[BUG] ESP32-S3 Stuck in adc functions

[CommanderRedYT]

Describe the bug
When trying for example the Inline current sensing test, the ESP32-S3 (esp32-s3-devkitc-1 v1.0), the esp32 is not printing anything. After some debugging, we discovered that it is stuck in adcRead (https://github.com/simplefoc/Arduino-FOC/blob/master/src/current_sense/hardware_specific/esp32/esp32s_adc_driver.cpp). We found that the while loop checking if the adc is ready is causing the problem (with our pin configuration, its the bottom one in line 241):

Arduino-FOC/src/current_sense/hardware_specific/esp32/esp32s_adc_driver.cpp

Lines 237 to 243 in 05954cb

	if(channel > 7){
	while (GET_PERI_REG_MASK(SENS_SAR_MEAS2_CTRL2_REG, SENS_MEAS2_DONE_SAR) == 0); //wait for conversion
	value = GET_PERI_REG_BITS2(SENS_SAR_MEAS2_CTRL2_REG, SENS_MEAS2_DATA_SAR, SENS_MEAS2_DATA_SAR_S);
	} else {
	while (GET_PERI_REG_MASK(SENS_SAR_MEAS1_CTRL2_REG, SENS_MEAS1_DONE_SAR) == 0); //wait for conversion
	value = GET_PERI_REG_BITS2(SENS_SAR_MEAS1_CTRL2_REG, SENS_MEAS1_DATA_SAR, SENS_MEAS1_DATA_SAR_S);
	}

When adding this line of code, I can see "reg: 80160000" inside the serial monitor:

Serial.printf("reg %x\n", READ_PERI_REG(SENS_SAR_MEAS1_CTRL2_REG));

Describe the hardware setup
For us it is very important to know what is the hardware setup you're using in order to be able to help more directly

• Which motor: None (devboard only)
• Which driver: None
• Which microcontroller: Espressif ESP32-S3-DevKitC-1
• Which position sensor: None
• Current sensing used? Yes

IDE you are using

• Visual Studio Code with Platformio
• espressif32@6.3.2 (latest)

[env:esp32-s3-devkitc-1]
platform = espressif32
board = esp32-s3-devkitc-1
framework = arduino
lib_deps = askuric/Simple FOC@2.3.0
lib_archive = false
monitor_speed = 115200
upload_speed = 921600

We tried this with multiple esp32-s3 boards, and did not get it to work. When using this platformio configuration for normal esp32 (not S3!), it works however (same code):

[env:esp32dev]
platform = espressif32@2.0.0
board = esp32dev
framework = arduino
lib_deps = askuric/Simple FOC@2.2.0
lib_archive = false
monitor_speed = 115200
upload_speed = 921600

[runger1101001]

Have you tried with the dev branch code of SimpleFOC and the most recent version of the espressiv32 framework?

[mailonghua]

Has this issue been resolved? I also encountered the same problem

[CommanderRedYT]

@mailonghua We still have to test it, haven't had time yet. I will post updates when we find out more

[mailonghua]

@mailonghua We still have to test it, haven't had time yet. I will post updates when we find out more

I think I have located this problem. This is a bug. Each ADC of esp32s3 has 10 channels 0~9, and the SENS_SAR_MEAS1_CTRL2_REG register shows the status of ADC1, and SENS_SAR_MEAS2_CTRL2_REG shows the status of ADC2. But when I use GPIO9, currently The judgment detects BIT16 of SENS_SAR_MEAS2_CTRL2_REG, which belongs to ADC2. When I change if(channel > 7) to if(channel > 9), my current program is normal.
But I think this method may not be appropriate for ADC2

[mailonghua]

@mailonghua We still have to test it, haven't had time yet. I will post updates when we find out more

I think I have located this problem. This is a bug. Each ADC of esp32s3 has 10 channels 0~9, and the SENS_SAR_MEAS1_CTRL2_REG register shows the status of ADC1, and SENS_SAR_MEAS2_CTRL2_REG shows the status of ADC2. But when I use GPIO9, currently The judgment detects BIT16 of SENS_SAR_MEAS2_CTRL2_REG, which belongs to ADC2. When I change if(channel > 7) to if(channel > 9), my current program is normal. But I think this method may not be appropriate for ADC2

