ESP32-S3: Interpretation of the ADC characteristics and calibration results
Posted: Thu Aug 28, 2025 7:46 am
Dear community,
The esp32-s3 adc's INL is specified as ±8 LSB, see below:
Is that specification based on the best straight line method or on the end point method?
I've applied a voltage ramp from 0 to 1.1V(=Vref), with 10ms stabilisation time foreach step and 10 samples for averaging, to the adc. The attenuation was set to 0 dB and I added a 100nF bypass-capacitor on the input. The INL I get, by using the end point method, is far worse... about 100 LSB worst case. Since the specified INL is symmetrical and my results are far worse, I'm guessing you've use the best straight line method...?
Furthermore, how do the INL and DNL specification harmonize with the ADC calibration results, see below:
The total error at 0dB attenuation is ±5mV. Considering that 1 LSB = 1.1V/4096 = 0.27mV, the total error = 5mV/0.27mV ≈ 19 LSB. The total error is the sum of all errors, isn't it ? So it should be:
total error = INL + DNL + OTHERS, where OTHERS = 19 - 8 - 4 = 7 LSB.
Where are these 7 LSB coming from ? The attenuation is set to 0 dB, so it should not influence the signal integrity. Is it coming from the limited resolution of the calibration calculations ?
I'd appreciate some feedbacks. And if I were to lack some understanding of the specifications, I'd be keen for some references I could read to educate myself deeper.
Best regards,
Helena
The esp32-s3 adc's INL is specified as ±8 LSB, see below:
- INL.png (20.45 KiB) Viewed 464 times
I've applied a voltage ramp from 0 to 1.1V(=Vref), with 10ms stabilisation time foreach step and 10 samples for averaging, to the adc. The attenuation was set to 0 dB and I added a 100nF bypass-capacitor on the input. The INL I get, by using the end point method, is far worse... about 100 LSB worst case. Since the specified INL is symmetrical and my results are far worse, I'm guessing you've use the best straight line method...?
Furthermore, how do the INL and DNL specification harmonize with the ADC calibration results, see below:
- cali.png (22.54 KiB) Viewed 464 times
total error = INL + DNL + OTHERS, where OTHERS = 19 - 8 - 4 = 7 LSB.
Where are these 7 LSB coming from ? The attenuation is set to 0 dB, so it should not influence the signal integrity. Is it coming from the limited resolution of the calibration calculations ?
I'd appreciate some feedbacks. And if I were to lack some understanding of the specifications, I'd be keen for some references I could read to educate myself deeper.
Best regards,
Helena