From: Dave H. <dhy...@gm...> - 2008-06-04 19:19:23
|
Hi theorifce, > I am also getting this problem very frequently and have yet to figure any > sort of solution. I have found that even rapid subsequent calls to > 'i2c-load 0x0b Info' can cripple the i2c. Unloading and Reloading the i2c > modules sometimes works, but often I need to do a hard restart. > > I'm attaching an image with the output of a scope on the i2c SDA and SCLK > lines for a transaction between the Verdex and the Robostix. The clock > looks majorly distorted with very slow rise and fall times although this > maybe because I am scoping from the pinouts on the robostix board rather > than from wherever the i2c lines are on the Robostix. I've scoped i2c > transactions between the verdex and other devices and the clock looks > nominal. Hmmm. If I remember correctly, the i2c spec uses fairly low value pullup resistors on the SDA and SCL lines, something like 1.5k. The robostix only has 10k on the 3.3v side and on the 5v side of the voltage converter. It would be interesting to parallel these lines up with a lower value resistor to see if that makes a significant improvement. >From looking at your scope, I'm assuming that the top one is SCL and the bottom one is SDA. SDA looks pretty good, but your SCL looks really crappy. Almost like it's picking up alot of capacitance from somewhere. If you swap probes, does the low rise/fall follow the probe or the signal? It would be interesting to see if the 3.3v signals coming from the gumstix look like that too (probably need a tweener to monitor these). -- Dave Hylands Vancouver, BC, Canada http://www.DaveHylands.com/ |