[SOLVED] Hi eff, hi freq bidirectional DC-DC buck-boost nonisolated converter topologies.

[Akanimo]

Hi,

Please what are the practical bidirectional topologies for non-isolated buck-boost converters for efficiencies above 95% at 1.5MHz switching frequency and output power above 65W?

A long list would be appreciated.

 

[KlausST]

Hi,

What have you done so far? And what are the results?

It sounds as if you want others to do your research job..

Klaus

 

[Akanimo]

Hi,

What have you done so far? And what are the results?

It sounds as if you want others to do your research job..

Klaus

I'm doing a lot of analysis here, trying out different configurations. Calculations are massive. Issues here are that the magnetic core is supposed to be very light, probably less than 5g and this means a very small window area. At the end the inductor is copper loss limited. I've gone past that point though so I came up with an optimum frequency around 1.5MHz giving me a winding resistance lower than 1ohm. So I now have to tackle switching losses. I'm practically leaving out the use of diodes to reduce switching losses and to ensure CCM so I'm going to use synchronous switching.

I have also looked up some topologies on the internet but the switching frequencies are quite low for this efficiency spec considering switching losses at 1.5MHz. The highest I found was about 1MHz but output power was way higher than 65+-5W indicating a possibility.

While I'm doing some more work here, I'd like some help.

I said that a long list would be preferred. All I meant was that I need as much contributions as I can get.

 

[mtwieg]

Hi,

Please what are the practical bidirectional topologies for non-isolated buck-boost converters for efficiencies above 95% at 1.5MHz switching frequency and output power above 65W?

A long list would be appreciated.

To be clear, a "topology" is the general converter circuit, which you've apparently already settled on (buck-boost). Sounds like you're looking for help with specific component selection, though without knowing the I/O voltage ranges it's impossible to start. Also the fact that you're aiming at such a high frequency suggests you're going for something very compact/lightweight at the expense of slightly less efficiency...

 

[Akanimo]

To be clear, a "topology" is the general converter circuit, which you've apparently already settled on (buck-boost). Sounds like you're looking for help with specific component selection, though without knowing the I/O voltage ranges it's impossible to start. Also the fact that you're aiming at such a high frequency suggests you're going for something very compact/lightweight at the expense of slightly less efficiency...

At first I was stuck in determining an arrangement that would be able to do the bidirectional requirement. I've been able to figure out an arrangement that I'm still trying out. The I/O voltage ranges from 0 to 5V.

When I calculated for the winding resistance of the inductor, I got a value a little below 0.8ohm. At first I was thinking it might suffice. Now, it doesn't seem like that anymore. It feels like I'll have to perform some magic to get it to that efficiency at that ohmic value [LAUGHS].

But then I'm getting 94.97% efficiency with simulation but this value is measured at saturation which leaves me with doubts. Makes me think I'd made a mistake while selecting the core. I think I have to go over the selection process again. I need some help, seriously.

I'm selecting the I/O voltage at operating points with Duty*Gcontroller. Is that correct?

I'll appreciate any help you offer me. I'm still on the power stage design.

- - - Updated - - -

...
without knowing the I/O voltage ranges it's impossible to start.
...

By I/O here, you mean the input and output ranges of the converter? I thought you were asking of the I/O voltage range of the power management system.

The converter input is 9.6 to 12.6V; output is 5V for buck operation and 20V for boost. Charging voltage is 20V.

I mentioned 0 to 5V earlier. That's the I/O range from the management system.

The proper voltage level from management system is supposed to be calculated for each mode of operation? I think that should be duty*controller_gain. Is that correct?

 

[asdf44]

Bidirectional -> Replace diode with a syncronous fet.

High Eff. + High Freq -> Look at gan fets.

Also look at off the shelf parts -> LTM8049



EPC sells half-bridge development boards. Buy one of those plus an L and a C, hook up a function generator to drive a variable duty cycle and you can have a test up-and-running in a day (speaking from experience)

The other thing to mention is ZVS. You can, with variable frequency control, soft switch a half bridge for almost any input or output. The downsides are the control complexity (calculating the right frequency on the fly) and the fact that you need much higher current ripple (you need negative current every cycle).

Vicor also makes buck boost converters which you should evaluate to see if there is a fit (they make bidirectional buck-boosts) but also because their app notes and datasheets typically discuss the topology - at least at a high level.


Add: Also why are you manually winding inductors. Look for off-the-shelf. Coilcraft has excellent website selection guides where you can input frequency and ripple and it spits out core and winding losses for you. They have millions of tiny inductors optimized for cell-phone applications etc that you're not likely to be able to beat.

 

[Akanimo]

Thanks asdf44, I checked out the references. They were a whole lot of leverage. I am taking special interest in the ZVS. Coil craft inductors are amazing, especially with the reported Rdc values. For the purpose of what I'm doing, though, I need to take the route that I'm taking.

Like I had suspected, I had made a mistake in my inductor design calculations. I went over it again, this time with 1MHz switching frequency and I got a 77mOhm for Rdc with a less-than-5gram core. My major challenge now is that the output is limited in boost mode. When simulating the buck part separately, it bucks as desired. Same goes for separate boost part. When I connect up the buck leg and the boost leg into a single system, boosting never gets to 10V when 20V is required.

From my observations, bootstrapping is the issue. The output seems to be limited by the input. Any ideas on how to go about bootstrapping this combination? Or do you think it's due to something else?

 

[asdf44]

I think you’ll have to share your sim or schematic to get much more help.

How do you plan on managing the ZVS control?

Personally I like my bets cashing in the efficiency of GANs and keeping control simple rather than embarking down the path of a complex ZVS control scheme (speaking from some experience on both fronts).

 

[Akanimo]

...
How do you plan on managing the ZVS control?

Personally I like my bets cashing in the efficiency of GANs and keeping control simple rather than embarking down the path of a complex ZVS control scheme (speaking from some experience on both fronts).

I obtained 96.6% efficiency. I just had to reduce the efficiency to 270kHz and increase the air gap and the number of turns. I obtained an Rdc of about 5mOhm with an EE19 core that weighs less than 5 gram.

I selected an input capacitor to obtain a desired input voltage ripple. The formula I used resulted in a capacitance that is greater than the value I needed as values substantially below the value I calculated still left me within the ripple spec. This was a good thing though but for the sake of optimal values, I need to have a way of calculating for the capacitance value that would land me at just about a desired input voltage ripple.

Any known formula, concept, idea?

Thanks in anticipation.

- - - Updated - - -

Any attempt is appreciated.

 

[Akanimo]

Hi,

So this was successful. Efficiency was 96.067% and inductor core weight was less than 5 gram.

It was good contribution from all that did contribute especially asdf44, KlausST and mitwieg.

Thanks.

 

[asdf44]

It's nice when people follow up on threads.

Can you post any more schematics or pictures?

 

[Akanimo]

...
Can you post any more schematics or pictures?

Unfortunately, it's not something I should post here. If it were, I would have posted while seeking for help.