We are currently laying out a new flex design, any thoughts on how we can “design” our costs lower?

You have already begun to lower costs as you realize cost control is born with the very first CAD command! Your question is also timely, as recently I have seen a rash of what I would describe as “intolerant tolerance.” Drafters and designers are an exacting bunch, a very good trait to have for the job. The problem begins when we set up tolerance blocks on a flex-circuit like we would use on a piece of tool steel. Flex materials are inherently unstable, changes in temperature and humidity can have a big impact on finished dimensions. Our experience has shown the potential for a mil or better of stretch or shrink per inch. I see many drawings with outside trim dimensions (edge to edge) held to +/- .005”. While this may be acceptable for aluminum, it is far beyond the ability of common flex materials. The very nature of a flex-circuit is to yield and bend. Super tight tolerances are rarely required.

So how does dimensional tolerance impact cost? When given a trim profile window of .010”, I, as the manufacturer, am “forced” into laser tooling. If we are lucky and the quantity justifies it, we can order a hard tool punch and die. Neither method of tooling would be considered inexpensive. Changing the trim tolerance to a .020” profile window can have a dramatic impact on cost reduction as it allows the use of “simple” chemically milled dies (CMD’s) or steel rule punches (SRD’s). If your flex-circuit grows much above 10” in any direction, you are better off going to a .030” trim profile window. Ideally, larger circuit trim dimensions should be “reference.”

Keep in mind, profile dimensions are not the only dimension that is often toleranced beyond the base materials inherent abilities. While hole sizing can be easily held to +/- .003,” “hole to hole” or “hole to edge” positioning cannot. Within a given hole pattern, cluster tolerances can be held tight, but “cluster to cluster” tolerances are limited by the same stretch/shrink issues as our trim dimensions. Demanding “hole to edge” tolerances are indeed possible but remember tooling and production costs will rise (compare minutes of laser machine time to sub-second punch cycle times).

Here is where the concept of “key characteristics” becomes our best hope for controlling cost. Examine your design for what features are truly important and call them out for special attention. If only one dimension is critical, then by all means ask for tight control, but do not pass the requirement on to other features that do not require it. An example of this would be a ZIF connector pattern. A ZIF requires a tight “etched feature to edge” tolerance, often as tight as +/- .002”. This is certainly achievable, but if the remaining trim profile were held to the same tolerance, price would grow at an alarming rate! Note: I find it strange that even though critical to function, I rarely see a ZIF pattern with any positional tolerance requested.

Feature size is next on the “top toleranced” list. There seems to be a propensity to design very narrow and complex “slots” in flex-circuits. In the example shown (Figure A; an excerpt taken from an actual drawing) we have a narrow slot with multiple champhers and tight tolerances. Can we build this? Sure, but at what cost? Does the design really require the exacting champhers? There are no cheap steel rule dies that can be made to punch a .015” slot. A punch and die is perhaps a possibility but fine feature size would make a multi-stage tool a necessity. The best tool for this job would be laser trimming ($$$). So what are our alternatives?

Figure B shows us a simplified profile easily made with an SRD or CMD. The “slit” continues to a strain-relief hole just as our original slot. The end result is a circuit that can bend and form exactly like the original but can do so for at least 10 percent less! “Slits” like this are not limited to just a single layer board but have been successfully used in boards up to 8 layers thick. After a review with our customer, we went with the “slit” option and a 20 mil profile. The parts have functioned as intended while saving the end user significant costs over time.

I wish I could recall who said it, but I am reminded of a comment I overheard at a recent design conference—“75 percent of your circuit cost is under the control of the designer.” From what I have seen after reviewing countless drawings, I certainly believe it. Don’t draw your way into a corner! n