CricketOne board: where it stands and how we get it made
4 November 2026. Before the 3D-printed parts, the board comes first. This page checks the board against everything a maker needs to build it, and sets out how we get boards in our hands quickly, to test them here.
1. Where the board stands today
What we have is a placement plan, not a circuit yet:
| Done | Not done yet |
|---|---|
| The board’s shape: rigid where chips sit, flex where it bends, checked in our stump CAD (rev T) | The circuit drawing (schematic): every pin of every chip and what it connects to |
| All 48 parts chosen, each at its real size and place, with 0 KiCad rule errors | The footprints with pads: the copper each part is soldered to |
| The design review: camera, battery for a 10-hour day, heat, shock, radio and software (0 fails) | The copper tracks: memory, camera and radio lines routed on the layers |
| The request for quote and the design pack | The stack-up: layers, materials and track widths for the fast lines |
That is why the 3D board looks like blocks on a board rather than a real board. The blocks are the right size and in the right place, but nothing joins them yet. Today’s quote asks the maker to draw the circuit and route it. Only a few makers do that for a chip like the RV1126B, it costs the most, and it is the slowest step.
2. What a maker needs to build a board
The full list of files a board maker and an assembler need. Blocks a quote means makers cannot price the job without it.
| # | File | Blocks a quote? | Status |
|---|---|---|---|
| 1 | Gerber (or ODB++) copper, mask, silkscreen and paste layers | yes (board) | outline only; copper to do |
| 2 | Drill files and drill chart | yes (board) | to do |
| 3 | Fabrication drawing: stack-up, materials, impedance table, finish, IPC class | yes (fine lines, rigid-flex) | to do |
| 4 | Parts list (BOM) with maker part numbers, DNP parts and approved alternates | yes (assembly) | test board: done (153 parts, maker part numbers); alternates to add |
| 5 | Placement file (centroid), same revision as the BOM | yes (assembly) | done (rev A-0) |
| 6 | Assembly drawing: polarity, special steps (underfill, coating, X-ray) | no | done (both sides) |
| 7 | Schematic PDF | no, but every design firm asks | test board: done, KiCad ERC 0 errors (the circuit); 8 items to confirm |
| 8 | IPC-356 netlist, for the bare-board electrical test | no | schematic netlist done; IPC-356 comes with the layout |
| 9 | KiCad source files | no | done (placement) |
| 10 | Test points, test list and fixture notes | no | test list done; test points to place |
| 11 | How to load the software: insert the card, switch on | no | done (section 3) |
| 12 | Rigid-flex extras: zone map, coverlay, stiffeners, bend radius, no vias in bends | yes (rigid-flex) | sections done; drawing to do |
4 Nov 2026: the test board’s circuit is drawn (153 parts, 119 nets, KiCad ERC 0 errors): the circuit. The gap now is the copper (1–3, 8): 15 footprints of our own, then placement and routing.
3. Starting from the microSD card
Put the card in, switch on, and it starts our program. On every board, test and sale alike. What makes it work:
- The card always wins. The RV1126B tries SPI flash, then eMMC, then the card, then USB. A written eMMC would start before the card, so the eMMC ships empty and the loader on our card is set to read the card first. (Until today our board notes said sale boards start from the eMMC; that is fixed.)
- The card circuit:
- its own power switch (U12), so a stuck card is restarted without restarting the stump;
- ESD protection (U13), because the slot is the one place a person touches the electronics;
- 22 Ω series resistors on the clock, command and data lines;
- a card-detect line;
- the card lines on SDMMC0 (GPIO2_A0–A5, 3.3 V). The same circuit is in Luckfox’s public RV1126B schematic.
- The card:
- an industrial pSLC microSD (Kingston Industrial or Swissbit S-56u), rated −40 to 85 °C, in a hinged, latched slot that a ball hit cannot open;
- a read-only system with a separate data partition, so a power cut while writing cannot break it;
- Wi-Fi updates write the second system copy and roll back by themselves.
- What is on the card: the start loader at sector 64, U-Boot at sector 16384, then the system and data partitions. One image file, written on a PC with balenaEtcher.
- Proved on the dev kit before we order: start 10 times in a row from power-off with the eMMC empty; restart the card with its power switch; 1,000 power cuts while writing.
Every board we design follows the same rule (lesson sdstart; all 24 boards checked, one switch and one ESD array per card socket: 0 fails). The ESP32-S3 boards cannot start from a card at all, so the maker writes our small loader into the chip’s own flash at test, and the loader runs the program on the card.
4. Two ways to get it made
Route 1 (recommended first): a test board made near us in Texas
One step changes everything: the RV1126B, its memory, its power and its eMMC come on a module that is soldered flat onto our board. The module is the Boardcon LGA1126B: 30 × 28.75 mm, 238 solder pads at 1.25 mm pitch on its underside, RV1126B with 2 GB LPDDR4X and 8 GB eMMC, 3.1–5.5 V in. It is soldered on, not plugged in, like the ESP32-S3 modules on our other boards.
