XMOS StartKit part 2: Hands On 

+Nabeel Fattah Hello there! I'm not sure I can give you good hints because I'm far from being an expert. I suggest to use the XMOS forums, when I had questions the folks there were very nice and gave good support. I am of course happy to give you the pointers I can: First, for UART you should check applications notes AN00158 and XM006381. Those should have example codes to fulfil task 1. For the GPIO part, you can check XM006034 and AN00166 but probably my videos suffice to fulfil this part. I would try to get those two running individually first. The tricky part is probably doing the double-buffer read/write access thing. I once considered doing part 3 of this series with exactly that, but it never happened. So I don't have first-hands experience in doing that, and since I'm not developing for XMOS right now I would probably give you bad advice. There is, however, good news: They have an application note that describes exactly the double buffer scenario that you need. It is described in note AN10011 with example code. I hope this gets you started! Best regards -- Matt

Hello pt2091. Sorry it's been a while since I did this, I don't remember dividing by 32. But I'm probably doing it somewhere... can you show me where in the code or in the video that is?

MrCircuitMatt 8:57 I just realized why now though haha

Ah! Thanks. To transmit a HIGH or LOW bit, we have to generate a waveform that has a duration of 1.25 microseconds according to the specification. Depending on whether it's HIGH or LOW, it's 600 ns of high voltage followed by 650 ns of low voltage (HIGH), or 250 ns of high voltage followed by 1000 ns of low voltage. This is somewhat confusing because the HIGH/LOW bits to be transferred are encoded with waveforms that toggle between high and low voltages themselves. Based on this, the smallest "pulse" we have to produce has a duration of 250 ns; it's the high pulse at the beginning of the waveform for LOW. By dividing the 100 MHz reference clock by 32, I end up with a resolution of 320 ns. This is not the 250 ns resolution that I need to generate that 250 ns pulse, but it's close enough. So I can approximate the waveforms with segments of 320 ns each. Does that make sense?