Skip to content
IEMI Labs
CricketOne

Our own circuit board, shaped to the stump

One board we design, with every chip soldered on it: the camera, the bail sensors, the call lights, Wi-Fi and Bluetooth, the RV1126B processor and the battery system. It is rigid where chips sit and flexible where it bends, so it folds into the stump in one piece. No plugs, ribbons or adapters.

Stage 1 of 6 done: shape and placementNot yet ready to manufactureUpdated 3 November 2026 · rev A-0

The board in 3D

Every part at its real size, in its real place, checked against the stump CAD. Turn it, zoom in, switch between the board in the stump and laid flat, and pick a part (on the model or in the list) to see what it is. The chip bodies are simple blocks until each part's own 3D model is added.

Loading the 3D board…

Drag to turn · scroll to zoom · right-drag to move · click a partDrag to turn · tap a part

Where everything is in the stump

Bail sensors705.5 mm

A Hall switch on a thin flex finger in each of the head's two Hall pockets, right under the bail grooves. When a bail lifts, its magnet moves away and the switch sees it.

Camera662 mm

The OV9281 global-shutter camera chip, soldered on the head in the CAD's camera slot, under an M8 lens holder with a Ø7 lens, tilted 12° to look out of the window at the far end.

Call lights600–642 mm

Eight side-facing LEDs on two flex wings, shining into the clear light strips on both sides of the head, driven by the light driver on the brain.

Wi-Fi and Bluetooth558–570 mm

The antenna at the top of the brain, above the metal heat bar, so the radio gets out through the glass-fibre tube.

Brain477–570 mm

The RV1126B processor pressed onto the aluminium heat bar, with 1 GB memory, 8 GB storage, the Wi-Fi module, motion sensor, security chip and ID chip.

Battery system24–234 mm

Three 3.4 Ah LiFePO4 cells, their tabs soldered to the flex tail: 13.8 h worst case for a 10-hour day. The base disc holds the charger, cell protection, battery gauge, tamper switch and charge pads.

Side view of the board folded inside the stump, with every key part labelled at its height

Heights are above the grass. Tap a drawing to open it full size.

Every part, both sides

Close-ups of each section with every part named. Side A faces the window and the heat bar; side B faces the back.

Close-ups of the head, call lights, brain and base disc from both sides, every part named
The board unfolded, and opened in KiCadThe board unfolded, both sides

KiCad 10 render, window side, then back side (whole board, left to right: bail fingers, head, camera flex, brain, 454 mm tail, base disc):

KiCad render, window sideKiCad render, back side

Design review: checked before we order

Every board-level parameter the calls depend on, checked against its target with the source of each number. It runs on every change (products/cricketone/hardware/pcb/board-review.mjs). Fixed means the review found a problem and the design changed: a third cell, shock-proof clocks, a latched card socket, the cool-blue finish. To prove means it passes on paper and is measured on the development kit or rev A before boards are ordered.

