TMS320F280021-Q1: EPWM for LLC Control Process,PWM outputs are both high level issue

update the pwm load code as bellow：

void loadpwm()
{
    F_SW = CC_PID.PIDU;         //PID return value is freq value
    F_SW = __fmin(150,F_SW);
    F_SW = __fmax(80,F_SW);                                 
    EPWMPeriod = 50000 / F_SW;
    EPWMPeriod_DIV2 = EPWMPeriod / 2;
    LLC_Hduty_EPWM =  EPWMPeriod_DIV2;

    if( LLC_Hduty_EPWM > EPWMPeriod_DIV2 - 25 )
    {
        LLC_Hduty_EPWM = EPWMPeriod_DIV2 - 25;
    }
    
    EPWMPeriod = 50000 / F_SW;
    EPWMPeriod_DIV2 = EPWMPeriod / 2;
    EPWMPeriod = EPWMPeriod_DIV2 * 2;
    F_Int = 50;
    T_Ctrl = 0.001f/F_Int;

    EPwm3Regs.TBPRD = EPWMPeriod;                           //Period of TB = 10kHz
    EPwm4Regs.TBPRD = EPWMPeriod;                           //Period of TB = 10kHz
    EPwm3Regs.CMPA.bit.CMPA = EPWMPeriod_DIV2 - LLC_Hduty_EPWM;
    EPwm3Regs.CMPB.bit.CMPB = EPWMPeriod_DIV2 + LLC_Hduty_EPWM;
    EPwm4Regs.CMPA.bit.CMPA = EPWMPeriod_DIV2 - LLC_Hduty_EPWM;
    EPwm4Regs.CMPB.bit.CMPB = EPWMPeriod_DIV2 + LLC_Hduty_EPWM;

    EPwm1Regs.GLDCTL2.bit.OSHTLD = 1;
    
}

We record the PWM register values(EPwm3Regs.CMPA.bit.CMPA,EPwm3Regs.CMPB.bit.CMPB and EPwm3Regs.TBPRD) for each cycle in a loop in a buffer, and after a short circuit occurred, we write them to the FLASH. We couldn't find any issues. Is there a BUG in the chip?

The specific data is shown in the table below:

The 32nd occurrence of a short circuit in the data record.

Hi Qin,

When are you updating the ePWM parameters? Are you using a timer ISR? What ePWM signals are the green/yellow, is it ePWM3's output? 

Best,

Ryan Ma

I updatae the ePWM parameters in a ADC ISR (PieVectTable.ADCA1_INT = &User_ADC_normal_isr).The PID and loadpwm()  function are in the same ISR.The ePWM signals is the ePWM3's output,the green signal is low drive,the yellow is high drive.Thanks!

Hi Qin,

Could you show me when the PWM CMP registers get updated and loaded by toggling a GPIO during your ADC ISR? According to your configurations is the following diagram representative of what is occurring before the short with AU = CMPAU, and BU = CMPBU, BD = CMPBD events? Is the highlighted red No. 32 row where the short occurs?

EPwm3Regs.AQCTLA.bit.CAU = AQ_SET;
EPwm3Regs.AQCTLA.bit.CBU = AQ_CLEAR;

EPwm3Regs.AQCTLA.bit.ZRO = AQ_CLEAR;
EPwm3Regs.AQCTLB.bit.CBD = AQ_SET;
EPwm3Regs.AQCTLB.bit.CAD = AQ_CLEAR;
EPwm3Regs.AQCTLB.bit.ZRO = AQ_CLEAR;

The data log you are capturing appears to have CMP values still even after the highlighted red portion, how come the PWM signals stop after the short? Can you verify that the TBCTR is continuously running after the short circuit occurs? When the short occurs, are you shutting off the PWM?

I believe there is something happening with the loading scheme of your setup, with an AQ event being missed because either CMPBU is greater than TBPRD or there is nothing in the shadow registers at the time the global load scheme occurs. Is there anything special about this last PWM period that is happening in the code?

