Releases: meshsat/meshsat-fieldkit
Release list
Rev A: carrier set as prepared for the JLCPCB order (3 Sep 2026)
Rev A of the MeshSat field-kit carrier set: the boards as designed and checked (six since 4 Sep), with the JLCPCB order set prepared. Nothing in this release has been fabricated or built yet; the order gate is the document check of the design record and the owner's release of the set (v2/BUILD.md, section 0).
| Board | Rev | Size (mm) | Layers | JLCPCB assembly |
|---|---|---|---|---|
| PCB-A POWER + I/O | A19 | 285 x 160 | 4 | top |
| PCB-B COMPUTE | B12 | 245 x 170 | 4 | top, bottom 1 |
| PCB-C CONTROL PANEL | C5 | 442 x 311 | 2, 2.0 mm, vias plugged | top (LEDs), bottom (cluster, connectors) |
| PCB-D APRS | D5 | 80 x 62 | 4 | top 78, bottom 1 |
| PCB-E1 DOCK STRIP | E4 | 278 x 60 | 4 | top |
| PCB-E5 DOCK BLOCK | E5 | 43 x 26 | 2, 2 oz | none |
Assets: one Gerber and drill zip per board (KiCad 9 export, every copper layer, Excellon drill), the complete JLCPCB order set (per-board upload folders with BOM, CPL and ORDER-NOTES.txt, the order log, the CPL rotation table), the review prints (1:1 sheets, assembly drawings, copper layers, schematics, renders), the 1:1 template for the case wall receptacles and the battery module models (STEP and STL). The KiCad sources, the generators and the design record are in the repository at this tag (v2/ecad/, v2/docs/).
Known gaps of Rev A are listed in v2/BUILD.md, section 9. Licence CERN-OHL-S-2.0; third-party vendor files under v2/vendor/ are excluded from it.
Update 4 Sep 2026: PCB-A A16 and PCB-B B10 are superseded by A17 and B11 (design record section 32): the X1202's true outline from Geekworm's drawing, and a node-fed 5 V module rail on PCB-A that takes every 5 V load but the Pi off the X1202. The A17 and B11 gerber assets, the order set and the review prints were refreshed; the other five boards are unchanged. Nothing has been fabricated yet.
Update 4 Sep 2026, early morning: PCB-A A17 is superseded by A18 and the RF junction strip E2 by E3 (design record section 32.10). The panel switches, the SMA couplers and the dock spring pins are now named parts, each from its manufacturer's document in v2/vendor/: C&K ATP19 and ATP16 pushbuttons, NKK M-series toggles, Amphenol Connex 132170 couplers and the Preci-Dip 813-S1-008-10-016101 spring-loaded connector. A18 changes only the eight J_DOCK holes (1.1 mm for the connector's solder tails); E3 changes only the D-hole flat (6.00 mm across, as the 132170 drawing asks). The A18 and E3 gerber assets replace A17 and E2; the order set and the review prints are rebuilt (the retired order folders sit under order/superseded/). Nothing has been built yet: the cart lines for A, B and E2 are rebuilt from A18, B11 and E3 before the order is placed.
Update 4 Sep 2026, the respin: after the owner rulings of 4 September (design record 32.13 to 32.22) the set was respun. PCB-A A19 carries the kit's own charger, fuel gauge, three 5 V converters, main power control and a seven-port USB hub, with blind-mate RF receptacles underneath; PCB-B B12 carries the Pi alone on three rails from PCB-A; the dock strip E4 grew to four layers with a solar tracker and the module entry; the new dock block E5 presents the contacts the stack lands on. The cells left the boards for a battery module on the case floor (models in the new asset), and the two external connectors became MIL-DTL-38999 receptacles on a plate over the case wall (template in the new asset). All seven boards route clean (0 unrouted, 0 hard DRC) and a cross-board contract check passes. The A19, B12, E4 and E5 gerber assets replace A18, B11, E1 and E3; the order set and the review prints are rebuilt. The JLCPCB cart still holds the earlier phases and is rebuilt before anything is ordered. Nothing has been built.
Update 4 Sep 2026, night: A19 pass 5 and the sealed panel C5. PCB-A A19 was rerouted on locked fine-pitch escapes after two I2C pins of the charger came back open (design record 32.33): 0 unrouted, 0 hard DRC, no isolated copper; the dock block E5 was rebuilt from it. The control panel became C5, the case's sealed weather face (design record 32.34, owner rulings of 5 September): no plated hole on the face but the LEDs and the frame screws, keyed holes for the sealed toggles, foam tape frames under the display glass and under a new polycarbonate lens over the e-paper window, a panel-mount IP68 sounder, SMD connectors on the underside, plugged vias, and a die-cut gasket ring between the panel and the frame; the spacer ring R1 is retired. The die-cut outlines ship as pcb-c-display-seals.dxf in the C5 deliverable. The C5 gerber asset replaces C4, the R1 asset is withdrawn, the A19, B12, E4 and E5 assets, the order set, the review prints and the case wall template (now with the seal parts named) are refreshed. The sealing is a construction, not a test result: the bench procedure of 32.34 has not been run on any panel. Nothing has been built. Six boards.
