Once in a while I sell one or two of the RGB LED RING boards. As not everybody feels comfortable with soldering SMD parts (and I’ve explicitly warned that it’s not that easy and that the RGB LEDs die easily) I do some ‘contract soldering’ from time to time as well.
Today I’ve killed 3 – THREE – of these friggin’ things, and I only have about 10 spares left. That’s a loss of more than 1€ of semiconductor. Time to get more spares it seems.
In this shot I managed to show what’s going belly up here. It’s always the green ones that die. As shown here, the bonding wire has been ripped off the left pad due to stress.
Looking at this image again a few hours later, it seems the whole pad has moved to the left opening a gap of maybe 0.1mm. What a heap of shit. Could the surface tension of the solder create such a big force and pull the pad out of the somewhat softened plastic body?
I’m trying to keep the heat down and be quick, but today was a very bad soldering day. I have a feeling they wanted to save time and material in the bonding process and the wires are a bit overstressed anyway. Maybe they also ‘tweaked’ the gold content a bit and it got too brittle.
If these LEDs are soldered in a reflow oven this problem might not occur at all, as the epoxy around the wires will soften a bit. Given the correct temperature-vs-time profile is used. Time to get an SMD toaster? Maybe. Any recommendations? Hotplate? And I don’t want to spend a fortune on it as well. Oh BTW, I hate things that suddenly just sublimate without any signs of warning! I’m looking for reasonably priced quality products here. Nothing fancy, but fully functional with a minimal amount of cursing.
Something completely different:
Today was also the very first time that I just couldn’t reliably upload code using a bootloader with the chip running on the internal 8MHz RC oscillator. Maybe this is due to the somewhat elevated temperatures we’re having here right now. It’s in the middle of the night right now and I’m still getting 26°C in my room. Or maybe this AVR’s factory calibration is just terribly wrong. Luckily I had previously assembled another one of the boards, so I just swapped them. The good one is sent away and I keep the rebellious one for further testing. I’ll put it into the freezer tomorrow and see what happens to it.
→ Update 1: 08.07.2010:
Well, deep freezing the chip didn’t change anything. I’m using the Moderndevice BUB FTDI adapter to upload code. It’s got the option to switch between 3.3V and 5V for the signal lines, the supply voltage stays at 5V at all times. Using 5V for the signal lines throws upload errors most of the time in the verification stage of the bootloader. Switching to 3.3V works most of the time. I’m puzzled. There should be enough decoupling capacitors on the board. This is the very first time I’ve encountered this problem, and I’ve built and spread more than 40 boards around the world by now. When I get access to a suitable storage oscilloscope I’ll have the board send a test message a look at the signals and measure the actual baud rate and see if it’s withing the error margin.
→ Update 2: 29.07.2010:
Well, today I plugged many of these ring boards (mostly old prototypes) onto the perfboard I made for my bike and both atmega chips got warmer than expected. I was using the “baby milk bottle” method (a.k.a. my lips) to sense the temperature. It was near to unpleasant, so I guess it must’ve been above 45°C. Then I did the same thing to the up-to-date board that showed uploading problems. Same old. Then I removed the jumper used to bypass the onboard voltage regulator and the thing was still running, which it shouldn’t have – or so I thought. Fast forward a few minutes. Hint: 2 input pins were now exposed to VCC and GND. And they have clamping diodes ;-)
Anyhow. The uploading problems started to get worse and worse. End of story: I’ve replaced the darn chip with a new one and everything works as expected.