Yellow glue, a problem waiting to strike your vintage electronics

Many in the retro-computer community are aware of the dangers of Varta rechargeable batteries found on many older computer mainboards. These permanently soldered-on components have a tendency to leak as they age and wreak havoc as the acidic fluid corrodes and dissolves the copper traces on the board. As I learned today, there’s another hidden foe that may be lurking in some of your devices, particularly power supplies. This evil substance is a type of yellowish glue.

Yellow glue isn’t silastic, the typically white non-conductive silicone-based adhesive found in some higher quality applications. It seems to be possibly some type of neoprene rubber adhesive that is cheap and sets quickly making it a popular choice for low cost manufacturing of electronics. It’s typically applied to glue components in place for assembly, insulation and protection from vibration. The down side is that this material ages poorly, especially in hot environments and becomes hygroscopic. Once this happens, the material becomes partially conductive as it absorbs water which can lead to some pretty dramatic failures, especially if the material is on a mains or high voltage part of a circuit board.

I discovered this issue on some power supply boards I was troubleshooting. I had been going through the process of reworking them and was checking all of the components prior to powering up. One of the probes to my multimeter slipped off the diode I was measuring, but to my surprise I was still getting a reading. I jammed both probes into the browning mess of old glue and what do you know, it was conductive! It measured over a mega-ohm, but it was consistent. I started checking the other spots of glue and any that looked darker and crusty were conductive.

The other downside of this material is that it’s very difficult to remove without damaging components. In my application, I was able to use some stiff plastic spudger tools to scrape it away. It was very labor intensive and didn’t entirely remove it. I may try something like acetone on individual components like inductive coils, (to attempt to dissolve and remove it) but unsure if this would be safe to just douse the board with.

Chamberlain Smart Garage Control teardown

With the recent Chamberlain API lockdown, I thought it might be interesting to tear down one of their controllers to see what makes it tick. Thankfully I happen to have a spare unit sitting new in the box. Why do I have a spare unit you might ask? The standard package comes with what Chamberlain calls a WiFi hub and a single door sensor. I have 2 garage doors, each with openers. At the time a spare door sensor was most of the cost of the whole package, so I just bought the whole thing again.

The WiFi hub is incredibly simple to disassemble. There are just 2 screws that hold the faceplate to the frame. Once you remove the faceplate, a single snap lever retains the main PCB. Once I had the main PCB separated from the frame, it was easy to pick out the core components.

The system is based on a PIC18 microcontroller. In this case the Microchip PIC 18F67J11-1/PT. The PIC18 is a 16-bit RISC CPU compatible with the earlier PIC16. It has 128KB flash onboard and 4KB of RAM. While it’s no powerhouse, the CPU does offer 4 interrupt lines, 5 PWM, Parallel and Serial ports with SPI and i2C support.

Linked to the PIC18’s i2C bus is a 24C16K SPI Flash ROM. I had a compatible harness for this and dumped the contents which were mostly empty. I’m assuming this rom is used for configuration data, possibly to store the opener codes programmed in during initial setup.

The remaining modules of interest were an Si4432 433mhz transceiver chip (left) and an Fn-Link 6220N-IS 2.4Ghz WiFi 5 module. (right) These enable comms with your network and most garage door openers.

So, what can we do with this knowledge? Well, given the CPU has a built-in flash, we could attempt to dump the contents of that storage and see if there’s anything useful we can decipher from it. There are test points on the board that appear to be connected directly to the CPU for programming. I’m going to attempt to verify this against the datasheet for the PIC18 and see if that’s the case. To be continued…