TMS320F280049: Adding HWPWM TBPHSHR to DAB converter causes loss of waveform symmetry

Jason Galtieri1

Part Number: TMS320F280049

Hello,

I would like to add HRPWM TBPHSHR capability to my dual active bridge. I am currently only looking at one side of the bridge, so I have two PWM channels with A and B outputs. The A/B outputs on each channel (EPWM1 and EPWM2) are run at 200kHz, ~50% duty (minus deadtime) and complementary to one each other. EPWM1 is the master leg and TBPHS is used to set a phase shift between EPWM1 and EPWM2. The converter works as expected using only TBPHS (no TBPHSHR), but I would like more resolution.

From my understanding if I use the deadtime Active HI complementary setting, I should still get a complementary signal even if HRPWM TBPHSHR is added. I need the deadtime, so I don't think I can use the SELOUTB setting to just output inverted "A" as B. My PWM code is setup below.

What I am noticing is when I added TBPHSHR, I lose waveform symmetry where HR phase is added to the negative cycle, but not the positive cycle. You can see this in the width measurement and its about 10 ns difference when I add the full TBPHSHR range. Autoconvert is on and the timing is correct but something in my setup is causing my PWM2 A/B to no longer be complementary when adding TBPHSHR. Any help is appreciated


data-mode="text"> // EPWM Initialization -------------------------------------------------- = 0; =   PWM_TBPRD;          // Set period for ePWMx =   0x0000;             // Clear counter =   TB_DIV1;            // Time-base Clock Prescale Bits =   TB_DIV1;            // High-Speed Time-base Clock Prescale Bits =   TB_CTR_ZERO;        // EPWMx is a "master", sync "down-stream" =   TB_DISABLE;         // EPWMx is a "master", Do not load the time-base counter at sync =   TB_COUNT_UP;    // UpCount Mode =   PWM_TBPRDdiv2;      // Start at a duty cycle of approximately 50% CC_CTR_PRD;    // Load at both the top and bottom of the carrier CC_CTR_PRD;    // Load at both the top and bottom of the carrier =   UPAQ;                //Set   PWMxA on event A, up count =   DNAQ;               // Clear PWMxA on zero =   UPAQ;                //Set   PWMxA on event B, up count =   1;                  // Half cycle clocking enabled. The dead-band counters are clocked at TBCLK*2 =   DB_ACTV_HIC;        // Active HI complementary =   DB_FULL_ENABLE;     // Enable Dead-band module on both rising and falling edges =   DEADBAND;           // Rising Edge Deadband =   DEADBAND;           // Falling Edge Deadband =   0; =   0;                  // Disable TZ1 as one one-shot trip source =   TZ_FORCE_LO;        // TZ1 forces A output lo =   TZ_FORCE_LO;        // TZ1 forces B output lo =   1;                  // Enable TZ interrupt =   1;                  // Enable SOC on A group =   ET_CTR_ZERO;        // Select SOC A on EPWM1 Timer Zero =   ET_2ND;             // Generate pulse on every event = HR_BEP;  // MEP control on falling edge = HR_BEP; // MEP control on falling edge = HR_PHS; = HR_PHS; = HR_CTR_ZERO_PRD; // Enable auto-conversion                                             // logic = 1; clock is already enabled =   PWM_TBPRD;          // Set period for ePWMx =   0x0000;             // Clear counter =   TB_DIV1;            // Time-base Clock Prescale Bits =   TB_DIV1;            // High-Speed Time-base Clock Prescale Bits =   TB_SYNC_IN;         // EPWMx is a "slave", sync flow-through =   TB_ENABLE;          // EPWMx is a "slave", Load the time-deg counter from TBPHS at sync =   PWM_TBPRDdiv2+2;             // Start phase shift aligned to EPWM1 =   1<<8; =   TB_COUNT_UP;        // Up-Down Count Mode =   PWM_TBPRDdiv2;      // Start at a duty cycle of approximately 50% =   PWM_TBPRDdiv2+5;    // Start at a duty cycle of approximately 50% CC_CTR_PRD;         // Load at both the top and bottom of the carrier CC_IMMEDIATE;       // CMPB is a safeguard value which we want updated ASAP, overwrites LOADBMODE CC_CTR_PRD;         // Load at both the top and bottom of the carrier =   DNAQ;               //Clear   PWMxA on event A, up count =   DNAQ;               //Clear   PWMxA on event B, up count =   UPAQ;               // Set PWMxA on zero =   1;                  // Half cycle clocking enabled. The dead-band counters are clocked at TBCLK*2 =   DB_ACTV_HIC;        // Active HI complementary =   DB_FULL_ENABLE;     // Enable Dead-band module on both rising and falling edges =   DEADBAND;           // Rising Edge Deadband =   DEADBAND;           // Falling Edge Deadband =   0; =   0;                  // Disable TZ1 as one one-shot trip source =   TZ_FORCE_LO;        // TZ1 forces A output lo =   TZ_FORCE_LO;        // TZ1 forces B output lo = HR_BEP;  // MEP control on falling edge = HR_BEP;  // MEP control on falling edge = HR_PHS; = HR_PHS; = HR_CTR_ZERO_PRD; // Enable auto-conversion = 1; = 1;
Zoomed Out figure showing symmetric waveforms when TBPHSHR = 0

Zoomed in when TBPHSHR = 0 

Zoomed in when TBPHSHR is close to max

over 3 years ago

0 Cody Watkins over 3 years ago

TI__Mastermind 42195 points

If i remember correctly i believe a few of the HR enhancements require you to remove one clock pulse from your standard setting. Meaning TBPHS should become TBPHS-1 and then add TBPHSHR. I don't know if that's your problem here being that your shift is being applied asymmetrically, but depending on your PWM configuration its could be true.

