Figure 1: Differences in logged times of user login events, when running from a lab power supply.As witnessed, in approximately 1 in 10 reported boots, the OS does not recognize the LORD Microstrain AHRS as a serial port on its USB interface. These tests are performed on a stack of a Congatec MA5 with a Connectech CCG020 carrier.
In the process of getting set up for these tests, the hard drive was corrupted in such a way that the system was not recoverable and was rebuilt, using a new 1TB SSD. It was discovered that the DC/DC Converter was undersized at 15-W and certain peripherals were increasing the consumption past the limit of the converter.
A script was constructed that simply logs all connected serial ports to a file, timestamps that file, and reboots the computer after a 2-minute reboot window. Effectively 1 test runs every 3 minutes, as their is 1-minute of boot/login. This was assigned using the task scheduler, and give 12-hours to run. The CPU system was powered by an external power supply set to 12-VDC and only connected by network, and with only the AHRS and NI-cDAQ attached via USB.
This test ran from 5PM to 5AM, and cycled 241 times, and all times the compass was appropriately mounted via the USB interface.
Figure 1: Differences in logged times of user login events, when running from a lab power supply.
Same test as above, for 12 more hours but powered through the PCB with a muRata UEI15-120-Q12P-C. Supply voltage was set for 15.15-VDC and a maximum current of 3.3-A. This test was run for 7 hours before the system was checked, and new tests begun in the same way at 5PM, albeit the input voltage has been lowered to 12V.
It was observed that the system, operated in this way, which is very close to the in-vehicle configuration, may be marginally compatible with the upstreams DC/DC converter. Current spikes grow near 1.1-A on the 15-V primary, in excess of the 15-W rating of the converter. Simply logging into the system from the Remote Desktop can increase power requirements enough to brown out the cpu depending if this login is attempted while other systems are still being brought online by the OS. However, the system boots fine and I have not witnessed any drops of the serial port.
After final inspection of the boot log, the serial port came up during each logged boot, 426 times in 23-hours. Dropping the voltage mid-way through the test had no effect, but more power dips were observed via an oscilloscope. I have not been able to reproduce the failure in the newly built system to presents. I will say that in this configuration, I have been unable to remote desktop into the machine, as this causes a brownout of the DCDC, so its possible immediate remote login attempts that may happen during boot cycles may be interrupting the enumeration of the USB port. This is extremely hard to prove.
Figure 2: Differences in logged times of user login events, when powered by the original Murata DC/DC converter.
This test changed the after-login batch script so that after 2-minutes, the system would instruct itself to shutdown instead of restarting. In between the power supply and the Murata DC/DC converter, I added a relay whose coil was programmed by an external PC. That coils was set to open and close at a duration longer than the boot/login/shutdown sequence that would happen on the embedded PC (~3-minutes).
This test also had 100% success for each successful boot, but I believe there were a few events where the system failed to boot. This will be studied further, but overall the failure of the USB device to enumerate was not reproducable under this setup as well. It is also possible that for a few events the PC-controlled relay failed to complete it's cycle, as this part is new, from Ali-Baba, and therefore not fully trusted.
Figure 3: Differences in logged times of user login events after power being completely removed and applied, through the Murata DC/DC converter.
Overcurrent shutdown of the DC/DC Converter may likely be causing some of the problems with the system at present. Being so near this margin is also not good for