A SCARA Arm is a 4-axis horizontal articulated robot with two rotary joints in the horizontal plane and a vertical Z-axis with end-of-arm rotation. Its joints are stiff vertically but compliant horizontally — so the arm absorbs lateral misalignment during insertion while staying rigid against gravity. We use it to place, screw, and insert parts at high speed with repeatability down to ±0.01 mm. A typical SCARA runs a 25 mm × 305 mm × 25 mm pick-place cycle in under 0.45 seconds, which is why electronics assembly lines lean on it heavily.
SCARA Arm Interactive Calculator
Vary link lengths, target X-Y position, and elbow configuration to see the SCARA inverse-kinematics joint angles and reach margin.
Equation Used
This calculator uses the standard planar two-link inverse kinematics described for a SCARA arm: the X-Y target and the two arm lengths define the shoulder angle theta1 and elbow angle theta2. The elbow direction input selects the lefty/righty solution.
- Planar 2-link SCARA arm in the horizontal X-Y plane.
- Joint limits, acceleration limits, payload deflection, and Z-axis motion are not included.
- Elbow direction s = +1 or -1 selects the two possible SCARA configurations.
- If the target is outside the reachable annulus, angles are calculated at the nearest kinematic boundary.
Operating Principle of the SCARA Arm
The SCARA Arm, also called the Selectively compliant arm (SCARA), gets its name from exactly that property — selectively compliant. Both shoulder and elbow joints rotate around vertical axes, so the arm is stiff in Z (gravity is taken by the joint bearings, not the motors) but yields slightly in X and Y when something pushes back. That yield is what lets it shove a connector into a socket without snapping pins. The Z-axis is a separate ball-screw spline, and a fourth rotary axis on the tool flange lets you orient parts.
The motion math is plain 2-link planar kinematics. You command an X-Y target, the controller solves for joint 1 (shoulder, θ1) and joint 2 (elbow, θ2), and the inverse-kinematics result gives you two solutions — "lefty" and "righty" elbow configurations. Picking the wrong configuration mid-cycle causes a 180° wrap that wrecks cycle time. Most controllers lock the configuration at program load.
Where it goes wrong is timing and backlash. Harmonic drive reducers in joints 1 and 2 typically hold ±30 arc-seconds of lost motion. If the harmonic drive wears or the preload drops, you'll see end-effector position drift of 0.05-0.15 mm — small, but enough to miss a 0402 component pad. The Z-axis ball screw spline needs preload too; if the spline nut loosens, the tool wobbles in yaw and your screwdriving torque readings go noisy. Tune the controller's acceleration profile too high and you'll excite the natural mode of the outer link — typically 25-40 Hz on a 600 mm reach arm — and the tooltip rings for 80-150 ms after every move.
Key Components
- Joint 1 (Shoulder, θ1): Rotary joint at the base, typically driven by a 400-750 W AC servo through a harmonic drive at 80:1 or 100:1 ratio. Carries the full inertia of the outer link plus payload, so it's the limiting axis for cycle time on long-reach moves. Repeatability spec usually ±0.01 mm at the tool, contributed mostly by harmonic drive lost motion of ±20-30 arc-seconds.
- Joint 2 (Elbow, θ2): Second horizontal rotary joint at the elbow, driven by a smaller 100-400 W servo through a harmonic drive. Determines the inner-radius dead zone — typically you can't reach within ~150 mm of the base column in righty configuration. Lost motion here adds directly to tool-tip error in the radial direction.
- Z-axis (Ball Screw Spline): Vertical linear axis using a combined ball screw and ball spline shaft — one shaft does both the lift and the rotational support. Stroke is typically 150-400 mm. Lead is 10-20 mm/rev, so a 3000 RPM motor gives 0.5-1.0 m/s vertical speed. Preload loss in the spline nut shows up as tool-yaw wobble during screwdriving.
- Joint 4 (Tool Roll): Final rotary axis on the tool flange, used to orient the gripper or screwdriver. Driven directly or through a small belt reducer. Resolution typically 0.01°. Speed up to 2000°/s on small SCARAs like the Epson G3 series.
- Controller and Drive Pack: Coordinates the four servos through inverse kinematics solved at 1-4 kHz. Holds tuning parameters, configuration choice (lefty/righty), and the path-blending profile. A poorly tuned acceleration limit excites the outer-link first mode and adds 80-150 ms of settle time per move.
Industries That Rely on the SCARA Arm
SCARA arms dominate jobs where you need fast vertical insertion at high cycle rate and tight horizontal repeatability — but you don't need full 6-DOF orientation. If the part goes in straight down, a SCARA beats a 6-axis arm on cost, speed, and footprint nearly every time.
- Electronics Assembly: The Epson G6 SCARA placing connectors and shielding cans onto smartphone PCBs at Foxconn, hitting 0.4 s cycle times with ±0.015 mm repeatability.
