Automotive Rain Sensor PCB: How Optical Paths, Potting, and Fabrication Quality Decide Field Performance

Production-End Testing, Sealing and Batch Consistency

In the years I have dealt with automotive-grade rain sensors, my biggest feeling is that this industry has too many people placing their attention on chip solutions while neglecting the board that carries the entire optical chain. Whenever the Automotive Rain Sensor PCB is mentioned, many people reflexively go to find an HDI PCB manufacturer who can do impedance control, thinking that high enough layer alignment accuracy is enough. In fact, what is truly thorny is not routing density but how you stabilize those few optical paths on the board.

I have seen several HDI PCB suppliers whose prototype boards passed all electrical tests but showed their flaws as soon as they were put on an optical bench. The substrate under the infrared emission and receiving windows warps at the micron level after several reflow cycles, invisible to the naked eye, but the light spot shift is enough to invalidate all calibration data. At that point you realize that when choosing suppliers, you cannot only look at whether he is a high speed PCB manufacturer or a high frequency PCB manufacturer; you have to see whether he has long made hybrid structures with optical windows and whether he understands how copper foil distribution affects local thermal expansion.

Last year one of our projects temporarily switched to another heavy copper PCB supplier, because in the original board the potting adhesive and the copper surface delaminated after thermal cycling. That board had a thick copper coil on the top layer for the heating function. An ordinary heavy copper PCB manufacturer only piles on copper thickness and ignores the current density concentration effect at the copper edge, and local overheating makes the potting silicone slowly lose its elasticity. Later we re-prototyped at a factory that was both a high frequency PCB supplier and a heavy copper solution provider. They designed the heating copper with a gradient width and used a low-stress solder mask opening around the optical window. After potting, we sampled fifty pieces and ran two rounds of shock from minus 40 to 105 degrees, and the optical transmittance attenuation did not exceed 2%.

Many people think that once the optical hurdle is passed, what follows is conventional EMC and vibration, but in fact the testing system at the end of the production line is the job that depends most on experience. Automated test fixtures are not something you can use just by setting up a pair of transceiver tubes. You have to simulate the real water film state on glass, reproducing everything from a thin mist to a continuous water layer, and the fixture itself must not be disturbed by ambient stray light. We paid for this early on. A bare board test was all normal, but after potting, the retest showed that sensitivity had drifted collectively by nearly 10%. After disassembly and analysis, we found that during potting a trace amount of adhesive had seeped into the micro-lens structure of one receiving window, changing the refractive interface. Later we changed the potting process to vacuum assist and added a window cleanliness check to the 100% test on the production line, and only then suppressed the consistency problem of the whole batch.

In the end, manufacturing opto-mechanical-electrical sensors like this is no longer simply an “assembly” concept. You cannot release a part once the electrical tests pass, as you would an ordinary automotive ECU. What causes mass incidents is often not design defects but a change in the flowability of a certain batch of potting compound, or a PCB supplier changing solder mask ink without notifying you. So now, when I evaluate a new supplier, I always pull out the optical function test data and look at three consecutive production batches, at the mean and at Cpk. That is the last valve of the “shutoff fault” at the end of the production line.

Manufacturing Stability Before Optical Concepts

I have done automotive electronics hardware design for several years, and one of the projects that gives me the biggest headache is the rain sensor. Many people, when talking about this topic, immediately focus on the optics, as if that tiny lens and infrared transceiver tube determined the success or failure of the whole product. I have a completely different view. The real lifeline of the Automotive Rain Sensor PCB is not the optics itself but the manufacturing stability and long-term material behavior of that board. You can finish the optical design in two months, using off-the-shelf lens groups with calibration algorithms, and basically nothing goes seriously wrong. But to produce the PCB, wandering among different HDI PCB manufacturers and HDI PCB suppliers, trying round after round of samples, you find that layer-to-layer registration of multilayer boards, thermal expansion of materials and moisture absorption are what keep you awake all night.

For a period, our team tried to find a high Frequency PCB manufacturer for prototyping, because we wanted to shorten the response time and push signal conditioning one level forward. We found that after the high-frequency material was selected, combined with the lamination of the multilayer board under the optical window, the deformation under thermal shock was simply unbearable to look at. That micron-level deformation might be acceptable for RF traces, but for an infrared optical path it is a disaster. Later we switched to another high Frequency PCB supplier specializing in hybrid lamination of Rogers and FR4. The boards it made had beautiful impedance control, but as soon as the optical window area got slightly damp, the sensitivity drifted unbelievably. There is something quite counterintuitive involved here: it is not that the more “advanced” the board material, the better it suits optical applications. Many high speed PCB manufacturers are used to the process of high-speed boards with more than 10 layers, which is over-design when applied to rain sensors, and a complicated stackup brings higher residual interlayer stress.

I have always felt that the factories that can make a good Automotive Rain Sensor PCB are often those that have dug deep into thick copper and reliability. We later dealt many times with two Heavy copper PCB suppliers, not simply because they can do thick copper, but because such factories are very sensitive to the internal stress distribution of multilayer boards after lamination. They know how to adjust the heating curve during feed and discharge stages, and are willing to spend time doing cross-section analysis to look at microvias and copper thickness uniformity. One Heavy copper PCB manufacturer even transplanted experience from automotive headlight control boards and did selective thick copper on local areas of the board to balance heat distribution. This idea suddenly pushed our failure rate in temperature cycling tests down a large step.

The hardest part of the manufacturing process is that you are satisfying electromagnetic performance, thermal performance and optical stability at the same time, and these three often work against one another. To reduce the parasitic capacitance at the bottom of the optical window, you have to pull the inner copper back farther, but once it is pulled back, local heat conduction is poor, and uneven heating and cooling during reflow soldering leads to board warpage in the optical window area exceeding 50 microns. Some well-managed factories intervene at the CAM stage and run thermal balance simulations for these areas. This is not high technology, but there are really few manufacturers willing to do this kind of simulation for small and medium batches of sensor boards. Very often, what you get are resources transferred from phone HDI product lines. They are precise indeed, but those HDI PCB manufacturers’ understanding of automotive-grade long-term reliability testing is another matter altogether.

The optical part ultimately has to return to the board for verification. When we did optical calibration, as soon as an outlier appeared, we had to go back and check this batch’s DFA report to see whether the solder mask opening size was at the lower limit, or whether the flatness of the surface finish layer exceeded the upper control limit. Gold plating that is too thick is not acceptable, and too thin carries a reliability risk. These details may be treated as fine-tuning items in ordinary multilayer board production, but here they become yield killers. So my thinking now is: do not chase the most advanced optical integration concepts from the start. First understand manufacturing stability thoroughly, find the right cooperating partner, and manage your board as an optical component, and the whole thing goes much more smoothly.