Skip to the content
4 layers130 × 64.3 mmMegabank, the printed circuit board in 3D

Battery-backed 4 MB expansion-bank RAM board for the Timex Sinclair 2068 (TS-Pico v3)

Size
130 × 64.3 mm
Copper layers
4
Thickness
1.6 mm
Parts
97
Nets
166
Pads
561
Tracks, vias
1,945, 355
Pours
9
Routed
Yes
KiCad
KiCad 10.0 (file 20260206)
Created
16 Oct 2026
Updated
7 Nov 2026

Overview

Megabank is an open-source printed circuit board by timex-sinclair-projects, from the timex-sinclair-projects/TS-2068-Megabank repository on GitHub, published under the MIT licence. It is a 4-layer board of 130 × 64.3 mm with 97 parts on 166 nets, routed with 1,945 tracks and 355 vias.

It is built around the RP2350B, 74LVC245APW and 74LVC1G06Z, with the 74LVC2G34GW, GD25Q128ESIGR and CY62167EV30LL-45ZXIT among its other parts. Its parts include 35 capacitors, 23 resistors, 16 ICs, 3 switches and 2 diodes. It belongs with the RP2350 boards of this library.

Main components

12

README

hardware/README.md

Megabank KiCad project

Status 2026-11-07: routed, DRC clean at every severity (no errors, no warnings, nothing unconnected), not yet ordered. The board has the TS-Pico v3's form factor and is built with v3's script pipeline (tspico-hardware, hardware/scripts/), copied here and renamed. Spec: ../docs/spec.md.

  • Schematic: four sheets, generated. ERC: one warning, expected (below).
  • Board: outline, gold fingers and mounting holes are v3's. Routed by scripts/flow.py with Freerouting 2.4.1 and no hand edits: 1945 track segments (1315 top, 630 bottom), 355 vias, all 0.6/0.3 mm, narrowest track 0.11 mm, nothing on the plane layers. Closed on the first attempt (re-routed 2026-11-07 for the core LDO, which replaced the RP2350B's switcher and L1, and the GPIO46/47 test pads). Silkscreen: all 90 designators placed.
  • Fab files: python3 scripts/gen_fab.py writes the Gerbers, BOM and CPL to fab/jlcpcb/, committed: megabank-gerbers.zip, megabank-bom.csv, megabank-cpl.csv, ready to upload as they are (the unzipped Gerbers are not committed); fab/NOTES.md lists what to check on upload.
  • Parts: every fitted part has an LCSC number. $26.63 a board in parts at 10 boards (jlc_price.py, 2026-11-07; it counts U1-U4 once, so about $27), plus $45 in extended-part fees once; the two SRAMs are $20.86 of it. The ME6211C12 LDO (32,146 in stock) is an extended part like the L1 it replaced.

Form factor

The same card as v3, so the two share the extender, brackets and standoffs:

ItemValue
Board130.00 x 64.25 mm, 1.6 mm, 4 layers (JLC04161H-7628): F.Cu, In1 GND, In2 +3V3, B.Cu
Body130.00 x 56.00 mm
Tongue83.66 x 8.25 mm, left edge 18.48 mm from the board's left edge, 1 mm chamfers
FingersTS2068_EdgeConnector_2x32, 32 positions at 2.54 mm, 31 a side, key slot at position 6
HolesM3 at (5, 5), (113.6, 4.2), (5, 51), (112, 51.5): v3's

The finger geometry and the side-A-is-top assumption are v3's; see v3's hardware/README.md, "Edge connector geometry". Check side A before ordering (spec, open question 6).

Sheets

SheetGeneratorContents
megabank.kicad_schby handRoot: four sheets, flat, global labels
interface.kicad_schgen_interface.pyP1 fingers; U1/U2 address buffers (always on: BA0-BA15), U3 data buffer (DIR = /RD, /OE = nDOE), U4 control buffer; U5 74LVC1G06 /BE driver; U6 74LVC1G06 /NMI driver; U7 74LVC2G34 /INT and /NMI in; pulls R1-R4
memory.kicad_schgen_memory.pyU10/U11 CY62167EV30 as 2M x 8; U12 /WE gate; ~CE1 pull-downs, WINH pull-up, BHE/BLE links
mcu.kicad_schgen_mcu.pyv3's RP2350B core, same designators, but DVDD from U23, a 1.2 V LDO (no L1); USB-C; RUN, BOOTSEL; SW5 write protect; TP4/TP5 on the spare GPIOs; the GPIO map
power.kicad_schgen_power.py+5V diode-OR (edge, USB), TLV1117LV33, TPS3839 -> CE2, LM66200 -> VSRAM, CR2032, battery monitor

gen_symbols.py writes lib/megabank.kicad_sym: the edge connector (v3's), the CY62167EV30 (from the Infineon datasheet, 002-24706 Rev *B, figure 2) and the LM66200 (TI SLVSG04, table 5-1). parts.py maps value and footprint to MPN and LCSC number; schlib.place() stamps them on each symbol.

