It is real
After months of schematic churn, opcode tables, and the occasional 40b detour, Tomato has left the screen.
See also: Designing additional boards · Falling back to 32b
Yesterday: bare boards on JLCPCB.
Tonight: every IC, cap, LED, and SIP resistor network on DigiKey — enough for two 07_alu slices.
The dual-LUT ALU is no longer a KiCad fantasy. Copper is being etched. Chips are in a warehouse in Minnesota. I am genuinely excited.
Figure 1 — What we are actually building: 07_alu — board render (left), routed top copper (right). DRC clean, 0 unrouted nets.
Fabrication — boards
I finally stopped polishing and hit submit. The existential “one more routing pass” voice is quiet for now — and honestly, it was not buying much anymore.
- Board:
07_alu— 8-bit dual-LUT ALU slice - Cost: under $6 for the PCB order (easy to reorder if something goes wrong)
- ETA: ~two weeks to Philadelphia
If there is a fab error, I can spin another batch without heartbreak. That was the whole point of keeping the board small and the order cheap.
Figure 2 — JLCPCB order — copper incoming.
DigiKey — parts (order 100884560)
Same night, different kind of commitment: 34× 74ACT151 muxes do not lie. This is a real machine, not a simulation.
I sized the cart for two boards — a full 16-bit datapath from day one. One board proves the slice; two boards is the actual Tomato width on the bench. The marginal cost of the second slice (~$20 in parts on top of ~$40 for one) was too good to skip.
| Line | Part | Qty | Role |
|---|---|---|---|
| 1 | CD74ACT151M96 | 34 | 8:1 LUT muxes — the heart of the dual-LUT plane |
| 2 | CD74ACT283M | 4 | 4-bit ripple adders |
| 3 | SN74ACT00DR | 4 | Quad NAND (glue) |
| 4 | SN74ACT86DR | 2 | Quad XOR |
| 5 | MC74ACT377DWR2G | 2 | Octal D-FF / flag latch |
| 6 | CD74ACT541M96 | 2 | Octal bus buffer |
| 7 | CD74HCT688M96 | 2 | 8-bit identity comparator (zero detect) |
| 8 | CC0603 100nF | 54 | Decoupling (50 + spares) |
| 9 | LTST-C170KRKT red LED | 130 | Debug wall (126 + spares) |
| 10 | 4610X-101-471LF SIP-9 | 14 | Bussed 470Ω LED limiters |
- Subtotal: $48.93
- Total (ship + tariff + tax): ~$61
- Availability: all lines immediate — no backorders
- BOM on disk:
07_alu_digikey_bom.csv - Order export:
100884560.csv(fabrication copy)
Swapped the backordered Samsung 100nF caps for Yageo 311-1344-1-ND. Picked CD74HCT688M96 over SN74HC688 for cleaner 5V ACT interfacing. Same footprint, better margins.
Figure 3 — DigiKey order 100884560 — 10 line items, two boards worth of silicon.
Still to source
DigiKey does not have the SIP-4 networks without a 2,000-piece reel trap, and I already have headers in the lab:
| Item | Qty (2 boards) | Refs |
|---|---|---|
SIP-4 bussed 470Ω (4604X-101-471LF or equiv.) | 4 | RN9, RN10 |
| 4-pin headers | 4 | J10, J11 |
| 8-pin headers | 14 | J5–J7, J12, J15–J17 |
Fallback for RN9/RN10: discrete 470Ω resistors, common on one side.
What happens next
- Wait — JLC boards (~2 weeks), DigiKey box (days).
- Grab SIP-4 + headers before the soldering marathon.
- Assemble board #1 — 131 placements, 63 LEDs, the whole debug wall.
- Bring up 8-bit — opcode walk, carry chain, flag latch, LED sanity.
- Assemble board #2 — cascade to 16-bit, wire slice-to-slice, and finally see the dual-LUT datapath breathe.
If something is wrong, I will know because 130 red LEDs do not lie.
This is the best part of the project — the part where the schematic meets flux and patience. Tomato is ordered. Now we build.