No check fails13 pass · 5 fixed by the review · 7 to prove before we order · 0 failDesign review of 3 November 2026
AreaCheckOur boardTargetStatus
CameraBall size at the far stumps (20.12 m)
The tracker needs about 3 px of ball to find it; the M8 2.7 mm lens gives 18.1 px per degree.
3.7 px≥ 3 px to trackpasses
CameraMotion smear at 144 km/h, 0.5 ms exposure
A global shutter freezes the whole frame at once; a rolling shutter would skew the ball about 1.4 m.
20 mm≤ 1/3 of the ball (24 mm)passes
CameraFrame rate the calls were simulated at
Accuracy (13 mm worst sure call) was simulated with a 1280 × 400 band at 200 fps. The OV9281 reads 120 fps at full height, so a 400-row window must reach 200 fps: a driver mode to prove on the RV1126B dev kit before rev A. Change: Bring up the 1280 × 400 window at 200 fps on the dev kit (driver register set); fall back to the 180 fps mode, simulated at 15 mm worst sure call.
200 fps (1280 × 400 band) and 180 fpschip: 120 fps at 1280 × 800, 180 fps at VGAto prove on the dev kit
CameraCamera data rate vs the 2-lane link
36 % spare on the link; the RV1126B camera input takes 2.5 Gbps per lane.
1.02 Gbps needed (10-bit, 200 fps band)≤ 1.6 Gbps (2 lanes × 800 Mbps, est.)passes
CameraImage processor load
One 1 MP camera at 120 fps is about a third of what the RV1126B image processor is built for.
123 Mpx/s≤ 360 Mpx/s (12 MP ISP at 30 fps)passes
CameraShutters in step across the six stumps
The six cameras fire staggered from one shared clock; the OV9281 frame-sync input is wired through the camera flex to a timer pin on the RV1126B. Needs a real-time (PREEMPT_RT) kernel, asked for in the quote.
frame-sync line on the camera flex; scheme B clock 0.03 ms p95≤ 0.05 mspasses
CameraAI engine time per frame
The model size is set by the tracking work; its run time on the RV1126B AI engine is measured on the dev kit.
3 TOPS availableball and bat model inside the 5 ms frame at 200 fpsto prove on the dev kit
BatteryRun time, 3 cells, worst case (aged to 80 %, 5 °C)
A 10-hour day (8 h of play, 700 balls, 2 h of warm-up and breaks) uses 14.2 Wh: 2.78 W in each delivery, 1.28 W between balls, 0.56 W in breaks. With the 2.4 Ah A123 cells, two give 6.8 h worst case and three 10.3 h, so the board carries 3 cells (the third bay of the rev S cell carrier) of a 3.4 Ah LiFePO4 26650 (counted at 3.2 Ah). Change: 3 cells in parallel (1S3) of 3.4 Ah: pads BT1–BT4 on the tail; between balls the camera drops to 60 fps and the AI engine sleeps.
13.8 h≥ 10 hfixed by the review
BatteryRun time, 3 cells, new, 20 °C
New cells at room temperature.
20.4 h≥ 10 hpasses
BatterySpare run time, worst case
With the 2.4 Ah A123 cells the spare would be only 3 %, so the production cell is a 3.4 Ah energy cell of the same size. The real draw is measured on the dev kit and rev A.
38 % over 10 h≥ 0 % (more is better)passes
BatteryCharge time
The BQ25070 charges at up to 1 A from the charging case. A completely flat pack takes longer than one night, but one night always puts back a full 10-hour day.
11 h from flat at 1 A; 9 h overnight puts back 100 % of a 10-hour daya full day back overnight (9 h)passes
BatteryPeak current from the cells
Well inside the cells and the converter.
1.23 A at 2.5 V≤ 3 A (buck-boost), ≤ 1 C per cellpasses
BatteryCold and hot charging
LiFePO4 must not be charged below 0 °C.
thermistor on the cells, charging held outside 0–45 °CLiFePO4 rulepasses
HeatProcessor temperature, full sun, 35 °C air
Our board averages 1.42 W, against 2.3 W for the Pi build, and the heat bar carries it to the tube wall (7.5 K/W). The sun, not the chip, heats the stump most.
69 °C≤ 85 °C (RV1126B rated −20 to 85 °C)passes
HeatCell temperature in full sun, 35 °C air
In the plain blue finish the cells reach 60 °C, the limit of the A123 cell. The IR-reflective "cool blue" finish keeps them at 52.5 °C. Change: The stump finish is the IR-reflective cool blue (rev T).
blue 60.5 °C → cool blue 52.5 °C≤ 55 °C (cell rating 60 °C)fixed by the review
HeatCamera temperature in full sun
Inside its rating; image noise rises when hot, so the dark level is measured hot on rev A.
66 °C≤ 85 °C operatingto prove on the dev kit
HeatMetal near the antenna
The aluminium collar sits just above the antenna end, so radio range is measured on rev A.
heat bar ends 5 mm below the antenna; collar 2.6 mm aboveno metal over the antennato prove on the dev kit
ShockPeak shock on the board
Quartz crystals and push-push card sockets are not rated for thousands of g. The clocks are now MEMS oscillators (rated to 50,000 g), and the microSD socket has a hinged, latched lid. Change: MEMS oscillators for Y1, Y2 and Y20; a hinged-lid microSD socket; underfill under the processor, memory, storage and Wi-Fi module; flexible-termination capacitors near the board edges.
4053 g (rigid-body upper bound)every part rated for itfixed by the review
ShockBall-grid chips
Underfill is asked for in the quote; the drop and ball-hit tests on rev A prove it.
underfill under U1, U2, U10, U5no cracked joints after the drop and ball teststo prove on the dev kit
SoftwareStarts from the microSD card on battery power
Owner requirement (4 Nov 2026): put the card in, switch on, and it starts our program.
J1 hinged, latched socket on every board; the maker delivers U-Boot and a device tree that start from the cardno setup on arrivalpasses
SoftwareThe card wins the start
Rockchip's start order is SPI flash, eMMC, card, USB: an image on the eMMC would start before the card. Proved on the dev kit: power on with a card in and the eMMC empty. Change: U10 now ships empty (4 Nov 2026; BOARDS.md said sale boards start from the eMMC).
eMMC U10 shipped empty; our loader reads the card firstthe processor starts from the card every timefixed by the review
SoftwareCard power switch and protection
The card slot is the one place a person touches the electronics, and a card that hangs must be restarted without restarting the board. Change: U12 and U13 added (4 Nov 2026).
U12 switched card supply; U13 ESD on 6 lines, 22 Ω series resistorsa stuck card can be power-cycled; the slot is ESD-safefixed by the review
SoftwareCard grade
Consumer cards wear out and fail hot. Proved on rev A: 1,000 power cuts while writing, and the ball-hit test with the card in.
industrial pSLC microSD (Kingston Industrial or Swissbit S-56u), −40 to 85 °C; read-only system with a separate data partitionrated for heat, cold, shock and writesto prove on the dev kit
SoftwareClip buffer for an unsure call
The last seconds are kept raw in memory and sent compressed (the RV1126B has an H.265 encoder).
4 s of 8-bit 200 fps video in 40 % of the 1 GB memory (the rest: Linux, the program and the AI model)≥ 3 spasses
RadioWi-Fi between the two ends
An easy range for Wi-Fi 5, but the antenna sits inside a tube, close to the metal collar: measured on rev A.
20.12 m, line of sightcertified Wi-Fi 5 module, chip antennato prove on the dev kit

