Let’s get this out of the way first…
Turns out that some units were accidentally shipped with a 3.3V regulator instead of a 5V unit:
The effect is not dramatic, in the sense that the RBBB will work just fine with either regulator. But as soon as you start hooking up stuff, you may get weird behavior. The worst case would be to hook up stuff with its own 5V power supply, since that means the attached device could potentially put a higher voltage on the IO pins than the 3.3V supply voltage of the ATmega.
So how do we fix this? Well, if you’ve got an RBBB from JeeLabs in the past months (probably anywhere from just before the summer break until now), please check the regulator voltage. Let me know if it’s the wrong one and I’ll send replacement 5V regulators.
(Note that the RBBB’s L4931 and the JeeNode’s MCP1702 regulators have different pinouts!)
As for the current batch of RBBB’s I still have here, the store gets a little bit more complicated: these are all wrapped and sealed by Modern Device, and the bags are too thick to be able to read that lettering on the regulator. I’ve opened up almost a dozen bags and found a few 3.3V units in the older batch, and none in the newer one. There’s not much more I can do than replace those I know about, and repeat that offer to replace any regulators you come across yourself.
FWIW, I suspect that this problem is gone by now.
But there’s a second member in the Murphy family, still dropping by occasionally, it seems…
As you may remember, I switched all SMD-based ATmega328′s over to use OptiBoot a couple of months ago. As it turns out, the standard OptiBoot loader included with the Arduino IDE 0022 isn’t reliable under Windows – my hunch is that there’s some timing problem with the combination OptiBoot <=> FTDI USB interface. It appears to be solved by changing the watchdog timer from 500ms to 1s, see the discussions here and here.
There’s now a new Arduino IDE 1.0beta2 out, and I’m getting some signals that we’re not alone with these OptiBoot troubles, so for now I’ll just stick to that modified 1s OptiBoot code which I’ve been using for several months now, and I’ll hold off switching to OptiBoot on the DIP version of the ATmega328, as used in JeeNode kits. Failing bootstraps are no fun if you don’t have an ISP programmer and no time to try out various tricks to emulate one.
In a nutshell: all SMD chips from JeeLabs use the 1s OptiBoot (which runs at 115,200 baud), while all the DIP chips get pre-loaded with the Duemilanove bootstrap (which runs at 57,600 baud).
I’ll revisit this when the dust settles.