ESP32-S3 : Problem with ADC in continuous mode

[ThomasESP32]

Good morning,

I am trying to use the ADC in continuous mode on an Esp32-S3 but I get a problem.
Can somebody help me please ?

Here is the explanation :

I am using an Esp32-S3-DevKitC-1 board for the development.
My goal is to sample an analog audio signal using the ADC1_7 pin of the board.
I would like to use a sampling frequency of 64000Hz and to record 4096 bytes (1024 32 bits samples) each time the ADC is started.

Before that, I just want to test the behaviour of the ADC with one 32 bits sample (4 bytes) recorded every 5 seconds.
So, I connect a Power Supply on the ADC1_7 so I canchange the value manualy and every 5 seconds, the ADC records one sample.

At the beginning, the Power Supply is set to 0v. So the ADC gives me 0v => Here everything is OK.
Unfortunately, after the first sample is recorded and displayed, if I set the Power Supply to 1,3V immediately,
the next sample is still 0v ( It should be 1,3V because I see using an oscilloscope that the value has changed), but the one after is 1,3V.

Maybe something is not configured correctly and I lose one sample.
Here is the configuration :

/* On efface le buffer qui contiendra les échantillons : */
memset(ptrSamples, 0x00, NBR_SAMPLES);

/* On initialise l'ADC : */
AdcHandle = NULL;
adc_continuous_handle_cfg_t AdcHandleConfig = {
.max_store_buf_size = NBR_SAMPLES,
.conv_frame_size = NBR_SAMPLES,
};

Err = adc_continuous_new_handle(&AdcHandleConfig, &AdcHandle);
if (Err != ESP_OK)
ESP_LOGI(TAG, "=====> AdMeasure : Adc Continuous Init Failed.");
else
{
/* On configure l'ADC : */
adc_continuous_config_t AdcConfig = { // @suppress("Invalid arguments")
.sample_freq_hz = 64000,
.conv_mode = ADC_CONV_SINGLE_UNIT_1, // On utilise uniquement l'ADC1.
.format = ADC_DIGI_OUTPUT_FORMAT_TYPE2, // On utilise le type 2. Pris dans l'exemple.
};

adc_digi_pattern_config_t adc_pattern[SOC_ADC_PATT_LEN_MAX] = {0};
AdcConfig.pattern_num = NBR_CHANNELS;
for (int ii = 0; ii < NBR_CHANNELS; ii++)
{
uint8_t unit = ADC_UNIT_1;
uint8_t ch = channel[ii] & 0x7;
adc_pattern[ii].atten = ADC_ATTEN_DB_11;
adc_pattern[ii].channel = ch;
adc_pattern[ii].unit = unit;
adc_pattern[ii].bit_width = SOC_ADC_DIGI_MAX_BITWIDTH;
}
AdcConfig.adc_pattern = adc_pattern;
Err = adc_continuous_config(AdcHandle, &AdcConfig);
if (Err != ESP_OK)
ESP_LOGI(TAG, "=====> AdMeasure : Adc Continuous Configuration Failed.");
else
{
/* Mise en place de la calibration permettant d'avoir les résultats en mV directement :
* On utilise le schéma de calibration de type Courbe (Le seul proposé par la puce). */
adc_cali_curve_fitting_config_t Curve_Fitting;
Curve_Fitting.unit_id = ADC_UNIT_1;
Curve_Fitting.atten = ADC_ATTEN_DB_11;
Curve_Fitting.bitwidth = ADC_BITWIDTH_12;

Err = adc_cali_create_scheme_curve_fitting(&Curve_Fitting, &AdcCaliHandle);
if (Err != ESP_OK)
ESP_LOGI(TAG, "=====> AdMeasure : Fail to create the Curve Fitting Scheme.");
else
{
/* On démarre l'ADC : */
adc_continuous_start(AdcHandle);
}
}
}

Every 5 seconds, the sample is recorded this way :

int Voltage = 0;
memset(ptrSamples, 0x00, NBR_SAMPLES);

MeasureResult = adc_continuous_read(AdcHandle, ptrSamples, NBR_SAMPLES, &RealLen, 150);

if ((MeasureResult == ESP_OK) && (RealLen == NBR_SAMPLES))
{
adc_digi_output_data_t *Output = (adc_digi_output_data_t*)&ptrSamples[0];
MeasureResult = adc_cali_raw_to_voltage(AdcCaliHandle, Output->type2.data, &Voltage);
if (MeasureResult == ESP_OK)
ESP_LOGI(TAG, "=====> AdMeasure : Value %u.", Voltage);
}

[ThomasESP32]

I have tried the same thing with 3 samples every 5 sec in order to see if only the first sample is not OK,
but I get the same result.

If I set the Power Supply to 0v, the first records are :
I (80869) AdMeasure: =====> AdMeasure : Value 124.
I (80874) AdMeasure: =====> AdMeasure : Value 163.
I (80879) AdMeasure: =====> AdMeasure : Value 39.

After the display, I change the Power Supply immediately to 1,9V but 5 seconds after the records are :
I (85872) AdMeasure: =====> AdMeasure : Value 130.
I (85877) AdMeasure: =====> AdMeasure : Value 0.
I (85882) AdMeasure: =====> AdMeasure : Value 0.

And 5 sec aftern the records are :
I (90873) AdMeasure: =====> AdMeasure : Value 1964.
I (90878) AdMeasure: =====> AdMeasure : Value 1877.
I (90883) AdMeasure: =====> AdMeasure : Value 1849.

So, it seems that all the values are not correct on the 2nd records. Using an oscilloscope, I am sure that the values have changed.

Moreover, If I use the function :
MeasureResult = adc_continuous_read(AdcHandle, ptrSamples, NBR_SAMPLES, &RealLen, 1000); => With NBR_SAMPLES = 12,

sometimes RealLen is 12, but sometimes it is 8 or 4. Could you explain me why ?
I have the same result if I try to check that the records are taken using a callback with a semaphore..

Best regards,

Thomas TRUILHE

[ThomasESP32]

Any Idea on the subject please ?

[ThomasESP32]

It seems that it is because of the RingBuffer used to store the result :
If after each Read in continuous mode, I do a :
- StopAdc
- Deinit Adc.

Then if the next time I :
- init the Adc
- StartAdc
- Read again

I manage to read each time 256 samples but it takes the Adc 14ms to init each time I want to read something what is impossible in my application.

I would like to record 256 samples each time, to read them and the next time I want to read the Adc, I have 256 new samples.
Do you know how to do it ?

[jtroncin21]

Hi Thomas,

Did you solve the problem ?
I try to do the same thing but encoutered the same troubles with start and stop and restart adc.... the number of read values is not stable.
The first read is good but after a stop and a start, it is smaller than expected.

Thanks.

Jannick TRONCIN

[snutw_]

When I had to add a delay between measurements of 5 s, the ADC continuous driver started having problems - even with 50 ms between them. It was like he started having a serious inertia: showing changes after a long time.
But yea that deinitialization and initialization every times helps. Thanks
(even though I probably should replace this whole continuous set up with usage of one shots measurements)