Comparing ADC Performance of Espressif SoCs · Developer Portal

[giscus[bot]]

Comparing ADC Performance of Espressif SoCs · Developer Portal

This article compares the Analog-to-Digital Converter (ADC) performance of different Espressif SoCs and shows the techniques to improve the ADC accuracy using software calibration. The results should help developers choose the most suitable SoC and optimize it for their application's accuracy and voltage range requirements.

https://developer.espressif.com/blog/2025/08/adc-performance/

[felixcollins]

Can you please do a follow up article on the unresolved issues with using the ADC?
For example: espressif/esp-idf#10612

[felixcollins]

Also: espressif/esp-idf#6588

[bubblesnake]

Thank you for follow-up.
#10612: we are working on a fix now.
#6588: we have provided further comment in the ticket.

[MitchBradley]

What were the test conditions for these measurements? In particular, was WiFi active? I have seen situations where, on an ESP32 with WiFi active and a constant DC voltage on the ADC input , the ADC count would suddenly change by about 100 mV, and then, after some time, suddenly revert to the previous value. Sudden large changes are difficult to filter out with common techniques. The hypothesis is that some electrical condition in the radio section is leaking into the ADC electronics.

Please repeat the ADC characterization tests with WiFi active, if you have not already done so.

[bubblesnake]

Thank you for reporting. This is a known issue due to high current during Wi-Fi Tx/Rx, the internal reference voltage may be unstable, or there may be conductive noise. Depending on the use case, if it is low-frequency sample (e.g. NTC temperature measurement, battery voltage monitoring), some outlier filtering can be applied based on the standard deviation of the data distribution. We are trying to develop a filtering algorithm in ADC driver to improve the condition. If you could log an issue in ESP-IDF and provide an example to reproduce spike/high deviation, it would greatly help our optimization.

[MitchBradley]

This problem was reported years ago on forums. It is easy to reproduce; just modify any wifi app to repetitively read the ADC. I would hope that Espressif would be able to create such an app from one of the many example apps. With a constant DC voltage on the ADC input, even with large value filter capacitors, the ADC count experiences large jumps at unpredictable times.

Most conventional filtering algorithms rely on some form of averaging. They will not be effective unless the averaging time is on the order of many seconds, too long for most uses. Even with very long filters, the result will depend on the duty cycle of wifi activity, which can change from minute to minute. One software fix might to couple the ADC driver to the wifi code. If the ADC driver knows when the wifi power is in the state that causes the disturbance, it might be possible to subtract the offset.

Another possibility would be to detect sudden jumps of the right approximate magnitude and apply the offset until the value jumps back. That would only work for input signals that change slowly, but a long averaging filter is even worse in that regard.