Keyboard backlight

Only the backlit keyboard variants (04X0177, 0C43982) have an LED strip. The non-backlit 0C44020 has none — on that keyboard the circuit below is simply left unpopulated and BACKLIGHT_ENABLE = no.

Circuit

Backlight MOSFET switch

A low-side N-MOSFET switch on GP26. The Pico cannot source the strip’s current directly (GPIO limit ~12 mA; the strip is ~100 mA on comparable ThinkPad arrays).

3V3 ──── R_series ──── LED strip anode (+)
                        LED strip cathode (−) ──── Drain
                                                   Source ──── GND
GP26 ──── 10 kΩ ──── Gate
                     Gate ──── 100 kΩ ──── GND

Part	Value	Why
Q1	2N7002 (SOT-23) or BS170 (TO-92)	Logic-level V_GS(th), turns on fully at 3.3 V
R_gate	10 kΩ	Limits the gate-charge current spike on each PWM edge
R_pd	100 kΩ gate → GND	Holds the gate LOW while the Pico boots and GP26 is high-Z: the backlight does not flash at power-on
R_series	100 Ω to start — set by measurement	The strip has its own internal resistors on most ThinkPads; measure before trusting any value
Supply	3V3 from the Pico	If the strip is dim at 3.3 V, move the anode to VBUS (5 V) and re-measure current

Measure before you trust it

Identify the backlight anode/cathode pins on the keyboard FPC (Phase 1).
With the MOSFET not fitted, put a multimeter in series and apply 3.3 V through 100 Ω.
Note the current. Expect tens of mA; delingren measured ~5.6 mA per LED × 18 LEDs ≈ 100 mA on a 04Y0819 with no series resistor.
If it is under ~20 mA and visibly dim, reduce R_series or move to 5 V. If it is over ~150 mA, raise R_series. The USB budget is 500 mA total; keep the backlight under a third of that.
Record the final values in this file.

QMK driver selection

This is where the earlier plan was wrong.

Driver	What it is	RP2040?
`pwm`	Hardware PWM via ChibiOS `PWMD`	Fails to compile on RP2040 — QMK #24470: `unknown type name 'PWMConfig'`, undeclared `PWMD4`
`timer`	Interrupt-driven pin toggling	Documented for AVR and STM32; not the RP2040 workaround
`software`	PWM emulated in the main loop	Works on RP2040 — the workaround recorded in #24470
`custom`	You write it	Unnecessary here

So: BACKLIGHT_DRIVER = software. The cost is some flicker sensitivity if the scan loop stalls; at our ~350 µs scan this is invisible. If a later QMK release fixes RP2040 pwm, switch and record it in CHANGELOG.md.

Reference: QMK backlight docs.

Configuration

/* config.h */
#define BACKLIGHT_PIN     GP26
#define BACKLIGHT_LEVELS  5
/* no BACKLIGHT_BREATHING — unsupported by the software driver */

# rules.mk
BACKLIGHT_ENABLE = yes
BACKLIGHT_DRIVER = software

FN + F11 (CK_BKLT) steps through the five levels. No breathing: backlight_software.c refuses BACKLIGHT_BREATHING at compile time, so it is not defined.

Troubleshooting

Symptom	Cause
Always off	Gate/drain/source mixed up (TO-92 pinout varies by manufacturer — check the datasheet, not the package shape)
Always on, even before firmware	Missing 100 kΩ pull-down
On but no dimming	Driver is not `software`, or `BACKLIGHT_LEVELS` is 1
Dim	R_series too high for 3.3 V — measure, then consider 5 V