48V to 400V, 1600W, isolated push-pull converter design (with Ltspice files)

[rodv92]

Hi. This is my first post here.
I am currently learning the intricacies of various SMPS technologies and I decided to design a high-power push-pull converter as an exercise.
The idea was to generate a design that could have practical uses, as it is intended to operate on a Telco (48V) SLA battery bank,
and get an output voltage of 400V DC at a nominal power of 1600W so it can drive a UPS circuit (using a control board like the EGS002)
400V is also the same voltage as PFC would output, so it could enable seamless load sharing/load switching from a PFC (powered by mains)
and the push-pull converter.

Keep in mind that I have more experience in design than in prototype realization, so the article and model mainly deal with the design
phase, not much with the prototyping phase. But overall, the simulation gives adequate results and moderate to good efficiency. (between 0.92 and 0.94)

I decided to test the use of non-linear inductors, which could be useful to test the operation over the upper design limits of output current (into core saturation)
In the linear region, I doubt that the Ltspice core Chan model will make a lot of difference.

The difference over the basic designs offered by Analog is that I changed it into an insolated design, so there is an optocoupler for the feedback network.
Since Ltspice complains when having a separate ground, I had to stitch the GND and COM nodes with a 0R resistor for the simulation to work.

I had also some doubts about using the Core Chan model using the Iron powder material specifications. Using them as is in the Core Chan model gave way too high (over 10000X)
Al values, So I had to get the datasheet values by increasing the Lg (gap length). I assume that the material datasheets are for the 'pure' iron powder composition,
and do not take into account the distributed gap of the finished pressed cores. If anyone could shed light on the issue, I would be grateful.
I used the Ltspice test bench from this thread to get the adequate Al value by increasing Lg :
https://www.eevblog.com/forum/projects/arbitrary-%28saturable%29-coupled-inductors-in-ltspice/


I did not include the design formulas are these are already available in most of the resources mentioned.
Errors or questionable design choices may linger, So feel free to comment.

There is also the optocoupler issue. I would like to get rid if possible of the Howland current pump by using a simple shunt resistor to limit current flowing into the
optocoupler diode, but that gave rise to unacceptable noise in the current waveform. Without the Howland current pump, there would be no need for a low DC power bus on the isolated
side, But it could be a moot issue since any decent design would probably need a 5V or 3.3V source to operate other logic on the secondary side.

As for power, the primary uses an LM317HV powered from the battery bank with a 60V Zener TVS protection.
The secondary DC power for low-level DC voltage components also uses a LM317HV, powered by an auxiliary winding.
The circuit takes around 60 ms to start switching

The model is attached.

The main article detailing the circuit and simulation operation:

https://www.skynext.tech/index.php/2023/06/11/ltspice-model-of-an-isolated-48v-to-400v-1600w-push-pull-converter-using-chan-model-inductors-using-the-ltc3721-ic/