WS2812B only works while I measure the voltage

Question

I'm going crazy, there is clearly something I'm missing here. I have a NodeMCU V3.0 and a 5V LED strip WS2812B with 10 diodes. I tested this first with a 5V 0.3A power supply, only powering 3 diodes, this worked fine.

Than I changed the power supply to a 5V 2A (Measured to 5.2V) and now the colours are way off, they even change everytime its updating, even though nothing is changed. I put my voltmeter across "D3" and "GND" and suddently the LEDs were perfect and stable.

If I remove it again, the LEDs change to random colours again, I tried pulling "D3" to ground with a 10 kOhm resistor, but this doesn't change anything.

What is going on here, why isn't it working, and why does it work with the voltmeter?

Code:

#include <ESP8266WiFi.h>
#include <FastLED.h>

#define LED_PIN     3
#define NUM_LEDS    10

CRGB leds[NUM_LEDS];
int R = 255;
int G = 90;
int B = 10;

//Set up MQTT and WiFi clients
WiFiClient client;

void setup()
{
  FastLED.addLeds<WS2812, LED_PIN, GRB>(leds, NUM_LEDS);
  Serial.begin(9600);

  //Connect to WiFi
  Serial.print("\n\nConnecting Wifi... ");
  WiFi.begin(WIFI_SSID, WIFI_PASS);
  while (WiFi.status() != WL_CONNECTED)
  {
    delay(500);
  }
  Serial.println("OK!");
  pinMode(D4, OUTPUT);
  digitalWrite(D4, HIGH);
}

void loop()
{
    leds[0] = CRGB(255, 0, 0);
    leds[1] = CRGB(0, 255, 0);
    leds[2] = CRGB(0, 0, 255);
    FastLED.show();
    delay(5000);
}

Answer 1

WS2812B LEDs are 5V devices. The NodeMCU is 3V. Your first 5V supply will probably have been somewhat under 5V once the NodeMCU had taken a big chunk of the current (see Voltage Droop). This would mean that the Logic High Threshold would be within the bounds of the 3.3V output by the NodeMCU.

However, with your new high powered power supply that voltage droop will be less prevalent (added to the fact that it's measuring 5.2V), and so the logic high threshold will be higher. At, or very close to, the 3.3V that is output by the NodeMCU. That makes it rather unstable at communicating, since the logic HIGH may not always be above that required HIGH threshold.

Adding your voltmeter obviously changes that logic signal so that it stabilizes above the required threshold (the volt meter input has capacitance. That will help hold the line higher for longer).

To drive 5V devices from a 3.3V MCU you really need to add a logic level shifter, or ensure that the power supply isn't too high.

The WS2812B has a logic HIGH of 0.7 × Vdd. At 5V that's (0.7*5=) 3.5V. That's close, but not quite there. Your original power supply was probably putting out more like 4.75V under load. That would be 3.325V for a logic high - just about manageable by a NodeMCU.

However at 5.2V the logic HIGH will be 3.675V, which is way above the 3.3V that the NodeMCU outputs.

So you either need to:

• Add a 3.3V to 5V logic level shifter, or
• Reduce your LED power input voltage to below 4.75V.

The latter can be done by adding a single series diode with a suitable current handling capacity to the VDD connection. That will (depending on the diode) drop around 0.7V taking you to about 4.5V (the actual drop usually depends on the current to a certain extent). The WS2812B will happily run from 3.5V to 5.3V.