We are developing an app that communicates with a custom LCD display via IIC (TWI in Nordic terms) so it needs to send a lot of data on the IIC bus. We have developed an interrupt driven power aware TWI driver. Since this driver needs to call the softdevice to access PPI and set up IRQs for its timeout timer (TIMER1), its IRQs are low priority. Other parts of the app use RTC1 for other timing tasks as well as the ADC (3 channels) all interrupt driven. In any case, with the softdevice enabled and the BLE stack initialized, but nothing BLE happening, all works correctly. Our app has run and been tested running this way for 36 hours without a problem. When I enable nonconnectable advertising, repeating every 2 seconds indefinitely it will crash sometimes in as little as a minute, other times it may take 15 or 20 minutes for the crash to occur. The crash always kills the TWI periphial. Watching the SDA and SCL lines with a storage scope, I see it actually stop clocking part way through transmission of the the address byte some times. Other times there will be a short (less than a microsecond) glitch on both lines just after the crash happens. My timeout timer correctly detects failure of the bus, however it seems that nothing I can do will recover without a cold reset. Once the crash has occurred, if I put a breakpoint in the routine that detects the bus failure, any attempt to trace the code from that point forward makes no sense. It's as if the JTAG interface has gotten messed up by whatever happened. I can place breakpoints wherever I need and trace without problems before the crash occurs. We have tested with first an eval kit attached to a custom pcb with our display mounted with jumper cables and also with our custom PCB with all hardware in place. We have tested with S110 5.2.0, 5.2.1 and the 6.0.0.1 alpha stack. One more item, have tried 3 different JLinks, same results with all. I am wondering if possibly we are seeing something similar to the post titled Spurious S110 crash noted in the link here. Or, is this a hardware issue, hopefully resolved in the next rev silicon we are all waiting for.
