by Leland Teschler, Executive Editor
Nancy Viter
Even smart engineers can make dumb mistakes when they create printed circuit boards. A few stories from the trenches illustrate how problems often arise in this interview with Nancy Viter, Director of Manufacturing, Sunstone Circuits, Mulino, Ore.
Some engineers think the design process is on the home stretch once the printed circuit board specs go to a board manufacturer. But surprising problems can arise in the process of getting a workable PCB out the door. We spoke to Sunstone Circuits’ Nancy Viter about such subjects as typical practices that make PCBs difficult to manufacture and why it may not be a good idea for engineers to blindly trust PCB layout software. Viter has been Sunstone’s Director of Manufacturing since 2008. Sunstone has been manufacturing PCBs since 1972.
Design World: It’s sometimes said that design engineers don’t make the best manufacturing engineers. Can you give us an example of that phenomenon when it comes to circuit boards?
Viter: The engineers we deal with who send us board designs really are brilliant. But we still find problems with what they send us. Our pre-engineering people look at designs to see whether they will make it through the manufacturing process and whether the finished board will work as intended. They often spot things that cause customer pain points.
In one case, we received a design that included surface-mount attachment points to an inner layer of the board. We weren’t sure the board designer understood what he’d done. As I mentioned, sometimes we deal with real rocket scientists. What sometimes looks like a mistake to us can actually be an example of cutting-edge technology that isn’t yet widely understood. So we weren’t sure whether we were dealing with a design defect or just something completely new. We questioned the engineer about how he would get surface-mounted components onto that inner layer and he insisted the design was right.
Well, we turned around the board and once it was in the engineer’s hands, it finally dawned on him that there was a problem. He had simply mislabeled one of the layers. On ensuing board designs he listened more carefully to us when we pointed out opportunities for improvement. It’s a lessen he’ll probably never forget.
DW: What happened to his board?
Viter: We ended up rebuilding it for him as a courtesy. We looked at it as a way of showing him he was in good hands with us.
But that little episode illustrates a point. Engineers should note how PCB fabricators handle problems when there are errors. Returns should get top priority because they’ve already been out the door once. Not all fabricators have this philosophy. It’s not uncommon for returns to go to the bottom of the list because the sale has already been made. Engineers should ask ahead of time what happens in the event of a return and whether it will take another two weeks to get a rebuilt board back.
DW: Nowadays, many engineers use PCB layout software to do the board design. Why do problems like this show up despite the routing automation?
Viter: You would think that automatic router software would inherently avoid putting traces in areas that would make the board hard to manufacture. But while the concept of an automatic router is good, the software is only as smart as the person who set it up. The design will only work if the right rules are in place.
If rules are not set correctly in PCB CAD software, the result can be spacing problems during auto-routing. This example (top) shows some 0.001 and 0.003 spacing violations on an otherwise simple board. To handle the cramped dimensions, a board fabricator might end up quoting this job priced for tighter technology pricing. A customer-centric board house will offer suggestions to reduce cost. In this case, a simple repositioning of the trace (bottom) changed this to an “easy technology” category, saving thousands of dollars in the transition from prototype quantities to production runs.
Part of the problem with routing software is that as it went into wide use for board designs, engineers using the software tended to get little or no training on how to lay out a board so it would be manufacturable. That training element is really lacking and it puts a big burden on the relationship between the engineer and the board houses.
DW: What problems have you seen arise because of PCB routing software?
Viter: It takes time for someone to get up to speed on how to produce a manufacturable design with it or even understand what they are seeing in a given design. The average engineer can look at a design on a 15-in. screen and it looks doable. A little 2-in. square board blown up on a big screen can give the false impression that there is plenty of space between the traces. But once the board is made, those spaces that looked as though a truck could drive through them can end up being about the width of a human hair.
That’s where the fabricator can work with the engineer to great effect. There may only be a couple of areas that are difficult to process. There may be other real estate on the board available. If the routing had been done slightly differently for just a few traces, the board could cost significantly less. You certainly don’t want to go into production with something that costs four times more than what it could. A fabricator can fine-tune a board design up front to keep this from happening.
DW: Do you see problems other than just the board design that arise because of short comings on the part of the engineer doing the design?
Viter: The generation of the board design file itself can cause manufacturing problems. The size of the file can be an issue. File sizes comprising several gigabytes typically cause problems in data transfer and manufacturing. We’ve noticed size problems becoming increasingly frequent as ordinary computers have become more powerful. It used to be that engineers were more sensitive to memory constraints. Now even laptop computers have 4 GB memories, so the mindset seems to be that any file size is okay, when in reality, a file that’s too big can slow or stop the manufacturing process.
