How to implement the design of a bidirectional buck-boost converter?

Question

I need your help in designing a bidirectional buck boost converter with the following requirements:

Input voltage: 100V-400V
Voltage range: 250V-1.2kV
Output current: 40-80A
Power dissipation of 24 kW.
Switching frequency: 20kHz

I did some designing on LTspice. I would like to know if this looks right.

That is not the MOSFET I am using for my design. I would be using either GaN or SIC MOSFET. I chose to use an inductor rather than a transformer. Please help me with the right direction to accomplish this design.

You need to add much more detail to your question or it may be flagged for closure. First question: please define what you mean by "bidirectional". — Dwayne Reid, Jun 03 '22 at 16:47
II want to be able to charge a battery in either boost mode or buck mode. So I can either step the outvoltage from the input to the load or step down from the load to the input. — victor_uk, Jun 03 '22 at 16:49
I am not using this mosfet, it is just an example. I am using something with a higher voltage and more power dissipation. I can also add an extra mosfet to share the current between the mosfet. I just need some help on how to go about this design. — victor_uk, Jun 03 '22 at 16:58
The topology is correct, but for a first contact with this design you should implement a 100-200 W converter to learn the details and the questions around them. D2 is a killer, remove it — Jens, Jun 03 '22 at 17:01
D2 was a mistake. I would take it off. I am trying to design this for my final year project. I need help with the right implementation. — victor_uk, Jun 03 '22 at 17:04
Andy Specifically, do you know which mosfet to use, or should I try to place more in parallel? — victor_uk, Jun 03 '22 at 17:06
Have you ever built a buck boost or other kind of switching converter at lower voltages? — bobflux, Jun 03 '22 at 17:10
Each of your gate drivers will require a separate, isolated 20V supply grounded to the source of the transistor it drives. Check out the appnote for your gate driver: https://www.analog.com/media/en/technical-documentation/user-guides/EVAL-ADuM4120EBZ-4120-1EBZ-UG-1109.pdf — Bryan, Jun 03 '22 at 17:10
Is this going to be a paper design, or are you actually going to build hardware? If the latter, it's not really a suitable project for someone without a lot of experience in working with power converters and high voltage circuitry. If it's just a paper design with simulation then you should be OK. — John D, Jun 03 '22 at 17:14
@bobflux No, I have not built a buck boost, but I have designed one separately. — victor_uk, Jun 03 '22 at 17:16
This is one of those rare questions where I hope it is a school assignment that will never really be built. — JRE, Jun 03 '22 at 17:20
@JRE if it's not going to be built. Why would I be asking the question here? That is something I feel very interested in doing, so why do I ask the question? — victor_uk, Jun 03 '22 at 17:24
I was asking because if you want to learn about a topology, you'll want to probe it (a lot), mess with it, measure stuff, etc. That would be a lot more user-friendly at non-lethal voltages, where you can use standard probes, etc. Not to mention, what kind of test bench will you use for this kind of power? — bobflux, Jun 03 '22 at 17:57
@bibflux Thanks for your response. I am a very high-voltage auxiliary power supply. We have it at my university because it has a high power board. I am being monitored by the lab technician. I really want to try the high-power design and I have been reading Christophe P. Basso's books, which give a better explanation of the design with less power. I have a big dummy resistor to use as the load and an oscilloscope for my measurement. — victor_uk, Jun 03 '22 at 18:12

Andy aka · Accepted Answer · 2022-06-03T20:42:45.313

2

You need to choose a switching device that looks like it has a prayer of working

From comments: -

G2R120MT33J That the Mosfet. I am using. – victor_uk

Here's the safe operating area from the data sheet: -

I've targeted two points on the curves (orange and red). Both points suggest a peak power of 5 kW i.e. the curve I've chosen is the maximum instantaneous power that this GeneSIC device is rated at and, in my experience of GeneSIC devices, they will die if you much exceed it.

So, the scenario where you get maximum instantaneous peak power dissipated in the MOSFET is when your MOSFET-bridge changes state and reverses the inductor's voltage polarity. You will find that for a short length of time the MOSFETs will have about half the line voltage across them at about half the current in the inductor (just a simple and approximate rule of thumb that is always best checked for in simulation).

So, do simulations for each scenario and check peak power dissipated, average power dissipated, peak voltage seen from drain to source and peak drain current. You can't simulate enough when it comes to converters of this sort of power. Leave no stone unturned.

As an example, a recent design I did was a 10 kW converter and it worked first time with no value changes. It's this sort of target you need to aim for because if it all goes up in smoke you have no chance of diagnosing what went wrong. Sure, I eventually had the odd MOSFET failure in later testing but, unless you get that confidence early on you'll start to feel the threat of that high-voltage and that'll eat you. So, "do sims" until you are sick of it. Then spend another 5 days doing more sims.

Going back to my approximate rule of thumb...

If the peak voltage is 1200 volts (as per your spec) and the peak inductor current is 80 amps (as per your spec), a reasonable estimate of peak power dissipated in the MOSFET is 40 amps × 600 volts = 24 kW and miles over what this device can achieve.

The problem with GeneSIC devices is that they just won't give you a peak dissipation figure for (say) 10 μs (unlike other suppliers). If you found a different supplier that gives a 10 μs figure, you'd find that the peak power dissipated might be 20 kW to 30 kW and then, you'd be in with a chance.

