UCC28780EVM-002: UCC28780 RDM RTZ

Hello Eleven,

Thank you for your interest in the GaN-based EVM for the UCC28780 ACF controller.

Values for RDM and RTZ resistors are calculated based on several circuit parameters (see datasheet equations 8, 9, and 10) specific to each particular design. The values used on UCC28780EVM-002 are appropriate for that design based on the other parameters of that design.

You mention that these values differ greatly from those of a Mathcad calculation, but you do not specify which design that Mathcad pertains to. Very likely it is based on a design with significantly different base-line parameters (such as magnetizing inductance and/or turns-ratio, etc.).

If you are referring to the Mathcad program that is down-loaded from the TI website (myportal.ti.com/.../dt ), then this program contains example parameters and value calculations based on an arbitrary design at the time that the program was produced. It is not the design of the GaN EVM and many of the basic parameters are significantly different than those of the EVM. Therefore the results for RDM and RTZ would also be significantly different.

It is intended that the user will change the various parameters as necessary to match the intended new design targets, each time.
I hope this clarifies the issue.

Regards,
Ulrich

Hello Eleven,

For your question 1:
Yes, it is not easy to make an accurate estimate of Csw. However some estimate does need to be made, because Csw and Lm determine the partial-resonant transition time of the switched-node voltage (Vsw) from high to low. This interval can be a significant portion of the total switching period, and the value of Lm is targeted from a desired minimum switching frequency. The estimate of Csw will be based on the (GaN or Si) MOSFET datasheet curves of Coss, plus estimated transformer winding capacitance and reflected rectifier capacitances from the various other windings.

These estimates will not be very accurate, but the overall error is usually small unless a very high minimum frequency is chosen. Typically, the resonant interval is about 10% of the period, and since resonance goes by the SQRT of LC, a 40% error in Csw results in a about 20% error in the 10% interval, so maybe 2% error in the total period. Of course, the error is worse if the resonant interval is a larger portion of the period.

RDM is not dependent on Csw, but it is dependent on Lm, Na, Np, Rcs, Rvs1 and Rvs2 (see datasheet equation 10). If any of these parameters are adjusted during prototype debug, RDM should also be re-adjusted.
RTZ is dependent on Csw, Lm, Ns, Np, Vout and Vbulk_max (see equations 8 & 9).

Q2: Correct, RDM sets the first TDM time in a burst in ABM. After that cycle, the internal tuner adjusts timing to optimize zero-volt- switching (ZVS). RDM and RTZ are set for the nominal case of all component parameters. The tuner has a plus and minus range to accommodate parameter variations in Lm and the resistors.

Q3: Yes, the timing is adaptive, but relies on RDM and RTZ to have been set for nominal conditions in order to be able to adapt sufficiently to variations in either direction.
If RDM is too high, the enforces high-side on-time will be too long. The tuner will be unable to compensate fully and too much negative current will be generated. This then increases the positive current to compensate for the additional power loss and it increases power loss even more.
If RDM is too low, the high-side FET may turn off before the transformer has fully demagnetized, interrupting the active-clamp resonant current and causing high switching losses in the high-side FET and SR FET. And it will disrupt the tuner operation as it hunts back and forth to try to restore the optimal switching points.
If RTZ is too high or too low, it will shift the turn-on point of the low-side FET to either side of the resonant valley of the drain waveform, leading to hard-switching instead of ZVS.

I hope this answers your questions.

Regards,
Ulrich