Gaudentiu Varzaru POLITEHNICA UNIVERSITY OF BUCHAREST
Not long ago, I answered Editor Andy Shaughnessy’s “Whose Fault is That Bad Board?” survey. When I answered the first question (“If a board fails in the field, whose fault is it, typically?”), I was disappointed that he used radio buttons instead of checkboxes. I did not want to blame only the designer for every bad board in the world. Did Andy want me to name the ones who are most often blamed? Who are these designers? PCB designers are like magicians; they can materialize an idea from a piece of paper, and many of them are also the creators of the product. And designers create many jobs. Their projects may have gaps, but I would not blame designers for all the bad boards. They are the first to be blamed because they take the first step in the product’s life cycle. They can make mistakes too, but sometimes their fault is having too much confidence in the people who follow up on their work. I know an American entrepreneur who went to Poland to open a PCB design bureau. He found painters and architects for hire, but not many engineers; he was very pleased to find many electronic engineers here in Romania. But are all engineers qualified to be great PCB designers?
Years ago, I held a position in an EMS company where projects were analyzed before sending them to be produced on the assembly lines. We found that even some of the best and most innovative circuits could not be manufactured. Why? Because the PCB designer, an electronic engineer, was not acquainted with the fabrication process. He had no idea about the technological requirements necessary for electronic production. Here is a funny story: I know of one designer who learned, finally, the importance of the thermal relief pad for heat restriction during reflow for good soldering. His response? “Oh, was that what they were for? And to think I worked so much to remove them!”
The CAD program itself had introduced thermal bridges where the pads were linked to large copper areas, but the designer’s eye did not like the way
Figure 1: V-cut panelized 10” x 10” circuit board.
they looked. This was a happy case because the designer had presented the project before sending the order for fabrication. But other times, matters were much more serious. When a board came in for assembly, it was necessary to manually heat the pad and the component with two soldering irons. Some designers understood this aspect (especially after they were walked through the factory to see the whole technological process), while others even got angry, yelling, “I will send the project to China and they will do it!” Yes, they will, but they will fabricate exactly what was sent, including the design errors. This was the case once when a designer forgot to send an Excellon file; the printed circuit board was manufactured without the holes for the 40-pin DIL package of a microcontroller.
Some designers will gladly fit the schematic on the entire sheet. One designer learned that, with the right modifications, the area of the printed circuit board could be reduced, and thus the cost of the board could be reduced. He replied, “Oh, it is for the Army, and they have enough money not to worry for the size of the board!”
I think such an army has lost the battle even from the design stage: More materials mean heavier weapons, more fuel consumption for transportation, larger pack sizes, fewer weapons, less money for further development, and so on. So, in order to be a professional designer of today’s electronics, it is not enough to be an electronic engineer; it is also necessary to have the proper technological knowledge or to team up with a technologist. This is one reason that we decided to include a DFM course in the electronics curriculum at my college.
Nowadays, printed circuit boards are more complex. Many PCBs now have controlled impedance requirements. Flex-rigid is more popular than ever, and we’re seeing glass circuit boards and many metal-backed PCBs. They have become dedicated passive components; some are even active, such as the embedding technology from Würth Elektronik, which can embed flip-chips within the multilayer structure. Moreover, an assembled electronic module may have several thousand solder joints, which are also components.
As my colleague, Ioan Plotog used to say, “Solder joints are living things.” They are not immutable; due to the environmental conditions, different kinds of stress (mechanical, thermal) and aging, they are changing outside (tin whiskers), and they are changing inside (microstructure), so their electrical, mechanical and thermal functionalities are affected. It is known that ESD issues may hit a long time after the product is launched into the market.
With circuit boards, one hand doesn’t always know what the other hand is doing. For example, did the component supplier or the electronic assembly provider respect all the preventive procedures against electrostatic discharge? I heard about a board that functioned on a rack but stopped functioning when it was removed and inserted into another rack several meters away. That was the day when the operators from the electro-mechanical telephone exchange discovered that walking on carpet charges you with several thousand volts and the circuit layers of the electronic module must not to be touched. The cleaning of the boards is another problem which may have a delayed effect due to electrochemical migration.
Figure 2: The component could have been damaged if a PCB depaneling machine were used.
Once, a customer brought in a panel board for assembly. The panel contained 100 pieces arranged in 10”x 10” small circuits for V-cut (Figure 1). In order to print the solder paste onto the board, a stencil had to be manufactured. The designer didn’t have the Gerber files for the panel, but for a single circuit. However, when the design of the panel was ready, it was observed that the dimensions of the panelized project did not fit the dimensions of the physical panel.