(this is also supported because your signal shifted by one SYSCLK pulse 10nS)

If this doesn't seem to help it could be an artifact of the RED and FED delays, i would recommending removing or disabling them to check if the deadband and phase shifting are having an unexpected interaction.

Regards,
Cody

0 Jason Galtieri1 over 3 years ago in reply to Cody Watkins

Prodigy 165 points

Hello,

Thanks for replying. The 10 nS is from applying the full TBPHSHR (close to 0xFF). If I apply 32768 I get 5 nS shift etc.

I don't know how I can test if its the deadband because I need the deadband to avoid shoot-through on my switches.

I guess the real question is what I am trying to do even possible? Am I able to get a HR phase shift on both A and B channels by relying on the activ high deadband setting (i.e something outside the HR framework) to correctly set HR B output? This section here states (pg 1866 in sprui33d.pdf) states I can output an inverted HR pwm as the B output, but I'd need the HR +deadband

0 Jason Galtieri1 over 3 years ago in reply to Cody Watkins

Prodigy 165 points

Put a probe on the pins to get the pwm signals directly and it looks like only one edge is moving with TBPHSHR even tho I have BEP setting.

TBPHSHR = 0

TBPHSHR = 32768

TBPHSHR = 64000

0 Cody Watkins over 3 years ago in reply to Jason Galtieri1

TI__Mastermind 42195 points

Can you confirm how you are implementing the HR phase shifting? Additionally please confirm what value is assigned to the HRPE bit?

https://e2e.ti.com/support/microcontrollers/c2000-microcontrollers-group/c2000/f/c2000-microcontrollers-forum/982965/tms320f28379d-tbphshr-register

Regards,

Cody

0 Jason Galtieri1 over 3 years ago in reply to Cody Watkins

Prodigy 165 points

HRPE is 0 on all PWMS

For these examples I am implementing HR control as

PWM1 is master, tbphs, tbphshr unused. sends out synco pulse on counter = zero

PWM2 is slave tbphs = 261 right now, (TBPRD = 499),

I am setting tbphshr from the expression window with the values you see above (typically 0,32768, 64000) for a 0%, 50%, !95%

0 Jason Galtieri1 over 3 years ago in reply to Cody Watkins

Prodigy 165 points

Maybe another clue, but writing to TRREM, instead of TBPHSH, a value 1-255 doesn't move either EPwm2A o EPwm2B. Writing to TBPHSHR seems to at least move EPwm2A

0 Jason Galtieri1 over 3 years ago in reply to Cody Watkins

Prodigy 165 points

Still haven't been able to get this to work as expected. Looks like the PWMA edges are moving correctly, but the PWMB edges are only moving on one side. Could this be running into some limit region HR doesn't like?

TBPHSHR 0 -> 64000

0 Cody Watkins over 3 years ago in reply to Jason Galtieri1

TI__Mastermind 42195 points

I know this sounds rudimentary, but I assume that you have read back the configuration or the BEP bit?

As for running into the keep out zone, are you within 3cycles of the beginning or end of a period?

This is very interesting that the edges aren't affected by writes to the TRREM register. Nima Eskandari Have you seen this before?

Regards,
Cody

0 Jason Galtieri1 over 3 years ago in reply to Cody Watkins

Prodigy 165 points

Verified the register values in the watch window. The tbphs values for what I am showing here are in the middle of the period 240-261 (for a period of 499), so I don't think I am running into anything due to placement at the beginning or end of the period.

0 Jason Galtieri1 over 3 years ago in reply to Cody Watkins

Prodigy 165 points

Looked at the probes with PWM1 and PWM2 on the screen, so I could use PWM1 as a reference. PWM1A is shown in magenta. Looks like both edges of PWM2A move when I change TBPHSHR from 0 to 64000 and neither edge of PWM2B move with a changing TBPHSHR. Attached a image of my EPWM2 register values where I believe I am correctly setting EPWM2B to be HR, but hopefully I am just making a silly mistake.

PWM2 TBPHSHR = 0

PWM2 TBPHSHR = 64000

+1 Jason Galtieri1 over 3 years ago in reply to Cody Watkins

Prodigy 165 points

Okay, I believe the problem is I am in upcount mode which always starts an edge at ctr=0 and messes up something about the HR edge placement, due to being in the deadzone

This talks about the problem but in cmpahr
e2e.ti.com/.../tms320f280049c-hr-dead-time-on-outb

C2000™︎ microcontrollers

C2000 microcontrollers forum

TMS320F280049: Adding HWPWM TBPHSHR to DAB converter causes loss of waveform symmetry