GPIO map

A PIO state machine sees 32 pins, from GPIO0 or GPIO16. The port program needs data, A0-A7 and the strobes; the memory program needs the chunk bits, the strobes, /BE, the ~CE1s and the page bits. Everything both need sits in 16-31:

GPIONet
0-7SD0-SD7SRAM-side data: IN/OUT to FCh/FDh
8-15BA0-BA7port decode
16-18BA13-BA15chunk (out pc, 3)
19-25nMREQ_L, nIORQ_L, nRD_L, nWR_L, nRFSH_L, nM1_L, nRESET_L
26nDOEdata buffer enable, pulled up
27BE_REQto U5, pulled down
28-29nCE1_0, nCE1_1SRAM enables, pulled down
30-37PA13-PA20page bits, contiguous with the ~CE1s
38WINHwrite inhibit, pulled up
39nWPwrite-protect switch, low = protected
40BAT_ADCADC0
41BAT_EN
42nBE_L/BE read back through U4's spare channel
43nINT_L/INT, through U7
44nNMI_L/NMI, through U7
45NMI_REQto U6, pulled down (R4)
46-47GP46, GP47spare, on test pads TP4, TP5 (silk "GP46", "GP47") for bench experiments

Decisions taken here, beyond the spec

  • USB-C, as on v3 (open question 11): firmware by BOOTSEL and the settings (BASE, SIZE) over USB. VBUS is diode-ORed with the edge +5V, so the board runs from USB alone on the bench. SWD pads TP1-TP3 too.
  • /BE read back: U4's eighth channel brings /BE in as nBE_L (GPIO42), for timing on the bench and to see v3 serve a chunk.
  • BT1 is a Keystone 3034TR (C5213768, 2026-11-07), the footprint's own part, as on the OpenSpand card, where JLCPCB placed it. The Q&J CR2032-BS-6-1 (C70377) made JLCPCB move the order to Standard PCBA (edge rails, which would sit against the fingers). Stock is thin (654); if it runs out, leave BT1 off and fit a Keystone 3034 from DigiKey (36-3034-ND) by hand.
  • R3, a /BE pull-up to +5V_EDGE, DNP (open question 13): the bus already has v3's 2k2; the pad is there if the prototype shows /BE rising too slowly.
  • An LDO for the core, not the on-chip switcher (2026-11-07): U23, an ME6211C12M5G-N (SOT-23-5, C236672), supplies DVDD at 1.2 V, as on the OpenSpand card. The switcher's inductor (v3's L1, Abracon AOTA-B201610S3R3-101-T) was the scarcest part on the board (3 at JLCPCB in October) and needed a pre-order or hand fitting. VREG_LX is open, VREG_FB is grounded, VREG_VIN (C22) and VREG_AVDD (R28/C39) stay powered (RP2350 datasheet 6.3.7). 1.2 V is what v3's firmware sets for 250 MHz; the firmware may put the unused regulator in HIZ to save about 400 µA. C40 (1 µF) at the LDO's input, C20 (4.7 µF) at its output; C21/C47/C48 stay at DVDD pins 51, 32 and 10.
  • Test pads on the spare GPIOs: TP4 (GPIO46) and TP5 (GPIO47), for bench experiments such as a second controller's handshake (spec, Bus mastering).
  • One 0 R each for BHE and BLE (R13, R14), shared by both chips.
  • Battery-monitor designators are R51-R54, C55 (the spec's R1-R4, C1).
  • SRAM pins may be swapped at layout: any permutation of A0-A20 or of I/O0-I/O7 is harmless, since both chips always see the same mapping. Only SDn must stay on GPIOn. Worth using if the router struggles.

The ERC warning

U5 pin 4 (open collector) and U4 pin 18 (tri-state): the /BE driver and the buffer that reads /BE back. That is the intent.

Layout

Coordinates in mm from the body's top-left corner (gen_layout.py, table P).

