My problem: stm32h743 ADC does not convert small signals(about 1-5mv) correctly

[MKhak]

I configure ADC1 and ADC2 in Dual mode, 16 bit resolution, differential-ended. I use CubeMx to initialize ADC. I want measure voltage and current. My project samples the signals and then sends them to the computer for plotting and logging by the WiFi module. When the amplitude of voltage and current are too small, the ADC does not convert the signals correctly. I have attached figures of a logged signal. In these figures the amplitude of voltage changes from 1mV to 9 mV. Figures show that, a part of the signal has not been converted correctly.

This behaviour is seen in other channel that you can see in below figure.

In my hardware, the signal first passes through a differential-ended amplifier (AD8132) and then enters the ADC. The common mode voltage at the amplifier outputs is Vref/2(There is some potential difference (about a few millivolts) between the amplifier outputs). Schematic of the amplifier is attached. 

ADCs are initialized by below functions. Sampling frequency is 12KHz.

static void MX_ADC1_Init(void)

{

 ADC_MultiModeTypeDef multimode = {0};

 ADC_ChannelConfTypeDef sConfig = {0};

 hadc1.Instance = ADC1;

 hadc1.Init.ClockPrescaler = ADC_CLOCK_ASYNC_DIV6;

 hadc1.Init.Resolution = ADC_RESOLUTION_16B;

 hadc1.Init.ScanConvMode = ADC_SCAN_ENABLE;

 hadc1.Init.EOCSelection = ADC_EOC_SEQ_CONV;

 hadc1.Init.LowPowerAutoWait = DISABLE;

 hadc1.Init.ContinuousConvMode = DISABLE;

 hadc1.Init.NbrOfConversion = 3;

 hadc1.Init.DiscontinuousConvMode = DISABLE;

 hadc1.Init.ExternalTrigConv = ADC_EXTERNALTRIG_T6_TRGO;

 hadc1.Init.ExternalTrigConvEdge = ADC_EXTERNALTRIGCONVEDGE_FALLING;

 hadc1.Init.ConversionDataManagement = ADC_CONVERSIONDATA_DMA_CIRCULAR;

 hadc1.Init.Overrun = ADC_OVR_DATA_PRESERVED;

 hadc1.Init.LeftBitShift = ADC_LEFTBITSHIFT_NONE;

 hadc1.Init.OversamplingMode = ENABLE;

 hadc1.Init.Oversampling.Ratio = 31;

 hadc1.Init.Oversampling.RightBitShift = ADC_RIGHTBITSHIFT_5;

 hadc1.Init.Oversampling.TriggeredMode = ADC_TRIGGEREDMODE_SINGLE_TRIGGER;

 hadc1.Init.Oversampling.OversamplingStopReset = ADC_REGOVERSAMPLING_CONTINUED_MODE;

 if (HAL_ADC_Init(&hadc1) != HAL_OK)

 {

  Error_Handler();

 }

 multimode.Mode = ADC_DUALMODE_REGSIMULT_INJECSIMULT;

 multimode.DualModeData = ADC_DUALMODEDATAFORMAT_32_10_BITS;

 multimode.TwoSamplingDelay = ADC_TWOSAMPLINGDELAY_1CYCLE;

 if (HAL_ADCEx_MultiModeConfigChannel(&hadc1, &multimode) != HAL_OK)

 {

  Error_Handler();

 }

 sConfig.Channel = ADC_CHANNEL_2;

 sConfig.Rank = ADC_REGULAR_RANK_1;

 sConfig.SamplingTime = ADC_SAMPLETIME_8CYCLES_5;

 sConfig.SingleDiff = ADC_DIFFERENTIAL_ENDED;

 sConfig.OffsetNumber = ADC_OFFSET_NONE;

 sConfig.Offset = 0;

 if (HAL_ADC_ConfigChannel(&hadc1, &sConfig) != HAL_OK)

 {

  Error_Handler();

 }

 sConfig.Channel = ADC_CHANNEL_3;

 sConfig.Rank = ADC_REGULAR_RANK_2;

 if (HAL_ADC_ConfigChannel(&hadc1, &sConfig) != HAL_OK)

 {

  Error_Handler();

 }

 sConfig.Channel = ADC_CHANNEL_4;

 sConfig.Rank = ADC_REGULAR_RANK_3;

 if (HAL_ADC_ConfigChannel(&hadc1, &sConfig) != HAL_OK)

 {

  Error_Handler();

 }

 HAL_ADCEx_Calibration_Start(&hadc1, LL_ADC_CALIB_LINEARITY, ADC_DIFFERENTIAL_ENDED); 

}

static void MX_ADC2_Init(void)

