Quick answer
Best first pilots on PCB: red glue, local underfill, TIM, selective coating—fixed paths, clear keep-outs, written acceptance
Material and consistency: published dispensing automation bands often cite ~15–30% less adhesive waste, manual variation ~±20% vs automated ~±5%, and ~8–15% capacity lost to manual takt jitter
Cobot vs gantry: cobots win on multi-SKU, frequent path edits, and tight line-side space; gantries often win on large flat runs at stable high takt
What makes or breaks a pilot: fixture and board Z tolerance, pump/valve match to adhesive, and whether cure—not arm reach—caps line takt
Payback band: collaborative dispense guides often cite ~8–12 months (costly TIM/underfill pays on material first); size ROI from your weigh data and AOI overflow, not the table alone
A slow corner piles adhesive; half a millimeter of overflow beside a keep-out means wash or a full board recheck. When the AOI light comes on, the question is rarely “can we extrude glue?”—it is whether tool speed and flow are locked to the same path.
Cobot adhesive dispensing uses a robot arm to carry a dispensing valve or needle along a programmed path and apply adhesive, sealant, or other fluid to a part. The arm controls four variables at once: path geometry, application speed, pressure/flow, and bead start/stop. If speed changes and flow does not, bead width drifts—operators almost always drift on corners; a programmed path can keep tool speed and output in sync so the cross-section stays consistent on straights and corners.
On PCB / PCBA lines that matters for red glue, underfill, thermal interface material (TIM), selective coating, and connector reinforcement: bead width/height, keep-out overflow, AOI/recheck rates. Half a millimeter past a keep-out is often a wash or a scrap lot; when the line accepts boards to IPC-A-610 or similar assembly acceptability standards, bead deviation often shows up directly in AOI or recheck—not “close enough.”
The International Federation of Robotics (IFR) updated position paper on collaborative robots notes that cobot use cases keep expanding—from handling and welding to painting, dispensing, and assembly—and that manufacturing sectors such as electronics were early adopters. This guide starts from that industry framing, then focuses on when a collaborative PCB dispense cell actually fits.
Manual vs cobot adhesive dispensing
| Factor | Manual | Cobot automation |
|---|---|---|
| Bead width / height | Drifts with speed, pressure, fatigue | Same path + pressure/speed recipe every cycle |
| Keep-out boundaries | Operator judgment; worse late shift | Programmed edges; corner slow-down cuts pile-up |
| Cycle time | Often jitters tens of seconds on the same station | Dispense segment becomes measurable for line balance |
| Material use | Operators pad volume for uncertainty | Metered to recipe; costly TIM/underfill waste is controllable |
| Skill dependency | Quality drops when a trained operator leaves | Recipes archive; training becomes teach + first article |
| Chemical exposure | Long hands-on time near mist/solvents | People move to nozzle change, load, and sampling |
What automation usually buys on a PCB dispense station
Bead consistency — sample or AOI limits can match a saved width/height instead of “feels right”
Less overflow and wash — fewer keep-out tails and over-dispense; expensive adhesives often pay back on material before headcount
Measurable takt — the dispense slice stops being “whoever is faster” and can enter line balance
SKU change by recipe — multiple paths on one station; nozzle swap + first article beats hard tooling rebuilds
EHS — less prolonged operator contact with epoxy/solvents; people own exceptions and inspection
IFR also publishes plant cases where a cobot runs silicone dispensing (and related assembly steps) and moves people off solvent contact—for example Creating Revolutions / Universal Robots. That path—“recipe dispense + humans on sampling”—is already in production somewhere; treat their reject/throughput numbers as that site’s result, and size your ROI from your own overflow, rework, and takt data.
Published evidence bands (common in dispensing automation guides)
| Metric | Published band | Source |
|---|---|---|
| Material waste reduction | About 15–30% (stronger on costly adhesives) | EasyRobotics dispensing guide |
| Capacity lost to takt jitter | About 8–15% vs a stable robotic dispense station | EasyRobotics dispensing guide |
| Volume / consistency variation | Manual ~±20% vs automated ~±5% (general robot dispensing comparison) | Dispense Robotics comparison |
| Dispense time per part (order of magnitude) | Manual ~20–60 s vs automated ~2–10 s (path-dependent) | Dispense Robotics comparison |
| Overfill waste | Manual ~20–30% overfill vs automated ~2–5% waste | Dispense Robotics comparison |
| Payback | Cobot dispensing often ~8–12 months | EasyRobotics dispensing guide |
| Path repeatability (as marketed) | About ±0.05 mm class | EasyRobotics dispensing guide |
Read the table carefully: some rows are general “dispensing robot” bands, not one PCB line. TIM/underfill material math is usually more sensitive than commodity structural glue. Size ROI from weigh-offs, AOI overflow, and your takt—do not paste the table as a quote promise.