Sorry, I may have forgotten to note a setting. In addition to the above modifications, you also need to read it before currency.init when using the analogread() interface. The current sensor will be initialized normally. The specific reason has not been determined yet, but I guess it may be Related ADC enable issues

[CommanderRedYT]

@mailonghua We still have to test it, haven't had time yet. I will post updates when we find out more

I think I have located this problem. This is a bug. Each ADC of esp32s3 has 10 channels 0~9, and the SENS_SAR_MEAS1_CTRL2_REG register shows the status of ADC1, and SENS_SAR_MEAS2_CTRL2_REG shows the status of ADC2. But when I use GPIO9, currently The judgment detects BIT16 of SENS_SAR_MEAS2_CTRL2_REG, which belongs to ADC2. When I change if(channel > 7) to if(channel > 9), my current program is normal. But I think this method may not be appropriate for ADC2

Sorry, I may have forgotten to note a setting. In addition to the above modifications, you also need to read it before currency.init when using the analogread() interface. The current sensor will be initialized normally. The specific reason has not been determined yet, but I guess it may be Related ADC enable issues

What exactly do you mean by currency.init? Also, isn't analogread Arduino.h?

[mailonghua]

@mailonghua We still have to test it, haven't had time yet. I will post updates when we find out more

I think I have located this problem. This is a bug. Each ADC of esp32s3 has 10 channels 0~9, and the SENS_SAR_MEAS1_CTRL2_REG register shows the status of ADC1, and SENS_SAR_MEAS2_CTRL2_REG shows the status of ADC2. But when I use GPIO9, currently The judgment detects BIT16 of SENS_SAR_MEAS2_CTRL2_REG, which belongs to ADC2. When I change if(channel > 7) to if(channel > 9), my current program is normal. But I think this method may not be appropriate for ADC2

Sorry, I may have forgotten to note a setting. In addition to the above modifications, you also need to read it before currency.init when using the analogread() interface. The current sensor will be initialized normally. The specific reason has not been determined yet, but I guess it may be Related ADC enable issues

What exactly do you mean by currency.init? Also, isn't analogread Arduino.h?

My phenomenon is that it is blocked during the initialization process of the InlineCurrentSense object, that is, current_sense.init();, which is ultimately blocked to the register location you mentioned above waiting for the ADC to complete.
I am currently using the PlatformIO platform. In addition to the modifications I mentioned above, I use the ardunio interface analogread() to read the two pins of the current detection before initializing the current sensor. Then the blockage of the current sensor initialization will no longer exist.

[CommanderRedYT]

Just as FYI, we still did not have time to test it, but we might be able to do it until new year. (Pending)

[runger1101001]

We have just merged this PR: #346 to the dev branch.

It is related to problems with ADC2. Perhaps it also solves your issue?

[CommanderRedYT]

We have just merged this PR: #346 to the dev branch.

It is related to problems with ADC2. Perhaps it also solves your issue?

Is possible, yes. I will try it out in the next few days and then report here if everything works

[SwapnilPande]

We have just merged this PR: #346 to the dev branch.

It is related to problems with ADC2. Perhaps it also solves your issue?

I've been following this issue on the ESP32-S3 too and trying to track it down. I mirrored the changes from this PR into the ESP32s driver with no success.

Digging a little deeper, it seems that the driver is not actually fully configuring the ADC correctly. Unless I'm misunderstanding, it looks like the driver for the ESP32-S* microcontrollers comments out the configuration for the ADC to avoid compilation errors (contributed in #198). It looks like these functions need to updated to match the new registers.

The code seems to be getting stuck waiting for an ADC conversion here. This seems consistent with incomplete ADC configuration. I'm going to try to track down a fix for this over the next few weeks, but help would be greatly appreciated. I'm not very familiar with the S3!