By Mark Thompson | Published on: July 17, 2017
I use the term “Design For profitability” as opposed to DFM (Design for manufacturing), DFT (Design for test) or DFA (Design for assembly) for a few reasons. If you practice and are conscious of all of the above it really becomes Design for profitability. In this article I am going to go over just a few of the things fabricators routinely run up against and what their solutions are...
I will start with Design for Manufacturing or DFM. Generally the first stage for prototyping and something that depends greatly on the capabilities of your chosen fab shop.
Some designs are finished with auto-routers after the critical traces have been hand placed. It is at this point that sometimes unintended issues can arise between Design and fab.
An example of this is same net-spacing violations where a track may “double back” near a surface mount creating same -net spacing violations. See example fig 1. Whereas the software does not see these as legit violations as they are same net, a fabricator knows that any features creating spaces less than .003 can easily flake off at the image stage and create havoc elsewhere in the form of shorts. Edit time must be taken at the fab stage when these same net spacing violations occur and the slivers eliminated. Some CAM software packages have a “sliver fill” option but again this requires additional edit time at CAM.
Other high edit time potentials are jobs that have been outsourced for design/layout and the contractors use an auto –router for routing non critical traces. These auto routers are notorious for making silly decisions like not centering the trace between pins creating potential gap violations.
An example of time critical jobs being slowed down would be a drawing with a specific callout for impedance's based on metric sizes but the Gerber files are typically output as “inch” units. These slight rounding errors many times mean a phone call is necessary to make sure we are dealing with the correct impedance line sizes.
This brings me to my next point. Many times in our industry an end user asks for PANELIZED data from the fabricator and uses this PANELIZED data to send to other fabricators. The thought process would be that borders with rails, tooling and fiducial placement will have already been established. This is a HUGE risk and is Not at all advised by the fabrication community for the following reasons...
One way of ensuring the data is as the customer expects it to be is to send check plots but again, Even “check plots” can be mis-understood as typically they are sent AFTER any fabrication mods have been done for both drill and image compensations. From the fabrication standpoint the purpose of the “check plots” is to verify things like locations and sizes of additional frame fiducials or tooling holes.
So How can a designer ensure the broadest range of fabricators can reliably build their products? Simple, design for maximum producibility, allow for process variances IN THE DESIGN.
Let me give you an example of this philosophy.. I get phone calls all the time from our customers asking for a minimum annular ring size for a Given hole size. If maximizing the producibility for the part, My answer is always the same, For Signal layers ( both internal and external ) make the pad size .010 larger than your desired finished hole size. This allows for .004-.005 drill compensation ( to plate back down to the intended Finished hole size.) and .002 per side annular ring prior to process to deal with any slight mis-registration. For Inner plane layers obviously a fabricator would like to see more distance between the edge of a plated hole and the adjacent copper pour for plane so they can allow for their machine tolerance and any drill mis-registration that may occur. ( with today’s drill machines this is less and less of an issue)
Lastly, Let’s talk a little bit about RF features at a Conventional fabricator. I will attempt to go over some common issues on artworks provided for RF jobs. First off, The very nature of RF is that the features must be tightly controlled. Typically RF features are Done as “Constructs” or “Islands” as using standard “drawn” features leaves undesirable radii. This Means a fabricator must be very careful when attempting an “etch compensation” for the know loss at An etcher based on starting copper weight. All features must be selected on the constructs or they will Not perform as expected. Net-lists are helpful but can also cause delays on RF jobs…Let me give you an Example:
If the net-list definition assumes a later connection to a surface feature say through a metal screw head Or plated half hole ( castellations ) you can get false Open nets that can cause delays.
A good Fabricator sees an RF device for what it is and understands certain “rules” apply in CAM. Things Like clipping back metal running to a part edge ( assuming no z-axis connection is to be made) Normally A shop may clip metal back anywhere from .004 to .010 so burring does not occur at final rout. Not so On RF type jobs, Trimming the RF leads more than .005 can sometimes result in performance issues and Is Generally not done unless the customer is aware and has approved.
Many RF features themselves can look very much like “stubs” or un-terminated traces and some CAM Systems may flag them as such. Even on predominantly Digital boards, RF features such as Antenna’s Can look like stubs or un-terminated traces.
Clear S/M seems to be a favorite of RF engineers who like to see that the feature geometries as Designed ( WYSIWYG- What-you-see-is-what-you-get)
Here at Prototron, We know that unlike conventional designs, RF designs have characteristics that Require an understanding of the intended performance of the product. Almost as Important, Communication With the customer is key when any modification is being considered. Even the slightest Variation in the End product from the design can yield performance issues.
Copyright © 2013 Prototron Circuits, Inc. All Rights Reserved.