TMS320F280039C: TMS320F280039C HRPWM PSFB assignment

Hi Gregg,

1. In many applications, two PWM outputs need to be complementary to each other (one goes low while the other goes high), often with some dead time in between when both outputs are low. The PWM module is designed to handle cases like this. It's dead-band module provides a reliable hardware based approach for producing a complementary B output based on the A output.

2. EPWM1-EPWM4 have HR capability, so any of the A or B outputs from these 4 EPWM modules can use HR.

Thank you,

Luke

Hi Luke, thank you for the prompt response.
I have further clarification:
Does this mean for complementary-with-dead-time PWM function, the A and B are purposely designed functions for PSFB application?
Is it also possible to jumble the HRPWMs by selecting for example PWM1_A and PWM4_B as complementary with dead time function?
Another jumbled selection but keeping the A and B pair,
PWM3_A and PWM1_B
PWM4_A and PWM2_B

Hey Gregg,

You can't directly derive the PWM4_B output from the PWM1_A output, however you could load both EPWM1 and EPWM4 with the same EPWM configuration and synchronize the PWMs during initialization. However this would result in an unused PWM4A and PWM1B output, which may create issues for you if you're using SysConfig.

Is there a reason you want 1A and 4B to be complementary instead of 1A and 1B? Is this due to the location of the pins that offer these functions?

Thank you,

Luke

Hi Luke,

Yes, we never had a hint at all regarding the A and B pair as being complementary during our power board hardware design. We thought we could just freely set any HRPWMs and drive them as PWMs for PSFB implementation. 
Our current hardware PWM selections are:
Supposed complementary Left Leg = PWM3_A (high side), PWM4_B (low side)
Supposed complementary Right Leg = PWM4_A (high side), PWM3_B (low side)
Please help to check whether this arrangement is possible and reliable.

If not possible then we have no choice but to make a cut and jump rework in the power board. 

Hey Gregg,

It may be possible to implement this using the CLB depending on your implementation. You have the capability to route the PWM inputs into the CLB and use the CLB outputs to override different stages of the PWM signal chain. Here's what you may be able to do:

Configure complementary left leg using 3A and 3B

Configure complementary right leg using 4A and 4B

Route the outputs from the deadband module from EPWM3 and EPWM4 into the CLB:

Swap 4B and 3B outputs by having the 3B signal going into the CLB override the 4B_DB output and vice versa:

Potential issues with this approach:

• A propagation delay will be added to the PWMs that are routed through the CLB, so you will need to compensate for this either by routing all of your PWMs through the CLBs, or using a phase shift to adjust for the prop delay after measuring on your scope.
• It's possible that overriding the deadband output will cause you to lose HR precision, since the CLB clock is slower than the length of a MEP step. The HR delay is supposed to be applied after the dead-band module, but it's not specified whether the output override from the CLB occurs before or after the HR delay is applied. If it is applied before, this approach should work.

Apologies for the confusion about the flexibility of the PWM outputs. Let me know if you have any questions about the workaround I've proposed above or need any further assistance.

Thank you,

Luke

Hi Luke,

Thank you for this advice. Looks like this approach and associated effect is formidable though for us given our little experience on this device.
I will ask our software team to try it out and see if can be accomplished the dead time and other timing requirement.
If not we'd better just go for the cut and jump rework.

Thanks a lot.

Hi Gregg,

I will be out of office today, I will be back tomorrow if you have any additional questions.

Thank you,

Luke

Hi Gregg, I am back in the office. It seems l like this issue is on hold until you get feedback from the software team, I will mark this thread as "Waiting for Customer".

--Luke

Hi Luke, Noted. It may take us some time and surely will update you once we have something decisive.

Sounds good, I will await your reply.

--Luke