Rendered from electronics/ELECTRONICS.md + docs/power-solar.md + docs/BOM.csv in the repository at build time. Estimates are labelled as such in the source; nothing on this page is a measured result unless it says so.

Versions A, B and C

All three share the heads, the XIAO ESP32C6 and the two ULN2003A drivers; they differ in how the controller box is powered. Compare them in the 3D model (switch A/B/C there).

Common to all versions: Seeed Studio XIAO ESP32C6 (pre-soldered headers); ULN2003A 7-channel Darlington array DIP-16; Electrolytic capacitor 470 uF 16 V 8 x 9 mm; Tact switch 6 x 6 x 4.3 mm; Perfboard FR2 70 x 50 mm; JST-PH 5-pin housing; JST-PH 5-pin header straight (top entry); JST-PH crimp contacts AWG28-22; 6-wire flat modular cable white 6.8 x 2.4 mm; 3M double-sided acrylic foam tape 12 mm; Printed cable clip (hardware/cad/cable_clip.scad); 28BYJ-48 5 V unipolar geared stepper (stock 5-wire); NBR O-ring 8 x 2 mm (70 Shore A); USB-C 5 V 3 A power supply with captive cable (white); Matt white PETG (close to RAL 9010) or ASA; TPU 95A.

From the electronics doc

0. Cost-down default build (round 5, 2026-10-17): 2x ULN2003A, versions A / B / C#

This section is the default build from round 5 on. §1-§2 and §4-§5 describe the TMC2209 build, kept as the quiet option (StealthChop, current limit, bipolar-modded motors). §10 is the old LiFePO4/BQ25570 study. Values are typical or estimated unless a source is given. Verify every board pinout against your parts.

Common to all three: Seeed XIAO ESP32C6, 2x ULN2003A (DIP-16, one per motor) soldered on the perfboard, stock 5-wire 28BYJ-48 (no bipolar trace cut), one 5-pin JST-PH header per head, 6-wire flat cable to the head.

Pin maps, wiring and power paths for each version →

Energy, from the power budget (estimates)

C1. Move energy and daily load (2000 mAh LiPo, Thread sleepy device 150 uA)#

The ULN build roughly halves battery-only life (version B). The cause is the 4x longer move at 3 rpm at full coil current, and the daily re-home at 2 rpm (about 30 % of the daily load is sleep, the rest is motion). If the coils really are 200 ohm (the motor listing's value) the motor power is about 0.16 W and the 2 x 4 load is about 26.0 mWh/day -> about 209 days. Measure one coil before trusting either figure. Levers if B is too short: re-home weekly instead of daily (saves about 12 mWh/day for 2 blinds), fewer moves, or the quiet TMC2209 option.

C2. Solar balance, version C (panel inter-pane on the outer pane = placement B, 2 x 4 moves, 47.6 mWh/day)#

December mean harvest into the cell minus load, and the lowest charge over 2015-2023 (PVGIS hourly):

• v0.3 (P124 + TMC2209) for comparison: South +248 mWh/day / 97 %, North 86 % unshaded.
• At 2 x 8 moves (66.9 mWh/day) half-shaded North runs empty about 28 days a year; East/West stay above 12 %.
• Room-side panel placement (behind the slats) is much weaker: South 77 %, East 29 %, West 23 %, North empty about 25 days/yr at 2 x 4 moves. Keep the panel between the panes.
• June surplus is large everywhere (about +0.8 to +1.7 Wh/day), so the cell refills over the summer; the risk is the December-January deficit on East/West/North windows.
• Cold: the charger only charges above 0 C (NTC on TS after the 6106 rework, ELECTRONICS.md §0.4). The cell stays in the room-side box, so the 0 C gate does not block winter charging.

Plot: docs/power_budget.png (v0.4). The 0.75x panel factor is the weakest number here: log one week of charge current on a south window before relying on the December figures.

C3. Fit (from the CAD echoes)#

Heights inside the box use two ASSUMED component heights (terminal block 8.5 mm, mated JST-PH + wire bend 10 mm): tb_h and jst_h in ebox.scad.

Full power budget →