  • Over the fingers, as v3: data U3 (x 40), high address U2 (x 63) over A15-A8, control U4 (x 80) over /MREQ to /RFSH, /BE driver U5 by finger A29. /INT and /NMI over their fingers (A14, A15) between C3 and U2: U7 (x 51.5), U6 and R4 (x 56.5). Low address U1 above U2.
  • RP2350B cluster at v3's coordinates with v3's designators (U20 at (62, 26), 90°), so v3's crystal and fan-out pre-routes carry over. Flash U21 to its left.
  • Core supply: with VREG_FB grounded, pin 51 is the only DVDD pin a track can reach from outside the pad ring (on v3 the switcher's output came in through FB). So the LDO U23 sits above pin 51, at (60.55, 14.6), pins 4/5 towards the chip: its output runs straight down to pin 51, 0.3 mm wide, past C21, with C20 off its side, and v3's DVDD spine under the package carries it on to pins 10 and 32. C40 above U23; plane vias for its input and ground. FB goes to a GND via where L1 was. TP4/TP5 at (53.5, 13.5) and (53.5, 16), up and left of the chip's GPIO46/47 pins. One pre-route changed: on v3, supply pin 29 dropped into the lanes of pads 27 and 28 to reach its via, because those GPIOs (26, 27) were spare. Here they are nDOE and BE_REQ, and the first run left both sealed in on all three attempts. The 24/29 supply link now carries on under the package to pin 15, whose fan via is already on the +3V3 plane.
  • SRAM U10 and U11 on the right, leads top and bottom; the /WE gate, the supervisor and the pulls in the strip between them and the crystal.
  • Power: diodes and LDO top left, LM66200 and the battery monitor in the free strip above the RP2350B, the CR2032 bottom left. USB-C on the left edge (v3's place). RUN and BOOTSEL on the top edge (v3's places). Write-protect slide switch on the right edge.

Next

  1. Check footprints against drawings: the MSK12C02's actuator should face the right edge; TSOP-I 48 against Infineon's 51-85183. Then check side A of the fingers (spec, open question 6).
  2. The SRAM rail with a fresh cell: the LM66200 passes whichever input is higher, and a new CR2032 reads up to about 3.3 V open circuit, against +3V3's 3.25-3.35 V. At SRAM standby currents the cell may carry the rail some of the time with the 2068 on. Harmless to the data; it costs battery life. Measure on the prototype (spec, hardware test 3); a Schottky in series with the cell would settle it if needed.
  3. Stock: 42 CY62167EV30 at JLCPCB on 2026-10-18, 40 on 2026-11-07, enough for 20 boards.

Routing

bash
cd hardware
PY=/Applications/KiCad/KiCad.app/Contents/Frameworks/Python.framework/Versions/Current/bin/python3
$PY scripts/flow.py                                   # jar from $FREEROUTING_JAR (or --jar); needs Homebrew's openjdk
scripts/check.sh                                      # ERC (one expected warning) and DRC
python3 scripts/gen_fab.py

flow.py regenerates the placement and discards all routing, then pre-routes, autoroutes, finishes and places silkscreen, up to three attempts; Freerouting does not give the same result twice. Each attempt took 4-8 minutes to route here. A failed attempt is kept as fab/route/attempt-N.kicad_pcb.

Regenerating

bash
cd hardware
python3 scripts/gen_symbols.py                       # project symbols
for s in interface memory mcu power; do python3 scripts/gen_$s.py; done
PY=/Applications/KiCad/KiCad.app/Contents/Frameworks/Python.framework/Versions/Current/bin/python3
$PY scripts/gen_layout.py                            # skeleton, netlist, placement, planes (discards routing)
$PY scripts/preroute.py                              # v3's locked pre-routes

Regenerating a sheet overwrites hand edits; once a sheet is edited in KiCad, retire its generator. The same goes for gen_layout.py once routing starts by hand.

BOM and price:

bash
cd hardware
KC=/Applications/KiCad/KiCad.app/Contents/MacOS/kicad-cli
$KC sch export bom --output bom/megabank-jlcpcb-bom.csv --fields "Value,Reference,Footprint,LCSC" --labels "Comment,Designator,Footprint,LCSC Part #" --group-by "Value,Footprint" --exclude-dnp megabank.kicad_sch
python3 scripts/jlc_price.py --boards 10 --bom bom/megabank-jlcpcb-bom.csv

(jlc_price.py counts a range like U1-U4 as one part, so its total runs a little low.)

