The bottleneck in high-mix, low-volume work is never the placement speed. A machine can put down 80,000 points an hour, but if the line changes over three times a day, yield is decided by the first three boards after each changeover.
Put 01005 chips and 0.4 mm pitch BGAs on the same panel and the problem gets harder: the two device families want almost opposite things from temperature, accuracy and solder volume.
An 01005 measures 0.4 × 0.2 mm and weighs about 0.04 mg. It is smaller than a grain of salt, polarity is almost invisible, and a touch of the tweezers sends it flying.
The difficulty concentrates in three places:
One point that is often overlooked: the joint strength of an 01005 is low, so how the panel is depanelised and handled directly affects the defect rate. Routed depanelling beats punching, and carrier support beats a free-standing stack.
The core problem with BGA is that you cannot see it. The joints sit under the body, AOI is helpless, and only X-Ray answers. Three things are judged: voiding, bridging and head-in-pillow.
QFN brings a different trouble: the thermal pad is large, so a little extra paste floats the part and a little less leaves voids. IPC-7093 points to controlling the aperture array on the thermal pad, holding voiding below 25% while keeping a visible side fillet.
This is the real dilemma in high-mix products:
| Requirement | What 01005 wants | What BGA wants |
|---|---|---|
| Peak temperature | Lower, to avoid tombstoning and overheating | Higher, to fully collapse the balls |
| Ramp rate | Gentle, to limit thermal shock | Per paste specification, not too slow |
| Cooling rate | Moderate | Fast cooling refines grain, but too fast cracks joints |
| Nitrogen | Helpful, not essential | Markedly improves wetting and voiding |
The answer is not a compromise but zonal control: use a carrier or a baffle to reduce measured airflow over the small-component area, so the actual heat input where the 01005s sit is lower than at the BGA; and set peak temperature at the lower limit that still fully collapses the BGA, rather than at a midpoint.
The reflow profile must be measured, never copied from the paste supplier's recommended curve — board thermal mass, copper distribution and component density all change the result. After every changeover, the first run is profiled again.
High-mix changeover cannot depend on the experience of a senior operator; it has to become a repeatable process. Our seven steps and target times:
| Step | Action | Target | Poka-yoke |
|---|---|---|---|
| 1 Clear the line | Remove reels, count, return to store, clear remnants | 8 min | Every reel UID scanned, book-to-physical reconciliation |
| 2 Remove stencil | Unload, clean, log | 5 min | Stencil ID bound to the work order |
| 3 Fit stencil | Install, check tension, align | 10 min | Tension measured before first article; below limit stops the line |
| 4 Load program | Retrieve program, verify revision and panel count | 6 min | Program revision force-checked against the work order |
| 5 Load material | Load per feeder table, scan-verify | 20 min | Wrong part or wrong slot locks the machine |
| 6 Set rails and pins | Rail width, support pins, edge clamping | 8 min | Pins placed per pin map; thin boards always on a carrier |
| 7 Trial run and first article | Three trial boards, profile measurement, first-article approval | 15 min | Release only after FA-600IV pass |
For a single-sided product under 100 placements, the whole-line changeover target is 75 minutes. That number matters more than how fast a machine can swap feeders — it decides whether small orders can be scheduled at all.
Only when all three pass does the batch start. If any fails, the line returns to the corresponding step — it does not "run a batch and see".
Three SMT lines run YAMAHA placement machines. Key parameters, as calibrated on our actual machine types:
| Item | Capability | Notes |
|---|---|---|
| Placement accuracy | ±0.05 mm (chip) / ±0.03 mm (QFP) | 01005 needs dedicated nozzles and vacuum sensing |
| Smallest component | 01005 (0.4 × 0.2 mm) | With Type 4/5 paste and laser stencil |
| Largest component | 55 × 100 mm | Tray feeding; odd shapes configured separately |
| Component height | ≤ 15 mm | Limited by nozzle and head geometry |
| Board size | Up to 510 × 460 mm | Larger needs a carrier or panelisation |
| Feeder positions | 120+ including tray | Sets the component mix per changeover |
| Accuracy verification | First article plus periodic CPK | Standard board verified quarterly |
Parameters are only the boundary. What decides yield is whether the seven steps above are executed, how often, and whether they are recorded. Equipment can be bought; process has to be practised.
This is exactly the work we do: 01005 through 0.4 mm pitch BGA on the same board, batches from tens to thousands, several changeovers a week. If you have a product looking for a line, send the Gerber and BOM and we will assess process feasibility first, then talk price.
