, velocity, acceleration, repeatability and the agreed motion profiles run at load, witnessed if you attend.
Station 02 is where a platform is won or lost
Joint preload is the least glamorous station on the floor and the one that decides how the finished machine feels. A parallel platform has thirty-six joint degrees of freedom between the base and the table. Every one of them with clearance in it contributes lost motion, and the contributions add rather than cancel.
Too little preload and the machine has backlash: reverse the direction and the table does nothing for a moment, which shows up as a dead band no controller can tune out, because there is nothing to control during it. Too much preload and friction rises, the low-amplitude behaviour turns sticky, and the joint runs hot and wears the very clearance you were trying to remove.
So the setting is a window rather than a target, it is set per joint with a torque figure and verified by a reversal check, and the sheet stays with the serial number. When a machine is serviced five years later, that sheet is the reason the replacement joint can be brought back to the same condition instead of to somebody’s judgement.
Lost motion is added at assembly and cannot be inspected out afterwards.
What station 01 produces, and why it is per cylinder
Every servo cylinder is run in on its own before it becomes a leg, and the acceptance figures are recorded per unit, not per batch. The servo cylinder class used across this range is specified in the 2024 product manual to a steady-state position error within 0.1 mm and a velocity accuracy within 10 mm/s, measured on the cylinder itself, not at the platform.
Recording it per unit matters for one practical reason. Six legs that each sit at the edge of a tolerance band in the same direction produce a platform that is out of tolerance; six that scatter produce one that is not. Knowing which cylinder is which lets them be paired and positioned instead of fitted at random.
Rendering of a servo electric cylinder on a run-in test bench with a monitor displaying encoder feedback and tracking error plots.
Station 03 is the one that separates a custom build from a configured one. On a series machine the joint positions are whatever the fixture holds them to. On a one-off they are measured after machining and fed into the kinematic model, so the controller solves the geometry of the machine that exists rather than the geometry that was drawn. That is also why the calibration data set is serialised: it belongs to one machine and is not interchangeable.
Want to see the records before you commit? Ask for a sample acceptance pack from a comparable build →
What each check catches, and what it cannot see
The 5-Gate Custom-Build Inspection Stack
Every inspection has a blind spot, and a supplier who lists only what their checks catch is telling you half of it. The right-hand column is the half that decides where problems actually escape.
The 5-Gate Custom-Build Inspection Stack Gate, method, what it catches, what it is blind to
| Gate | Method | What it catches | Blind to |
|---|---|---|---|
| G1 Incoming | Certificates and dimensional check on machined and bought-in parts | Wrong material, out-of-tolerance bores and faces, missing traceability | Anything that only appears once parts are assembled together |
| G2 Cylinder run-in | Full-stroke and reversal run against the cylinder’s own encoder | Steady-state error, velocity accuracy, early bearing and screw defects | How that cylinder behaves as one of six in a coupled mechanism |
| G3 Assembly | Joint preload torque, reversal backlash check, as-built joint coordinates | Lost motion, geometry drift from the drawing, misassembly | Anything load-dependent — deflection under a payload that is not there yet |
| G4 Unloaded calibration | Measured pose against commanded pose across the workspace | Kinematic error, sign and origin mistakes, encoder offsets | Structural deflection, thermal drift and everything the payload changes |
| G5 Loaded FAT | Full ballast at your mass, CG height and inertia; agreed profiles run and logged | Real travel, velocity and acceleration at load; repeatability; thermal behaviour over a run | Your site — its floor, its supply, its ambient, and how your host software behaves |
Gate 5 is the one worth attending or witnessing. Everything before it is a check on our own work; gate 5 is the first time the machine does something recognisably like the job. Third-party witnessing is accepted, and an inspection body appointed by you is welcome on the floor for it — the requirement is notice, so the ballast and the profiles are ready when your inspector arrives rather than being built while they watch.
