The stickiest engineering problems Iíve encountered involve pushing the limits of the materials. In this case it was thin-walled injection molding and the dielectric strength of the material.
When I worked in higher-speed electronic connector development for the connector industry, we were designing electronic connectors with tighter circuit densities, higher I/O count, and ground structures designed to improve signal isolation. Some of these had very low impedance sections. If those lengths were short enough with respect to the wavelength, and adjacent to a higher impedance section of the circuit path, the averaging seemed to minimize the reflections.
On one particularly high-speed (for that time) cable system running around 300MHz, we started experiencing high-resistance shorts cropping up.
Based upon TDR examination of the defective cables, a bad location could be located. The suspect area had a 0.0035 inch nominal thickness molded wall in LCP adjacent to a metal ground plane.
We found the specific circuit problem area was best located by high-potential testing at around 1,200V ac. Cracks in this connector area could form by any flexing of the plastic housing prior to assembly into its stronger external structure. These cracks would then have only air as the dielectric, which was weaker than the plastic, and would eventually form a silver dendrite short, bolstered by a carbon arc track.
Through careful DOE optimization of the injection molding process, the insulator poor knit/flow lines could be reduced without injection pressures that would destroy the fragile core pins for the tiny terminals. We also established better packaging and handling methods to reduce mechanical damage to the fragile component. We thought we had the problem solved, but problems still persisted at a lower occurrence rate.
Further cable assembly troubleshooting with these same TDR and High Potential tools uncovered that we had carbon lumps in the plastic wall, causing a high resistance short through this thin wall.
The root cause of this carbon was found to be two-fold: A) If the black pigment for the plastic was not ground finely enough, or it lumped back together, the thin wall could be bridged with carbon. B) If the material degraded in the molding press barrel at any time, specks of degraded and carburized material could shed and come through with the other raw material. Either source could cause the failure mode.
Our solution was to convert from black to natural (cream-colored) neat plastic material. The pigment was then eliminated. Also, degraded specks could be visually separated, and if the barrel required disassembly and cleaning, it was readily apparent. Additionally, a high-potential test of the raw connectors with probes at the higher 1,200V AC voltage prior to termination was added as an additional testing net to capture any subtle stragglers with carburized specks.
In summary, we had two problems in the thin insulator wall: 1) cracks, and 2) carbon lumps. The second problem had two root-cause sources contributing to the problem. 2A) black pigment, and 2B) degraded plastic carbon specks in the molded part. The problem incidence rate was about 80 percent from cause 1, about 16 percent from cause 2A, and about 4 percent from cause 2B.
This shows there can be multiple concurrent problems in the same area. These problems may be independent or interdependent. A way to monitor the performance is required, and sufficient perseverance must occur to resolve all the root causes. Just because you found a root-cause source of the problem doesnít mean the problem will totally go away. Conversely, just because the problem didnít totally go away doesnít mean you didnít fix a root-cause source of the problem.
David T. Humphrey is a manufacturing engineer.
This entry was submitted by David T. Humphrey and edited by Rob Spiegel.
Tell us your experience in solving a knotty engineering problem. Send to Rob Spiegel for Sherlock Ohms.