An eight-bar linkage is a planar mechanism built from 8 rigid links connected by 10 revolute joints, giving a single degree of freedom when the Grübler count works out. Where a 4-bar can only trace simple ovals and a 6-bar handles moderate curves, an 8-bar carries enough joints to synthesise complex coupler paths — flat-bottomed walking gaits, dwell-rise-dwell motions, near-straight lines over long stroke. Engineers reach for it when no shorter chain can hit every required pose. Theo Jansen's Strandbeest legs and the Klann walker both run on 8-bar geometry.
Eight-bar Linkage Interactive Calculator
Vary the link and joint counts to see the planar Grübler mobility and an animated eight-bar teaching diagram.
Equation Used
Grübler mobility counts how many independent inputs a planar linkage has. For the article example, n is total links including ground, j1 is the number of revolute pin joints, and j2 is any higher-pair contact constraint.
- Planar rigid-link mechanism.
- Revolute joints are one-DOF lower pairs.
- No redundant constraints or special singular geometry.
- Ground link is included in total link count n.
How the Eight-bar Linkage Works
An eight-bar linkage is a kinematic chain of 8 rigid bars joined by 10 pin joints, with one bar grounded and one bar driven. Plug those numbers into Grübler's equation — DOF = 3(n−1) − 2j₁ − j₂ — and you get DOF = 3(8−1) − 2(10) = 1, which is exactly what you want for a mechanism driven by a single crank or motor. The remaining 6 bars float as a coupler network, and that's where the magic happens. Each floating point on each link traces its own coupler curve, and the designer picks pivot positions until one of those curves matches the required output path.
Why 8 bars and not 6? Because a 6-bar Stephenson or Watt chain runs out of free parameters once you ask for more than about 5 precision points on a path. An 8-bar gives you roughly 16 independent dimensions to tune, which is enough to plant a foot flat, lift it cleanly, swing it forward, and set it down again — the four-phase walking gait Theo Jansen worked out in 1990. Get the link lengths wrong by even 1-2% and the foot either scuffs during what should be the swing phase, or it lifts too early and the body rocks. The Klann walker is more forgiving because it uses fewer precision points, but it still demands link tolerances inside ±0.3 mm on a 100 mm leg before the gait starts to limp.
Failure modes are almost always geometric, not structural. Pin-joint slop above about 0.1 mm radial play stacks across 10 joints into millimetres of foot wander. Pivot misalignment out of the plane causes the bars to bind near singular configurations — the points where the linkage briefly loses or gains a degree of freedom and the input torque spikes. If you feel a hard pulse twice per crank revolution, you're either crossing a singularity or one of your revolute joints has gone tight from a bent pin.
Key Components
- Ground link (frame): The fixed bar that anchors two of the 10 pin joints to the chassis. Its length and the spacing between the two ground pivots set the scale of the entire mechanism — change ground-pivot spacing by 5% and every coupler curve in the network distorts.
- Crank (driver link): The single input link, usually rotated by a gearmotor at 30-90 RPM. Crank length must be shorter than the shortest coupler bar by Grashof's rule, otherwise the linkage locks up before completing a full revolution.
- Coupler links (×6): The floating bars that carry the output path point. On a Jansen leg, three of these form the upper triangle and three form the lower triangle — link lengths of 38, 41.5, 39.3, 40.1, 55.8, and 39.4 mm in Jansen's published 'holy numbers' produce his characteristic flat-bottom gait.
- Rocker link: An oscillating bar pinned to ground at one end and to the coupler network at the other. Provides the constraint that converts crank rotation into bounded oscillation upstream of the output point.
- Revolute joints (×10): Pin joints with one rotational degree of freedom each. Radial slop must stay under 0.1 mm or stacked play across 10 joints destroys path accuracy. Bronze bushings or PTFE-lined bearings are standard; ball bearings only where the load justifies the cost.
- Output point (foot or tool tip): A specific point on one of the coupler links whose traced curve is the design objective. Move this point 2 mm along the coupler bar and the foot trajectory shifts noticeably — it's the single most sensitive parameter in the whole linkage.
Who Uses the Eight-bar Linkage
Eight-bar linkages show up wherever a designer needs a complex, repeatable path that no shorter chain can produce, and where the cost of a CNC or servo solution can't be justified. The classic uses are walking robots and walking sculpture, but the geometry shows up in suspension systems, packaging machines, and surgical instruments too. Anywhere you need a flat-bottomed dwell, a near-straight stroke, or a controlled rise-fall-rise profile from a single rotating input, an 8-bar earns its place.
- Kinetic sculpture: Theo Jansen's Strandbeest series — wind-driven walking sculptures on the Dutch coast, every leg a planar 8-bar with the published Jansen 'holy numbers' for link lengths.
