boost + buck-boost vs Full Bridge - price, eficienci, etc. at kW levels

[Miyuki]

Hi folks,
I want to get positive and negative 350V for the AC output inverter, from a 120V DC system
Desired power level is in 3-6kW levels.
It is just a personal project to try something new.
I can't decide what will be a better approach. I do not need isolation when it will have DC negative tied to AC neutral.

If it will be better to use interleaved boost and buck-boost to generate separate positive and negative lines.
It can be 3 phases interleaved, which will give me small components and almost no ripple current.
And it is really easy to design and use off-the-shelf inductors and so.

Or to go way with isolated topology, something like a Full-Bridge with a beefy transformer, 3kW level does not seem undoable on the paper with single-stage and still can offer reasonable simple paralleling for higher power

I do not think LLC will offer reasonable benefits for me and it is a rabbit hole to design.
So ill keep it boost+buck-boost vs hard switched full bridge

[NiHaoMike]

I want to get positive and negative 350V for the AC output inverter, from a 120V DC system
Desired power level is in 3-6kW levels.
It is just a personal project to try something new.
I can't decide what will be a better approach. I do not need isolation when it will have DC negative tied to AC neutral.

Just buy a Prius inverter (about $100) and replace the logic board with your own. Well worth it for the IGBTs alone, then you also get the gate driver board, capacitors, boost converter inductor, and a few other useful extras.

[Miyuki]

Just buy a Prius inverter (about $100) and replace the logic board with your own. Well worth it for the IGBTs alone, then you also get the gate driver board, capacitors, boost converter inductor, and a few other useful extras.

An interesting piece of tech, I'll consider using it when I might do some EV conversion in the future
But not what I need here

[NiHaoMike]

An interesting piece of tech, I'll consider using it when I might do some EV conversion in the future
But not what I need here

If you replace the logic board with your own, you basically get a general purpose inverter.

The Prius inverter can handle 10-20kW continuous and on the order of 50kW peak. Just the parts to build an equivalent power stage would cost a lot more than $100 even if you somehow got it working perfectly the first time around.

[Wolfram]

I have been involved in using the Prius inverter (Gen. 2 and 3) for other applications before, and I would definitely not recommend it for this purpose. The MG1/MG2 IGBTs are dimensioned for hundreds of amps, and they are pretty slow. To keep switching losses manageable, you need to operate in the 5 - 20 kHz range, which means large magnetics, and efficiency won't be amazing in any case. Also, you would be using the phase current transducers at a small fraction of their intended range, exaggerating the effects of offsets and noise. Thirdly, you need water-cooling which is a large burden for a project in the sub-10 kW class. I would design something from scratch here, which might be a bit more effort but also a valuable experience, if you are comfortable working with the voltages and currents involved.

Boost followed by an inverting buck-boost would be my suggested approach, but a full-bridge with a center-tapped voltage doubler on the output side (with separate inductors per rail) is also an option, especially if the cross-regulation is not extremely critical. The latter options will likely be easier to compensate too, due to a lack of any right half plane zeros in the plant transfer function and the lack of cascaded converters.

I'll add a few notes based on my own experiences. Generally, I'd recommend fully galvanically isolated drivers for both low and high side devices, they are much less sensitive to switching node undershoot and ringing than the common high/low side driver options. Also observe the CMTI rating of any high side drivers, though with only a few hundred volts on the primary side this is less likely to be an issue. Hardware protection (especially during prototyping) is often a lot cheaper than even a single set of new power devices and gate drivers. And at these voltages and power levels, failures tend to be pretty catastrophic. I've still lost some hardware through stupidity (and most of the time, the hardware ends up being a total writeoffs), but the hardware protections have saved us around a hundred times as much carnage. Current mode control is nice and simplifies a lot of things, while overcurrent protection is mandatory. Also output overvoltage protecion on converters with a boost-type transfer function.

[Miyuki]

... a full-bridge with a center-tapped voltage doubler on the output side (with separate inductors per rail) is also an option, especially if the cross-regulation is not extremely critical. The latter options will likely be easier to compensate too, due to a lack of any right half plane zeros in the plant transfer function and the lack of cascaded converters.

I'll add a few notes based on my own experiences. Generally, I'd recommend fully galvanically isolated drivers for both low and high side devices, they are much less sensitive to switching node undershoot and ringing than the common high/low side driver options. Also observe the CMTI rating of any high side drivers, though with only a few hundred volts on the primary side this is less likely to be an issue. Hardware protection (especially during prototyping) is often a lot cheaper than even a single set of new power devices and gate drivers. And at these voltages and power levels, failures tend to be pretty catastrophic. I've still lost some hardware through stupidity (and most of the time, the hardware ends up being a total writeoffs), but the hardware protections have saved us around a hundred times as much carnage. Current mode control is nice and simplifies a lot of things, while overcurrent protection is mandatory. Also output overvoltage protecion on converters with a boost-type transfer function.

A full-bridge with voltage doubler sounds like a good option, just two windings, and two diodes
But how does it cross regulate? It can end up with an asymmetrical load 

Good point with isolated drivers, Infineon has a nice line of isolated drivers.
Current mode control seems safer and easier in almost all aspects.
I plan to do most of the testing on a power supply with current limited to avoid spectacular failures

[Benta]

I'd always go for the full-bridge when talking those kind of power levels.
Simply because the amount of ferrite and copper you need is much lower than with storage-based magnetics. Do the calculations yourself.
You'll be surprised at the amount of inductive storage you'll need for 3...6 kW.
Even a capacitor-splitter half-bridge is a better choice.

[uer166]

I'd always go for the full-bridge when talking those kind of power levels.
Simply because the amount of ferrite and copper you need is much lower than with storage-based magnetics. Do the calculations yourself.
You'll be surprised at the amount of inductive storage you'll need for 3...6 kW.
Even a capacitor-splitter half-bridge is a better choice.

Eh, in my mind it's easier to design it based on normal boost (or inverting buck boost, which has the same transfer function) topology. I've seen many PFC front-ends in the 10-40kW range and the size is not that ridiculous, in fact comparable to a LLC converter that sits downstream of the PFC. I'll second the isolated gate drivers, even if technically not required.

[Benta]

I love "I've seen".
Did you ever do the maths? And where did I use the word "ridiculous"?

[uer166]

I love "I've seen".
Did you ever do the maths? And where did I use the word "ridiculous"?

Why so confrontational? All I'm saying is I think at this power level a non-isolated topology works just fine and would be easier. And yes, "I've seen", as in, the team I work in parallel with designed 40kW converter modules that include the above mentioned power stages. I have designed interleaved boost converters in the 2kW range, where each inductor is wound on a ETD54 core, and loads up about 1kW. So uh, yes, I have absolutely done the math: all the energy is stored in a the air gap, so if done right, it's not much bigger than say a LLC stage.

[NiHaoMike]

I have been involved in using the Prius inverter (Gen. 2 and 3) for other applications before, and I would definitely not recommend it for this purpose. The MG1/MG2 IGBTs are dimensioned for hundreds of amps, and they are pretty slow. To keep switching losses manageable, you need to operate in the 5 - 20 kHz range, which means large magnetics, and efficiency won't be amazing in any case.

For this application, you can switch the IGBTs at a rather slow 60Hz rate and shape the output waveform by varying the rail voltage using the buck/boost converter. Without further filtering, it will be somewhere between modified sine and pure sine, good enough for most large loads.

Also, you would be using the phase current transducers at a small fraction of their intended range, exaggerating the effects of offsets and noise.

Remove the bus bars going through the sensors and replace them with several turns of magnet wire.