I am starting the design of a 750W power supply to charge a battery or ultracapacitor bank. In this application the power supply limits the current for a long time, than limits the voltage to finish the charge (this is a simple charge algorithm, known as CCCV, or constant current, constant voltage).
For the rectifier/PFC stage, we’re designing a 2-phase boost converter with TI UCC28063 controller. The PFC output is 380Vdc.
For the DC-DC stage we choose to use a half-bridge converter. This topology was chosen only because it is simple and fast to build and the cost is reasonable. But, as it well known, the voltages of the capacitor divider can get unbalanced when the half bridge operates in current control. I found an IC that promises to solve this problem, the LM5039 (Half Bridge controller with average current limit). Unfortunately, the drivers of the LM5039 are for 100V only. Still, I intend to use this controller with an external 600V driver.
I would like to ask the group if someone uses the LM5039 with current limiting with good results and if someone knows of a similar IC for higher voltage, so I can avoid using the external driver.
Some other comments on this project? Thanks in advance!

Just a question appeared in my mind when i saw that topic. Can we try asymmetric HB topology here? Then you may not need split capacitors, transformer will not drastically change and you will have ZVT???

Why not use a 2-switch forward converter? Much simpler solution. I have seen the 2-switch forward used in shielded arc welder power supplies with success.

I thought the half-bridge had problems with just peak current mode control. I haven't looked into any problems with the half bridge and average current mode control, which is what you would be using with your battery charger. You can eliminate the top capacitor if you want...the bottom one then acts as a DC blocking capacitor for the transformer primary. Either way, I would use the 2-switch forward if I were in your shoes.

That's right - the 2-switch forward converter is simpler and very rugged, and you can use peak current control. But the transformer is utilized in only one quadrant, which means it must some times be one size bigger. Also, the output inductor must normally be larger because duty cycle (on the output) is limited to 50% - with a 1:1 demagnetization winding at least. And you must still use transformer drive or a high side driver for the upper fet.
The choice is really a matter of individual preference and experience. Both solutions + other solutions can do the job.
However, I think I would say that the 2-switch forward is more "simple and fast to build" than the half bridge.

One thing to keep in mind.... Ripple current can heat up batteries during charge cycle...
It can also alter the batteries aging... as well as disturb the battery chemistry to the point of providing a false dynamic SOC reading .. Just to take note....

Before current mode came along, the half-bridge was quite popular. Due to this problem mentioned, it almost disappeared in a period of a year or so.

A few companies clung to the topology with the magnetics fix mentioned, but it is yet another complication to the circuit. Added to shoot through possibility, active current balance, etc.

If you need speed of design, the two switch forward is much better. If you want to push frequency a little harder, and get the size down some (not twice), the half bridge is OK but be prepared for many more failure modes along the way, and a longer development time.

I saw it used one time in a welding application at 5 kW. 40% of the converters were failing during burn in, basically because no current limit was in place. Once the current got out of control, they couldn't prevent shoot-through. Doesn't mean you shouldn't use it - just be prepared for complications that are not in the papers or app notes.

Once you go up the technology curve further, the half bridge comes into play again with the LLC but that has the coupling cap to balance

Well - if you omit the current limit then you can hardly blame the topology. That said you are quite right to warn against a topology with inherent instability.

For old fashioned HB designs another source of instability (possible cause for shoot-thru) was the gate drive transformer - that can get saturated if symmetry is lost.

LLC does give a much more efficient solution, but not many engineers seem to be able to add a constant current mode to that one - do you have any ideas for that?

Tonu, TKS for referencing the article I coauthored. Didn't know if anyone out there ever read it after publication. This was a 120W half-bridge peak current mode design implementation currently in use on the space station main communication antenna.

Felippe, the 120 W design approach was inspired from the TI (Unitrode) paper that had been published. Unfortunately, it wouldn't work as is, because the balancing winding current influenced the current feedback loop information (I went into details of this issue in my article) and things became 'unbalanced' when operating in current limit mode. Once this the balancing winding information is compensated for in the feedback loop (reference the Fairchild article), the half-bridge peak current mode topology works very well.

Now this is quite true at the power level for which I designed the power supply (120W). There is always the possibility of unforeseen issues interfering with this method at 750W (i.e. a lot more balancing current may be required, will the balancing winding have time to do its thing?). The balancing winding current is not instantaneous due to inherent winding inductance (and wire resistance).

I appreciate all the interesting comments that this question has generated.