- Educational robotics: The Klann linkage walker kits sold by EK Japan and various STEM suppliers — a Stephenson-derived 8-bar built into classroom-scale 4-leg robots that walk over rough terrain without needing a microcontroller.
- Automotive suspension: Multi-link rear suspensions on cars like the BMW 7-series E65 use planar-projected 8-link kinematics to decouple wheel toe, camber, and bump steer across the suspension stroke.
- Packaging machinery: Carton-erecting machines from Bosch and Marchesini use 8-bar mechanisms to fold flat blanks into 3D boxes in a single rotation of the input crank — the dwell phase holds the carton stationary while glue sets.
- Mining equipment: Walking draglines like the Bucyrus 8750 use 8-bar walking-shoe mechanisms to lift and shift the entire 7000-tonne machine 2.4 m per stride across the mine floor.
- Surgical robotics: Remote-centre-of-motion (RCM) instruments for laparoscopic surgery use 8-bar parallelogram chains to keep the tool tip pivoting about a fixed point at the trocar, regardless of how the surgeon moves the input handle.
The Formula Behind the Eight-bar Linkage
The single most useful number for an 8-bar linkage is its degree of freedom, calculated by Grübler's equation. You want exactly DOF = 1 — anything else means the chain either locks up or flops. The formula tells you whether your link and joint count is even viable before you start CAD work. At the low end of typical practice (8 links, 10 revolute joints, no higher pairs), DOF lands at 1 and the mechanism behaves predictably. Add a single redundant joint without an extra link and DOF drops to 0 — the linkage becomes a static structure that can't move. Drop a joint and DOF jumps to 2, which means you'd need two motors to control it. The sweet spot is the exact 8-link, 10-joint, 1-DOF configuration that defines the eight-bar family.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| DOF | Degrees of freedom of the mechanism — how many independent inputs are needed to fully define motion | dimensionless | dimensionless |
| n | Total number of links including the ground link | count | count |
| j1 | Number of lower-pair (1-DOF) joints — revolute or prismatic | count | count |
| j2 | Number of higher-pair (2-DOF) joints — cam or gear contacts | count | count |
Worked Example: Eight-bar Linkage in a beach-cleaning walking robot
A coastal-cleanup nonprofit in Lisbon is building a 4-leg walking robot to drag a sieving rake across sandy beaches at low tide. Each leg is a planar Jansen-style 8-bar with link lengths scaled 1.5× from Jansen's published holy numbers, giving a stride of approximately 180 mm. The crank is driven by a 24 V brushed gearmotor and the team needs to know whether the proposed link/joint count is kinematically valid, and how forward speed scales across the 20-100 RPM crank range the gearmotor allows.
Given
- n = 8 links
- j1 = 10 revolute joints
- j2 = 0 higher-pair joints
- Stride length = 0.180 m per crank revolution
- Crank speed range = 20 – 100 RPM
Solution
Step 1 — verify the linkage has the correct degree of freedom using Grübler:
Good. Single input, single output — exactly what an 8-bar should give you. If you'd accidentally added an 11th joint without an extra link, DOF would crash to −1 and the linkage would be over-constrained and immovable.
Step 2 — compute forward speed at nominal crank speed of 60 RPM. One full crank revolution moves the body forward by one stride length:
0.180 m/s is a brisk human walking pace scaled down — the robot covers a metre roughly every 5.5 seconds, which feels purposeful but not rushed when you watch it cross a beach.
Step 3 — at the low end of the gearmotor range, 20 RPM:
At 60 mm/s the robot looks like it's deliberating each step — useful when raking heavy wet sand because each foot has time to sink and grip before the next leg-pair takes load.
Step 4 — at the high end, 100 RPM:
0.300 m/s is theoretical. In practice, above about 75 RPM the swing-phase time on a Jansen leg drops below the leg-clearance window, the foot scuffs the sand instead of lifting cleanly, and the body starts pitching fore-and-aft on each step. The gait sweet spot for a 1.5×-scale Jansen leg sits around 50-65 RPM.
Result
DOF resolves to 1 as required, and nominal forward speed at 60 RPM is 0. 180 m/s. That's a comfortable walking-sculpture pace — fast enough to be useful for beach-coverage, slow enough to look mechanical and intentional rather than frantic. Across the 20-100 RPM range the speed scales linearly in theory from 0.060 m/s to 0.300 m/s, but real-world performance degrades above roughly 75 RPM as the foot starts dragging during what should be the lift phase. If the team measures forward speed 30% below predicted, the most likely causes are: (1) cumulative pin-joint slop above 0.1 mm radial across the 10 joints, which lets the foot wander vertically and shortens effective stride; (2) link length tolerance error greater than ±0.3 mm on the critical lower-triangle bars, which distorts the flat-bottom portion of the gait; or (3) a bent crank pin causing the linkage to bind at one specific crank angle, producing a once-per-revolution speed drop the operator may misread as low average speed.
