I have a small circuit that uses an AS3935 lightning detector, connected via I2C with 3.6V/4k7 pull-ups to an ATmega8. I figured it would be easy to replace that "other" MCU with an MSP430G2553, but immediately ran into an issue with I2C connectivity. While the 'G2553 USCI/I2C is working correctly with a DS1307, attempts to access the AS3935 would fail START with a NACK. I probed the circuit (built on a solderless breadboard with associated stray capacitances) and it started working; I soon realized the small additional capacitance on the SDA line would make it work; removed the probes and found 10p on SDA to ground would make it work. The 'G2553 was placed on a solderless breadboard; later experiment contained the 'G2553 on a LaunchPad. In both cases, the 'G2553 Vcc/Gnd should be adequately bypassed.
Driver code is polled and available upon request, but this issue is entirely with the START condition generation in hardware.
So, of course, I took a picture - upper trace is SDA, lower trace is SCL - note the sharp spike in SDA immediately following the START condition as SCL drops:
This spike is not the result of open-drain I/O - it has a linear rise time on the order of ~60nS to ~1.5V, representing a logic '1'. It's far too early to be clocked in as the MSB of the address/RW byte, but I suspect it's causing the AS3935 to disregard the START condition (which seems like a violation of I2C timing/filtering, sure). I also note smaller SDA glitches on later falling SCL, but this should be benign.
Is this a known behavior of the MSP430Gx2xx USCI in I2C mode? Is there some work-around to remove this glitch, other than putting a small capacitor on the SDA line? The small capacitor surely resolves this by delaying the leading edge of the glitch just a little bit (small number of nS).
Cheers,
Dana
