Issue link: https://iconnect007.uberflip.com/i/1126810
12 DESIGN007 MAGAZINE I JUNE 2019 difference between a successful design and a quagmire of self-created issues, such as com- mon-mode emissions that can make or break passing EMI testing. I even learned a few things about via fill- ing with silver-epoxy or copper-epoxy for ther- mal applications. Remember, if you are using a via fill for conductivity, you are missing the point; the electromagnetic fields are generated around the barrel of the holes, not the materi- al filling the hole. In some cases during Rick's presentation, I wanted to slap myself in the forehead and say, "I should have had a V-8!" Rick showed one slide with four examples to be rated from least to most of radiated EMI. Fortunately, they are the same as a firing order in most four-cylinder combustion engines (1, 3, 4, 2) and pretty much everyone understood not to rout a trace over a split or a slot. On many occasions during Rick's presenta- tion, I began to make sense of some of the cra- zy things we have been asked as a fabricator over the years. Eight years ago, one customer asked us to mill a channel in a thick PCB using an end mill to make the corners as square as possible (not routed with a radii) and plate the trough and cap the whole structure with what appeared to be a wide plated "trace" running the entire length and width of the channel. At either end was a small hole where I was told they would inject a liquid. Then, I saw Rick's slide saying, "Apply a wave whose half wave- length is equal to distance X. We have a perfect balance and a true waveguide." We were basically creating an acoustic wave- guide through the Z-axis of a PCB. A 1995 arti- cle by L. Dwynn Lafleur and F. Douglas Shields of the Acoustical Society of America noted, "It is shown that certain combinations of material properties can produce a mode, which in zero frequency limit, has plane-wave motion in the liquid." This is presumably why the customer was injecting liquid into the waveguide. Great stuff! I did know a few things that Rick covered in his class, like routing longer differential pairs and clocks inside the PCB and keeping the short- er traces to the outer layers. But I only thought I knew why; when solving for impedances in- ternally, the differential pairs and single-ended structures are more forgiving. The electromag- netic wave field is not completely contained in the substrate as with a stripline used internally where the EM field is contained within the sub- strate. This is at least partially due to the Dk of air over the trace on an outer layer. Great Fabrication Starts With Great Design Again, I remember when I was asked to re- view a design for a customer from the stand- point of manufacturability. It was an 18-layer board where all layers had impedances. The layers had reference planes like a checkerboard above and below signals alternating back and forth on every layer. From a pure design perspective, it was a work of art. But could it be produced? It turned out that all the traces were laid out with the same trace width and different spaces (for multiple differential pair scenarios of 90, 100, and 120 ohms) and ground separations on every lay- er. In conjunction with that, all of the copper pour used the same trace width they had used for the signals—both single-ended and differ- ential pairs. There were five different scenarios with the same trace width, making it virtually impossi- ble to tweak any of the signal traces, and all of this had to be done through changes to dielec- trics and material type. I spent the better part of a day looking at what could be done and came up with a specific dielectric and material combination to make it all work without hav- ing to change the trace widths at CAM. With five different scenarios, that would have been a nightmare. Along with the stackup and material type, I said that the design would benefit from chang- Remember, if you are using a via fill for conductivity, you are missing the point.