Powering the Strip

WS2812-class LEDs draw about 60 mA each at full white, so a long strip at full brightness is an amp-scale load. The ESP32 devkit adds roughly 250 mA with WiFi up and peaks near 500 mA on transmit.

What EdS actually draws

Brightness is applied before the gamma 2.6 table, which means current falls away far faster than the brightness figure suggests. Measured against the firmware’s own lookup table:

The first row is so small because the idle scene’s 0.15 brightness on (255, 170, 90) becomes PWM duties of 2, 1 and 0 out of 255. Almost all of that current is the controller ICs idling. It is what a stand is sent while the turntable is on and silent. At power on the strip is dark (ADR-0029), and a stand that cannot reach the broker lights one LED, red, in blinks.

Bench testing on USB

16 LEDs is a comfortable working figure: even at the breathe peak that is ~164 mA of strip and ~420 mA total.

8 LEDs is unconditionally safe — worst case is 488 mA of strip even if a scene arrives at full white and brightness 1.0, so nothing published can overload the port.

Past ~30 LEDs, use an external 5 V supply with its ground tied to the ESP32.

The host’s USB port is rarely the fragile part — it current-limits and disables on overload. The devkit is: powering a strip from its 5 V pin routes all of that current through the USB connector and a small onboard diode often rated around 500 mA to 1 A.

Wiring

One 5 V supply runs both the strip and the board:

5 V supply (+) ──┬── strip +5V
                 └──▶|── ESP32 5V / VIN pin
                    1N5817, stripe towards the board

5 V supply (−) ──┬── strip GND
                 └── ESP32 GND pin

ESP32 GPIO 18 ──[330–470 Ω]── strip DIN

Data is GPIO 18. That is LP_LED_GPIO, whose default is 18; it is changed in menuconfig under Strip → Data GPIO. It is the only GPIO the strip needs: a WS2812 has no clock line.

Branch at the supply or at the strip’s input, so the strip’s current does not pass through the board. The board takes its 5 V on the pin labeled 5V, VIN or VBUS, which feeds the onboard regulator. Never the 3V3 pin.

For the stand’s 18 LEDs the worst case is 18 × 61 mA, about 1.1 A, plus the board: a 5 V 2 A supply covers it with room to spare.

The diode is for reflashing in place. With USB plugged in and the supply off, the strip would otherwise draw from the USB port, backward through the board’s 5 V pin. The diode lets current into the board and none out, so USB powers the board alone. Use a Schottky (1N5817 to 1N5819): it drops about 0.3 V, which still leaves the regulator enough. A 1N4001 drops too much.

It does not protect the other direction. What keeps the supply out of the USB port is the devkit’s own diode between the port and the 5 V pin, which most have. On a board without one, lift the wire from the 5 V pin before plugging USB in and leave ground attached.

• 330–470 Ω in series on the data line, at the ESP32 end.
• 1000 µF across 5 V and ground at the strip input, to absorb the inrush when many LEDs switch at once.
• Common ground is mandatory when using an external supply.
• The ESP32 drives 3.3 V data into 5 V-logic WS2812s. This usually works and is technically out of spec; flicker on the first pixel is the symptom.

Limiting count in firmware

LP_LED_COUNT is 18, the number that fit across the front of the stand. It is compile-time — main.c declares static lp_rgb_t frame[LP_LED_COUNT] — so changing it needs menuconfig, a rebuild and a reflash.

A strip longer than the configured count is safe and useful for bench work: the driver only ever writes max_leds pixels, so a 60-LED strip configured as 16 lights the first 16 and leaves the rest dark and idle.