Our Visual Camera (the Basler ace acA5472-5GC) interfaces to our desgined-in-house PCB (more detail to follow in a later post) and to our payload computer (the BeagleBone Black Enhanced Industrial). It takes power via a Harwin Gecko connector from the PCB, and the data interface is Gigabit Ethernet on the payload computer.
Out of the box, the camera uses an RJ45 jack, and a 6 pin I/O connector. Now obviously, these aren’t space rated – so we wanted a more robust interface, with the added issue that the RJ45 connector on the payload computer didn’t fit in the stack. This lead to step one – desoldering both of these connectors.
We did this in stages, testing first the I/O cable – which was changed to a Harwin Gecko – was capable of delivering power. Only the 12V and Ground pins were connected to the camera, to avoid any potential issues of floating values from the other 4 connectors on the PCB (in previous testing, a camera was lost, we suspect due to this).
We then desoldered the RJ45 connector, instead soldering a Cat5 Ethernet cable onto the board of the camera (taking care to match the colours to those we identified). We plugged this into the payload computer and ran a simple connection test on the camera – success! But, this only solved half the issue – the Beaglebone still had the bulky RJ45 socket on it and we had to remove this for integration of the PCB.
So we removed the port, and soldered the Ethernet cable into the correct pins – we turned it on, and then the BeagleBone we were using broke (note: at this time in testing we were using a ‘normal’ BeagleBone Black, which only supports 10/100 Ethernet, the principle of operation is the same however, just with fewer lines needed). Following some reading, we identified what we thought was the cause. The RJ45 connector uses several 1:1 transformers for isolation. On the Basler side, these were separate from the port as a discrete component. On the BeagleBone however, these are integrated as a part of the port.
To test whether this was indeed the cause, we salvaged the integrated magnetics from an old port, connected the cable to it, and connected it to the board. This connected, and so we slowly began to strip away the plastic casing – testing to make sure the camera connected as we went. Every test was successful, until one of the thin wires in the magnetics broke. We started the process again with another integrated magnetics from yet another port, this time being especially careful around all the wires (indeed, the one we chose to work on next had a much better coating of the black plastic glue seen above). In addition to this, we wanted the camera to be able to be unplugged from the board, so put a D-sub 9 into the cable, connecting the colours to each other on either side of this. There was one final issue before we had a working solution – the integrated magnetics are as tall as the port – so had to be bent flush with the board. We carefully managed this, and with it had a working solution.
This solution, while viable, had 2 problems – the D-sub is heavy (and would have to be epoxied to the structure somewhere), and the integrated magnetics from the stripped-down port were very fragile – not ideal for a space application. Now we knew it would work, our electronics lead set about designing a PCB to sit on the top of the BeagleBone, under the main PCB, to allow integration of a discrete magnetics component.
Simultaneous to this, and while waiting for this to arrive, we modified the Basler further, adding a micro D-sub 9 to the back of the housing, allowing a removable connection. This tested successfully with the fully stripped-down solution from before, so we were confident that it would be the same on the Ethernet adapter when it arrived.
The board arrived, was integrated into the fix, and the camera connected – and indeed allowed images to be returned. We also swapped out the Ethernet franken-cable for Alpha Wire vacuum-compliant wire in twisted pairs.