Originally having given up on my Raspberry Pi from RS, a very helpful post from Dilligaf at the Raspberry Pi Foundation’s Forums gave me this advice:
If you’ve given up on the warranty you could always try baking it in the oven at 400 for 10 minutes, or heat it with a heat gun. This may have a detrimental affect on the polyfuse but you could always bypass it or power from the gpio. Baking/heating it should reflow the solder if it’s a bad solder joint. You’ve really got nothing to lose if returning it isn’t an option.
Indeed, he was right. But these things take a little thinking time for me – will things fall off the board? What temperatures/times would I need? Ramp speeds? What might happen to all the connectors and plastics? Would the fan in my oven blow components off? Will there be many fumes which could poison us all? Luckily one problem was already solved for me, I voided my warranty by taking off their defective polyfuse, so that won’t pose me future issues.
From the title, you can already guess what happened. Yes, I baked a Rapberry Pi.
If you have a problematic board, feel free to try this or a variant of this, but at your own risk of course. (I felt it was probably good to get a feel of it, because in my near future for my PhD, I will need to “ghetto” reflow a few PCBs.)
A domestic home oven was used reasoning that fumes and outgassing should be low given it has already been reflowed at least once, it reaches sufficient temperature, and the fan shouldn’t blow the components off if I place the board near the door – away from the fan. A K-type thermocouple attached to a multimeter with thermocouple support for temperature readout was used to get an idea of the temperature – the thermocouple was placed on the aluminium foil . A baking tray was lined with a piece of foil and the board was placed upside down, resting on connectors and scrunched up balls of foil. The choice to place the board upside down was made so as to get the maximum clearance around all sides for the airflow to ensure even baking, and resting on connectors would avoid any possibility that components might get “soldered” onto the foil. I reasoned that, providing no bumps occur during the process, the soldered components should hopefully resist gravity by virtue of the surface tension of solder.
So, turn on the oven and start baking. But it’s not as simple. Underbake and there will be no reflow action, or poor reflow action. Overbake and your connectors will melt off, and the board may be permanently damaged. The technical term for this is reflow time-temperature profile and JEDEC (the guys behind the memory standards) have a standard reflow profile which most people in the industry use. Note that lead-free solders used in RoHS devices have a higher reflow temperature requirement, and have more stringent requirements on temperature and time – i.e. a smaller process window. Here’s a generic reflow profile – the peak temperature should be about 245-260 for lead free solder, and the sustained temperature within 5 degrees C of the peak should be about 5-30 seconds (depending on the sensitivity of the components). Note how that there’s a slow ramp-up in order to pre-heat the board and reduce thermal stresses.
But don’t let the onerous reflow profile requirements scare you off. Ultimately, I wasn’t going to build a controller and try to do it perfectly. Instead, I did as people do when cooking food – go with my intuition … and my thermocouple readout!
We’re bake bake baking away … I decided to level off at 100 degrees C for a while, then at around 160 degrees C, then it was charge right up to 245 degrees C and I held that for a minute. Note that due to thermal mass and also uneven heat distribution that while the thermocouple might have read 245 degrees C, the center of the device might not be at that temperature – hence why I held the temperature for twice as long as would really be recommended.
So we must begin rampdown, but not too quickly so that the solder can form a good joint. Power off the oven, leave the oven door ajar for a bit … then open it completely. Don’t disturb it at all until cool. So after baking, this is what it looked like (note thermocouple wire gone brown, and nothing’s fallen off).
For people that think this might be a joke – well, it’s not. You can tell because the serial number label has now shrivelled into a mess near the SD socket (in hindsight, should have taken that off), and the plastic surrounding the header pins has gone oily and somewhat goopy in appearance (but still that’s okay). Not pictured is the boxy audio jack which now has slightly concave walls instead of straight walls.
And in the end, this really DID solve the problem, suggesting they may have had soldering issues/reflow issues (reminiscent of the Xbox red-ring-of-death problems). I spent the rest of the day working on a small side project of mine using the Raspberry Pi that had been baked – no problems with X, apt-get, and configuring servers. No unexpected black-screens – finally! Thanks Dilligaf! (And no thanks to RS who sent me this half-baked Pi.)