Before we order the first boards, we prove these on the RV1126B development kit or in the rev A tests:

  1. Frame rate the calls were simulated at
  2. AI engine time per frame
  3. Camera temperature in full sun
  4. Metal near the antenna
  5. Ball-grid chips
  6. Card grade
  7. Wi-Fi between the two ends

Where the design stands

  1. 1Shape and placementdone

    Board outline in the stump's shape, all 46 parts placed at their real size, checked in the stump CAD (rev T), KiCad (0 violations) and the design review (no check fails).

  2. 2Schematicnext

    Every wire between the chips, from Rockchip's RV1126B reference design.

  3. 3Footprintsnext

    The exact solder pads under each chip.

  4. 4Routinglater

    Copper tracks on all layers, incl. the processor's memory lines; reviewed by a Rockchip-experienced engineer.

  5. 5Files for the makerlater

    Gerbers with copper, drill, mask, paste, the rigid-flex drawing and the assembly parts list. These go to the board maker.

  6. 6Rev A boardsJan–Feb 2027

    Made, assembled and tested on our fixture.

Design files (rev A-0)

These are the stage 1 files: the board's shape and where every part goes. A board maker needs the stage 5 files to build it.

Who can make it

The board is one rigid-flex piece: an 8-layer HDI brain with ball-grid chips, joined by flex sections. It needs a maker who builds rigid-flex boards and assembles ball-grid chips with X-ray checks. We plan three quotes once the stage 5 files are ready: two from China for price, and one from the US for speed.

MakerWhereRigid-flexHDI / layersAssemblyFit for us
PCBWayfirst quoteShenzhen, Chinayes, 6 to 16 layersHDI up to 32 layers, 0.35 mm BGA pitchyes, turnkey, incl. BGA and X-rayBest fit for rev A: the rigid-flex board and its assembly in one order
NextPCBsecond quoteShenzhen, ChinayesHDI, 6 to 32 layersyes (PCBA)Second quote, to compare price and lead time
ALLPCBChina (a network of partner factories)yes, through partner factories1 to 30 layersyesThird quote; check which factory builds it
JLCPCBShenzhen, Chinano (flex only: 1, 2 and 4 layers, up to 490 mm)HDI added Sep 2026yes, incl. BGANot for the one-piece board today. Good for the test fixture, simple boards, and the fallback of separate rigid boards joined by soldered flex
Sierra Circuits (ProtoExpress)US quoteSunnyvale, California, USAyesyesyes, in-house, incl. BGA; turnkey in about 5 daysUS option: fastest turn and engineering help, at a higher price
CirexxSanta Clara, California, USAyesyesyes (turnkey or consigned)Second US option
Advanced CircuitsAurora, Colorado, USAyescheckyesUS quick-turn; check its rigid-flex assembly
EpecNew Bedford, Massachusetts, USAyescheckyes; also builds battery packsUS option that can also make a certified LiFePO4 pack

From each maker's own capability pages, checked 3 November 2026. Limits change, so each quote confirms them for our exact stack-up.

Full design notes

All 40 parts with their heights, what is on each section, the camera and battery system, heat, the factory test, cost and timeline.