ADC conversion mixing channels values

[GVassis1999]

I'm working on a project in speed control of a pmsm and i am using ADC1 and 7 channels of the ADC to sample 7 different inputs. The ADC gets triggered from TIM2 and i'm doing the conversions every time i get a callback. At first the adc works fine with the V/F control scheme, but when i switch to the closed loop control adc starts mixing up the sampled values though the sevens channels. 

The callback code is the following: 

void HAL_ADC_ConvCpltCallback(ADC_HandleTypeDef* hadc){

HAL_ADC_Start_DMA(&hadc1,(uint32_t*)adc_value,7);

    Ia= I_coef*adc_value[1]-  I_const; 

Ib = I_coef*adc_value[0]-  I_const; 

            Ic= I_coef*adc_value[2]-  I_const; 

ha= Hall_Binary_Conversion(adc_value[3]); 

hb= Hall_Binary_Conversion(adc_value[4]); 

hc= Hall_Binary_Conversion(adc_value[5]); 

Vdc = 5.5*V_coef*adc_value[6]; 

wmeas =Speed_calculation(ha, hb, hc, Ts);

th = Angle_calculation(ha, hb, hc,wmeas,Ts);

if (Ia > 5 ||Ia < -5 ||

Ib > 5 ||Ib < -5 ||

Ic > 5 ||Ic < -5){

svm_sectors(0,0);

svm_swinst_calc(Vdc);

}

else{

// V/f code----------------------------------------------------------------------

if(f<15){ 

f= 0.7957* wref;

Vrms = 0.24*f ;

Va_vf = 1.4142*Vrms*cos(2*3.14159265358979*f*(cnt_i*Ts));

Vb_vf = 1.4142*Vrms*sin(2*3.14159265358979*f*(cnt_i*Ts));

svm_sectors(Va_vf,Vb_vf);

svm_swinst_calc(Vdc);

}

//--------------------------------------------------------------------------------

else{

Park_components_calculation(Ia,Ib,Ic,th);

PIController_Update(&p1,wref,wmeas);

Iqref_calculation(&p1);

PIController_Update(&p2,Iqref,Iq);

PIController_Update(&p3,0,Id);

Va_Vb_calculation(&p2, &p3,th, wmeas);

svm_sectors(Va,Vb);

svm_swinst_calc(Vdc);

}

}

}

[TDK]

What leads you to believe they're being mixed up? What are the values normally and what are they when they are "mixed up"?

What is your sample rate?

[Andrei Chichak]

CubeIDE has a bunch of good examples. I've settled on a conglomeration of a couple of examples like ADC_TripleModeInterleaved and ADC_TriggerMode to come up with a chunk of code that uses a timer to trigger a bunch of ADC conversions from 1 - 3 ADCs simultaneously (depending on which processor I'm currently using). When a bunch of results are ready, I get an interrupt from the DMA system, then I can pull the data out of the ping-pong buffers, convert them to engineering units with the latest value of Vref, average and post them for consumption.

I only run my ADCs at 200Hz, but that's what I need.

The examples show how to un-"mix-up" the results. Remember, if you're using DMA without ping-pong (circular) buffers, the buffers will be updated as you are using them, that would give weird results.