Placement Machine Accuracy: What the Specifications Mean

A placement machine is usually selected on the basis of a specification sheet, and the specification sheet is usually written in a way that makes comparison difficult. Accuracy, repeatability, resolution and speed are quoted in different units by different manufacturers, at different confidence levels and under different conditions. Two machines that both claim twenty five micrometre accuracy can behave very differently on a real production line.

The purpose of this article is to explain what the numbers actually describe, how they are measured and which of them matter for a given product. That understanding is what allows a purchasing decision to be based on the requirement rather than on the largest number on the sheet.

Accuracy, Repeatability and Resolution

Accuracy is the difference between where the machine was told to place a component and where the component actually ended up, averaged over many placements. It includes the systematic errors of the machine: the calibration of the axes, the accuracy of the vision system and the offset introduced by the nozzle. Repeatability is the spread of the results around that average, and it describes how consistently the machine returns to the same point.

Resolution is the smallest increment the machine can command, which is a property of the encoder and the drive rather than of the mechanical system. A machine can have a resolution of one micrometre and an accuracy of thirty, and quoting the resolution as though it were accuracy is a common way of making a specification look better than it is. The distinction matters because repeatability is usually the more useful figure for production, and it is often better than accuracy by a factor of two or more.

How the Figures Are Measured

Accuracy is normally measured by placing a set of components on a glass plate or on a specially prepared board and measuring the offsets with an optical system. The result depends on the component, because a machine that holds a large chip accurately may not hold a small one as well, and it depends on the speed at which the measurement was made. A figure quoted at reduced speed is not achievable at full throughput.

The confidence level and the number of measurements also matter. A specification of twenty five micrometres at three sigma means that almost all placements fall within that distance, while the same figure quoted as a maximum over a small sample means something weaker. Where the manufacturer does not state the conditions, the number should be treated as indicative, and a demonstration on the actual component mix is more informative than any data sheet.

Vision Alignment and Component Presentation

Modern machines align components with a camera, either looking up at the part while it is on the nozzle or looking down at it in flight. The vision system removes the angular error of the feeder and the pick position, so the effective placement accuracy depends on the resolution of the camera, the lighting and the algorithms used to find the leads. For a fine pitch device the vision system is often the limiting element rather than the mechanical axes.

Component presentation sets the floor for the whole chain. A part that is not sitting flat in its pocket, or a tape whose pockets are not consistently positioned, introduces an error that the vision system has to correct and may not be able to. Feeder quality therefore affects placement accuracy as much as the machine does, which is why a machine demonstration using the customer own components and feeders is the only meaningful test.

Process Capability Rather Than Machine Capability

What production actually needs is a process that places every component inside the tolerance the design requires. That is expressed as a capability index, which relates the spread of the placement to the tolerance band and to the centring of the distribution. A machine with modest accuracy that is perfectly centred can produce a better capability index than a machine with a tighter spread that is offset from the target.

Measuring capability requires data from the line rather than from a demonstration. Placing a few hundred boards and measuring the offset of a representative component on each gives a distribution, and the distribution shows whether the errors are random or systematic. Systematic errors can usually be corrected by calibration, which is inexpensive; random spread cannot, and it is the real limit of the equipment. Any serious evaluation of a placement machine ends with that measurement.

Throughput and Its Measurement

The headline placement rate is measured with the simplest possible conditions: small chips, a full feeder bank, no vision alignment and continuous operation. The rate achieved on a real product is always lower, because large components are placed more slowly, vision alignment takes time, and the machine stops for feeder changes and for board transfer. The realistic figure is obtained from the component list of the actual product and the machine cycle time for each package type.

Where several machines are compared, the comparison should be made with the same product. A machine that is fast on chips and slow on connectors may be the wrong choice for a board that is mostly connectors, and the opposite is equally true. The useful metric is the time to build one board of the real product, including the time lost to feeder changes over a shift, rather than the chips per hour figure from the brochure.

What to Specify for a Given Product

For a board whose smallest pitch is half a millimetre, a machine with fifty micrometre accuracy and good repeatability is adequate. As the pitch falls towards three hundred micrometres and the packages become thinner, both the accuracy and the vision resolution have to improve, and the machine needs a nozzle and a feeder system that can handle the component without damage. Matching the machine to the finest component on the roadmap, and not only to the current product, avoids a premature replacement later.

Finally, the specification should be written in terms of the outcome rather than the equipment. A requirement expressed as a placement offset distribution on the actual component set is testable, comparable between suppliers and directly connected to the yield of the assembly. It also makes the acceptance criteria for a new machine objective, which is the only way to resolve a disagreement about whether the equipment meets the requirement. That discipline is the same whether the equipment is a placement machine or any other step in pcb manufacturing processes.

Process Control and Verification

On a design of this kind, placement accuracy is the item that decides how the rest of the board is arranged. The measurements that matter are the repeatable ones: conductor width and spacing, annular ring, finished hole size, plating thickness and surface finish are all verifiable on a coupon that travels with the panel. Running a first article through the same checks as the production panel confirms that the two agree, and that comparison is the cheapest form of process control available at prototype stage.

Related reading: our fabrication notes, board quality and design release notes cover the same ground.

FAQ

Is a smaller accuracy figure always better? It is better, but only if it is quoted under the same conditions. A machine with a slightly larger figure that achieves it at full speed and with vision alignment is usually the better production tool.

Why is repeatability more useful than accuracy? Because systematic offset can be calibrated away, while random spread cannot. Repeatability measures the part of the error that the process cannot correct.

How should a machine be evaluated before purchase? By placing a few hundred boards of the real product with the real feeders and components, then measuring the offset distribution of the finest pitch part.