Best,

Ryan Ma

Hi Ryan:

The following  are normal charging process ePWM3 and GPIO toggling waveform.yellow and blue are ePWM,Green is GPIO.When the Green signal rise is begin to load PWM CMP registers. You can see the detail code as bellow.

void PWM_load() is use in the ISR.

void PWM_load()
{
    static Uint16 CountNum = 0;

    LED1on();                   // test GIIO go high
    
    F_SW = CC_PID.PIDU;         //PID return value is freq value
    F_SW = __fmin(150,F_SW);
    F_SW = __fmax(80,F_SW);                                 
    EPWMPeriod = 50000 / F_SW;
    EPWMPeriod_DIV2 = EPWMPeriod / 2;
    LLC_Hduty_EPWM =  EPWMPeriod_DIV2;

    if( LLC_Hduty_EPWM > EPWMPeriod_DIV2 - 25 )
    {
        LLC_Hduty_EPWM = EPWMPeriod_DIV2 - 25;
    }
    
    EPwm3Regs.TBPRD = EPWMPeriod;                           //Period of TB = 10kHz
    EPwm4Regs.TBPRD = EPWMPeriod;                           //Period of TB = 10kHz
    EPwm1Regs.CMPA.bit.CMPA = EPWMPeriod_DIV2 - LLC_Lduty_EPWM;
    EPwm1Regs.CMPB.bit.CMPB = EPWMPeriod_DIV2 + LLC_Lduty_EPWM;
    EPwm2Regs.CMPA.bit.CMPA = EPWMPeriod_DIV2 - LLC_Lduty_EPWM;
    EPwm2Regs.CMPB.bit.CMPB = EPWMPeriod_DIV2 + LLC_Lduty_EPWM;
    EPwm3Regs.CMPA.bit.CMPA = EPWMPeriod_DIV2 - LLC_Hduty_EPWM;
    EPwm3Regs.CMPB.bit.CMPB = EPWMPeriod_DIV2 + LLC_Hduty_EPWM;
    EPwm4Regs.CMPA.bit.CMPA = EPWMPeriod_DIV2 - LLC_Hduty_EPWM;
    EPwm4Regs.CMPB.bit.CMPB = EPWMPeriod_DIV2 + LLC_Hduty_EPWM;

    EPwm1Regs.GLDCTL2.bit.OSHTLD = 1;

    WriteFlashBuf[i] = EPwm3Regs.CMPA.bit.CMPA;     // Save pwm value every cycle
    WriteFlashBuf[i+512] = EPwm3Regs.CMPB.bit.CMPB;
    WriteFlashBuf[i+1024] = EPwm3Regs.TBPRD;

    if(CountNum < 511)
    {
        CountNum++;
    }
    else
    {
        CountNum = 0;
    }

    WriteFlashBuf[1536] =  CountNum;   // Save occur short-circuit count num
    LED1off();
}

      The diagram what I show is save in a buffer every ISR,when the short circuit occurs,the hardware will occurs a TZ module protect and stop the PWM, and then write the buffer data to the use flash area.After a short circuit occurred, I read Flash data and CountNum variables,Through the count number I know when the error happen.The CountNum may not  be the counting at that time,but it must be one of the previous times.

      Yes,I suspect that scertain register is not set correctly or miss some seting.But I don't know.

     On the other hand，I found a iInteresting things,when the ePwm frequency is change from high drop to low over 2K,it maybe occurs short circuit.The following table shows the data when errors is occur.The red line is the final CountNum record.

So,I write a program to simulate a frequency jumps.It run on a demo board.It can reproduce the phenomenon of short circuits.

The program is as follows:

***********************************************************************/
void loadpwm()
{
    volatile static float Temp_CC_PID_Gain = 1;
    static Uint16 u16TurnPhaseFlag = 0;
    static Uint16 u16adjfreqNum = 0;
    static float Fnum = 0;

