TMS320F28377D: ePWM lose pulse

Joe Ji

Tool/software:

Hi expert

Customer use 28377 in CLLLC control in OBC.
PWM Configurations are below:

CC: load COMPA=0 and COMPB=1/2*TBPRD to generate 50% duty.
AQ: Counter=COMPA set high, Counter=COMPB set low.
Enable register shadow mode PWM1 counter = 0 event set as pwmsyncout, all pwmsyncin group signals are linked to PWM1.
In each PWM interrupt, TBPRD will be modified, the shadow register will set to active, and new TBPRD value will load in the shadow register to realize frequency variation.
In the test, customer feedback 2 COMPB event will be ignored, the PWM waveform is shown below:

if change AQ to Counter=COMPA toggle, Counter=COMPB toggle, the waveform will change below:

the issue will happen randomly, issue channel is PWM1. The TBCLK data Format is checked equal to 16bit.
Are there any further suggestions to debug this issue?

Thanks

Joe

over 1 year ago

0 Allison Nguyen over 1 year ago

TI__Mastermind 25280 points

Hi Joe,

Can you please have the customer check these related threads to see if the issue is similar to their case? I provided an example of a workaround for this in the second link.

Best Regards,

Allison

0 Joe Ji over 1 year ago in reply to Allison Nguyen

TI__Expert 3990 points

Hi Allson

I also looked into the E2E thread when debug with customer, but my issue is related to CMPB, not CMPA.

I suggest customer to use TBCRT = 0 event instead of CMPA=0 to trigger AQ, but there is no help for this issue. So may be CMPA bug in sync is not the root cause?

customer will use F280039 for mass production, i would like to check whether this device has the same issue?

Thanks

Joe

0 Allison Nguyen over 1 year ago in reply to Joe Ji

TI__Mastermind 25280 points

Hi Joe,

F2837xD and F28003x both use the Type-4 EPWM module. There are some small differences between the two modules, which are outlined in the Peripherals User Guide section 3.3.8 Enhanced Pulse Width Modulator (ePWM) Module- but I don't believe these small differences would change the behavior.

A few questions:

Is the PWM waveform from the oscilloscope images EPWM1? Is EPWM1 taking in any SYNC-IN signals (is phase shift enabled?)? Or is EPWM1 only providing the SYNCOUT source for other PWM modules?
Can you share the basic initializations for the problematic PWM (EPWM1)?
When is the TBPRD register shadow-to-active loading? From my understanding, the ISR to update TBPRD occurs on CTR = 0, but is shadow loading occurring on the next CTR = 0, or some other event, or is immediate load mode being used?
Is there any correlation between the PWM parameters (counter value/period value) and the occurrence of the long pulse?

Best Regards,

Allison

0 Joe Ji over 1 year ago in reply to Allison Nguyen

TI__Expert 3990 points

Hi Allison

Sorry for the late reply, the root cause is that the shadow to active action is not synchronized with PWM interrupt cycle, that will lead to the case which shown in Table 15-7. Behavior if CMPA/CMPB is Greater than the Period, which leads to the PWM pulse missing.

Thanks

joe

0 Allison Nguyen over 1 year ago in reply to Joe Ji

TI__Mastermind 25280 points

Hi Joe,

Glad you were able to find the cause! Thanks for posting it on this thread

Best Regards,

Allison

C2000™︎ microcontrollers

C2000 microcontrollers forum

TMS320F28377D: ePWM lose pulse