A cobot cell does not remove oven/UV bottlenecks, but it turns teachable segments into saved recipes. Count full takt as cure + load/unload + dispense—see How to calculate cobot cycle time.
Which PCB dispense jobs to automate first
Automate stations with stable paths, stable chemistry, clear keep-outs, and written acceptance:
Red glue / SMT structural dots — short, high-repeat paths
Selective underfill — when fixtures hold and paths can be taught; full-area high-rate micro-fill often needs dedicated or jet systems
TIM strips — costly material; tight width tolerance; less overflow pays fast
Selective coating / connector reinforcement — clear keep-outs favor programmed boundaries
Board-adjacent seal beads — continuous perimeters where corners must not void
Requiring “zero human glue on day one” usually explodes scope. Freeze a list of 100% programmed paths vs human touch-up points.
What a collaborative dispense cell includes
| Element | Role |
|---|---|
| Cobot + controller | Repeatable path and pose; line I/O |
| Valve / pump | Time-pressure, volumetric/piston, auger/gear, or jet—set by viscosity and tolerance |
| Fluid supply | Syringe, pot, or drum; two-part needs metering/mix |
| Board fixture | Datums for XY and plane height; board Z tolerance fails needles before rated reach does |
| Vision (optional) | Locate, correct, or inspect—scope before the quote; do not add “100% vision” after signature |
Valves, needles, flange I/O: How to choose a cobot end effector. Footprint and human zones: First cobot workcell layout.
Cobot vs Cartesian gantry vs dedicated dispenser
| Option | Fits | Weak fit |
|---|---|---|
| Six-axis cobot | High mix, changing paths, side-of-line retrofit, humans change nozzles/load boards | Sub-ms multi-valve sync, ultra-high-rate micro-dispense lines |
| Gantry / Cartesian | Flat boards, rigid paths, long stable volume | Frequent SKU change, tight door-side space |
| Dedicated dispenser | Micro-volume, fast two-part, takt locked on dispense | Pilots while fixtures and acceptance are still moving |
The same IFR paper notes the usual cobot trade-off versus traditional industrial arms: for safe, lightweight collaboration, high payload and high speed jobs often stay with industrial robots or dedicated machines. Most PCB pilots still start on a cobot—faster teach, cheaper recipe change than hard tooling—then revisit gantry or dedicated platforms after material, takt, and acceptance freeze. People beside the cell still need a written risk assessment; collaborative does not mean zero safety work → Safety & I/O scope guide (often under ISO 10218 / ISO/TS 15066).
How to specify the cell (material → pump → fixture → arm)
1. Material and cure — MSDS, viscosity window, one-part vs two-part, ambient/UV/thermal cure. Fast-cure grades punish valve timing.
2. Pump/valve — use the table below; follow adhesive vendor guidance, not only “cheap to mount.”
3. Fixture and Z — target width, nozzle standoff, board height tolerance in writing; no fixture budget means no micron-height promises.
4. Arm and envelope — valve + needle + small cartridge is often about 0.5–2 kg. Tight cells can start with r-Lite; larger workspace → r-Core. Put both in Side-by-Side Comparison; if the scenario is still fuzzy, run Product Advisor then lock numbers in Comparison.
5. True-scale footprint — check door-side reach on phone before a demo: True-scale AR preview.
Pump / valve shortlist
| Type | Better for | Watch-outs |
|---|---|---|
| Time-pressure | Mid/low viscosity, fast pilots | Drift with viscosity/temp—calibrate |
| Volumetric / piston | Tight volume, thicker materials | Higher cost and maintenance |
| Auger / gear | Continuous beads, wider viscosity | Match vendor guidance |
| Jet | Micro dots, non-contact speed | Costly; common on dedicated PCB lines |
Process knobs
| Parameter | Effect |
|---|---|
| Pressure / flow | Width and height; re-check on lot change |
| Path speed | Corner pile-up; skin-over if too slow |
| Dwell / valve timing | Start/stop tails with Z lift |
| Nozzle standoff | Driven by fixture height tolerance |
| Cure time | Real takt and WIP |
Pilot sequence
One-page process freeze: material/MSDS, cure, target width, keep-outs, sample or AOI rules
Real boards + real adhesive (no “similar” chemistry)
Lock fixture and Z tolerance; layout: workcell guide
Teach primary paths: corner slow-down/dwell; save recipes
Lot-change first article: weigh or caliper sign-off
Written human touch-up list to stop scope creep
After PO, align milestones with RooollTrack
Production care (nozzle, lines, calibration) via Support / Roooll Care
When not to start with a cobot (or when to buy dedicated)
Glue type/points change every board—no recipe stability
Sub-ms multi-valve timing beyond the collaborative cell design
Board height tolerance exceeds standoff and there is no fixture budget
Contract requires closed-loop vision but the budget only covers a teach station