The package is a tad weedy too. I've used similar dies from ON Semi packaged in TO-247 and in SMD (like yours) and the TO-247 is always going to muscle through without failure compared to the SMD package. Apart from anything else, you are going to need a substantial heatsink for this level of power throughput and the SMD parts just won't compare the heatsinks for the TO-247.

Your MOSFET gate drivers will not work

They need to be isolated drivers that can push maybe an amp or two into the MOSFET gates. There are devices available of course.

Static MOSFET power dissipation

If the average current in the MOSFET is (say) 40 amps, then there is a power dissipated in the MOSFET of \$40^2 \times R_{DS(ON)}\$. \$R_{DS(ON)}\$ is quoted as being 0.12 Ω hence there might be a static power of 192 watts. This maybe halved to 96 watts for 50% duty cycle but, it's still a lot of power to dump from a TO-263 device.

edited Jun 03 '22 at 20:42

answered Jun 03 '22 at 18:29

Andy aka

434,556
28
351
777

Thanks so much for this very helpful start. I'd have to look for a different mosfet, and due to the semi conductor short, it's difficult to find something better. I am only using the gate drive on LTspice for simulation to see the response of the circuit. This is the type of gate drive I am using UCC21521. – victor_uk Jun 03 '22 at 18:38
That looks a good choice on the face of it. – Andy aka Jun 03 '22 at 18:41
I believe this is a better mosfet NTBG020N120SC1. – victor_uk Jun 03 '22 at 18:48
@victor_uk yeah I use 40 milli ohm versions of that ON Semi device but, the voltage rating is only 1200 volts. I didn't want to tell you what I used because you should find a device by yourself. I still think you might be on the edge though. but maybe OK if you can live with a max working voltage no greater than 1,000 volts. Anyway, you know how to choose a decent MOSFET now and, that's the important thing. [Please take the 2 minute tour to find out why folk give free advice](https://electronics.stackexchange.com/tour). – Andy aka Jun 03 '22 at 19:01
Thanks again. I saw something much better. Does the schematic look okay to achieve this output? – victor_uk Jun 03 '22 at 19:40
It would look better with the right part on it hence, why don't you add an update section to your question. You WILL need to simulate this to death and then simulate again and again if you want to get the details right. Spend much more time testing scenarios in a simulator than anything else. You might not see it that way but I do know the truth about this and there ain't a good circuit on paper for a power converter that hasn't been strongly tested in simulation. Leave absolutely nothing to chance. – Andy aka Jun 03 '22 at 19:55
This is what I am trying to design here with my specification for the buck boost. – victor_uk Jun 09 '22 at 10:29
@victor_uk unfortunately you have overwritten your original question with a new schematic that now makes my answer look silly. What you should do is add the new circuit diagram and notes as an amendment section below the original text/schematic. There is no particular format for this i.e. just leave the original question intact and add a heading below it with the new details following that heading. – Andy aka Jun 09 '22 at 10:43
You must not have the extra source pins on the upper MOSFETs connected to ground. If you don't respond I will roll-back the question to how it stood prior to today to make my answer match. Alternatively, you might consider asking a new question with this new circuit but make it clear that the new circuit is a follow-up to this Q and A session and the advice you got from this session. – Andy aka Jun 09 '22 at 10:49
I am connecting it between two capacitors in series to the common of a DC/DC isolator (MEJ2D1209SC). I can not find anything close to this on LTspice or if there is any way I can connect. Are my inductor and output capacitor values right? – victor_uk Jun 09 '22 at 11:00
@victor_uk I'm sorry but I'm going to roll-back your question to how it originally stood when I answered it on June 3rd. Here is your new image should you want to use it again: https://i.stack.imgur.com/pRcU4.png - if you want help you must respect the site rules and practices. – Andy aka Jun 09 '22 at 11:13
Yes, I should not have done that, and I do know why I deleted the old one. Sorry! If it is a complicated thing, I will keep to the rules and practises here. – victor_uk Jun 09 '22 at 11:30
Maybe it might be easier to ask a follow-on question because the circuit has evolved somewhat and most folk here hate questions that evolve because it means more work in modifying the answers and, there is a danger that a good original question and answer become lost in the amendments. Remember that this is a Q and A showcase so, maybe ask a new question. You should also "complete this session" as per the 2 minute tour information I mentioned on June 3rd. – Andy aka Jun 09 '22 at 11:33

score 0 · Answer 2 · answered Jun 03 '22 at 17:15

I can show you a 1 GW SMPS design to solve your problem which is around 100 kW and get > 92% efficiency. You're are getting probably <50% efficiency with 24 kW power dissipation. (Guestimate from VI range)

Please understand, this will be much more complicated than your model.

You need to define all sources of voltage error, and regulation errors fron source impedance , stored energy impedance (sqrt L/C vs Load Z(f) ) and load variation (step, ramp, and steady) to examine all the losses of regulation error.

If you can add clarity to your design expectations for stability and efficiency with all the above parameters, then a solution might be fitting to your problems with your design specs and component choices for Q, DCR, ESC and impact on all the above in configuration choices.

Do not get lost on component choices until you have done the above. Then we can break it down into more details like Q, Bode Response, phase margin and power loss budget with stability.

Please understand this has been done before if you follow this design process.

How to implement the design of a bidirectional buck-boost converter?

2 Answers2