Wednesday, June 20, 2018
Home/Current Issue >  Industry Articles >  SMT and Production > 

Why Internal Metrology Doesn't Work
The International Bureau of Weights and Measures in Sevres, France, is the principal repository for metric standards. See note at bottom. (Photo from the NIST archives).

Doing more with less has been the standard operating procedure in manufacturing for the past ten years. Everyone looks for where they can cut corners. Many placement machines have the ability to self-calibrate and then provide their users with capability numbers. In an attempt to save resources, manufacturers are using these values in place of true capability studies. That leaves two questions that need to be answered. The first is: How valid are the internally measured numbers? Secondly, if the numbers are not valid, do they still have some value? The simple answer to these questions is they are not valid to predict yield, but they can still have great value to the customer.

The capability indexes of Cpk/Ppk are used to determine if a process is capable of producing a product without excessive scrap. The created scrap drives up the poor quality costs and also increases the probability that the customers will receive a defective product. This could possibly cause the loss of customers and sales, which in turn will drive down profits. When used properly, the indexes can predict the expected yield of the process. A six sigma process (Cpk 2.0) expects 3.4 parts per million defect rate, which will reduce the cost of poor quality to less than 5 percent of sales. To say that the internally measured Cpk was a valid number the yield expectation would have to match that of a Cpk measured with an external metrology machine. This is not true in most cases due to the way the Cpk numbers are determined and therefore, the Cpk values are not valid.

The key concept in metrology is traceability. Wikipedia states the definition as "the property of the result of a measurement or the value of a standard whereby it can be related to stated references, usually national or international standards, through an unbroken chain of comparisons, all having stated uncertainties." If this is done properly, the measured value can be compared to one taken at a previous time or anywhere in the world. The IPC9850 standard defines the way this is typically done in the electronics industry. This standard states that using an optical coordinate measuring machine (CMM) is the primary way to determine placement accuracy of a placement machine.

Coordinate Measuring Machine
The CMM is calibrated using standards/jigs that have been certified by National Institute of Standards and Traceability (NIST) in the USA or other standards agencies like National Physical Laboratory in the UK. All of these national institutes trace their standards back to the International Bureau of Weights and Measures in Paris established in 1875. This allows every CMM to be calibrated to the same standard — meaning a measurement of 0.005mm could be reproduced anywhere in the world to some level of stated uncertainty. Without this level of traceability the numbers received can not be considered metrology.

According to the IPC9850 the inspection process also needs to pass standard gauge repeatability and reproducibility (GR&R) tests 4 times a year with a PT ration of less than 25 percent error. That means the numbers measured should be repeatable and reproducible. If the board was measured multiple times and the same results were received every time, the process is repeatable.

Reproducible means different operators at different times would also get the same results. This error is defined as the process to tolerance ratio (PT ratio). It compares the percentage of error measured between runs to the process tolerance. If the PT ratio was 100 percent there would be as much error in the measurement as the process and the results would be meaningless. Besides having acceptable levels of GR&R error, the CMM also has to pass an accuracy test. This test measures the CMM to a standard that was certified by the appropriate agency. Both the GR&R and Accuracy test need to be passed to be considered an acceptable inspection system.

Ready to Measure
Once the inspection machine has proved that it can measure points accurately, within the stated uncertainties, it is ready to inspect placements from the process. These measurements become the last link in the chain of measurements that go all the way back to France and the International Bureau of Weight and Measures. In most cases, the inspections measured by the machine itself will not pass any of these tests.

The placement machine's inspections are not calibrated to national standards, and GR&R and accuracy tests are not performed on the inspection aspect of the placement machine, so the internal inspection cannot be considered a reliable metrology system. There is no guarantee that a number received by that machine can be reproduced in any other machine or at any other time. That means in most cases the internal measurement cannot be trusted to predict the yield for the process.

Another important aspect in a metrology process is that the discrimination of the inspection machine should be greater than the discrimination of the process it is inspecting. According to the IPC9850, the inspection machine should have a discrimination of 10 times greater than the process machine. When the machine is doing its own inspection there is a 1-to-1 correlation between the placement and measurement system. Beside the discrimination error, the gauge repeatability and reproducibility (GR&R) error should be a small percentage of the total error of the system.

A GR&R tries to determine the amount of system variations coming from the inspection machine and how much comes from the placement process. For a good metrology process these numbers need to be as low as possible.

Acceptable Error
According to IPC9850 acceptable error from the gauge is 25 percent with less than 10 percent being excellent. If the inspection system is the placement machine, resolving the percentage of error coming from the inspection is very difficult. The error reported by the machine cannot be attributed to placement or measurement. It is just as likely that the parts were placed perfectly, and the inspection caused all the error as the other way around — that the inspection is perfect and the parts were placed with all the errors. In reality the situation is somewhere in between those two extremes. Thus, internal inspection is not a good metrology system because it is impossible to tell whether the placement or the inspection is causing the measured deviations.