- Pharmaceutical Lab Automation: Yamaha YK400XR loading 96-well plates into Hamilton liquid handlers — the selective compliance absorbs plate-edge misalignment without crashing the deck.
- Automotive Sub-Assembly: Denso HS-series SCARAs driving screws into ECU housings on Toyota lines, where the Z-axis torque feedback verifies each fastener seats correctly.
- Food Packaging: Stäubli TS2-60 stainless-wash-down SCARAs picking individual chocolates into tray pockets at Lindt, running 120 picks per minute on belt-tracking.
- Consumer Goods Kitting: Adept Cobra s600 (now Omron) loading razor cartridges into multi-pack trays at Gillette, using vision-guided pick from a flexible feeder.
- Semiconductor Wafer Handling: Brooks Automation cleanroom SCARAs transferring 300 mm wafers between FOUP cassettes and process tools, where the horizontal compliance prevents wafer-edge chipping.
The Formula Behind the SCARA Arm
The number practitioners actually care about is cycle time for the standard pick-place move — the so-called "adept cycle" of 25 mm up, 305 mm across, 25 mm down, then back. Cycle time is dominated by the longest single move and the settle time at each end. At the low end of the typical operating range (long reach, heavy payload, conservative tuning) you'll see 0.7-0.9 seconds per cycle. At the nominal sweet spot — 2 kg payload on a 400-600 mm reach SCARA — you hit 0.40-0.50 seconds. Push to the high end (light payload, aggressive accel, well-tuned path blending) and a top-end Epson or Yamaha will close the cycle in 0.28-0.35 s. Beyond that the outer-link resonance dominates and you lose more in settle time than you gain in move time.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| tcycle | Total round-trip pick-and-place cycle time | s | s |
| tz | Time for one vertical Z-axis move (up or down) | s | s |
| txy | Time for the horizontal traverse between pick and place points | s | s |
| tsettle | Mechanical settling time after each move before the gripper can actuate | s | s |
Worked Example: SCARA Arm in a cosmetics-bottle capping cell
We're sizing a SCARA on a cosmetics-bottle capping cell at a contract filler in New Jersey. The job: pick a 12 g aluminium cap from a vibratory feeder, traverse 320 mm to the bottle nest, drop 30 mm onto the bottle neck, and drive the cap with the Joint 4 servo. The customer wants 100 bottles per minute on a single SCARA. We're evaluating a 600 mm-reach Epson G6-651S with 6 kg rated payload.
Given
- Reach (horizontal traverse) = 320 mm
- Z stroke = 30 mm
- Payload = 0.012 kg
- Z lead screw lead = 20 mm/rev
- Z motor max speed = 3000 RPM
- Required throughput = 100 parts/min
Solution
Step 1 — required cycle time from throughput. 100 parts/min means each cycle must close in 0.60 s or less.
Step 2 — at nominal tuning on a G6-651S with light 12 g payload, the Z-axis runs near its rated 1.0 m/s peak. The 30 mm Z move takes about:
Step 3 — the 320 mm horizontal traverse on a 600 mm SCARA at nominal acceleration runs around 0.18 s including blend, and settle time at well-tuned nominal is roughly 0.04 s per endpoint.
That meets the 0.60 s target with margin. Now check the range. At the low end of practical tuning — conservative accel for a customer who hates noise and vibration — txy climbs to about 0.26 s and settle to 0.07 s, giving tcycle,low ≈ 0.77 s. That misses the 100-bpm target, you'd land around 78 bpm. At the high end with aggressive accel and path blending, txy drops to 0.13 s and settle to 0.025 s, giving tcycle,high ≈ 0.42 s — 143 bpm theoretical. But above ~130 bpm on this arm, the outer-link first mode (around 32 Hz on a 600 mm reach) starts ringing the screwdriver bit, and your capping torque traces go ragged.
Result
Nominal cycle time lands at 0. 55 s, comfortably under the 0.60 s budget — you'll hit 100 bottles per minute with about 8% headroom. The range tells the real story: dialled back for quiet running you drop to 78 bpm and miss the target, while pushed hard you can theoretically hit 143 bpm, but above ~130 bpm the outer-link resonance starts blurring the cap-drive torque signature and you'll see false reject spikes. If your measured cycle is 0.70 s instead of the predicted 0.55 s, check three things in this order: (1) the inverse-kinematics configuration may be flipping between lefty and righty mid-program, adding a 200-400 ms reconfiguration move, (2) the harmonic drive on Joint 1 may have lost preload, forcing the controller to add settle time to compensate for the extra lost motion, or (3) Joint 4 may be commanded with a default low-speed profile that's masking the gain you got on Joints 1, 2, and Z.
