Choosing Test Methods From IPC-TM-650 for PCB Quality Control
Most of the numbers that appear on a printed circuit specification, from peel strength to moisture absorption, are produced by a named test method. The method fixes the coupon, the conditioning, the sample size and the way the result is calculated, and two laboratories that follow the same method should agree. When they disagree, the cause is almost always that one of those four elements was left to interpretation.
The IPC-TM-650 collection is the library of those methods for the printed circuit industry. Using it well is not a matter of quoting a number on a drawing, but of understanding what the method fixes and what it leaves to the user. This article describes the parts that most often differ between suppliers and customers, and how to write them down so that a result is comparable.
What IPC-TM-650 Is For
The collection defines how a property is measured, not what value is acceptable. The acceptable value is a specification limit that comes from the material standard, the customer drawing or an internal requirement. Separating the two is important, because a method is chosen for its repeatability and its relevance, while a limit is chosen for the duty of the product.
The methods cover visual, dimensional, chemical, mechanical, electrical and environmental tests, and they are numbered in families that reflect those groups. A method is a procedure with an apparatus list, a specimen description, a conditioning step, a measurement sequence and a reporting format. Reading the whole method rather than the title is what prevents an argument later.
Reading a Method Number
The numbering identifies the family and the specific test, and the revision matters as much as the number. A method that has been revised may have changed the specimen size, the load or the conditioning, and a result produced under the old revision is not directly comparable with one produced under the new. The revision should be quoted with the number on the drawing and in the report.
The family tells the reader what kind of measurement to expect. Mechanical methods in the 2.4 series include thermal stress and peel strength tests, electrical methods in the 2.5 series cover resistance and dielectric behaviour, and environmental methods in the 2.6 series cover moisture absorption and conditioning. Knowing the family is enough to know which laboratory and which equipment will be involved.
Conditioning Before Test
Most methods require the specimen to be conditioned before it is measured, and conditioning is the step that is most often skipped or shortened. A laminate that has been dried behaves differently from one that has equilibrated in a humid room, and a peel strength measured on a dry coupon can differ from one measured after humidity exposure by more than the tolerance being checked.
The conditioning is specified as a temperature, a relative humidity and a duration, and all three have to be held and recorded. Where the method allows a choice, the choice should be written into the purchase order, because a supplier who conditions for the minimum time and a customer who expects the full duration will disagree about the same material. The chamber used for the conditioning has to be calibrated, and its log is part of the test record.
Coupon Design and Where It Comes From
Many methods require a specific coupon, and the coupon has to be produced by the same process as the product. A peel strength coupon that was laminated separately from the board it represents says nothing about that board. The coupon is usually placed in the process edge of the panel, and its position and orientation are recorded so that it can be traced to the panel it came from.
The coupon also defines the direction of the measurement. Copper foil has a rolling direction, and glass fabric has a warp and a fill, so a peel or a flexure result depends on the orientation of the specimen relative to those directions. The method states the orientation, and the coupon design has to make it possible to cut the specimen that way.
Sample Size and Sampling
The method states how many specimens make a test, and the specification states how many tests make a lot. Those are different numbers. A peel strength test may require three specimens to give one result, and a lot may require one test per panel or per shipment depending on the sampling plan.
Reporting the mean without the individual values hides the case that matters, which is the weakest specimen. Most acceptance decisions are made on a minimum rather than on an average, and a report that shows only the mean cannot support that decision. The values should be reported individually together with the mean and the range.
Test Method Versus Specification Limit
A specification limit without a method is not enforceable, and a method without a limit is a measurement rather than a requirement. Both belong on the drawing, and where a customer standard already names a method, the drawing should not name a different one. Where two documents conflict, the conflict is resolved before production rather than at the first failed test.
Where a limit is derived from a design requirement, the derivation should be recorded. A minimum peel strength of a given value that has been tied to a thermal cycle result is a defensible requirement; the same number copied from a previous project is not, and it may be either too tight or too loose for the product in hand.
Reproducibility Between Laboratories
Two laboratories following the same method can still differ because of apparatus and operator effects. The load cell of a peel tester, the rate of travel and the way the specimen is clamped all influence the result, and the differences are larger for methods that involve a subjective step such as judging the mode of failure.
The practical control is a round robin or a split sample. The same material is sent to both laboratories, and the results are compared before a production lot is judged on them. Where the laboratories differ systematically, the cause is usually an apparatus calibration or a conditioning difference rather than a material difference, and identifying it once saves the argument on every subsequent lot.
Writing the Method Into a Purchase Order
A complete specification names the method and its revision, the conditioning, the coupon and its source, the sample size, the acceptance limit and the reporting format. It also names who performs the test and where the record is kept. That is a paragraph on a drawing, and it prevents most of the disputes that arise around incoming inspection.
For a material qualification the same information is needed at a larger scale, with the thermal and the environmental tests that represent the assembly process. Where a supplier proposes an equivalent method, the proposal should be evaluated against the purpose of the original rather than accepted on the basis that the number is similar. The reflow and thermal stress side of these tests is described in the notes on steam ageing and solderability.
Process Control and Records
In production the records are the test report, the conditioning log, the coupon identity, the instrument calibration and the operator. The coupon identity is the field that is most often missing, and it is the field that allows a failed test to be traced back to a panel, a lot and a process change.
Where a test is used for process control rather than for acceptance, the same discipline still applies, with the addition of a trend. A peel strength that is drifting down over several lots is a process signal even while every result is inside its limit, and the trend can only be seen if the conditioning and the method are constant. The solderability methods that share this discipline are described in the notes on wetting balance testing, and the surface cleanliness tests in the notes on solvent extract testing.
FAQ
Who chooses the test method, the supplier or the customer? The customer specifies it where the result is a contractual requirement. The supplier proposes it where the requirement is stated as a property rather than a method.
Can a result from a different revision be used? Only if the revision has been shown to give the same result for that property. Otherwise the test is repeated under the revision named in the specification.
Why is the weakest specimen more important than the mean? Acceptance limits are usually minima, and the mean can pass while a single specimen falls below the limit that the design depends on.