Date note, 4 Sep 2026, 22:40: the updates above labelled 5 September happened on 4 September (the repository's commit clock); the design record carries the same correction (section 32, date correction note).
Update 5 September 2026: the Compute Module 5 set (A20, B13, D6)
The compute side is respun around a Raspberry Pi Compute Module 5 on two board-to-board receptacles on PCB-B (B13): no USB socket, plug or cable between the compute and the kit, the module's 5 V on the board, the USB hub on the carrier for the SDR receptacle and the wall port, a mini PCIe socket for the LTE card, and the GNSS (u-blox NEO-M9N on I2C with its pulse), LoRa (Seeed Wio-SX1262 on SPI) and ZigBee (Ebyte E72-2G4M20S1E, CC2652P, on a UART) modules soldered to the board. PCB-A becomes A20 (its hub and the four dongle channels are gone, the wall-port channel stays, a transistor pin-order defect of A19 is corrected). PCB-D becomes D6 (a WM8960 codec on I2S replaces the USB sound path). C5, E4 and E5 are unchanged. The design record is appendix 32.35 and 32.36 of v2/docs/MESHSAT-709-geometry-appendix.md; the assets of A19, B12 and D5 are replaced by A20, B13 and D6 below. Nothing has been built; the three-module cluster carrier of the same study waits until B13 is proven.
| Board | Rev | Size (mm) | Layers | Assembly |
|---|---|---|---|---|
| PCB-A POWER + I/O | A20 (replaced by A21 below) | 285 x 160 | 4 | top 148 |
| PCB-B COMPUTE | B13 (replaced by B14 below) | 245 x 170 | 4 | top 181, bottom 1 |
| PCB-D APRS | D6 (replaced by D7 below) | 80 x 62 | 4 | top 62, bottom 1 |
Update 5 September 2026 (evening): A21, B14 and D7
Three boards are respun on top of the Compute Module 5 set. PCB-A becomes A21: the net classes now really apply to the labelled nets (a pattern defect had left every labelled net in the default class on every board), and the rail and boost-input currents no longer travel in router tracks at all: top-side islands at each converter output, locked stitch vias, bottom-layer collector bands and risers to the VH connectors, boost feeds on the inner layer and the bottom layer in parallel with tap islands at the inductors, all gated on the finished board for continuity. PCB-B becomes B14 with an M.2 E-key socket on the module's PCIe lane for an AsiaRF AW7915-AED WiFi card (kit-to-kit link without an access point, two end-wall antenna sites, a 3.3 V buck on the Pi rail with its own current monitor). PCB-D becomes D7, fitted to the NiceRF DMR858M V1.0 board as delivered (carrier detect and the channel knob read on an expander, the control UART on an SC16IS740 bridge, the microphone divider corrected, the module's heatsink cleared by 38 mm board spacers). The design record is appendix 32.35 to 32.39 of v2/docs/MESHSAT-709-geometry-appendix.md; the assets of A20, B13 and D6 are replaced by A21, B14 and D7 below. Nothing has been built. The B14 PCIe pairs are length-matched by post-route meanders (transmit and receive 0.00 mm, clock 0.25 mm), since the router does not length-match on its own.
| Board | Rev | Size (mm) | Layers | Assembly |
|---|---|---|---|---|
| PCB-A POWER + I/O | A21 | 285 x 160 | 4 | top |
| PCB-B COMPUTE | B14 | 245 x 170 | 4 | top, bottom |
| PCB-D APRS | D7 | 80 x 62 | 4 | top, bottom |
Update 6 September 2026: the kit moves to the Peli 1450, with an aluminium face, a six-layer compute board and a four-layer backer
The 1520EU bought on 3 September arrived and is twice the kit (it was sized for the earlier Pi 5 stack). Measured on Peli's own CAD (design record 32.41), the Peli 1450 with the 1450PF panel frame is the smallest panel-frame case whose window passes the board stack, so the kit moves there (32.40 items 4 to 9, 32.42). Only the panel is reshaped; PCB-A A21, PCB-D D7, the dock strip E4 and the block E5 are unchanged.
- The face is a 3 mm aluminium plate (365.5 x 249.5, ten M3 at the frame's inserts, black anodised, laser-marked legends and logo) instead of a PCB, by the owner's ruling that "the aluminium plate with a backer board is the safer face for anything that leaves the bench". The plate carries the Touch Display 2 aperture and its glass pocket, the e-paper window and its lens pocket, the keyed holes for the sealed switches, the sounder hole and sixteen 2.6 H7 holes for Mentor 1282.5004 IP68 light guides. New asset
meshsat-fieldkit-revA-case-face-plate.zip: STEP, DXF (outline, through cuts, 1 mm pockets, standoff holes), the marking SVG and a 1:2 drawing with the cut-out table. - PCB-C becomes C6, the backer board: a U-shaped 1.6 mm four-layer board (F.Cu, In1 routable under a ground pour, In2 signal, B.Cu; the two-layer backer left connections open in every route, and with In1 reserved as a plane no router completes the backer either, so C6 is the recorded exception to the In1 plane rule, design record 32.47) that hangs 10 mm under the plate on six self-clinching standoffs and carries the panel electronics, the LEDs under the light guides and the lands for the switch leads; the switches and...