The blind column also explains the one thing a factory acceptance test cannot settle. Site conditions are outside it by definition. Where a project has a floor, a supply or an ambient condition that is genuinely marginal, that belongs in the design inputs at the start, not in a site acceptance argument at the end.
Have an inspection body or a witnessed-test requirement in your procurement rules? Send the requirement and we will write the FAT procedure around it →
The commercial half, stated rather than discovered
What a custom platform costs that a catalogue platform does not
A bespoke machine carries costs a catalogue machine has already absorbed, and the ones that cause arguments are the ones nobody named at the start. Eight lines, with what we do about each.
The Custom Cost Ledger Where the money and the risk actually differ
| Line | On a catalogue platform | On a custom platform | How we bound it |
|---|---|---|---|
| Engineering / NRE | Already spent, spread across a series | Real, and spent before anything is manufactured | Quoted as its own line before design starts. Where the work produces something reusable, it is lower, and we say which case applies |
| Lead time | Short — the design exists | Design, review, non-stock procurement, build and loaded test, in that order | A week band per stage in the quotation, with the dates you have to hit marked as clearly as the ones we do |
| First-article risk | Absorbed by the series before you saw it | Genuine: the first machine is the delivered machine | Four models run before cutting, a design review gate where change is still cheap, and a loaded FAT before shipment |
| Change after freeze | Not applicable | Expensive, and more so the later it lands | An interface and geometry freeze with a date on it. Changes after it are quoted rather than quietly absorbed into the schedule |
| Spares | Catalogue part numbers | Some parts exist only on your machine | Commercial components wherever the design allows; a recommended spares list at delivery; bespoke part drawings held against your serial number |
| Documentation | A generic manual | Machine-specific: as-built geometry, calibration data, acceptance record | Serialised and delivered with the machine as standard, not produced later on request |
| The second unit | Identical price | Cheaper than the first — but only if the first was built under configuration control | Decided at the first unit, not at the second. See the repeat-supply section below |
| Obsolescence | Handled by whoever owns the catalogue | Yours, across a service life of ten to twenty years | Standard drives and motors, published fieldbus protocols, and a controller written in this company so it can still be rebuilt when the hardware moves on |
Two questions the industry avoids answering
Line 04 deserves one more sentence, because it is the line that damages projects. A geometry change costs almost nothing during concept, something during detail design, a great deal after long-lead parts are ordered, and more than the change is worth after assembly. This is not a commercial rule; it is the shape of manufacturing. Publishing the freeze date early is the only mechanism that helps.
Need an engineering budget figure before you can open a project? Ask for the sizing study priced separately, with its output defined →
For OEMs and equipment builders
Making the second one identical is a decision taken at the first one
Most repeat-supply problems are not manufacturing problems. They come from a first unit built as a one-off, with substitutions made on the floor and never written down, so unit two is a different machine wearing the same drawing number.
An OEM buying a platform as a component of their own product needs something a project buyer does not: the machine that arrives in year three has to behave like the one their integration and their documentation were built around. That is a configuration-control problem, and it is cheap to solve at unit one and expensive to solve at unit five.
The Repeat-Supply Freeze List What gets locked at first article, and what happens when it has to move
| What is frozen | Why it matters to your product | If it has to change |
|---|---|---|
| Bill of materials, to the part number | A different bearing or drive can change feel, noise and service parts without changing any published figure | Notified before the affected unit, with the reason and any characteristic that moves |
| As-built geometry and tolerances | Your mounting, your enclosure and your kinematic assumptions all depend on it | Treated as a new first article, with its own inspection report |
| Controller firmware and configuration | Behaviour on fault, on timeout and at limits has to stay identical across units | Version-stamped; the previous version stays available for units already in the field |
| Coordinate convention and command set | Your software is written once, not once per unit | It does not change. A new command set is a new interface document |
| Acceptance procedure and pass criteria | The same test on every unit is what makes the records comparable | Agreed once with you and re-issued only by agreement |
| Marking, labelling and documentation set | Your product carries your identity, not ours, where that is the arrangement | Your artwork and your document template, held against the configuration |
What we need from an OEM that a project buyer never has to give us
- A payload tolerance, not a payload. Your product will change over its life. Tell us the band you intend to stay inside and the machine is designed for the band.
