I have a couple more questions about tcpborphserver3. 1) Is there any way to deprogram the ROACH2 FPGA (e.g. to save power) via KATCP? I couldn't find any, so I modified progdev to deprogram the FPGA if no filename is passed. See:
https://github.com/david-macmahon/katcp_devel 2) Once the FPGA is programmed, the raw.current.1v5 sensor shows more than 9A of current (confirmed by measuring the voltage across the sense resistor) and it is marked as "error": #sensor-value 1376626889641 1 raw.current.1v5 error 9652 But looking at the sysfs files for this sensor, it looks like the value is less that the critical threshold: # head /sys/bus/i2c/devices/0-0051/in3_{input,crit} ==> /sys/bus/i2c/devices/0-0051/in3_input <== 819 ==> /sys/bus/i2c/devices/0-0051/in3_crit <== 5538 It seems like the scaled value (9652) is being compared with the UNscaled critical threshold (5538) and since 9652 > 5538 it is shown as an error condition. If that's true, it seems like an easy enough bug to fix, but where do the critical threshold values come from? Are they written into the monitoring chip's flash during the ROACH2 assembly process? Scaling the presumably unscaled critical threshold value of 5538 by a factor of ~11.68 leads to a scaled critical threshold value of over 60A which is more than twice the max output current of the DC-DC converter used for the 1.5 V supply, so 5538 seems like a not so useful (unscaled) value. It would be nice to make these limits be useful. Thanks, Dave