Parts

97
QtyValueFootprintReferences
26100nC_0603_1608MetricC1, C2, C3, C10, C11, C12, C21, C23, C24, C25, C26, C27, C28, C29, C30, C31 +10
1010kR_0603_1608MetricR1, R2, R4, R10, R11, R12, R21, R22, R50, R54
474LVC245APWTSSOP-20_4.4x6.5mm_P0.65mmU1, U2, U3, U4
4MountingHole_3.2mm_M3MountingHole:MountingHole_3.2mm_M3H1, H2, H3, H4
333R_0603_1608MetricR26, R27, R28
34.7uC_0402_1005MetricC20, C22, C39
3Fiducial_1mm_Mask2mmFiducial:Fiducial_1mm_Mask2mmFID1, FID2, FID3
20R_0603_1608MetricR13, R14
2100kR_0603_1608MetricR51, R52
210uC_0603_1608MetricC50, C51
215pC_0603_1608MetricC35, C36
21kR_0603_1608MetricR20, R23
21uC_0603_1608MetricC40, C53
25k1R_0603_1608MetricR24, R25
274LVC1G06SOT-353_SC-70-5U5, U6
2CY62167EV30LL-45ZXITTSOP-I-48_18.4x12mm_P0.5mmU10, U11
2SS34D_SMAD1, D2
112MHzCrystal_SMD_3225-4Pin_3.2x2.5mmY1
11MR_0603_1608MetricR53
12k2R_0603_1608MetricR3
174LVC2G34SOT-363_SC-70-6U7
1AO3400ASOT-23Q2
1AO3401ASOT-23Q1
1BOOTSELSW_Push_1P1T_XKB_TS-1187ASW4
1CR2032BatteryHolder_Keystone_3034_1x20mmBT1
1GD25Q128ESIGRSOIC-8_5.3x5.3mm_P1.27mmU21
1GNDTestPoint_Pad_1.5x1.5mmTP3
1GP46TestPoint_Pad_1.5x1.5mmTP4
1GP47TestPoint_Pad_1.5x1.5mmTP5
1LM66200SOT-583-8U32
1ME6211C12M5G-NSOT-23-5U23
1RP2350BQFN-80-1EP_10x10mm_P0.4mm_EP3.4x3.4mmU20
1RUNSW_Push_1P1T_XKB_TS-1187ASW3
1SN74LVC1G32DBVRSOT-23-5U12
1SWCLKTestPoint_Pad_1.5x1.5mmTP1
1SWDIOTestPoint_Pad_1.5x1.5mmTP2
1TLV1117LV33SOT-223-3_TabPin2U30
1TPS3839G33SOT-23U31
1TS2068_Edgemegabank:TS2068_EdgeConnector_2x32P1
1USB-CUSB_C_Receptacle_HRO_TYPE-C-31-M-12J1
1WPSW_SPDT_Shouhan_MSK12C02SW5

Similar boards

  • 4 layers130 × 64.3 mmTspico v3, the printed circuit board in 3D

    Tspico v3

    timex-sinclair-projects/tspico-hardware

    Status 2026-10-19: prototype 3 (v3.0-proto3, changes below). Proto1 (v3.0-proto1) was built and brought up; proto2 added its fixes (below).

    156 partsCERN-OHL-S-2.0
  • 4 layers57.9 × 59.4 mmGW4302, the printed circuit board in 3D

    GW4302

    garrettsworkshop/GW4302

    Prototype RAM cartridge compatible with Commodore 1700/1750/1764 RAM Expansion Unit (REU)

    100 parts11
  • 2 layers54.9 × 95 mmBoard v5, the printed circuit board in 3D

    Board v5

    brev-dev/another_esp8266_sensor_board

    PCB layout and instructions for a multi-purpose esp8266 sensor device, with optional LORA connectivity.

    86 parts5GPL-3.0
  • 4 layers157 × 105 mmFLoggy, the printed circuit board in 3D

    FLoggy

    markovejnovic/FLoggy

    Open source/hardware data collection platform

    212 parts
  • 2 layers63.9 × 15 mmESPower, the printed circuit board in 3D

    ESPower

    nagius/ESPower

    Motherboard for ESP8266-01 with voltage regulator, voltage protection and Li-Ion battery charger integrated.

    38 parts
  • 4 layers70.8 × 110 mmThors Arsenal, the printed circuit board in 3D

    Thors Arsenal

    shannon-greenlight/Thors-Arsenal

    Eurorack Module

    311 parts1GPL-3.0
  • 4 layers45 × 49 mmSwarm drone, the printed circuit board in 3D

    Swarm drone

    aniranaway/Drone_Swarm

    A fully custom, low-level embedded hardware and firmware platform designed for autonomous multi drone swarms. The system integrates custom STM32-based flight controller PCBs, coreless DC motor drive arrays for a micro…

    108 parts

Questions

How big is the Megabank PCB?

It measures 130 × 64.3 mm and has 4 copper layers, 1.6 mm thick.

What parts does the Megabank use?

97 parts, built around the RP2350B, 74LVC245APW, 74LVC1G06Z and 74LVC2G34GW. The full list of its parts is on this page, with links to each part's symbol and footprint where PCBKit's library has them.

Where can I download the Megabank KiCad files?

From the timex-sinclair-projects/TS-2068-Megabank repository on GitHub, at the commit this page was made from: the board is hardware/megabank.kicad_pcb and its schematic hardware/interface.kicad_sch. PCBKit opens the board in the browser, with nothing to install.

Can I use the Megabank design in my own project?

Its repository publishes it under the MIT licence: you may reuse it on that licence's terms (keep the author's notice, and for a copyleft licence share your changes alike).