What that gives us:
| Today’s design (route 2) | Test board (route 1) | |
|---|---|---|
| Our board | 8-layer HDI rigid-flex, laser vias, 0.65 mm-pitch chips | plain 4-layer board, normal vias, no part finer than 0.5 mm |
| The hard part (processor + memory layout) | ours to design and route | done by the module maker, tested |
| Who can make it | a few specialist makers | any of the local shops below, in days |
| Board and assembly time (est.) | 4–8 weeks | 1–2 weeks (Round Rock, Austin, Houston) |
| Design work left | full schematic and layout, about 8–12 weeks (est.) | the carrier: camera, power, card, Wi-Fi, lights, sensors |
How it fits the stump:
- The module is 28.75 mm across. In the 31.75 mm bore that leaves 6.7 mm each side of the centre line for the board and parts (on paper; checked in the CAD next, rev U).
- The camera chip (OV9281) stays on our own camera head.
- On the test board the battery tail is a short soldered lead, not the 453 mm flex. The base disc, cells and charger stay the same.
What it costs: the module adds a few millimetres of height and a part we buy, and the module maker’s part must stay available (Boardcon also offers an “open BOM” design for when we go to the full board).
Public references for the circuit:
- Luckfox’s full RV1126B schematic (GitHub, no NDA);
- Boardcon’s LGA1126B manual (pins, power, card and debug circuits);
- Forlinx’s RV1126B hardware manual (start order, card circuit). Its own carrier board is a plain 4-layer board.
Route 2: the full stump board, chips straight on it
This is today’s design: RV1126B and memory on our own 8-layer HDI rigid-flex board, the shape we sell. We do it after the test board proves the circuit, the camera modes, the card start and the battery life. By then the schematic is done and tested, and only the processor and memory layout is new. Two ways in:
- a design firm (Texas or Boardcon’s open-BOM service);
- the same Texas maker that builds HDI rigid-flex (Blind Buried Circuits, Stafford).
5. Who can make it
We are asking many makers at once, local ones first. The quote requests are written and waiting for the owner to send. We contact no one without the owner.
Austin and Central Texas (drive over during bring-up):
| Maker | Where | What | Why them |
|---|---|---|---|
| East/West Manufacturing | Round Rock | board fab + assembly + X-ray | prototypes in 1–5 days, one supplier for both |
| BSU Electronics | Austin | prototype assembly, 5D X-ray, BGA rework | an engineer reviews every order |
| Kodiak Assembly Solutions | Austin | SMT assembly, 3D inspection, X-ray, DFM review | prototype to production |
| Four Points Platinum | Pflugerville | turnkey SMT, fine pitch, X-ray | 20 minutes away |
| On The Spot Assembly | Granger | small fine-pitch prototypes, rework | no minimum |
| FirstView Consultants | Round Rock | schematic, layout, BGA, Linux and bootloader | nearest design firm with both |
| ATX Embedded | Austin | design review, layout, firmware | a low-cost local design check |
Rest of Texas:
| Maker | Where | What | Why them |
|---|---|---|---|
| MacroFab | Houston | online turnkey, 0.4 mm BGA, X-ray, HDI | instant quote, no minimum |
| Blind Buried Circuits | Stafford | fab to 50 layers, laser microvias, rigid-flex, turnkey | the Texas maker for route 2 too |
| Austin Circuit Design | Richardson | design + build, Linux-board background | design and build under one roof |
| Lone Star Circuits | Wylie | fab incl. flex and rigid-flex | a second Texas rigid-flex fab |
US prototype services: Screaming Circuits (Oregon), Advanced Assembly (Colorado), Bay Area Circuits and Laritech (California).
Design + build overseas:
| Maker | Where | What | Why them |
|---|---|---|---|
| Boardcon | Shenzhen | the LGA1126B module; custom carrier design; open-BOM | the module maker; also the bridge to route 2 |
| Forlinx | Baoding | RV1126B modules, custom carrier design | a second module source |
| Bittele (7PCB) | Toronto | design, HDI fab, turnkey, X-ray on every BGA | North American contact |
| Makerfabs | Shenzhen | 1–6 layer layout, turnkey | low-cost layout of the 4-layer test board |
Also on the list from before: PCBWay, NextPCB, Sierra Circuits and Cirexx (the full rigid-flex board).
6. Next steps, in order
The circuit (schematic) for the test boarddone 4 Nov 2026: the circuit, every pin and value, KiCad ERC 0 errors, 8 items to confirm.- Footprints with pads and the connections in our KiCad board. The board then shows real pads and the lines to route.
- Check it in the stump CAD (rev U) with the module in place.
- Copper: route the 4-layer board ourselves, or with a local design firm, and get a design review from them.
- Quotes back from the local shops for 6 and 12 boards. Order when the owner decides.
- Test here: card start, camera modes, battery life, heat, the ball-hit test. Then the 3D-printed parts.