- An indicative annual quantity. It changes tooling decisions and long-lead purchasing, and it is the difference between a sensible price for unit two and a repeat of unit one.
- Who owns the safety case. A platform integrated into your product is your machine when it leaves your building. That has to be stated in the contract, not assumed by both sides in opposite directions.
- Your document template and language. Re-issuing manuals into an OEM template after the fact costs more than doing it once at the start.
Building a platform into a product you sell? Send the product context and the quantity you expect, and the first reply covers configuration control →
Six shapes the work keeps taking
The custom configurations that recur, and what was non-standard about each
Customer names are not published here. What is useful without them is the shape of the work: the sector, the load band, the axes and the specific thing that could not be met from a catalogue. If one of these looks like your project, the conversation starts several steps further along.
The 6-Class Custom Configuration Map Sector, band, and the constraint that made it custom
| Class | Sector and job | Gross moving load band | What made it non-standard |
|---|---|---|---|
| K1 Instrument head | Sensor, antenna and optical pointing on a moving base | Under 300 kg | Angular accuracy specified at the aperture rather than the table, so the motion reference point moved onto the instrument |
| K2 Cabin and cockpit | Flight and vehicle simulation, training devices | 0.5–3 t | Occupant centre of gravity high above the table, and a tilt envelope driven by cueing rather than by geometry |
| K3 Test article fixture | Industrial and laboratory testing of a component under motion | 0.5–5 t | A profile to be reproduced faithfully at a stated frequency band, which set the structure before the payload did |
| K4 Research rig | University and institute platforms, human-factors and control research | 0.2–2 t | An open interface and a changing payload — designed for a load band and a decade of different experiments |
| K5 Compensation base | Equipment held still while its support moves | 0.3–10 t | The requirement stated as residual motion at the payload, which makes the sensor and the latency budget part of the mechanism |
| K6 Heavy motion base | Large vehicle and structural simulation, entertainment installations | 10 t and above | Floor loading, transport dimensions and installation sequence constraining the design as much as the motion did |
Each class links to the page that goes into the job rather than the machine: motion compensation, flight simulation, driving simulation, industrial testing and research platforms. The load bands describe the class, not a single delivered machine; project references with their actual parameters are available under a confidentiality agreement.
Recognise your project in one of the six classes? Name the class in your enquiry and the first reply starts from a comparable configuration →
The inputs that arrive last and decide the most
Four things that should be in the first email and almost never are
Mass, stroke and axis count arrive with every enquiry. These four arrive at the design review, by which point they have already changed the answer. None of them is difficult to produce; they are simply not the things a specification template asks for.
01 · The duty profile, not the peak
Motors are sized on two numbers. Peak torque sets what the machine can do for a moment; continuous torque sets what it can do all day, and continuous torque is governed by the root-mean-square of the duty cycle, not by its highest point.
The arithmetic is unforgiving in a useful direction. A profile at full torque for a fifth of the time and 30 % for the rest has an RMS demand of about 52 % of peak — so a machine sized on peak alone is roughly twice the motor the job needs. Read the other way: a machine sized on a duty cycle that was never mentioned will run into thermal limiting on site, and there is no adjustment that fixes it.
What we need is an amplitude against time, or a representative recording. Ten minutes of the real signal beats a page of adjectives about it.
02 · The frequency content, not just the amplitude
A 50 mm move at 0.5 Hz and a 50 mm move at 8 Hz are different machines. The first is a stroke problem; the second is a structure problem, because it lands against the first natural frequency rather than against the actuator.
Where a project has a spectrum — a sea state, a road profile, a flight or vibration spectrum — send the spectrum. Reproducing a defined spectrum is a specific engineering task with a specific acceptance test attached, and it is a very different order from “fast and smooth”.
