Skip to content
Snippets Groups Projects
README.md 3.02 KiB
Newer Older
  • Learn to ignore specific revisions
  • Jake Read's avatar
    Jake Read committed
    # Machine Kit Router
    
    
    Jake Read's avatar
    Jake Read committed
    This is a message passing / compute center for machines, robotics.
    
    Jake Read's avatar
    Jake Read committed
    
    
    Jake Read's avatar
    Jake Read committed
    ![board](/images/fab-front-rpi.jpg)
    
    ![board](/images/fab-front.jpg)
    
    ![board](/images/fab-back.jpg)
    
    Jake Read's avatar
    Jake Read committed
    
    
    Jake Read's avatar
    Jake Read committed
    ## Status
    It works! If you have one, it should forward packets no problem. Let me know if you find weird errors...
    
    ## Next Steps
    Did a redesign for this using the ATSAMD51 and USB-Serial on-chip conversion so that networks can run at 1-3MBaud, not 0.1MBaud. That board is out to the house.
    
    
    Jake Read's avatar
    Jake Read committed
    # New Board
    
    - is out to fab
    
    ![routed](/images/rpi-routed.png)
    
    ![schematic](/images/rpi-schematic.png)
    
    
    Jake Read's avatar
    Jake Read committed
    # Board Design
    
    I use the ATSAMS70 micro, which is very fast but still simple enough. 300Mhz and the I/O we need.
    
    That said, setting up the power schematic is messier than I would like. There's an internal regulator on the chip that is externally connected to a few 'vddcore' input pins (5), each of which needs a bypass cap. In addition, there are 5 vddio pins that supply voltage to the i/o lines. These also get bypass caps. This is 10 bypass caps. That is a lot to place right off the bat. In addition, there are some pesky vddpll lines that ask for filtered inputs from either the main line or the regulator, meaning I have now inductors and more bypass caps. Oh my. Here is the schematic checklist diagram:
    
    ![schematic-power](https://gitlab.cba.mit.edu/jakeread/mkrouter/raw/master/images/schematic-power.png)
    
    
    Jake Read's avatar
    Jake Read committed
    To do nice ADC things I want a well stabilized vrefp and vrefn, I will neglect doing this with this chip.
    
    Jake Read's avatar
    Jake Read committed
    
    
    Jake Read's avatar
    Jake Read committed
    Setting up the USB successfully is nontrivial, as it can be a properly high-speed differential line. I have bricked a few of these chips trying to implement the usb operation, so I'm going to skip most of this as well... Maybe breaking out the USBDM and USBDP lines just in case I get bored later, but I kind of doubt that.
    
    Jake Read's avatar
    Jake Read committed
    
    So, here it is. I didn't ask for this complexity. *grumbling*
    
    ![schematic-power](https://gitlab.cba.mit.edu/jakeread/mkrouter/raw/master/images/schematic-power-eagle.png)
    
    
    Jake Read's avatar
    Jake Read committed
    Less talk more routing ... one note - I am putting some ground plane between all of the RX and TX lines, which I hope to be driving around-and-above 3MHz. So, I want them to not be coupling into eachother. Leaving some ground plane between each trace is a bit like shielding a wire... Here you can see two of the ports taking 'the dive' to the ground plane, where they have lots of room, and then being routed with some space between. I should update other boards using similar rules as well - I'm re-doing my BLDC driver and have started a stepper board that I'll be bringing up at the same time, so I can do that. Nice.
    
    Jake Read's avatar
    Jake Read committed
    
    ![routing-gnd](https://gitlab.cba.mit.edu/jakeread/mkrouter/raw/master/images/routing-ground-uart.png)
    
    OK, here it is:
    
    ![schematic](https://gitlab.cba.mit.edu/jakeread/mkrouter/raw/master/images/schematic.png)
    
    
    Jake Read's avatar
    Jake Read committed
    ![board](https://gitlab.cba.mit.edu/jakeread/mkrouter/raw/master/images/board.png)
    
    Jake Read's avatar
    Jake Read committed
    
    # Notes
    
    PP7 is a miss, check others as well.
    
    Silk kind of messy.
    
    for all, make QFN64 footprint rock: the QFN for those tmc chips are awesome, follow suit