Lattepanda Sharp IR Collision Jitter

Yes - your understanding of ARC’s ADC scaling is correct:

• ADC ports in ARC read 0-255 on the Leonardo/ARC connection
• That corresponds to roughly 0-5V
• Your GP2Y0A21YK0F readings sound reasonable for a Sharp IR sensor
• The noise you’re seeing is also very common with Sharp IR sensors, especially in sunlight and on darker/less reflective surfaces

What’s happening

The Sharp IR Collision skill is designed as a simple threshold-based stopper. It does not currently provide configurable:

• moving average
• hysteresis
• N-of-M confirmation
• debounce delay

So if the sensor value is bouncing around the minimum distance threshold, the Movement Panel can stop/start/stop unexpectedly.

Recommended approach

For this sensor, I would not rely on the built-in Collision skill alone if you need stable behavior. The best practice is to:

1. Read the ADC in a script
2. Apply smoothing / confirmation
3. Stop the Movement Panel only after several consecutive "hits"
4. Optionally add hysteresis so it doesn’t re-arm until the reading rises well above the stop threshold

That will be much more reliable than a single raw sample.

A better JavaScript approach

Your example is close in spirit, but the loop needs a small fix and a little more structure.

Here’s a simple N-of-M confirmation example using a rolling average:

var adcPort = ADC0;
var stopThreshold = 100;      // stop if average falls below this
var rearmThreshold = 115;      // must rise above this before re-arming
var sampleCount = 6;
var hitCountRequired = 4;

var armed = true;

while (true) {
  var sum = 0;

  for (var i = 0; i < sampleCount; i++) {
    sum += ADC.get(adcPort);
    sleep(20);
  }

  var avg = sum / sampleCount;
  print("ADC avg = " + avg);

  if (armed && avg <= stopThreshold) {
    Movement.stop();
    armed = false;
    print("STOP triggered");
  }

  if (!armed && avg >= rearmThreshold) {
    armed = true;
    print("Re-armed");
  }

  sleep(30);
}

Why this helps

• Average reduces noise
• Hysteresis prevents rapid retriggering near the boundary
• The robot won’t keep stopping from a single bad reading

If you want, I can also give you a version that uses a true N-of-M vote instead of averaging.

About your current threshold

Your threshold selection sounds in the right range, but for a Sharp sensor you should really calibrate using the exact mounted robot geometry:

• measure the sensor value at your desired stop distance
• then set the threshold a little before that
• add hysteresis of about 10-20 ADC counts to start

Because Sharp sensors are nonlinear, the spacing between values changes with distance, so a threshold that works at 20 cm may not behave the same at 10 cm or 40 cm.

Why darker objects cause late stops

That’s also expected:

• Sharp IR sensors depend on reflected IR light
• black/dark/angled surfaces reflect less
• sunlight can swamp the sensor and reduce reliability

So late stops on dark objects are a sensor limitation more than an ARC issue.

One important note about your snippet

This line:

for(var i=0;i= 6) {

should be something like:

for (var i = 0; i < 6; i++) {

Also, if you are using a script to control stopping, you’ll usually get more predictable results by calling the movement command directly from the script rather than relying on the Sharp Collision skill at the same time.

Best options, from most reliable to least

1. Scripted smoothing + hysteresis
2. Scripted N-of-M confirmation
3. Built-in Sharp IR Collision skill alone

Helpful docs

• Sharp IR Collision skill:
  https://synthiam.com/Support/Skills/Infrared-Distance/Sharp-IR-Collision?id=16122
• ADC Graph skill for visual tuning:
  https://synthiam.com/Support/Skills/Adc/ADC-Graph?id=16091

If you want, I can write you a complete ARC JavaScript monitor that:

• samples ADC0,
• filters noise,
• stops the movement panel,
• and re-arms with hysteresis

for your LattePanda Leonardo setup.