Problem with ADC

[mikemike9141]

Posted on March 18, 2015 at 13:42

I am currently stuck with a strange phenomenon, and I am not sure anymore if what I do is how I should do it...

In my appliaction with the ST FOC library 4.0, I need to do some additional A/D conversions. This needs to be done as fast as possible. First I tried to use MC_RequestRegularConv/MC_GetRegularConv. This seemed to work, but was too slow as the conversions take place in TSK_SafetyTask at a 2kHz rate only.

I figured that the code there is using PWMC_ExecRegularConv to actually do the conversion. So I wrote the following function:

uint16_t doADC( uint8_t u8Channel )

{

    ADConv_t ADConv_struct;

    ADConv_struct.Channel = u8Channel;

    ADConv_struct.SamplTime = ADC_SampleTime_56Cycles;

    PWMC_ADC_SetSamplingTime( oCurrSensor[M1], ADConv_struct );

    return PWMC_ExecRegularConv( oCurrSensor[M1], u8Channel );

}

I call this function from within main() in a loop for several different channels:

#define ADC_I_BAT   ADC_Channel_8  // PB0

#define ADC_T_MOTOR ADC_Channel_9  // PB1

#define ADC_AIN_1   ADC_Channel_10 // PC0

#define ADC_AIN_2   ADC_Channel_11 // PC1

#define ADC_AIN_3   ADC_Channel_12 // PC2

#define ADC_T_BOARD ADC_Channel_13 // PC3

#define ADC_V_BAT   ADC_Channel_15 // PC5

while (1)

{

    u16Throttle = doADC( ADC_AIN_1 );

    u16Brake    = doADC( ADC_AIN_2 );

    u16Temp     = doADC( ADC_T_BOARD );

    fBoardTemp  = calcNTC( u16Temp );

    u16Temp     = 65535 - doADC( ADC_T_MOTOR );

    fMotorTemp  = calcNTC( u16Temp );

    fBattCurrent= ((float)doADC( ADC_I_BAT ) - 32768.0) / 65536.0 * 3.3 / 55e-3;

    fBattVoltage= (float)doADC( ADC_V_BAT ) / 65536.0 * 3.3 / 0.0279;

    ...

}

There are no calls from other locations, and I do not use the MC_RequestRegularConv/MC_GetRegularConv mechanism anymore.

Basically this works fine, but there are some irregularities. On one of the channels I sometimes (maybe once out of 100 times) get a wrong result. Instead of the real value (a few 100 counts) I get something in the range of 42000. Unfortunately this is the throttle input, meaning my motor receives an audible current ''kick'' from this. I checked carefully, the signal is not present on the analog input.

I know that the phase current measurement is performed simultaneously, and basically asynchronously to the above code, using regular / injected conversion ''preempting''.

Furthermore, I have read ST's errata sheet for the used STM32F405RGT6, which explains something exactly in this direction, but I do not understand what is exactly meant and if this applies to my problem.

So my question: is my implementation correct, is this the intended use of these FOC library calls, is maybe that device errata causing trouble here? I also wonder why only that first channel is affected.

Regards

Frank

[Gigi]

Posted on March 20, 2015 at 17:20

Ciao Frank

Sorry for the delay

The problem that you have is due to the interaction between the ADC conversions done for motor phase currents sampling and the regular conversion and between your ''regular'' conversion and the MC ''regular'' conversions.

We get the same issue and for this reasone we managet to execute the user conversion inside the Safety Task, so at least syncronous with the MC ''regular'' conversions.

Saying that, if you need to execute the ADC conversion at a fater rate and in a safe way I think that is necessary to add extra injected conversions ad the end of MC injected conversion sequence used for motor phase current sampling delaying a little bit the execution of the FOC.

You have to modify the PWMC class (in particular the derived class dedicated to the specific micro) in order to add this extra injected sampling.

Or probably you can just disable the mc ''regular'' conversions (BUS voltage and Temperature) and use your approach. Maybe this also works.

Ciao

Gigi

[mikemike9141]

Posted on March 23, 2015 at 10:52

Hi Gigi,

the project is setup without bus voltage and temperature sensing. (This is done in my code, in a more complex way.) I checked carefully, there are no MC-library related calls of regular conversions. I have now moved the conversion code into the high frequency task code, where the FOC algorithms are executed. This runs better so far. My suspicion is that the problem may be related to the STM32F405. The previous design used a STM32F305, this problem did not appear with this chip.

Regards

Frank