{

 ADC_ChannelConfTypeDef sConfig = {0};

 hadc2.Instance = ADC2;

 hadc2.Init.ClockPrescaler = ADC_CLOCK_ASYNC_DIV6;

 hadc2.Init.Resolution = ADC_RESOLUTION_16B;

 hadc2.Init.ScanConvMode = ADC_SCAN_ENABLE;

 hadc2.Init.EOCSelection = ADC_EOC_SEQ_CONV;

 hadc2.Init.LowPowerAutoWait = DISABLE;

 hadc2.Init.ContinuousConvMode = DISABLE;

 hadc2.Init.NbrOfConversion = 3;

 hadc2.Init.DiscontinuousConvMode = DISABLE;

 hadc2.Init.ConversionDataManagement = ADC_CONVERSIONDATA_DMA_CIRCULAR;

 hadc2.Init.Overrun = ADC_OVR_DATA_PRESERVED;

 hadc2.Init.LeftBitShift = ADC_LEFTBITSHIFT_NONE;

 hadc2.Init.OversamplingMode = ENABLE;

 hadc2.Init.Oversampling.Ratio = 31;

 hadc2.Init.Oversampling.RightBitShift = ADC_RIGHTBITSHIFT_5;

 hadc2.Init.Oversampling.TriggeredMode = ADC_TRIGGEREDMODE_SINGLE_TRIGGER;

 hadc2.Init.Oversampling.OversamplingStopReset = ADC_REGOVERSAMPLING_CONTINUED_MODE;

 if (HAL_ADC_Init(&hadc2) != HAL_OK)

 {

  Error_Handler();

 }

 sConfig.Channel = ADC_CHANNEL_18;

 sConfig.Rank = ADC_REGULAR_RANK_1;

 sConfig.SamplingTime = ADC_SAMPLETIME_8CYCLES_5;

 sConfig.SingleDiff = ADC_DIFFERENTIAL_ENDED;

 sConfig.OffsetNumber = ADC_OFFSET_NONE;

 sConfig.Offset = 0;

 if (HAL_ADC_ConfigChannel(&hadc2, &sConfig) != HAL_OK)

 {

  Error_Handler();

 }

 sConfig.Channel = ADC_CHANNEL_5;

 sConfig.Rank = ADC_REGULAR_RANK_2;

 if (HAL_ADC_ConfigChannel(&hadc2, &sConfig) != HAL_OK)

 {

  Error_Handler();

 }

 sConfig.Channel = ADC_CHANNEL_2;

 sConfig.Rank = ADC_REGULAR_RANK_3;

 if (HAL_ADC_ConfigChannel(&hadc2, &sConfig) != HAL_OK)

 {

  Error_Handler();

 }

 HAL_ADCEx_Calibration_Start(&hadc2, LL_ADC_CALIB_LINEARITY, ADC_DIFFERENTIAL_ENDED);

}

[waclawek.jan]

Did you perform the differential-mode calibration, as outlined in Calibration (ADCAL, ADCALDIF, ADCALLIN, ADC_CALFACT) subchapter of the ADC chapter in Reference Manual?

I don't use/understand Cube/HAL.

JW

[MKhak]

Thank you for your attention to my problem. I perform the differential-mode calibration. This line does it.

 HAL_ADCEx_Calibration_Start(&hadc2, LL_ADC_CALIB_LINEARITY, ADC_DIFFERENTIAL_ENDED);

I test my Program with the differential-mode calibration and without it. In both cases there is a problem mentioned above.

[Ozone]

> I have attached images of a converted signal. In these figures the amplitude of voltage changes from 1mV to 9 mV.

> Figures show that, a part of the signal has not been converted correctly.

I don't understand how you got to that.

The converted signal would numerical values in memory or a register, not a scope-able voltage curve.

Re-converting it via DAC would introduce further artifacts.

[MKhak]

Excuse me for the incomplete description. My project samples the voltage and current and sends sampled data to the computer using the WiFi module. In PC, a software plots data.

[Chris1]

Are your results worse than the claimed performance in the datasheet? There's some significant error numbers shown there, for example +/-10 or +/-16 LSB total unadjusted error, +/-4 or +/-6 LSB linearity error, etc.

[MKhak]

I don't think this problem is related to unadjusted error or linearity error. I added another figure to my description. for 1mV amplitude the result is good. when the amplitude of signal is increased, a distortion is occured in sampled data.

[Uwe Bonnes]

Did you have a look with the scope on the signals after the AD8132? Maybe the problems are before the STM32H7?

[waclawek.jan]

IMO this is an offset error - i.e. input of +-0mV is not converted to 0x0000 but to some small offset, e.g. +-0x0010 - compounded with an offset in center of the sinus, whether its source is internal or external.

This is IMO specific to the differential mode; so you maybe might want to convert the singal to single-ended by the external circuitry and convert it as such; and then remove the DC component by some suitable method in software.

JW

[waclawek.jan]

Which revision of the chip are you using?

JW