We searched for the cause, and we concluded that the PCB manufacturer did not make the V-cut panelization according to IPC standard recommendations; instead, the board shop took the V-cut spacing from the useful area of each circuit. This led to a manual depanelization in order to avoid component cutting by the depaneling machine.
This is an example that can be classified as a fault on the side of the PCB manufacturer because their operators did not follow the IPC-7351A “Generic Requirements for Surface Mount Design and Land Pattern Standard” recommendations. Therefore, we have decided that our faculty should include IPC standards (especially 2221, 2222, 600, 610, 7351) in the curriculum package for our engineering students. However, as I used to say to the young student designers, “Do not let others decide for you. Send complete information!”
Here is an example of a chain of mistakes featuring three actors. The electronic module was not working properly at the final test made on the customer’s premises.
After investigating the situation, it was found that certain PCBs presented defects like interrupts and badly plated vias. The boards did not present any trace of test needles, so it was concluded that the manufacturer of the printed circuit board did not perform the electric test, and some boards passed through, although they were defective. Later, the manufacturer agreed that their secondary facility in China did not do the job properly. Here is the first mistake: The PCB manufacturer did not follow test procedures.
The electronic assembly manufacturer did not check the bare boards at the incoming quality department because he had too much confidence in the quality of the Korean PCB company’s products. That was the second mistake: The electronic assembly manufacturer did not follow test procedures again. He assembled all the boards and after a routine optical inspection, and he concluded that all the components were correctly soldered.
The PCB designer did not supply the test procedure for the assembled electronic module, so the EMS company delivered the boards without testing them, except for the optical inspection of the solder joints. This was the third mistake: the customer did not deliver to the EMS company any test method for the assembled boards. The bad boards could not be repaired satisfactorily, so all the electronic components (a total cost of over US$1,200) were lost.
Yet another chain mistake example: When trying to fix the BGA capsule in a socket, it detached very easily from the PCB (Figure 3). Of the 12 sockets on the same electronic module, four had this problem. Going back on the route traveled by the electronic module, the following mistakes and miscues were found: A customer (also a designer) ordered the manufacture of a very expensive board to an EMS company, providing all the components except for one – a 1,156-ball FPGA, which the distributor could not deliver together with the whole bill of materials (BOM). This was the first mistake: Assembling a board without a complete BOM.
Figure 3: The BGA socket.
Figure 4: The solder paste melted and formed the joint with the balls of the socket.
Having in mind that a third reflow process would be necessary for FPGA assembly, the technicians from the EMS company assembled one side of the board with lead-free solder paste, and the other with tin-lead. They assumed that during the second reflow process, the solder joints from the first side would not melt and the component would not drop in the oven.
This was the second mistake: They mixed the solder alloys SAC with SnPb; the FPGA contained SAC alloy balls. In order to solder the new FPGA onto an 18-layer FR-4 board with components on both sides, a lead-free soldering thermal profile was used. The thermal process took place on an SMT rework and repair station using hot air for upside heating and infrared radiation for backside heating.
Inherently, the heat affected the components in the vicinity of the FPGA package, several BGA sockets among them. According to the datasheet, these sockets had SAC balls, but they were previously soldered with an SnPb alloy, too. At first glance, the solder paste melted and made joints with the socket balls (Figure 4). However, it was very curious to see that almost not a single ball remained on the socket’s spring (Figure 5). It looks as if, instead, the ball was not properly attached to the socket’s spring. Could it be a fake component (Figure 6)? This problem of distribution is another cause of bad boards, which seems to never stop.
Figure 6. Is this new component good or fake?
Figure 5. Where are the balls?
When a chain mistake occurs, the rule of 10 comes into play. The Latin expression “quod erat demonstrandum” is illustrative. You may think that none of this provides an answer to Andy’s simple question because in none of these cases did the board reach the client; this all happened inside the production stage. Who was at fault when a defect of the airbag system forced Nissan to withdraw cars from the market due to a bug in the cruise control deactivation switch manufactured by Texas Instruments? The PCB designer? The electronic module manufacturer? The car manufacturer? Everyone?
Why, despite the many DfX initiatives (including Design for Zero Defects and Design for Six Sigma), do these errors still happen? Basically, rules are for humans. I read an opinion column recently that described robotization as the single best way for America to bring back electronics products from China. If this happens, will the mistakes disappear? No, not if the robots are controlled by humans. As another old Latin saying goes, “Errare humanum est.”
Gaudentiu Varzaru is a researcher at the Politehnica University of Bucharest and a show manager for the TIE PCB design conference.