DW: Are there any board features that tend to be problem areas?
Viter: Logos on the board are habitual hiccups. PCB CAD programs let the designer create a logo for the board, but the problem is that the design must typically be output to a Gerber language. Gerber is a vector language that serves as the de facto standard for describing PCB features, but it is just not a smart language. Logos often come out garbled or with pieces missing.
In one board for a military application, the only way we could get a logo to appear correctly was by splitting it so one segment of it went on each of 99 different board layers. In a prototyping shop, that kind of issue can severely slow the board manufacturing process. Again, collaboration with the fabricator can prevent manufacturing hold-ups and the disappointment of getting a logo that doesn’t look right.
DW: Is the problem only with logos?
Viter: Converting design files to Gerber can cause other problems as well. Uncommon shapes (called apertures in PCB parlance) can come through differently. For example, a board might need a pad shaped like an octagon for some reason. When it is put on the board, its rotation may change slightly so the points on the octagon create a space violation or a potential short. Twisting it a few degrees might solve the problem. That’s where the fabricator should talk with the engineer and ask a few questions.
Solder masks created by PCB CAD programs are rarely problem free. In this case, (left) the solder mask openings are so expanded that the clearances overlap, obliterating any mask dams between surface mounts. A lack of mask dams can cause significant bridging during the assembly process if left as-is. One way to fix this problem is to either adjust settings in the PCB CAD program to create a global expansion equalling half the distance of the design’s smallest circuitry space, or create a global “no swell” (pad-master). Then the board house can adjust according to its capabilities to provide coverage on the board where needed and appropriate openings for solderability.
It isn’t just Gerber files that can output incorrect information. PCB CAD packages are notorious for not producing solder masks the right way. Solder masks can be a touchy situation because engineers often don’t think too much about them. You have to make sure the mask leaves the pads open so they’ll take solder while leaving protection between the thin lines of solder. That’s tricky where there are tight clusters of surface-mounted components. The trouble is that PCB CAD packages don’t provide a lot of help with getting the solder mask right. It is another area where the board fabber can help.
A free or inexpensive Gerber viewer, or one of the many free dfm services, can help catch Gerber errors such as improperly shaped pads. Pads with special shaping are notorious for being misinterpreted, so it is best to use them only when absolutely necessary. A 0.002-in. space changes into a 0.006-in. space when output correctly.
DW: Are there things design engineers do that aren’t particularly good practices for PCBs?
Viter: Yes, they sometimes specify super-tight dimensional tolerances for no good reason. We often see templates with numerous tolerances called out, which are probably not necessary for the project at hand. Positional deviation, board thickness, board size…if you over specify tolerances on any of these factors it affects manufacturing yield. You end up over-building to get enough boards with the right tolerances.
DW: Where do you find the most problems with overly tight tolerances?
Viter: Hole position is a good example. We might see one called out with a true positional tolerance of 2 mils (0.002 in.). But 5 mils is the standard on Class 2 PCBs as defined by IPC-6010, the standards document spelling out PCB topologies. And Class 2 is the norm for industrial products that don’t need extreme high reliability. The thing to note is that board manufacturers are constrained by the size of the drill — the smaller the drill size, the more deflection there will be. It is difficult to control the deflection beyond a certain amount. Few manufacturers can hold a 2-mil tolerance reliably, so they will over-build to make sure they produce enough good boards. Also, you have to ask whether you will even be able to measure tolerances that tight. The whole exercise might just be a case of wasted dollars on everyone’s part.
DW: Are there any other examples of misdirected practices that engineers should be aware of?
Viter: We are seeing a lot of engineers ganging four or five designs together on one build. That sounds efficient, but it usually causes more problems than it cures. For example, often some of the board designs have been done with different CAD packages or they are different types of boards, so they can’t easily be done in one batch. You might have one fine-lined board with 4 and 5-mil traces, another that is basically one big ground plane, a third with spots for ball-grid arrays and specific kinds of surface-mount components, and yet another with embedded ground planes and traces.
Each of these designs must be treated differently during manufacturing and with varying process constraints and handling practices. It is okay to gang boards together when their characteristics match, but ganging dissimilar boards together can delay the manufacturing process or give yields that will be disappointing.
DW: Are there any other trends in PCB fabrication that should be mentioned?
Viter: We are finding that PCB buyers are starting to appreciate companies that treat their own employees well and that encourage good work-life balance. We’ve had a number of conversations with firms that want low-cost boards, but are willing to pay a little more to work with companies having good core values.
Sunstone Circuits
www.sunstone.com
Filed Under: ELECTRONICS • ELECTRICAL
Tell Us What You Think!