diff --git a/frontend/src/__tests__/sensor-parts.test.ts b/frontend/src/__tests__/sensor-parts.test.ts
index f5369c24..25387add 100644
--- a/frontend/src/__tests__/sensor-parts.test.ts
+++ b/frontend/src/__tests__/sensor-parts.test.ts
@@ -152,6 +152,43 @@ describe('ntc-temperature-sensor — attachEvents', () => {
     // ADC should remain zeroed
     expect(adc.channelValues[0]).toBe(0);
   });
+
+  /**
+   * Issue #233. The divider used to be computed against a hard-coded 5 V rail,
+   * so on a 3.3 V board it handed the ADC a voltage above full scale: 25 C came
+   * back as ~1.4 C and everything below ~11 C saturated at the same reading.
+   * The rail now follows the board, so the midpoint of a 10k/10k divider is
+   * half of THAT rail.
+   */
+  it('scales the divider to a 3.3V board instead of assuming 5V', () => {
+    const logic = PartSimulationRegistry.get('ntc-temperature-sensor')!;
+    const injected: number[] = [];
+    // A simulator exposing setAdcVoltage is the ESP32 bridge shim.
+    const sim = {
+      ...makeSimulator(makeADC()),
+      setAdcVoltage: (_pin: number, volts: number) => {
+        injected.push(volts);
+        return true;
+      },
+    };
+
+    logic.attachEvents!(makeElement(), sim as any, pinMap({ OUT: 34 }));
+
+    expect(injected).toHaveLength(1);
+    // R_ntc(25 C) == R_pull == 10k, so OUT sits at exactly half the rail.
+    expect(injected[0]).toBeCloseTo(1.65, 2);
+    expect(injected[0]).toBeLessThan(3.3); // never above the board's full scale
+  });
+
+  it('keeps the 5V midpoint on an AVR board', () => {
+    const logic = PartSimulationRegistry.get('ntc-temperature-sensor')!;
+    const adc = makeADC();
+    const sim = makeSimulator(adc);
+
+    logic.attachEvents!(makeElement(), sim as any, pinMap({ OUT: 14 }));
+
+    expect(adc.channelValues[0]).toBeCloseTo(2.5, 2);
+  });
 });
 
 // ─── Gas Sensor ──────────────────────────────────────────────────────────────
diff --git a/frontend/src/data/examples.ts b/frontend/src/data/examples.ts
index 2ed10293..7d2c2612 100644
--- a/frontend/src/data/examples.ts
+++ b/frontend/src/data/examples.ts
@@ -7921,13 +7921,11 @@ const float NOM_TEMP_K  = 298.15;   // 25 °C in Kelvin
 float readTempC() {
   int   raw  = analogRead(NTC_PIN);
   float v    = raw * (VCC / 1023.0);
-  // Voltage divider topology used by standard NTC modules and by Velxio's
-  // wokwi-ntc-temperature-sensor: VCC → R_NTC → A1 → R_pull (10k) → GND.
-  // Higher temperature → R_NTC drops → V rises, so the NTC resistance is
-  //   r = R_pull * (VCC - v) / v
-  // Using the inverted form (r = R_pull * v / (VCC - v)) gives the wrong
-  // sign and Steinhart-Hart returns negative temperatures for hot inputs.
-  float r    = SERIES_R * (VCC - v) / v;
+  // Divider modelled by Velxio's ntc-temperature-sensor:
+  //   VCC → R_pull (10k) → A1 → R_NTC → GND.
+  // Higher temperature → R_NTC drops → V FALLS, so
+  //   r = R_pull * v / (VCC - v)
+  float r    = SERIES_R * v / (VCC - v);
   // Steinhart–Hart simplified equation
   float st   = log(r / NOM_R) / B_COEFF + 1.0 / NOM_TEMP_K;
   return (1.0 / st) - 273.15;
@@ -8256,8 +8254,8 @@ const float NOM_TEMP_K = 298.15; // 25 °C
 float readTempC() {
   int   raw = analogRead(NTC_PIN);
   float v   = raw * (VCC / 1023.0);
-  // VCC → R_NTC → A1 → R_pull → GND  (standard NTC module topology)
-  float r   = SERIES_R * (VCC - v) / v;
+  // VCC → R_pull → A1 → R_NTC → GND (the divider Velxio models)
+  float r   = SERIES_R * v / (VCC - v);
   float st  = log(r / NOM_R) / B_COEFF + 1.0 / NOM_TEMP_K;
   return (1.0 / st) - 273.15;
 }
@@ -9954,7 +9952,9 @@ const float B_COEFF    = 3950.0;
 
