Skip to content

Memory card v1.3 — the BUILT card (Fusion 360 export, 2026-09-24)

This folder is the record of the memory card that is in the machine: the design JLCPCB fabricated on 2025-06-27 (order 2000765A, 4-layer FR-4). It was exported from Fusion 360 on 2026-09-24 in Eagle format and it was proven against that order (below). It was filed first as eagle/v1.3-fusion-export-2026-09-24/ and renamed to eagle/v1.3/ the same day. The folder that used to be eagle/v1.3/, now ../deprecated/v1.3-do-not-use/, is an earlier save of the same revision (notes 2025-03-06, board file 2021-03-17) and is not the built card: do not use it. See "How it differs from the earlier save" below.

File What it is Original name
Memory V1.3.sch schematic, Eagle 9.7.0 XML (Fusion export) Memory_V1.3.sch
Memory V1.3.brd board, Eagle 9.7.0 XML (Fusion export): 177.8 x 114.0 mm, 4 layers (1 Top, 2 = GND plane, 15 = VCC plane, 16 Bottom), 53 parts, 1,263 tracks, 114 vias Memory V1.3.brd
fab/jlcpcb-order-2000765A-Y42.zip JLCPCB's order archive, downloaded as it came. YG/PCB_2000765A_Y42.zip inside is the gerber set JLCPCB received (Fusion CAM job "Memory V1.3 v4", 2025-06-27: 15 files, gerbers + drill + part list + pick-and-place); YG/4te.json has the order parameters (4 layers, FR-4); ok/ holds JLCPCB's processed CAM data Memory V1_2025-06-27_Y42.zip
pdf/ schematic and per-layer board PDFs (generated by tools/sch_to_pdf.py / tools/brd_to_pdf.py) -
FABRICATED marker (generated by tools/gen_fabricated.py from tools/fabricated.py) -

The KiCad conversion (proven, schematic vs board) is ../../kicad/v1.3/.

Proof that this is what was fabricated

tools/verify_fab_vs_brd.py compares the board file with the order's gerber set without using KiCad (it parses the Eagle XML, the Excellon drill file and the RS-274X copper layers itself). Re-run it from the repository root:

python3 tools/verify_fab_vs_brd.py "hardware/cards/memory/eagle/v1.3/Memory V1.3.brd" \
    hardware/cards/memory/eagle/v1.3/fab/jlcpcb-order-2000765A-Y42.zip \
    --same-as "hardware/cards/memory/eagle/deprecated/v1.3-do-not-use/fab/Memory V1_2025-06-27.zip"

Result (2026-09-24, and again after the rename): PASS

  • Holes: the board has 695 (579 pads, 114 vias, 2 mounting holes); the drill file has 695. Every one matches in position (0.01 mm) and diameter; nothing extra on either side. Sizes: 0.254 mm x 114 (vias), 0.813 x 456, 1.016 x 123, 2.794 x 2.
  • Tracks: all 576 Top and all 687 Bottom wires (all 0.1524 mm = 6 mil) are strokes of the same width in copper_top_l1.gbr / copper_bottom_l4.gbr. The gerbers' remaining strokes (407 per side) all lie inside pads: that is how Eagle draws the oval DIP pads. No other stroke.
  • Planes: the board's only inner-layer copper is one polygon per layer, layer 2 = GND and layer 15 = VCC; the inner gerbers are filled regions.
  • Parts: the fab part list names the same 53 parts as the board, including IC15 (74ALS11N) and IC26-IC29 (74LS374N); the pick-and-place puts all 53 at the board's coordinates.
  • The gerber set in the order archive is byte-identical, all 15 files, to ../deprecated/v1.3-do-not-use/fab/Memory V1_2025-06-27.zip (which the tree already had; that zip belongs to this design, not to the earlier save it is filed with).

Control: the same check against the earlier save (../deprecated/v1.3-do-not-use/Memory V1.3.brd) FAILS (178 of its holes are not in the drill file, 226 drill hits are not on that board, hundreds of tracks differ; the part list has IC15, the board does not).

How it differs from the earlier save in ../deprecated/v1.3-do-not-use/

Schematic (the only circuit change): IC15, a 74ALS11 triple 3-input AND. Gate A (pins 1, 2, 13 -> 12) = AND(-LO-RAM, -HI-RAM, -ROM-CS) drives net N$21 -> IC5 pin 19, the enable (-G) of the 74LS245 data-bus buffer. In the earlier save IC5 pin 19 sits on -VMA. With IC15 the buffer is enabled only while one of the card's own chips (low RAM, high RAM or ROM) is selected, so it no longer drives the data bus for a block the card leaves undecoded: v1.3 added the IC7 jumper block so any 4K block can be taken out of the map for memory-mapped I/O, and the video card answers in the $D000 block; with -VMA as the enable the memory card's buffer would fight the video card on every read there. Gates B and C have their inputs grounded. The jumper-block nets are also named in the export (SEL0-7 = IC7 outputs, RAM0-7, ROM0-7 = the jumper returns; they were N$5..N$44), same connectivity.

Board: IC15 is placed (151.13, 53.34) and routed; the TMP registers IC26-IC29 (74LS374, 154.94 / 105.41-67.16) and the resistor networks RN5/RN6 are on the board and routed (in the earlier save they sat off the board with their connections unrouted: its KiCad conversion lists 100 unconnected items). All of this board's copper is in the order's gerbers (above).

Last updated 2 October 2026