Someone dropped their adhd and autism meds and needed some way to put them back into their respective containers

You could do the funniest thing. Read up on Canada’s role in global access to cryptography at a time when the USG had strict export controls on ciphers. OpenSSH for example was created in Canada by Theo of OpenBSD, which would have been near impossible to do in the US at the time. And we have exceptionally cheap hydro, and a lot of cold vacant space.

Crosspoint switches and CV DACs are successfully under control of the FPGA over Ethernet, with a MIDI controller mapped to issue peripheral CSR updates over etherbone. I made some Litex Migen modules for both the CH446Q switches and AD5628 DAC with state machines that handle the chip updates from register writes on a wishbone bus, so there’s no softcore required at all. A small Ratatui TUI app runs with background threads for the etherbone bridge and MIDI control input, and moves messages over channels to issue wishbone bus reads/writes on the FPGA.

A nice advantage of milling prototypes is when you make a mistake or want to make a change, you can quickly spin a new revision. The reason I put the DAC chip itself on its own little board that plugs into 1.27mm headers was because of the 0.5mm pitch and cost (~$20 each) knowing that I’ll likely want to reuse it on different variations or projects without having to desolder it with hot air every time, and made the module a kicad footprint and 3D model so I can just drop it into a different design any time. I spun this revision of the DAC adapter/bias board in about 45 mins to fix a mistake with flipped inverting/non-inverting inputs on the differencing amps, and with the bismuth solder it was a breeze to grab and reuse all the SOIC chips from the bad board.

Those are some funky looking SMT 0402s

DAC board is done. Tomorrow I’ll write up a little HDL peripheral that manages the protocol in hardware, so channel voltages can be written to memory mapped CSRs. A hardware FSM will manage SPI transactions into chip shadow registers, and a control register will strobe the LDAC pin to apply all 8 channels simultaneously.

Pulling out the 0.1%’s for all the dividers on these differencing amps

I get asked the question a lot about what bits I use to get such sharp traces on my CNC machined PCBs. Typically all of the bits I use are from PreciceBits precisebits.com/applications… The specific bit I use depends on the pitch, how much copper area clearing I'm doing, and how nice I want it to come out. Using less expensive bits tends to leave more burrs, but a quick pass with a fine sanding sponge cleans it up and saves tool life on the more expensive bits. In this case when I'm routing edge connectors to slide into these PCIe sockets, I would rather have nice crisp edges for the contact pads without any burrs to catch on the contacts, so I chose a better bit. For this board, it's straight off the machine with no post sanding/processing using an EM3E8-0071-15VC which is a 0.18mm zirconium nitride coated 3 flute 15 degree tapered stub. I do a 0.085mm DOC with 15k RPM speed, 680mm/min feed, and 250mm/min plunge. If I'm doing finer 0.5mm or closer coplanar trace to copper clearance, I go to a 0.13mm tapered stub. For the contour cuts including mount holes larger than my shank I use 1.59mm 3 flute soft media cutter, which are also ZrN coated. MM3I8-0625-031FC. Routing tiny slots like the mechanical through hole tabs on USB-C connectors I use 0.56mm 3 flue soft media cutter MM3I8-0220-009FC

Turns out the @Terasic_FPGA Atum A3 Nano PMOD connectors aren’t spaced to Digilent specs

Made a PCB for my experimental analog platform to connect a SPI DAC to the FPGA with 8 analog outputs into the crosspoint switch fabric. Just waiting on parts to arrive today before populating. A set of differencing opamps will take the 2.5V internal reference of the DAC, and subtract it from each output to have a -2.5V to 2.5V swing Intending this to be for slow DC coupled control voltage outputs that can be routed/switched by the crosspoints into other modules plugged into the board to set initial conditions, comparator thresholds, etc

Logistics never fails to amaze me. Twice in a row my DigiKey order is on a truck for delivery less than 24 hours after ordering to an island in Canada that’s only accessible by sea or air

LiteX gateware with a VexRiscv core is running on the Ultrascale A DAC cable to a 10GbE switch is showing a physical link using GTY from liteeth and an in progress minimal PCS for RX slip, 64b/66b encoding/scrambling. I also have some x8 PCIe connectors on the way to make some more custom PCBs to breakout the 8 GTY serdes to see if I can do something for fun with a pair of BC-250 boards. The BC-250s are stripped down PS5 APUs with 16GB GDDR6 per board, 6 Xen cores, and 24 RDNA CUs. No external high speed interfaces on those, but I did add some DisplayPort connectors to the DigiKey order.. about to find out if I can get DisplayPort RX going on an FPGA for a unidirectional pipe up to 25Gbps from each board

Popped the heatsink off to double check this board has the FFBV676 package. I noticed they have a PCA9535 I2C expander on there.. why, when you have 280 IO pins available that are barely used

Custom made JTAG adapter is working on the ultra cheap alibaba Ultrascale I added a board to openFPGAloader, and created a spiOverJtag bitstream for the package to allow it to access the SPI flash (32MB MX25U25645G) I soldered a pin strip onto an unpopulated connector next to JTAG to see if I can find the pads and repurpose those signals as a UART connected to the second channel of the FT2232. DAC SFP cable connected to a 10GbE switch to see if we can get liteeth and etherbone going on this thing.

Surface mount parts on the JTAG board are populated

Paste application on 0.5mm pitch