Please make E.getAnalogVRef() for esp32!

[Quishish]

I am doing my school project on the esp 32 with espruino on it. Love espruino and don't want to change it, but my project depends on analog sensors and I need E.getAnalogVRef() very much!

[fanoush]

If you need it then adding it youself could be nice part of your project :-)

Anyway, why do you need it? Maybe the voltage is fixed and known in your situation so you can use some constant like 3.3 instead?

[gfwilliams]

I'd second this - every ESP32 I've seen has a 3.3v regulator on it, so unless you're hoping to see if it ever drops (eg due to a flat battery) it may not be that useful.

Or were you after something else? Normal analog inputs work fine on ESP32.

[fanoush]

I was checking more about analog on ESP32 and here are some links
https://docs.espressif.com/projects/esp-idf/en/v4.4/esp32/api-reference/peripherals/adc.html#adc-calibration
https://www.reddit.com/r/esp32/comments/j2fodo/what_are_the_reasons_why_the_internal_adc_of/
https://www.esp32.com/viewtopic.php?t=4354
https://github.com/G6EJD/ESP32-ADC-Accuracy-Improvement-function/blob/master/ESP32_ADC_Read_Voltage_Accurate.ino

so looks like

• there is some nonlinearity so in some areas it is not very accurate
• the internal reference voltage is about 1.1V but varies and with most ESP32 chips the calibrated value is stored in efuses
• there is esp_adc_cal_raw_to_voltage that tries to convert value to voltage according to those calibrated values stored

Espruino does not use esp_adc_cal_raw_to_voltage, it just reads 12bit value and divides by 4096

I am not sure how useful is that but as per those links above I tried to add returning the efuse value via E.getAnalogVRef() here
espruino/Espruino@master...fanoush:Espruino:f-esp32-analogvref
but did not test it. if it returns nonzero value then it is the internal calibrated value. Anyway it is easy to fork it on github and the ESP32 build with your changes is automatically made for you so you can get this build from ESP32 artifact downloadable on the bottom of this page https://github.com/fanoush/Espruino/actions/runs/13456269481
or you can add your own esp_adc_cal stuff there and get your own build.

And btw there is also
https://docs.espressif.com/projects/esp-idf/en/v4.4/esp32/api-reference/peripherals/adc.html#_CPPv416adc_vref_to_gpio10adc_unit_t10gpio_num_t
which brings internal voltage reference to some analog pin for your own measurement/calibration

[fanoush]

OK so had a bug there with float conversion and the value was truncated, here it is fixed https://github.com/fanoush/Espruino/actions/runs/13472822941
Tested it on two boards, first old one has nothing stored in efuses so there it returns 0 but on the other one it has the value stored

> espefuse.exe -p COM4 adc_info
espefuse.py v4.8.1
Connecting.........
Detecting chip type... ESP32

=== Run "adc_info" command ===
ADC VRef calibration: None (1100mV nominal)

> espefuse.exe -p COM6 adc_info
espefuse.py v4.8.1
Connecting.........
Detecting chip type... ESP32

=== Run "adc_info" command ===
ADC VRef calibration: 1114mV

and now E.getAnalogVRef() gets it as 1.114. But still it is not very useful since the analogRead uses ADC_ATTEN_DB_11 to scale the range to 3.3V. So maybe the result could be scaled by 1.1/1.114 but not sure it makes sense.

What feels a bit more practical to me is to

• expose the adc_vref_to_gpio so one could measure the real vref
• using the esp_adc_cal_raw_to_voltage to convert to milivolts by using the calibration data because as I understood it the single point vref calibration is not the only way and other esp32 chips may use two point calibration data so there the vref value is zero again. And looks like there are additional lookup tables to correct the nonlinearity Inconsistent values when using "analogRead()" espressif/arduino-esp32#92 (comment)
  https://github.com/espressif/esp-idf/blob/master/components/esp_adc/esp32/adc_cali_line_fitting.c#L89