03 · The site, in dimensions
Ceiling height, pit depth if there is one, door and crane access, floor construction and its allowable loading, electrical supply and what the room does with heat. These constrain a custom design as hard as the motion does, and unlike the motion they cannot be negotiated later.
The closed height of a platform grows with its stroke, so a low ceiling is not a detail — it is an input to the actuator selection. On the heavy classes, transport and installation sequence usually enter the design before the motion specification is finished.
04 · The interface datum
Which point on the machine your numbers refer to, and which way is positive. A pose means nothing until the origin and the sign of every axis are agreed, and the cheapest moment to agree them is before either side writes software.
Where your organisation already has a convention — a vehicle standard, an aerospace frame, a house rule — send it and we adopt it. Where there is none, we supply ours. What costs money is discovering at commissioning that both sides had one and they disagreed. The written convention lives on the controller and API page.
The 4 Late Inputs What each one costs when it arrives at the design review instead of the enquiry
| Input | Where to get it | What it decides | Cost of it arriving late |
|---|---|---|---|
| Duty profile | A recording, or an amplitude-against-time sketch of a normal working hour | Continuous motor and drive rating, thermal design, cooling | Re-sizing the drive train, or a machine that thermally limits on your floor |
| Frequency content | The spectrum or signal you have to reproduce, in the units it is defined in | Structural stiffness target and the first-mode separation | A structural redesign, which is the most expensive change on the list |
| Site dimensions | A marked-up general arrangement of the room, with supply details | Closed height, footprint, anchoring, transport split, cabinet placement | A machine that fits the specification and not the building |
| Interface datum | Your existing convention document, or ours if you have none | Origin, axis signs, units, command set, fault semantics | Commissioning time spent on an argument neither side can win |
The full list of what a quotation needs — commercial as well as technical — is on the Custom & OEM overview. The four above are singled out here because they are the ones that change the machine rather than the paperwork.
Have some of these but not all? Send what exists. An incomplete brief comes back as four questions, not a delay →
The pages this one deliberately leaves alone
Where to go next, depending on what is still open
Four pages answer questions this one raises: how the project runs, what the standard machines already cover, who owns the software, and what the trade-offs cost in numbers.
Process
How a custom project actually runs
Routes, review stages, scope of supply, the interfaces defined with you and what a quotation needs. This page is the machine; that one is the project around it.
Range
What the standard machines already do
Six-axis, three-axis and Stewart platforms with published envelopes. Worth reading first — a configured standard machine is faster and cheaper than a designed one.
Trade-off
Payload, stroke and acceleration
The same coupling this page opens with, worked through in numbers instead of in consequences — and why the duty profile changes a sizing answer more than a peak figure does.
Software
Controller and API integration
Where your host attaches, who owns the kinematics, the coordinate convention, and what happens when your software stops talking to the machine.
If the mechanism is still moving, this page and the overview are the right pair. If the geometry is settled and only the software is in question, go straight to the controller page — nothing on this page needs to be decided first.
Asked before a custom project is opened
Custom & OEM motion platform FAQ
What actually makes a motion platform custom rather than configured?
The line is whether the structure changes. Changing a table plate, a mounting pattern, a paint specification or a command set is configuration: the mechanism underneath is a machine that already exists and has already been proven. Changing the stroke, the payload band, the joint layout or the actuator frame size is custom, because it moves the natural frequency, the force envelope and the safety factors, and all three have to be re-established by analysis and re-proven by test. Most projects that arrive asking for a custom platform turn out to need a configured one, and that is the cheaper and faster answer wherever it is available.
How much payload can a custom platform carry?
The electric range runs from 20 kg to 35 t of gross moving load, and gross moving load means everything above the actuators — the table, any fixture and your article. Above that band it becomes a project figure, not a catalogue one, and the honest answer depends on the stroke and the frequency asked for alongside it: a heavy platform with a short stroke and a low bandwidth is a straightforward machine, and the same mass with a long stroke at 10 Hz may not exist at any price. Send the three payload numbers and the motion, and the answer comes back as a machine rather than as a maximum.
