Design007 Magazine


Issue link:

Contents of this Issue


Page 44 of 64

February 2014 • The PCB Design Magazine 45 the bare (board) truth UNDERSTANDING THE TyPICAL CAM PRoCESS continues Exception to this rule is quarter oz copper foil where no etch compensation needs to be added. Quarter-ounce foils are typically only used for outer or surface layers. Again, your design should be able to accom- modate these compensations. Example: If your design is .003"/.003" and you are asking for 2 oz. copper you will get a phone call from your fabricator. 3. Panelization If you are ordering parts as an array either as a tab rout, score or combination of both you will want to provide a sub panel drawing with any particulars your assembler may need like additional tooling, fiducials, specific areas where not to place a tab due to part overhang issues at assembly, etc. If you have no preferences and your chosen assembler does not need anything special typi- cally the board fabricator can come up with a good sub panel that works well for both manu- facturing and assembly. For instance: Even if no tooling or fiducials are specified for the given ar- ray, many times fabricators will add them to the array anyway. They are generally of no harm at that point and can be quite helpful if tooling and fiducials were not foreseen. 4. A 1-up compare and net-list compare The fabricator will perform a 1-up compare at the CAM stage even if no IPC netlist is pro- vided. This is to ensure the addition of the drill and etch compensations does not create manu- facturing capability violations. A 1-up compare is only that; it is not a netlist compare. If an IPC- 356 file is provided, the fabricator will also run the provided IPC netlist against your provided exported Gerber data. Remember this is a de- sign vs. exported Gerber file comparison. If you do not provide an IPC netlist for a class 36012 job, your fabricator will ask for one. I have been asked many times over the years to simply gen- erate a net-list based on their exported Gerber data. If a netlist is generated by the fabricator based on the Gerber data, at what point would we ever find a mismatch? So what is my point? If you would like us to verify your design itera- tion against your exported Gerber data please provide an IPC netlist. A brief note about IPC net-lists and their gen- eration: Avoid creating test points for things that will be connected only after the devices are load- ed, such as surface mounts or edge-plated castella- tions. This will avoid a phone call from your fab- ricator with false or erroneous "broken" or open nets. Likewise, many times in a design an AGND to DGND short is designed in and is intentional. Make note of these intentional net-list anomalies. Throughout this article I have used the term "artwork" but today you have fabricators that can send the image data from CAM directly to a direct image device, eliminating the use of films and thus another transgression of the artwork image. Every copy you make loses a little in the translation. If you can image the cores and out- er layers directly, this loss is minimized. Common Fabrication CAM Issues 1. RF parts RF type applications typically require a spe- cific design tool suite that does not always deal with things like RF impedance controlled traces in a conventional manner. For example: RF traces are often drawn as a composite of many smaller line sizes to be able to deal with the unique shapes and reflections needed for RF work. This is a nightmare for a CAM operator. The size being controlled may be specified as .025", but when the CAM operator queries the apertures used for .025", they do not exist. Again, this is because they have been drawn as a conglomerate or composite of many smaller trace widths. 2. Same size trace width for single- ended impedance structures, differential pairs and copper pour This does not allow the CAM operator to eas- ily select one or the other for any unique re-siz- ing that may be necessary for meeting or getting closer to the customers' desired impedances. For years, I have advocated using a tenth or a hundredth of a mil to differentiate between single-ended traces, diff pairs and copper pour that are using the same draw size. Fabricators cannot resolve this small difference but it al- lows the CAM operator to select and deal with those structures without affecting the others. I am happy to see this is becoming the norm.

Articles in this issue

Archives of this issue

view archives of Design007 Magazine - PCBD-Feb2014