TMS320F28335: F28335 first pulse lost

Sammi,

the PWM output remains low for an entire period, correct? This means that your "Set" event was missed. There are a few reasons why the pulse could be missed.

Are you generating both PWMxA and PWMxB from PWMxA using the deadband module?

Are you adjusting the phase?

Are you using Shadow loading for all registers?

Do you have a typo below?

Sammi Shang said:

EPMW1 A  CAU= 1

EPMW1 A  CAD= 2

EPMW1 B  CAU= 1

EPMW1 B  CAD= 2

Hello  Cody：
    this question is from me  ,  Smmi help me to write it here.

   the PWM set is below:
    EPMW1 A CAU= 1
    EPMW1 A CAD= 2
    EPMW1 B CAU= 1
    EPMW1 B CAD= 2
    DB. INMODE =2
    DB. POLSEL= 2
    DB. OUTMODE= 3
     DB.RED=300
    DB.FED=300

CMPCTL.SHAOWAMODE  =0 

CMPCTL.SHAOWBMODE  =0 

TBSTS.ALL= 0 

TBPRD = 18750 ;//2KHZ

TBPHS..HALF, TBPHS = 0 

TBCTL.BIT, PHSDIR = 1

TBCTL.BIT.SYNCOSEL =0

TBCTL.BIT.prdld =0

TBCTL.BIT.PHSEN=1

TBCTL.BIT.CRLMODE=2

CMP.BIT.LOADAMODE =1

when  CTR=PRD  , the CMPA would be changed by new value .   I only used CMPA to get  EPMWxA &B action, did not use CMPB

and the final output is that PWM1B response to U phase T1, PWM1A response to T3. PWM2B response to U phase T2, PWM2A response to T4

when the phase-voltage    turn from negative pulse to positive pulse, the three-phase UVW will always lost the first pulse.   forexample ,PHASE U , before  this point time  , the T1(EPWM1B always was  low ,becuase the  CMPA=  prd),   then on this point (CTR =PRD), the new CMP would be load  and it  is   a new value ,it would be smaller than PRD , and on this point ,  the PWM would be doing a toggle  because the CTR was bigger than new CMPA value , but the PWM did not do this action

   it losed this pulse ,but after this point , other pulse were all correct action,  only that pulse was loased

we used TMS320F2407 ,the same soft ,it did not this error, but  F28335 have this problem