Why do you need the centre-of-gravity height when the mass is already known?
Because the mass sets what the platform holds up and the height sets what it has to hold still. The same 500 kg mounted with its centre of gravity 1.2 m above the table instead of 0.3 m produces four times the overturning moment under the same acceleration, and adds nearly five times the payload's own rotational inertia through the parallel-axis term. Sizing from mass alone therefore means guessing a height, and the guess has to be conservative, which means quoting a bigger machine than the job needs. One dimension off a general-arrangement drawing removes that.
Can you build a platform to match the force of a hydraulic rig in the same envelope?
Usually no, and it is better said now than at acceptance. Hydraulic actuation carries more force per unit of actuator volume than an electric screw does, so an electric machine that matches a hydraulic force in the same physical envelope is often not available. What electric wins on is control resolution, quiet running, no fluid, far lower standby power and much simpler maintenance — which is why every platform sold through this site is servo electric. Where force density genuinely is the requirement rather than a habit inherited from an older rig, we say so and we are not the right supplier.
Will one custom unit be quoted, or is there a minimum order?
The minimum order quantity is one, and a single machine is a normal order, not an exception. What changes with quantity is the amortisation of the engineering work, and that appears in the quotation as its own line so you can see it rather than infer it. Where a first unit is expected to be followed by more, say so at the enquiry: a first article built under configuration control costs a little more and makes every unit after it cheaper, and that decision cannot usefully be taken later.
How do I know the machine will meet the specification before it ships?
By a loaded factory acceptance test against a procedure agreed before manufacture, not by a certificate written afterwards. Ballast is built to your mass, centre-of-gravity height and inertia, and not stacked as flat plates, the agreed profiles are run, and the channels are logged and delivered as files. You are welcome to attend, and a third-party inspection body appointed by you is welcome on the floor — the only requirement is notice, so the ballast and the profiles are ready when your inspector arrives. What a factory test cannot settle is your site: its floor, its supply and its ambient conditions are outside it by definition, which is why they belong in the design inputs at the start.
What happens if our payload changes after the design is frozen?
It depends entirely on when, and the shape of that curve is worth knowing before you need it. During concept a change costs almost nothing. During detail design it costs drawing time. After long-lead items are ordered it costs money and schedule together. After assembly it can cost more than the change is worth. So the freeze has a date on it, that date is in the quotation, not mentioned later, and changes after it are quoted rather than absorbed silently into a schedule that then slips for reasons nobody has written down. Where a payload is genuinely still moving, the better answer is to design for a stated band rather than a number.
Why is your quotation higher than a supplier offering the same axis count and payload?
Ask the other supplier five things and most of the gap explains itself. Does their figure refer to gross moving load or to payload alone? At what centre-of-gravity height was it sized, and does the quotation say? Is the stated acceleration a single axis or a combined case? Is a loaded acceptance test included, or is the machine tested empty and commissioned at your site? And who wrote their controller — because that decides whether a change request three years from now has a price or an apology. Those five account for most of the difference, they are all cheap to ask, and every one of them is expensive to discover after delivery.
Commercial questions — payment terms, warranty, shipping, installation and training — are answered in the full FAQ.
Send the load and the motion. The first reply says what it takes.
Three payload numbers, the motion you need and the room it goes in are enough for a first pass. Where the right answer is a standard machine with a different table on it, that is what comes back — it is a shorter lead time and a smaller invoice, and it is the answer more often than this page might suggest.
What comes back first
- Whether this is a configured machine or a designed one, and why
- Which line of the propagation grid your request sits on, and what it takes from the others
- An indicative envelope — travel, velocity, acceleration and first-mode target at your payload
- What is quoted as engineering, what is quoted as machine, and the freeze date the schedule depends on