2.2 Program Creation & Tuning

A Pick & Place machine’s physical performance is entirely governed by the software program that drives it. From the moment CAD data is imported, the engineering decisions made regarding rotations, vision setup, feeder layout, and placement sequence determine whether the line achieves peak theoretical throughput or suffers from unpredictable micro-stoppages. Establishing disciplined, standardized program creation ensures consistency across all operating shifts and protects the capital investment from inefficient utilization.

Data Integrity: Standardized Inputs

Before the nozzle touches a single part, a foundation must be built on clean, sound data. Errors at this stage force operators into manual, error-prone overrides on the factory floor.

To ensure input standardization, all data imports should use consistent units, typically millimeters. A single zero-origin point for both top and bottom sides must be established, and a standard Counter-Clockwise (CCW) rotation convention must be adhered to.

It is also important that all imported package names and physical Z-heights identically match the placement machine’s internal library names. Using one-off aliases that require manual cross-translation by an operator must be avoided. Global and local fiducials should be clearly defined in the Golden Data Pack and used consistently for board alignment and large component positional correction.

Ultimately, the final, verified set of placement data, machine offsets, and rotation tables must be archived as the single source of truth. This forms the baseline for the production run. If a program regularly requires manual bulk rotation or coordinate offsets immediately after import, it usually indicates that the central library conventions or the upstream CAD export settings must be corrected at the source, rather than patching the program itself.

Vision and Component Definition

Machine vision component recognition should be robust and instantaneous to prevent frustrating vision retries and cycle slowdowns.

Exactly one golden reference image per package family must be defined and locked in the machine library. This image should feature high-contrast pin-1 or polarity markers and sharp body edges. It must be ensured the nozzle pick-up point is precisely centered on the flattest, most repeatable surface of the component. Programming pick points over embossed logos, mold ejection marks, or irregular domes must be actively avoided.

When configuring the vision settings, the simplest algorithm capable of achieving stable, locked recognition must always be chosen. Reliability and speed generally outrank highly complex recognition algorithms, which can trigger false failures over slight batch-to-batch component lighting variations.

Feeder Optimization: Minimizing Travel Mathematics

The physical arrangement of feeders plays a significant role in actual machine speed. Feeder optimization is primarily an exercise in minimizing the travel distance geometry of the placement head.

First, high-runner components, like common passives, must be identified and assigned to permanent, fixed feeder slots across every product program on the floor. This eliminates a substantial portion of changeover time. Similarly, the components with the highest hit-count must be placed in the feeder slots located closest to the head’s home position, or close to the center mass of the board’s placement area.

Feeders should also be heavily clustered by the required nozzle size and type. This minimizes the number of tool changes the head must perform during the placement sequence, boosting mechanical efficiency.

Large BGA components, QFNs, and odd-form parts sourced from matrix trays require significant amounts of time and head travel. These parts must be sequenced last in the program to ensure the high-speed placement of standard chips remains uninterrupted.

Path Sequencing and Load Balancing

The compiled sequence of part placements determines the overall cycle time and dictates whether parallel machines operate at peak efficiency.

As a general rule for placement order, small chips must be placed first for dense, fast field moves. This must be followed by standard ICs, and the tall or odd-form components must be placed last to ensure they do not interfere with the travel path of the head.

When running tandem flexible mounters, the placement program must be sliced to ensure the cycle times of both sequential machines are balanced to within ±10%. It is important to slice this by effort—for example, allocating fast chips on one machine and complex, slower ICs on the other—rather than simply by component count, to prevent bottlenecking.

First Article and Program Freeze

The First Article procedure is a short, structured engineering trial used to prove the program’s integrity before initiating volume production.

The procedure must begin by teaching the three global fiducials and mechanically verifying the global X/Y and theta alignment on the bare panel. Next, a few witness parts for every unique package family, such as a few chips, a QFN, and a connector, must be placed near the board edge. A microscope must be used to confirm that orientation, polarity, Z-height compression, and X/Y offset are perfectly dialed in.

The machine pickup logs must be reviewed carefully. A high rate of pick misses or vision retries usually indicates a flawed vision model or an incorrect nozzle parameter, which should be investigated and fixed. Following a clean verification, the machine recipe and program must be digitally locked, and a high-resolution Golden Board photo set must be captured. This freeze helps prevent unauthorized tweaks mid-shift that can negatively impact production yields.

Key Performance Metrics

These key metrics must be monitored on the machine dashboard to track true efficiency and identify mechanical drift.

• Placement Time Per Board: The raw time consumed purely for the placement routine, excluding conveyor tracking or buffer delays. This is an excellent measure of the program’s efficiency.
• Miss/Retry Rates: The count of vision evaluation failures or physical pickup errors, segregated by part family. Spikes in this metric typically point to a failing feeder or a corrupted vision library.
• Nozzle Swaps Per Board: This measures the effectiveness of the feeder grouping logic. High numbers may indicate unoptimized clustering and wasted travel time.
• Actual CPH vs. Theoretical: By constantly mapping the achieved Components Per Hour against the machine’s advertised maximum theoretical CPH, the true underlying asset utilization can be calculated.

---

Recap: Program Creation & Tuning

Parameter	Requirement	Value / Criterion	Action / Condition
Data Input	Standardized units & origin	Millimeters; single zero-origin point for top/bottom sides	Use consistent CAD export. Archive verified data as baseline.
Component Definition	Library & vision integrity	Package names & Z-heights match machine library. One golden image per package family.	Lock golden image. Center pick point on flat, repeatable surface.
Feeder Optimization	Minimize head travel	High-runner components in fixed slots. High hit-count parts closest to head home/board center. Cluster by nozzle type.	Sequence large BGAs/QFNs/tray parts last.
Line Balancing	Tandem machine synchronization	Cycle times balanced to within ±10%. Slice by operation effort, not component count.	Program must achieve this balance to prevent bottlenecking.
First Article Verification	Program integrity proof	Teach 3 global fiducials. Place witness parts per package family. Verify orientation, polarity, Z-height, X/Y offset under microscope.	Lock program after clean verification. Capture Golden Board photo set.