/**********************debug************************/
    if(u16TurnPhaseFlag)
    {
        F_SW = F_SW + 0.01f;
        if( F_SW > 100 )            // max freq is 100K
        {
            u16TurnPhaseFlag = 0;
        }
    }
    else
    {
        F_SW = F_SW - 0.01f ;
        if( F_SW < 80 )            // min freq is 80k
        {
            u16TurnPhaseFlag = 1;
        }
    }

    if(TimerFlg.bit.TimerPWM1msFlag)  // 1ms timer flag
    {
        TimerFlg.bit.TimerPWM1msFlag = 0;

        u16adjfreqNum++;
        if(Fnum < 7.0f)             // 
        {
            Fnum = Fnum + 0.1;      // change freq
        }
        else
        {
            Fnum = 0;
        }
        if(u16adjfreqNum > 1000)
        {
            u16adjfreqNum = 0;
            F_SW = F_SW - Fnum;
            LED1on();            // GPIO tes signal，load epwm
        }

        if(u16adjfreqNum > 500)
        {
            LED1off();
        }
    }


/**********************debug************************/

    F_SW = __fmin(150,F_SW);
    F_SW = __fmax(80,F_SW);                                 // 限上下限
    EPWMPeriod = 50000 / F_SW;
    EPWMPeriod_DIV2 = EPWMPeriod / 2;
    LLC_Hduty_EPWM = EPWMPeriod_DIV2;

    if( LLC_Hduty_EPWM > EPWMPeriod_DIV2 - 25 )
    {
        LLC_Hduty_EPWM = EPWMPeriod_DIV2 - 25;
    }
    LLC_Lduty_EPWM = LLC_Hduty_EPWM*0.9f;                               // 开启同步整流后低压MOS PWM为高压MOS的90%


    EPwm3Regs.TBPRD = EPWMPeriod;                           //Period of TB = 10kHz
    EPwm4Regs.TBPRD = EPWMPeriod;                           //Period of TB = 10kHz

    EPwm3Regs.CMPA.bit.CMPA = EPWMPeriod_DIV2 - LLC_Hduty_EPWM;
    EPwm3Regs.CMPB.bit.CMPB = EPWMPeriod_DIV2 + LLC_Hduty_EPWM;
    EPwm4Regs.CMPA.bit.CMPA = EPWMPeriod_DIV2 - LLC_Hduty_EPWM;
    EPwm4Regs.CMPB.bit.CMPB = EPWMPeriod_DIV2 + LLC_Hduty_EPWM;

    EPwm1Regs.GLDCTL2.bit.OSHTLD = 1;
}

The short-circuit waveform is as follows:yellow is low drive,green is high drive,blue is GPIO signal(ePMW load signal).

Do you have any good suggestions?

Thanks!

Hi Qin,

Can you please clarify if green is ePWM3A and yellow is ePWM3B?

EPwm3Regs.GLDCTL.bit.GLDMODE = 6;

Here you're global loading on a SYCNEVT correct? 

EPwm3Regs.TBCTL.bit.PHSDIR = TB_UP;

After the sync event you count up, what is your TBPHS at the time? It may get updated to a value greater than TBPRD causing a missed AQ.

Best,

Ryan Ma

Hi Ryan:

The green is ePWM3A and yellow is ePWM3B,it is correct.

So How do we optimize the program for testing and verification？

Thanks!

Hi Qin,

I believe it is a loading scheme issue with the counter compare registers. Can you try loading on CTR == PERIOD or ZRO for ePWM3?

Best,

Ryan Ma

Hi Qin,

If you're using ePWM1 to sync the other ePWM modules, why are you changing ePWM3's and ePWM4's TBPRD. This will cause the ePWM's to become out of sync to each other, and we do not recommend this as the sync scheme. Could you synchronize ePWM3 and ePWM4's modules together since their TBPRD will be aligned with each other?

Best,

Ryan Ma

Hi Ryan:

"I believe it is a loading scheme issue with the counter compare registers. Can you try loading on CTR == PERIOD or ZRO for ePWM3?"

We had set the global Register GLDCFG and GLDCTL.The CTR == PERIOD or ZRO for ePWM3,is it right? So short circuits still occur.

EPwm3Regs.GLDCFG.all = 0x07FF;
EPwm3Regs.GLDCTL.bit.GLDMODE = 6;

"If you're using ePWM1 to sync the other ePWM modules, why are you changing ePWM3's and ePWM4's TBPRD."