Choosing the Eight-bar Linkage: Pros and Cons
Eight-bar linkages compete with simpler 4-bar and 6-bar chains on one side, and with cams or servo-driven CNC motion on the other. The choice comes down to how complex the required path is, how much you can spend, and how much accuracy you need.
| Property | Eight-bar Linkage | Four-bar Linkage | CNC servo + cam |
|---|---|---|---|
| Path complexity (precision points achievable) | Up to ~9 precision points | Up to ~5 precision points | Unlimited (programmable) |
| Typical operating speed | 20-150 RPM | 20-300 RPM | 1-3000 RPM |
| Path-tracking accuracy on 100 mm output | ±0.5 to ±1.5 mm | ±0.2 to ±0.5 mm | ±0.01 to ±0.05 mm |
| Build cost (educational scale) | $50-300 | $15-80 | $2,000-15,000 |
| Power input required | Single rotating shaft | Single rotating shaft | Servo + control electronics |
| Maintenance interval before slop affects gait | ~500-1000 hr (10 joints to wear) | ~2000 hr (4 joints) | ~10,000 hr |
| Best application fit | Walking gaits, dwell-rise-dwell, complex closed paths | Simple oval/D-shaped paths, slider-cranks | Arbitrary paths needing reprogramming |
Frequently Asked Questions About Eight-bar Linkage
This is almost always tolerance stack-up on the lower-triangle bars, not assembly error. Jansen's holy numbers are quoted to one decimal place but the gait is sensitive to about ±0.2% on the 55.8 mm bar specifically — that's ±0.1 mm. If you waterjet-cut your bars and one is 55.9 and its mirror is 55.7, the two legs trace measurably different curves.
Diagnostic check: clamp the chassis upside down, drive the crank by hand, and trace the foot path on paper for both leg-pairs. Overlay the tracings. If they differ by more than 1 mm anywhere, re-measure your six critical link lengths with calipers and find the outlier.
You almost certainly have a redundant constraint somewhere. The classic mistake is using two parallel bars between the same two pivot points — geometrically that's two links and two joints, but kinematically the second bar contributes zero new freedom because it's already constrained by the first. Grübler counts it but the real DOF is one less than the formula predicts.
The fix is to recount your independent constraints. If your linkage looks like it should be a Stephenson III but you've added a parallel reinforcing bar 'for stiffness', remove it and DOF will jump back to 1.
Klann is more forgiving and easier to build. It uses larger link-length ratios so a ±0.5 mm fabrication error barely affects the gait, and the foot path has a more pronounced lift that handles uneven ground better. Jansen produces a smoother, flatter gait that looks more lifelike but demands tighter tolerances and produces less ground clearance.
Rule of thumb: for a classroom build with hand-cut bars, choose Klann. For a sculpture or showpiece where the visual smoothness matters and you can hold ±0.1 mm on link lengths, choose Jansen.
You're crossing or approaching singular configurations. At a singularity the instantaneous mechanical advantage approaches zero or infinity, and the input torque required to maintain motion either spikes (near a dead-centre) or the linkage briefly gains a parasitic degree of freedom and shudders.
For an 8-bar, this typically happens when three of the coupler bars become collinear at a specific crank angle. Check your design in a free kinematic simulator like Linkage or PMKS — plot input torque versus crank angle and look for asymptotes. If you find them, shifting one ground-pivot location by 3-5 mm usually moves the singularity outside your operating range.
You have two practical options. The first is precision-point synthesis — pick 5 to 9 points your output must pass through, write the loop-closure equations for the 8-bar topology you've chosen (Stephenson I/II/III or Watt I/II), and solve the resulting nonlinear system. This is what Sandor and Erdman's textbook covers and it's solvable in MATLAB or Python with a reasonable initial guess.
The second is optimisation-based synthesis — define a cost function as the integrated error between your target path and the traced coupler curve, then run a genetic algorithm or Nelder-Mead solver over the 16-ish design dimensions. Tools like MotionGen and the open-source PyLinkage do this. Optimisation handles 9+ points where precision-point synthesis often has no real solution.
Plain bronze or PTFE bushings are fine at all 10 joints for any 8-bar running below about 100 RPM with light loads — that covers virtually every walking robot, sculpture, and educational build. Ball bearings only earn their cost when you're running continuous duty above 200 RPM or when joint loads exceed about 50 N, where bushing wear becomes the dominant lifetime limit.
The real watch-out is consistency. Mix bronze bushings on some joints and ball bearings on others and you'll get uneven friction torque around the crank revolution, which shows up as a wobbling gait. Pick one bearing type and use it everywhere.
References & Further Reading
- Wikipedia contributors. Linkage (mechanical). Wikipedia
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