 float readTemperature() {
   int raw = analogRead(NTC_PIN);
-  float resistance = SERIES_R * (1023.0 / raw - 1.0);
+  // R_NTC sits on the LOW side, so raw/1023 = R_NTC / (R_NTC + SERIES_R):
+  //   R_NTC = SERIES_R / (1023/raw - 1)
+  float resistance = SERIES_R / (1023.0 / raw - 1.0);
   float steinhart = resistance / NOM_R;
   steinhart = log(steinhart) / B_COEFF;
   steinhart += 1.0 / (NOM_TEMP + 273.15);
diff --git a/frontend/src/simulation/parts/SensorParts.ts b/frontend/src/simulation/parts/SensorParts.ts
index a1bb7da1..758f851c 100644
--- a/frontend/src/simulation/parts/SensorParts.ts
+++ b/frontend/src/simulation/parts/SensorParts.ts
@@ -18,7 +18,7 @@
  */
 
 import { PartSimulationRegistry } from './PartSimulationRegistry';
-import { setAdcVoltage, emitPropertyChange } from './partUtils';
+import { setAdcVoltage, emitPropertyChange, analogRailVolts } from './partUtils';
 import { registerSensorUpdate, unregisterSensorUpdate } from '../SensorUpdateRegistry';
 
 // ─── Tilt Switch ─────────────────────────────────────────────────────────────
@@ -64,11 +64,17 @@ PartSimulationRegistry.register('tilt-switch', {
  *
  * The injected OUT voltage uses the SAME β-model voltage divider the circuit
  * (and the example sketch) assume: a 10k pull-up from VCC to OUT and the NTC
- * from OUT to GND, so V_OUT = 5 · R_ntc / (R_ntc + R_pull) with
+ * from OUT to GND, so V_OUT = VCC · R_ntc / (R_ntc + R_pull) with
  * R_ntc(T) = R0 · exp(β · (1/T − 1/T0)).  This makes the injected ADC voltage
  * decode straight back to the slider value (set 50°C → read 50°C).  When the
  * electrical (SPICE) engine is active it drives A0 from the ngspice solve
  * instead, using the matching topology in componentToSpice.ts — both agree.
+ *
+ * VCC is the rail of the board the sensor is wired to, not a constant 5. The
+ * divider used to assume 5 V everywhere, which on a 3.3 V board put the node
+ * above the ADC's full scale: the slider read 1.4°C where it said 25, and
+ * everything below ~11°C pinned at the same saturated value (issue #233). The
+ * SPICE path already used the board's rail, so the two only agreed on AVR.
  */
 PartSimulationRegistry.register('ntc-temperature-sensor', {
   attachEvents: (_element, simulator, getArduinoPinHelper, componentId) => {
@@ -77,9 +83,10 @@ PartSimulationRegistry.register('ntc-temperature-sensor', {
     const NTC_R0 = 10_000;
     const NTC_BETA = 3950;
     const R_PULL = 10_000;
+    const vcc = analogRailVolts(simulator);
     const tempToVolts = (temp: number) => {
       const rNtc = NTC_R0 * Math.exp(NTC_BETA * (1 / (temp + 273.15) - 1 / 298.15));
-      return Math.max(0, Math.min(5, 5 * (rNtc / (rNtc + R_PULL))));
+      return Math.max(0, Math.min(vcc, vcc * (rNtc / (rNtc + R_PULL))));
     };
 
     // Room temperature default
diff --git a/frontend/src/simulation/parts/partUtils.ts b/frontend/src/simulation/parts/partUtils.ts
index 3dce0826..d22bab7a 100644
--- a/frontend/src/simulation/parts/partUtils.ts
+++ b/frontend/src/simulation/parts/partUtils.ts
@@ -36,6 +36,25 @@ export function getADC(avrSimulator: AnySimulator): any | null {
   return (avrSimulator as any).getADC?.() ?? null;
 }
 
+/**
+ * Supply rail the analog front end of THIS board runs on, in volts.
+ *
+ * A part that builds a resistive divider has to use the rail it is actually
+ * wired to. Assuming 5 V everywhere is not a small error on a 3.3 V board: it
+ * pushes the computed node voltage above the ADC's full scale, so the reading
+ * saturates and a whole stretch of the control slider reports the same value.
+ * (An NTC divider assumed at 5 V reads 1.4 C on an ESP32 when the slider says
+ * 25 C, and pins at full scale for everything below ~11 C — issue #233.)
+ *
+ * The split matches setAdcVoltage's own branches: AVR converts against a 5 V
+ * reference, RP2040 and the ESP32 bridge against 3.3 V.
+ */
+export function analogRailVolts(simulator: AnySimulator): number {
+  if (typeof (simulator as any).setAdcVoltage === 'function') return 3.3; // ESP32 bridge
+  if (simulator instanceof RP2040Simulator) return 3.3;
+  return 5.0; // AVR
+}
+
 /**
  * Write an analog voltage to an ADC channel, supporting AVR, RP2040, and ESP32.
  *