ADC

• ESP32 has 18 ADC channels.
• When Wi-Fi is disabled and ADC DMA is used, the theoretical sampling rate does not exceed 2 MHz. In practice, a lower rate is recommended.
• With Wi-Fi running normally, the rate can reach about 1000 samples per second.
• The internal effective ADC resolution is 12 bits.

The ADC input voltage must be greater than 0 V and lower than the upper limit supported by the chip, often up to about 3.3 V depending on the device. A floating input is undefined, so the measurement result is not meaningful.

The ADC input is capacitive, so its effective resistance can be considered very high.

The ESP32 ADC reference voltage is 1100 mV. The ADC range can be extended by configuring internal attenuation. Refer to the ADC chapter of the ESP32 Datasheet. If the required voltage still exceeds the supported range, an external divider should be used.

The theoretical maximum sampling frequency is 2 MHz.

Yes.

The supported frequency range is 611 Hz to 83333 Hz.

No. ESP32 ADC does not support simultaneous sampling of multiple channels. If multiple channels are needed, they must be read in a polling sequence.

By default, the difference between ADC measurements on ESP32 chips is about ±6%. See the ESP32 Datasheet.

A single ADC cannot read multiple ADC channels at exactly the same instant. You can poll the two ADC channels sequentially.

When ADC1 or ADC2 on ESP32-S3 is configured with ADC_ATTEN_DB_12, the measurement range is nominally 0 ~ 3100 mV. However, some chips may measure a maximum voltage lower than 3100 mV. The issue can be addressed in either of the following ways:

• Option 1: avoid operating near the upper boundary. Use an external divider to keep the input voltage closer to the middle of the range for better accuracy and consistency.
• Option 2: use the software ADC range extension solution to extend the maximum measurable voltage to 3300 mV. This solution is already supported for ESP32-S2 and ESP32-S3 in ESP-IDF v4.4.8 and v5.3.1, and can be ported to other ESP-IDF versions if needed.

I (455346) EXAMPLE: ADC2 Channel[1] Raw Data: 4095
I (455346) EXAMPLE: ADC2 Channel[1] Cali Voltage: 4985 mV
I (456346) EXAMPLE: ADC2 Channel[1] Raw Data: 4095
I (456346) EXAMPLE: ADC2 Channel[1] Cali Voltage: 4985 mV

• The raw ADC reading is normal. The converted voltage becomes 5 V because the valid measurable range of ESP32-S3 ADC is 2900 mV. See the effective measurement range corresponding to ESP32-S3 ADC attenuation levels.
• An input voltage above 2900 mV is outside the defined valid range, so this kind of result can occur. If you need to measure voltages above 2900 mV, use a voltage divider or adopt the ESP32-S3 ADC range extension solution.

Wi-Fi also uses ADC2 and has higher access priority. Therefore, while Wi-Fi is active, application calls to adc2_get_raw() may fail. It is recommended to check the return value and retry the measurement if necessary.

ESP32 chips use reference-voltage-based ADC calibration by default at the factory. If better ADC consistency is required, the calibration scheme in eFuse can be changed to the two-point calibration scheme. However, eFuse changes are irreversible and should be made with caution. A software-based two-point calibration solution is also recommended: adc_tp_calibration.

The following example uses battery voltage measurement.

1. First calculate the actual voltage seen at the module ADC pin:

$$ V_{\text{adc}} = 3.7\text{V} \cdot \frac{R_2}{R_1 + R_2} $$

Use the calculated $V_{\text{adc}}$ to look up the ADC error table in the datasheet and obtain the ADC measurement error at that input voltage, denoted as $\mathrm{Atten}(V_{\text{adc}})$.

2. The application usually needs to convert the measured ADC pin voltage back to the battery voltage:

$$ V_{\text{bat}} = \frac{V_{\text{adc}}}{R_2 / (R_1 + R_2)} $$

If $V_{\text{bat}} / V_{\text{adc}} = 4$, the divider ratio is $1/4$, so the ADC error needs to be multiplied by $4$:

$$ \mathrm{Atten} = \mathrm{Atten}(V_{\text{adc}}) \cdot 4 $$

If $V_{\text{bat}} / V_{\text{adc}} = 2$, the divider ratio is $1/2$, so the ADC error needs to be multiplied by $2$:

$$ \mathrm{Atten} = \mathrm{Atten}(V_{\text{adc}}) \cdot 2 $$

Yes, this is normal. The 0 ~ 2500 mV value mentioned in the documentation is the recommended valid measurement range, not a hard upper limit.

In other words, 0 ~ 2500 mV is the range in which Espressif guarantees measurement accuracy after hardware and software calibration. It does not mean the ADC will definitely stop working or immediately saturate above 2500 mV. Beyond that range, the input enters an undefined or accuracy-not-guaranteed region, so variation between chips is expected. For example, some chips may reach nearly 2.9 V, while others may only reach a little over 2.7 V.

If you need to measure 3.3 V, it is recommended to scale the input down to 0 ~ 2.5 V with an external divider, then read it with attenuation and the calibration driver.