When to Use a SCARA Arm and When Not To
The SCARA sits between the cheap-but-slow Cartesian gantry and the flexible-but-pricey 6-axis arm. Selectively compliant arm (SCARA) machines win on vertical insertion tasks; they lose anywhere the part needs to come in at a non-vertical angle. Here's how the three stack up on the metrics that actually drive selection.
| Property | SCARA Arm | Cartesian Gantry | 6-Axis Articulated Arm |
|---|---|---|---|
| Repeatability | ±0.01 mm | ±0.02 to ±0.05 mm | ±0.02 to ±0.06 mm |
| Standard cycle time (25-305-25 mm) | 0.30-0.50 s | 0.60-1.20 s | 0.45-0.80 s |
| Payload range | 1-20 kg | 1-100 kg | 0.5-500 kg |
| Workspace shape | Annular (donut), fixed Z height | Rectangular box, full Z | Spherical, full orientation |
| Best application fit | Vertical insertion, screw-driving, pick-place | Long-axis dispensing, large-format pick-place | Welding, painting, multi-orientation assembly |
| Typical capital cost | $15k-$40k | $8k-$25k | $35k-$120k |
| Setup and integration complexity | Low — 4 axes, fixed orientation | Lowest — orthogonal axes, simple kinematics | High — 6-DOF singularities, harder to program |
Frequently Asked Questions About SCARA Arm
Repeatability spec is measured at one fixed pose with a calibrated probe, no payload, and the arm fully warmed up. Your production number includes thermal drift in the harmonic drives (typically 0.02-0.04 mm shift between cold start and 30 minutes of running), payload-induced deflection of the outer link (a 2 kg load on a 600 mm reach deflects the link tip by 0.01-0.03 mm), and any tooling stack-up between the flange and the actual part feature.
Run a thermal soak for 20-30 minutes before the first qualification cycle and re-zero. If you still see drift, instrument the harmonic drive housings — anything above 45 °C steady-state means the duty cycle is too aggressive for that gearbox size.
Deltas win below about 1 kg payload and when the parts arrive randomly on a moving conveyor — their low moving mass gives them 200+ picks per minute and they're easier to belt-track. SCARAs win above 1 kg, when you need significant Z-stroke (deltas typically have 100-200 mm of Z, SCARAs do 300-400 mm), or when the task requires a screwdriving or insertion force the delta's parallel linkage can't apply cleanly.
Rule of thumb: if your task is "grab from a moving belt and drop in a bin," go delta. If it's "pick, traverse, insert with force, screw, release," go SCARA.
You're exciting the outer-link first natural mode, typically 25-40 Hz on a 400-700 mm reach SCARA. The controller's commanded acceleration profile contains frequency content overlapping that mode, so the arm rings every time it stops. The mass of your end effector matters here — a heavy gripper drops the natural frequency further and makes it worse.
Two fixes: enable the controller's input shaping or notch filter (Epson calls it "vibration suppression," Yamaha calls it "residual vibration control") tuned to your measured ringing frequency, or reduce commanded jerk by 30-50%. The cycle time penalty for input shaping is usually 10-20 ms, way less than the 100+ ms you lose to ringing.
The inertia seen by Joint 1 is the payload mass times the square of the distance from Joint 1 to the payload center of mass — not just to the tool flange. An offset gripper that puts the payload 80 mm forward of the flange on a 600 mm-reach arm raises the effective Joint 1 inertia by roughly 25-30%. The motor sizing tools in Epson RC+ and Yamaha YRC bake this in if you enter the offset, but most integrators leave it at zero by default.
If you're already running and the arm trips overcurrent on long traverses, the offset CoG is the first thing to check. Re-measure your end effector and re-enter the values; the controller will adjust its accel limits.
Yes — those are two names for the same architecture. "Selectively compliant arm" describes the mechanical property (stiff vertically, compliant horizontally). "Horizontal articulated robot" is the ISO/industry classification by joint configuration. "SCARA" is the acronym that stuck commercially after Hiroshi Makino's group at Yamanashi University coined it in the late 1970s.
You'll see all three terms in datasheets and standards documents. They refer to the same 4-axis machine with two horizontal-plane rotary joints, a Z-axis, and a tool roll axis.
Two common causes. First, the target is inside the inner-radius dead zone — the area too close to the base column where the elbow can't fold tight enough to reach. On a 600 mm-reach arm this is typically the inner 150-180 mm radius. Second, the target requires switching elbow configuration (lefty vs righty) and your program is locked to one configuration. The arm sees the target as unreachable in the current config even though the other config could hit it.
Run a quick check: command the target with the opposite elbow configuration manually. If it reaches, your program needs a configuration switch in the path. If it still won't reach in either configuration, you're outside the envelope and need a longer-reach model.
Hollow-Z (hollow ball-screw spline) lets you route pneumatic lines, vacuum, and signal cables down through the center of the Z shaft. The cables don't flex with every Z move, so cable life jumps from 1-2 million cycles on an external dress pack to 10+ million on internal routing. The penalty is cost (15-25% more on the arm) and slightly reduced Z payload capacity because the shaft wall is thinner.
For any application above 5 million cycles per year — most electronics assembly — spec the hollow Z. The cable replacement downtime alone pays for it inside 12 months.
References & Further Reading
- Wikipedia contributors. SCARA. Wikipedia
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
