Control Plan Development for PCB Assembly
A control plan is the document that says which parameters matter, how they are monitored and what happens when they drift. It is required by automotive and medical customers, and it is useful everywhere else, because it forces a production line to decide in advance what it will watch rather than reacting after a batch has been rejected.
What a Control Plan Contains
The plan lists each process step, the product and process characteristics that affect it, the specification and tolerance, the measurement method, the sample size and frequency, and the reaction when a limit is exceeded. It also names who is responsible for each action.
Critically, it distinguishes between product characteristics, which are features of the finished board, and process parameters, which are the settings that produce them. Controlling the parameter is usually cheaper and faster than inspecting the product, and a good plan shifts the emphasis that way.
Identifying the Parameters That Matter
Not every setting deserves monitoring. The parameters worth controlling are those that vary, that affect a customer requirement and that the process cannot correct on its own. A solder pot temperature qualifies; a conveyor speed that is mechanically fixed may not.
The identification process starts from the requirements and works backwards. For each requirement, ask which process step can violate it and which parameter drives that step. The result is a short list that can actually be monitored, which is far more useful than a comprehensive list nobody follows.

Setting Monitoring Frequency
Frequency follows the rate at which a parameter can drift and the cost of a bad batch. A parameter that drifts slowly can be checked at the start of a shift, while one that can move within an hour needs a check within the hour. Capability data makes this concrete: a process running close to a limit needs more frequent checks than one in the centre of its window.
The frequency should also account for change events. A paste change, a stencil change, a feeder refill or a profile adjustment all invalidate the previous check, so the plan should specify that a verification follows each of them rather than waiting for the next scheduled interval.
Measurement Method and Equipment
Each parameter needs a measurement method that another person could repeat. Stating that a temperature is checked is not enough; the plan should name the instrument, the location of the sensor and the acceptance band. Where the measurement itself has significant uncertainty, that uncertainty belongs in the decision.
Calibration status is part of the method. A measurement taken with an out of calibration instrument has no standing, and the plan should make the check of calibration status part of the routine rather than an audit discovery.

Reaction Plans
A reaction plan states what happens when a limit is exceeded: stop the process, quarantine the material produced since the last confirmed good check, notify a named person, and record the action. Without a reaction plan, a measurement that is outside limits becomes an observation rather than a control.
The quarantine boundary is the part most often omitted. It has to be defined by the last known good check, which is why the frequency and the reaction plan are two halves of the same decision. Widening the frequency automatically widens the amount of material at risk.
Capability and Continuous Improvement
A control plan maintains a process; it does not improve one. Capability indices computed from the monitoring data show where the process is barely able to meet the requirement, and those are the candidates for improvement. A process that is capable needs looser monitoring, which frees effort for the ones that are not.
The same data supports the analysis described in our article on first pass yield analysis, where a yield change is traced back to a parameter rather than to a person. The two systems share their data and their conclusions.
Documenting and Auditing
The plan should be a controlled document with a revision history, and the shop floor should have the current version at the point of use. A plan that exists only in the quality office is a document, not a control.
Auditing checks that the checks were performed, that the results were recorded and that the reaction plan was followed when it was needed. The most informative audit question is not whether the log is complete but whether anyone has acted on it in the last month. Where actions are absent, the limits are probably set too wide.
Keeping the Plan Alive
Processes change and plans fall out of date. A review at each product revision, at each process change and after any significant quality event keeps the document relevant. Where a review is not scheduled, the plan tends to grow rather than to change, and the monitoring burden rises without any gain in control.
Where a parameter is removed, the reason should be recorded. Deleting a check because it never failed is only defensible if the capability data supports it, and the record is what allows a later engineer to make the same judgement with the same evidence.
Process Control and Verification
A stack-up that is drawn rather than described removes most of the ambiguity from a quotation, and it lets the fabricator price the board against the dielectric and copper weights that will actually be used. Where the requirement is not written down, the shop supplies its own default, and the default is chosen for the process rather than for the design.
Related reading: our fabrication notes, board quality and design release notes cover the same ground.
FAQ
How many parameters should be monitored? As few as will control the requirements. A short list that is followed beats a long list that is signed without being performed.
Does a small company need a control plan? The formal document is a customer requirement in some industries, but the discipline of naming parameters, frequencies and reactions is valuable at any size.
What happens if a check is missed? The plan should say. Usually the material produced since the last good check is treated as unverified until it can be shown to meet the requirement.