Finally, errors that are inherent to the system will not be detected when the same system places and measures the parts. All placement machines work off of a system of encoders. Whether these encoders are linear or rotary, they count pulses and determine how many pulses the machine needs to advance the motor to move the part from the camera to placement. These encoders are just scales and when these scales get dirty or damaged, the machine doesn't place the part exactly where it was expecting to place it. Placement machines have offsets and calibrations to take care of these and other inaccuracies. But what happens when the machine is asked to calibrate and measure its own errors? These errors are masked — if the system added pulses when it placed the part it will add the same pulse when it goes to inspect the part; this will give the operator the false result of a perfect placement.

Here is a simple non-machine example of this principle. Say you were going to cut a string to 9mm but the scale that you were using was defective and went from 6mm to 8mm. The string that was supposed to be 9 mm would be 1mm short. If that piece of string was inspected by the same scale it would be classified as within specifications. However, if the string was inspected by a more accurate scale, it would be determined to be 1mm out of spec. This is yet another reason why a system measuring itself is not a good solution for metrology.

In summary, a placement machine measuring itself cannot be considered metrology. It is best described as an internal measurement system, or in some cases, an internal calibration system. Even though these machines may report such statistics as means, standard deviations, and Cpk, these numbers can not be trusted as metrology. They should not be used as part of a quality system measuring the capability of an entire factory. Internal measurement systems do have value, however this value does not estimate yield or compare line quality levels.

Internal Measurement Value
Even though internal measurement is not a valid way to determine capability, it still has value. Internal measurement or calibration systems allow machines to self-calibrate to a baseline level. In most cases internal measurement calibrates a machine to its own specifications. In the early 90s gantry machines only had 1 or 2 nozzles per gantry and there were very few offsets. Today machines have 12 or more nozzles per gantry and four or more gantries per machine and, in some cases, multiple cameras. Some of these machines have more than 600 placement offsets. This quantity would be too difficult to calibrate by hand. Internal measurement systems can be used to set these offsets quickly and efficiently and get the machine running to specifications.

The statistical numbers reported after these calibrations are only a rough estimate of how well the calibration procedure did its job. If the values received from the internal measurements are low and cannot be improved, there is some mechanical problem with the machine. So these internal numbers can give some confidence of mechanical reliability and repeatability of the machine. If a machine shows low standard deviations using internal measurement there is a high probability that the machine can be calibrated to an even higher level using an external measuring device. So, internal measurement is a good first-pass tool that can determine if external measurement is even needed.

External Measurement Value
External measurement's value, if done correctly, is true metrology in every sense of the word. So a Cpk measured in France will have the same meaning as one measured in the United States. Factories and facilities around the world can now compare statistics and benchmarks against one another. Customers use metrology for the larger purpose of using statistics to predict factory yields. This allows manufacturers to understand their processes and know when something changes before defects are created. The numbers can be correlated directly to DPMOs (Defects Per Million Opportunity) measured on the factory floor, which is tied directly to the cost of that facility's poor quality. Statistics measured using external metrology allow a user to ensure that all the machines are tied to the same datum point, so programs can be transferred line to line without any issues.

If each machine is self-calibrated there is no guarantee that this will be possible. Being able to easily transfer products from line to line makes a company more flexible and maximizes its output — ultimately helping its bottom line. This increased capacity allows companies to do more with less. Having all the machines tied to the same standard also enables prototypes to move directly to production with very little additional work — again saving time and money. Using external metrology also allows the maintenance departments to trouble shoot problems quicker and easier and get to root-cause analysis. When using internal measurement, internal problems can be masked. These problems are easier to identify when compared to an outside source. Maintenance can also calibrate machines to a higher level than the internal calibrations. They can run with higher Cpk values and better quality than without external metrology. This in turn will reduce the defects on the factory floor and the cost of poor quality and increase profit. These types of external measurement devices will pass the ISO standard 9001:2000 section 7.6 for controlling monitoring devices, whereas an internal measurement system will not.

When a machine measures itself it cannot be considered metrology, and so the benefits associated with metrology are not realized by that measurement. Metrology, when done correctly, can increase the profit in a facility dramatically. This profit comes from either reducing the costs of poor quality or increased capacity of the facility through better overall machine utilization. When manufacturers use internal measurement as metrology they are costing themselves money without realizing it. For metrology to be done correctly, it needs to be done using an external source.

Note — The Home of Our Meter and Kilogram
The metric system was based on a natural concept which assumed the meter to be one ten-millionth part of a meridional quadrant of the earth. Further investigation showed that this concept was wrong and that the original metric standards, kept in the Archives of France, had not been constructed with the precision possible three quarters of a century later. With the United States participating, a series of international conferences were held in the early 1870s to construct new metric standards. A graduated line standard, equal in length to the Metre des Archives at O°C, was chosen as the new basis for the system. Rejection of a natural basis for the meter made international agreement necessary in order to maintain the validity of the artificial meter. The Treaty of the Meter (1875) mandated the establishment of a permanent International Bureau of Weights and Measures, to be located at Sevres, France, which would not only keep custody of the new prototype meter and kilogram when constructed, but make comparisons between them and the fundamental standards of nonmetrical weights and measures in other countries.

Contact: Panasonic Factory Solutions Company of America, 5201 Tollview Drive, MSC13, Rolling Meadows, IL 60008 847-637-9600 fax: 847-637-9601 E-mail: Web:

search login