Yes.We want using ePWM1 to sync the other ePWM modules.The ePWM1's,ePWM2's,ePWM3's,ePWM4's TBPRD are change as the same time.I removed the TBPRD  loading for ePWM1 and ePWM2 in the code I showed you.All ePMW's TBPRD are load as the same time in the ISR.If it cause the ePWM's to become out of sync to each other?

void PWM_load()
{
    static Uint16 CountNum = 0;

    LED1on();                   // test GIIO go high
    
    F_SW = CC_PID.PIDU;         //PID return value is freq value
    F_SW = __fmin(150,F_SW);
    F_SW = __fmax(80,F_SW);                                 
    EPWMPeriod = 50000 / F_SW;
    EPWMPeriod_DIV2 = EPWMPeriod / 2;
    LLC_Hduty_EPWM =  EPWMPeriod_DIV2;

    if( LLC_Hduty_EPWM > EPWMPeriod_DIV2 - 25 )
    {
        LLC_Hduty_EPWM = EPWMPeriod_DIV2 - 25;
    }
    
    EPwm1Regs.TBPRD = EPWMPeriod;                           //Period of TB = 10kHz
    EPwm2Regs.TBPRD = EPWMPeriod;                           //Period of TB = 10kHz
    EPwm3Regs.TBPRD = EPWMPeriod;                           //Period of TB = 10kHz
    EPwm4Regs.TBPRD = EPWMPeriod;                           //Period of TB = 10kHz
    
    EPwm1Regs.CMPA.bit.CMPA = EPWMPeriod_DIV2 - LLC_Lduty_EPWM;
    EPwm1Regs.CMPB.bit.CMPB = EPWMPeriod_DIV2 + LLC_Lduty_EPWM;
    EPwm2Regs.CMPA.bit.CMPA = EPWMPeriod_DIV2 - LLC_Lduty_EPWM;
    EPwm2Regs.CMPB.bit.CMPB = EPWMPeriod_DIV2 + LLC_Lduty_EPWM;
    EPwm3Regs.CMPA.bit.CMPA = EPWMPeriod_DIV2 - LLC_Hduty_EPWM;
    EPwm3Regs.CMPB.bit.CMPB = EPWMPeriod_DIV2 + LLC_Hduty_EPWM;
    EPwm4Regs.CMPA.bit.CMPA = EPWMPeriod_DIV2 - LLC_Hduty_EPWM;
    EPwm4Regs.CMPB.bit.CMPB = EPWMPeriod_DIV2 + LLC_Hduty_EPWM;

    EPwm1Regs.GLDCTL2.bit.OSHTLD = 1;

    WriteFlashBuf[i] = EPwm3Regs.CMPA.bit.CMPA;     // Save pwm value every cycle
    WriteFlashBuf[i+512] = EPwm3Regs.CMPB.bit.CMPB;
    WriteFlashBuf[i+1024] = EPwm3Regs.TBPRD;

    if(CountNum < 511)
    {
        CountNum++;
    }
    else
    {
        CountNum = 0;
    }

    WriteFlashBuf[1536] =  CountNum;   // Save occur short-circuit count num
    LED1off();
}

"Could you synchronize ePWM3 and ePWM4's modules together since their TBPRD will be aligned with each other?"

How should I set it up?Could you show me the setting code？

On the other hand,We found some clues to this issue today. I test the frequency jumps code,and turn out all pwm.When the frequency change,the ePWM3A's period change first,but the ePWM1 is not change.Next cycle ePWM3B and ePWM1B are change,this time may causing a missed AQ.It caused a short circuit.

Do you have any idea?

As follow waveform:Yellow is ePWM3B,Green is ePWM3A,Blue is ePWM1A,Purple is ePWM1B.

ePWM3 period change:

ePWM1 period change:

Thanks!

Hi Qin,

We will do some testing here to produce the same results based on the configurations your team sent over. Once we have this set up we can see what configuration issue may be.

Thank you for your patience.

Best,

Ryan Ma

Thank you for your support! We are testing the solution.

Hi Qin,

I am preparing a test case for our design team to simulate.

Best,

Ryan Ma