This has to be 50V for the input supply. This particular TDK is automotive rated, good for vibrations and temp.
2.2uF in the highest voltage you can easily get, which is 16V. Again, automotive grade.
Surprisingly, this is the highest voltage and current rated diode in an 0603, which is uncomfortably close to it's actual rating. On the other hand, this is used during debugging only, so we don't care.
L1, L4 --- Ethernet common mode choke, 90 ohms, 400 mA -- TDK ACM2012-900-2P-T002 (DK 445-2207-1-ND)
Very, very small ethernet magnetics. Pioneered for the rocketnet hub.
TODO: Not sure the specs here: 0402 6.3 nH.... and? Random guess: TDK MLG1005S6N2ST000 (DK 445-3054-1-ND)
For filtering out the PHY's internal 3 V supply. Chosen based off of the old Olimex PHY chip using a similar part. Why 22 uF? We'll never know, but now we use them. :/
TODO: Not sure the specs here: 0402 6.3 nH.... and? Random guess: TDK MLG1005S9N1JT000 (DK 445-3058-1-ND)
We need to balance insertion loss with off-GPS-band attenuation. There are lots of very low insertion loss SAW filters, but they have weeny attenuation, especially at 2.4 GHz where we'll be nuking everything with our DxWiFi system (1W 802.11b @ 2.422 GHz using BPSK). So we chose this one, with a higher 1.2 dB max insertion loss, but 40 dB attention around 2.4 GHz. L8 between LNAOUT and MIXIN on the MAX2769: certainly the right place for a SAW filter. L6 before any LNA: Not so sure! However, we've experienced swamping out LNAs with our 2.4 GHz radiation, so it's probably waise to add a SAW filter even before our first LNA.
We could go for a u.fl but they're obnoxiously small and don't lock, and an end-fire SMA doesn't take much vertical height.
This is a double use header. First and foremost, it's a cable that connects to our COTS GPS board (the Hemisphere Crescent module), grabs the binary data it's spewing (as programmed in it's flash settings), packetizing it, and sending it off via Ethernet to the FC. This essentially makes the STM32 a UART to Ethernet converter, which is fine, that should be easy and fast with DMA and whatnot. But, we also have to power the Crescent board with 3.3V So I put the switching 3.3V supply on pin 2 of the 6 pin header.
The second use is that this just happens to be the correct pinout for a 3.3V FTDI cable. 3.3V FROM the FTDI is sent to a test point and will NOT run the board (that would be weird), and the 3.3V from the SPS goes to the RTS# line, which is an input for the FTDI. Thus, nothing should blow up no matter what is powered when.
This allows the ChibiOS monitor to run on the full speed USB peripheral. Bonus points: it also should power the whole board because of D2 and the fact that LMZ is pretty low drop out. So yay, you can program and debug without a high voltage power supply, which is exactly what we want. Note that we've backed it away from the edge of the board, which isnt how it's supposed to be. We just don't want things randomly sticking out the side of the board, so we're hoping that's not bad. Drawings suggest it's fine, but then why so cantilevered in the first place?
Power LED.
Just standard debugging LEDs, nothing special. Also used for Link/activity light on Ethernet PHY.
Standard PSAS RGB LED just because we have to.
Standard, in-house P-channel power switch
Smallest pulse transformers, evar. Pioneered on the rocketnet hub.
The whole reason we made this board!
- ANTBIAS isn't connected to anything, because we're powering our onboard LNA right from the analog supply (no reason to go through the Max, is there?). The EVM bypasses this without anything connected, which scares the bejezus out of us, so we bypass it anyway.
- ANTFLAG and LD are probably useless, but we read them in the STM32 anyway.
- !SHDN and !IDLE are both pinned out as well, although we'll probably only ever use !SHDN as a low power reset state.
Same as in the rocketnet hub, but we're using a different symbol. It's the coolest uC on the block.
Best PHY chip ever. We'll soon learn why it sucks. All configuration options are their default values, because again, best chip ever. Takes in 25 MHz clock from STM32 (MCO1) and PLLs it to 50 MHz which is uses for the REF_CLK. Has it's own LDOs. Does the dishes, too.
Main clock for the 2769. Dude, check out the ppm. Shamelessly copied from the Piksi. The drive capacitor is a bit unclear; the output drive has a 10kohm/10pF rating, but that doesn't seem to stop the EVM from using 10nF, the Piksi from using 47 nF, and the interwebz from generally suggesting 1 nF. Datasheet says 10 nF, so be it. Bypass wise, the EVM has 10uF + 100pF, Piksi has 47 nF. We'll give it a standard 100n + 100p, and throw a 1 uF in the larger package just for fun since we have room. Also, the 2769 apparently doesn't need a bias network for the REF/XTAL input, so none given.
This part we shamelessly borrowed from the Piksi. It seems like a very nice part: noise figure ~ 0.69 dB, and has a gain of 15 dB. It takes Vdd = 1.26 V, so we need to put in a resistor for bias; at 2.85 V (hey, it has that in the datasheet, so other people do this?) it's 470 ohms. The cap stays at 10 nF (although that seems weeny for a bypass cap) as per the datasheet. They claim they want the TDK GLQ1005 type of inductor for the 9.1 and 6.2 nH inductors.
Drop the 3.3V digital down to 2.85 V in order to filter out noise. We'll only get about - 30 dB on the switching noise because it's way up at 400 KHz, but otherwise it's pretty good.
U8 -- SPS 20V input 1A buck monolithic voltage regulator -- TI LMZ12001TZ-ADJ/NOPB (DK LMZ12001TZ-ADJ/NOPBCT-ND)
Laziest power designers in the world. Takes 4.5 V (USB debugging cable) to 20 V (rocket shore power), switches and inductors built in, Should Just Work(tm). Outputs 3.3V for digital stuff. Leave EN pin floating so it's always on (and we can use 4.5V USB power). According to the datasheet, at 20Vin and 3.3V out, choose Ron = 61.9k. This gives a switching frequency of 410.1 KHz... a little high for the linear supplies, but fine for everything else. Usually we like higher frequencies, but not in the case of GPS radios. To double check, ton = 547 ns @ 14.7V and 400 ns @ 20 V, well above the 150 ns limit. Css is set to a simple 100n, which gives us a 10 ms soft start time (2 ms is the recommended minimum). Feedback resistors are taken right from the datasheet for 3.3V.
25 MHz crystal to run the STM32. We choose 25 MHz in order to generate the 25 MHz for the Ethernet PHY chip via the MCO1 clock output pin. Note that our PHY then generates the 50 MHz necessary for the RMII interface between the STM32 and the Ethernet PHY. Since the dividers in the STM32 bring the 25 MHz down to 1 MHz internally, we can then PLL up to important other frequencies, like the 48 MHz USB peripheral clock, and the 168 MHz core clock. We used this crystal on the rocketnet hub, so might as well re-use it.