Runo, I read your paper about the HB with current signal injection and I can see that you have studied a lot about this question. I also took a look at the Unitrode paper and some other papers. But they all imply some additional circuit that will increase the development time (because I would probably have to do it more than one time to get it right). My hope was that the LM3039 had solved this question in a simple (if not “transparent”) way to the power supply designer. But, considering the comments above, I think now this is not the case.
You are right, it is wise to avoid saying that it is “simple and fast to build”. I also had my share of blowups to teach me that. I just thought that the HB would be simpler compared to other choices. In fact, I would like to build an LLC, just because I have never built one, so I would like to try it. But in this project I have to deliver the prototypes in a short time, so I have to stick to a more familiar solution.

Based on all the comments, I will change the topology. Luckily, I am starting the design, so this will not be a big issue. I just don’t want to risk a project delay because it is hard to design the HB with current control, and, as Alain said, “unforeseen issues” can interfere with the methods used to balance the capacitors at 750W.

It seems from the comments above that the favorite topology for this project would be the 2-switch forward. The asymmetric HB has also been suggested. I am considering also the full bridge, because it is more familiar to me. At this point, it is more important to reduce the development time than to reduce the cost of the product. The FB will increase the cost of the semiconductors, but it uses the transformer better. It is possible that in the future I’ll have to design a similar battery charger with higher power (say, 1.5kW), and the FB seems to be better suited to higher power levels than the 2-switch forward.

What about interleaving two 2-switch forwards....You could use a simple PWM IC like the SG2524, where both channels are limited to less than 50% duty. I've seen some welders use this IC. You could easily push some serious power with that approach (if you wanted to).

Would there be any way to use a buck current-fed full bridge (or push-pull)? Easy control with serious output power capability and no need to worry about transformer saturation. But then you would have a PFC boost, a buck stage, and then a push-pull or full bridge stage...all in one charger...kind of a bunch of stuff.

Hi Jay, what advantage the interleaved 2-switch forward would have over the phase-shifted full bridge? Probably that's only because I am not used to it... but I don't like that switch floating above the transformer in the 2-switch forward... The full bridge (conventional PWM or phase shifted) uses a common structure for the switches and uses better the transformer. So at this moment it seems to me a better choice...

Yes, adding one more stage would be too much stuff... In fact I already have 3 stages in this project (PFC + DC-DC + a buck converter that goes after the battery or ultracap).

Hi Tonu, I didn't know about the voltage spikes at the output. The controller has a synchronous rectifier capability. If I use mosfets at the output rectifier, I will reduce this problem?

Two switch forward advantage:

Much faster to develop due to less failure modes that can happen.

- no flux control needed in the transformer, naturally resets.
- no shoot through possible with slight timing errors in drive, or noise.
- simpler transformer

It doesn't mean you shouldn't shoot for a full bridge phase shift if you have the time, just be aware that there are many more creative ways to blow it up. The stress testing phase of the project when applying short circuits, transients, etc., can be very time consuming with a bridge.

The high side drive on the forward in no big deal - same as the full bridge, but no concerns about cross conduction so it is less critical.

If you want denser, more efficient, and quieter, the full bridge is the way to go. But if you want it cheaper and by tomorrow, the forward.

When you start using sync rectification with off the shelf controllers then you can find yourself in a spot of trouble if you are not very careful - even more so with a load like battery that can supply practically limitless reverse current to your output.
Maybe I am doing TI and Intersil an injustice but a few years ago most sync rectification controllers still had undocumented failure modes.
Also the resonant part of ZVS FB will need some attention.

So if you do not have lots of time - forward is a safer way to go. Pricewise the higher cost of transformer and choke will be balanced by the lower number of mosfets and lower price of a controller.

Probably TI and Intersil are amongst the best of the bunch. The thing is, almost every controller out there has undocumented regions of operation that you often find the hard way. Some of the controllers are so complex now, it is just about impossible to document them in any reasonable-sized datasheet.

Keeping it simple and sticking to the rugged and well-known controllers used to be a good approach, but the green power mandates are forcing everyone to the latest control chips.

I quite agree that TI and Intersil are among the best control circuit vendors. I am not questioning that at all.
The problem here are the principles of sync rectification. When it is done in a buck regulator then most controllers monitor output current and keep reverse current under tight control. For an isolated converter many vendors still propose sync rectification schemes with no output current monitoring. This seems to be also the case here (UCC1890). In such a situation the design engineer needs to check all possible error situations very carefully.

Thank you all who helped with this discussion. The comments were really interesting and useful.
After pondering the advantages of each topology and a meeting with the client in which we postponed the project deadline, I decided to follow on with the phase-shifted FB. As pointed in this discussion, it will take longer and may cost a little more but we expect better efficiency, smaller size and lower EMI.