Other Parts Discussed in Thread: UCC24610, , TPS23731
Tool/software:
The related thread touches this topic. For a given design ( Using PSR-control using Aux winding and synchronous rectification seems to suffer of increasing output voltage on low loads) I opted to specify no longer the -1-suffix, but the -2-suffix. The reason for this: on the first prototypes the on time was terminated too early (partially due to -9 mV threshold and low MOSFET Rds on). Given the switching frequency of approx. 230 kHz, and the longer guaranteed on-time of the -2-suffix-type, I selected this device. Especially also because table 6.3 says that the maximum switching frequency range of the -1-suffix-device is highest, and far beyond the switching frequency of the application.
In the course of the second prototype iteration test, I also see that the delay from VD on threshold to gate driver on is 180 ns. But that's quite significant with a PWM period at 800 kHz! That could result in on third of the secondary conduction time interval going with body diode conduction (which however is less a problem on certain topologies). When the secondary current is highest just after the secondary is ought to carry current to the output, this makes the given max. frequency interval quite absurd. However, these days I also saw the graph as cited in the related thread: Here additionally the delays are specified (as well as a topology list). So, suffix -2 is abandoned, because 180 ns results in too much loss in a flyback design. Even 5% of the secondary current ramp down time is unnecessary long (taken 2600 ns secondary current cycle part that would be 130 ns) results in unnecessary extra loss.
Additionally, I regard it as curious when newer parts (compared to the older UCC24610, which specifies a tough 44 ns delay) is faster by a double magnitude or more! Therefore these devices are abandoned for the next iteration, where I'll switch to MPS or onsemi devices. They're faster (30 ns) and the latter also does not switch the Mosfet off at -9 mV (typ). but at 0 mV, which allows me to select a Mosfet with lower Rds on.
Another problem is the given minimum VD-voltage according to absolute maximum ratings: may not be lower than -0.7V. In the xxx ns time interval that the gate is not driven, the full body diode voltage is exposed to VD which may well become 1 V (its typical value given a current is not specified in the datasheet). Well, the datasheet states that the voltage on VD may be -1 V as long as no more than 10 mA is drawn from the device. That suggests that without any series resistance to VD no robustness of the application can be guaranteed. I'd need to insert a series resistance of at least 30 ohms. But such a resistor is not even found in the ucc24612-1EVM. Anyway, it would be helpful if it would be specified whether or not 10 mA max. is guaranteed also without series resistor. Having said this I now realize that a series resistor would just limit the voltage to -0.7V when the drain voltage would be -1V. So this -1V/10mA spec is a bit weird... This value is NOT weird if the VD pin would include a 30Ω-resistor. But in that case, the -0.7V would be superfluous and be replaced by a -1.0V spec without accompanying current specification.
Luckily, I have designed in a series resistor, but its primary purpose is in a helper circuit that should support full CCM-behaviour from low to full load (as is the only mode that is supported by the tps23731).
