This is a couple of minutes traffic from a LoRa node building meetup.
The top graph is the current spectrum around 868 MHz, the lower graph is the history. The fuzzy red rectangles are spread spectrum LoRa transmissions.
Shopping on internet shows two different PoE solutions, the cheap (and slightly sloppy and maybe dangerous) way is to use unused pairs in the Cat 5 cable and use a passive injector and splitter. The problem with that solution is the dissipation and loss in the network cable at higher powers. The nicer and safer solution is the official IEEE 802.3af solution that I wanted, and I wanted this to work with thin 2-pair Cat 5 cable. Regular Cat5 cable has 4 pairs. I’ll only be using the IEEE 802.3af standard here.
I bought a (second hand and very loud) Cisco Home Office PoE switch to try this all out.
The insides seems legit,an actual PoE controller chip, isolating transformer and Ethernet transformer. The parts used look like the application note with one difference : the application note describes a 10/100/100 MB/s solution with 4 wire pairs, this splitter only supports 2 pairs.
The track clearance is not great though. And there is something going on with the connections between two cable pairs but this configuration seems to be used in other setups as well ( figure 4, figure 1.1 ).
One sore point seems to remain : Apple HomeKit certification apparently not only applies to the Hue bridge but also to the lamps connected to the Hue Bridge :
Before the 1.11 software update, a bug in the Philips Hue system allowed some non-Apple HomeKit certified lights to work with Apple HomeKit. Our 1.11 software update removed the bug with the result that non-Apple HomeKit certified lights no longer work with Apple HomeKit. This remains the case.
The first batch was a new run of the LivingColors Arduino shield.
The second batch was a breakout board for the NiceRF SX1276 LoRa module. Here i experimented a little with contour routing.
The third batch was an adapter board to use no-name CC2500 modules in boards designed for the Quasar QFM-TRX1-24G. The boards are small (about 20 mm x 25 mm) and the minimal size for dirtypcbs PCB’s is 100 mm x 100 mm. Here I experimented with breakout panels, putting 4 PCB’s in one design.
The end result worked out nicely, two things I’d do better now are :
Position the “mouse bites” differently so the rough edges won’t protrude, see this post.
Copy and fixate the silkscreen text to a new layer before copying and rotating the sub-PCB’s. If you look closely at the PCB’s you’ll see that the text didn’t rotate with the rotated copies of the traces. That’s what you want for a big PCB but not for a breakout PCB. There is a ULP for that and a nice tutorial.