• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 82 subscribers
  • Views 1487 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TMS320F28379D: PWM Phase Shift too large if "High-Resolution"-Register (TBPHSHR) is used

Matthias Ampferl
Prodigy 160 points
Part Number: TMS320F28379D
Other Parts Discussed in Thread: CONTROLSUITE

Hello!

I want to use the F2837xD LaunchPad to create four synchronized 100kHz, 50% duty cycle PWMs, each phased shifted with respect to the others. Each phase shifted signal should lag with respect to the reference PWM (Signal 1 of 5).

The scheme I'm using is the following:
I use ePWM1 to create the first PWM signal, which is also used as master for all other PWMs.
For the other PWMs, I use ePWM2 and ePWM4 to 6 as slaves. They load their respective phases value when the Master generates a SYNCOUT (in the event of master counter == 0).


I managed to get the ePWMs running using only the regular ePWM. If I want to use the HRPWM, I suddenly get big errors in the phase shift. For clarification, find below my register values and the measured phase shifts. Do you see an error with this or is there an altogether easier solution for realizing a lagging phase shift?

  phase lag in clock cycles calculated phase value in clock cycles TBPHS TBPHSHR Δt target Δt actual
ePWM1 0 0 0x0000 0x0000 reference PWM reference PWM
ePWM2 4.00 998.00 0x03E6 0x0000 40.00ns 39.7ns
ePWM4 4.25 997.75 0x03E5 0xC000 40.25ns 57.3ns
ePWM5 4.50 997.50 0x03E5 0x8000 40.50ns 54.3ns
ePWM6 4.75 997.25 0x03E5 0x4000 40.75ns 51.8ns
note: In this case, one PWM period has 1000 clock cycles, 1 clock cycle equals 10ns. The phase value is "split" between TBPHS and TBPHSHR with the HR-Register  containing the fractional part.
The phase values are calculated as follows:
PHS = (RELOAD+2 -PHASE) % RELOAD
whereby:
RELOAD = T - 1 = counter reload value used by ePWM
T = PWM period in clock cycles
PHASE = lagging phase shift expressed in clock cycles (can have fractional part)
PHS = absolute phase shift in clock cycles (the value that has to be assigned to TBPHS and TBPHSHR)
2 = correction of master-to-slave time lag (in this case 2 clock cycles)
% RELOAD ensures that no invalid values (bigger than RELOAD) can be applied.




Sincerely

Matthias
Student (University of Applied Sciences Ingolstadt)

  • 0
    TI__Guru 60865 points
    Hi Matthias,

    It seems like the regular phase delay is working, but only the HR portion is not? The HR step size is ~150ps and there are 255 steps, so you can affect the phase by as much as 38ns. You are off by ~10-17ns, so it seems like maybe you are writing an incorrect value to the HR portion?

    Are you using the AUTOCONV bit to automatically apply the SFO scale factor? Are you running the SFO library to populate the scale factor?

    It may be worth double checking your HR scaling against the examples in the TRM section "Scaling Considerations" in the HRPWM chapter.
  • 0 in reply to Devin Cottier
    Prodigy 160 points

    Hi Devin,

    Yes, I am using the AUTOCONV bit.

    Also, my code updates the MEP scale factor periodically. For this I use the SFO() function from the library. The library included in the makefile is C:/ti/controlSUITE/device_support/F2837xD/v190/F2837xD_common/lib/SFO_v8_fpu_lib_build_c28.lib

    I'm not using the scale factor directly, though. According to page 1849 in the TRM, in case of autoconversion usage, only the hardware uses the Scale Factor. The code that is periodically called to calculate a new TBPHSHR value looks like this (It doesn't use the scale factor): 

    //variable that holds the phase shift in clock cycles, e.g. 4.25 clock cycles = 42.5ns
    172   real_T TBPHSf = Converter_DHB2_PWM_LaunchPad_B.Switch2;

    173   TBPHSf -= EPwm2Regs.TBPHS.bit.TBPHS;
    174   TBPHSf *=65536;
    175   EPwm2Regs.TBPHS.bit.TBPHSHR = (uint16_T)TBPHSf;

    PS: The code isn't entirely written by me. I use automatic code generation of the C2000 block set of Simulink. Which is a bit of a hassle at times.

  • 0 in reply to Devin Cottier
    Prodigy 160 points
    Hi Devin,

    can you explain to me why the HR Phase shift register introduces additional phase shift that is negative compared to the "macro" phase shift?
    Is there are setting that makes it behave like the "macro" phase shift? (i.e. HR phase shift increases when "macro" phase shift increases and vice versa).