TMS320F280049: High resolution duty cycle and dead band?

Hi Mike,

I believe that on devices newer than F28M35x you can do HR edge control on both ePWMA and ePWMB + HR deadband. 

The differences between peripherals will depend on which specific devices you are looking at.  Check out the following document:

If you are looking at F28004x vs F2807x vs F2837xS vs F2837xD, things should be pretty similar.

This contradicts most of the other info I've already found. First of all, the TRM for the TMS320F28004x says it has a type 2 HRPWM (never head of type 4...). Also mentions in the intro "Enables high-resolution output on EPWMxB signal output via inversion of EPWMxA signal output," which would imply that HR operation of channels A and B is not really independent. As in, yes, channels A and B support HRPWM, but not independently at simultaneously.

Also I should clarify that I want to set this up such that I don't need to reconfigure registers every PWM period, as is the case with the C2000ware example hrpwm_deadband_sfo_v8. I don't need to control the period or phase offset with high resolution. Also I'm trying to use up-down count mode, if possible. Can you recommend an initialization which allows this without using ISRs?

Regards,
Mike

Here's a plot demonstrating roughly what I'm trying to achieve. This example uses a very coarse timebase (TBPRD=12), just to make it easier to see here. In reality it would be more like TBPRD=250.

The dashed purple trace (or the reference, as I'm calling it) represents the ideal waveform I'm trying to generate on my half bridge (ignoring dead time). The reference is center-aligned, and has duty cycle 0.5208 (12.5/24 TBCLK). Orange/yellow is the complementary pair I'm trying to produce, with deadbands inserted symmetrically around the edges of the reference (RED=1TBCLK, FED=1.5TBCLK). 

There are a couple problems I can see with generating this set of waveforms (with one ePWM submodule, that is).

1. Based on my understanding of the TRM, when using high resolution deadband generation, the deadband has a minimum duration of 15ns (TBCLK/2*3). This is unacceptably long for may application (need to go down to at least 5ns). So I can't make use of the high resolution deadband function.

2. If the two deadbands are not equal, then the PWMA/B waveforms won't be equal aligned to each other. I don't see how they can both be generated from a common timebase...

Can someone tell me if I'm off-track with my reasoning? Unfortunately none of the examples I've seen seem to demonstrate both duty and deadband control with high resolution. Some other threads have referenced the project hrpwm_deadband_sfo_v8 as an example of doing high resolution control of duty/deadband/period/phase at the same time, but from what I can tell it relies on a great deal of CPU overhead to do so (and I'm not sure it handles different deadbands for RE and FE). It's not clear how much high resolution control is actually possible without CPU overhead.

HI Mike,

Can you confirm that you have the most recent documents from here:

I see the following for the datasheet and TRM respectively:

I'll ask someone to take a look at your specific use case too.

Yes I know the ePWM is type 4, but the TRM specifies type 2 HRPWM:

Most documents seem to treat the ePWM and HRPWM as effectively separate peripherals, which may have different type. For example, table 10 in SPRU566M lists a type 4 epwm and type 2 HRPWM for the F2807x.

In any case the distinction between type 2 and type 4 isn't made very clear in any documents I've seen. I also don't see how this distinction relates to my question.

Regards,

Mike