DC/DC Boost converter with 95% eff.? GaN MOSFETs

[samuelr]

Hi

I am trying to build a boost converter with 95% eff. I have looked around for some solution but most of the them seem to be around 90% - 92% eff. I am thinking of using GaN MOSFETs to try to push the eff. up, I have not used these before not sure if they might help much. Other things like inductor (low series resistance) and low ESR tant. cap. I have already put into consideration. I like to do a 20V - 30V boost to 150V Giving me 150W. If there is a regulator out there that can give me this eff. I would be very happy with that. I am now considering if I should just use an MCU to do the switching using a GaN on a boost circuit. Not sure if this would have any added benifit.

 

[Warpspeed]

That is a fairly high step up 20v to 150v.
The problem with that, is the switching device sees both the high peak current at 20v and the 150v peak voltage.
Peak current will be at least 30 Amps, probably higher.

High voltage mosfets have higher rds on, which means multiple devices.
But even then, the conduction losses will be relatively high.
Once you get much over 90% it gets really hard to scrape each extra percent.

A different solution to this may be a flyback supply. The peak current will be the same, but you can then use much lower voltage mosfets to switch the primary which will dramatically lower conduction loss.

 

[samuelr]

That is actually a very good answer. I guess it doesnt really matter if I use an MCU with feedback etc. or a boost regulator. It would not mean I will get alot of improvement cause of the external components. I will try out some circuit and see what happened.

Thanks,.

 

[treez]

The "evil" that is diode reverse recovery will be your downfall on efficiency, because silicon power schottkys don't extend up to 150v...well, they do, but at that high voltage, they actually have a reverse recovery.
I would do a full bridge if you want efficiency.

 

[Warpspeed]

If its run in discontinuous mode, the secondary current ramps right down to zero before the next switching cycle.
In that case violent reverse recovery never occurs.

As the voltage is fairly high at 150v, normal fast recovery diodes with a slightly higher forward voltage drop are not as big a disadvantage as they would be at lower dc output voltage.
Its certainly something to be aware of, but in this type of supply the real killer of efficiency will be conduction losses in the switching device at the low dc input voltage side.

 

[treez]

very true, you could even use a chip that detects discharged inductor to start the next cycle, then you don't go too deEp into DCM. I AM SURE ANY Boundary conduction mode flyback chip could do it for you, in boost converter mode instead but.
As you know, DCM doesn't raise the inductor RMS current much either in boost, as a 0 to 10A BCM inductor current has rms of 5.77amps, and average of 5 amps...not that different.

 

[samuelr]

As for the diode I am thinking of using SiC (silicon carbide) and the MOSFET GaN. This would prehaps help is increasing the eff. slightly. Yes 150Vdc is higher than most common boost voltages. I was seriously considering doing the control using an MCU do handle the switching I still unsure if this might significatly increase the eff.

 

[Warpspeed]

I looked at doing something very similar myself not so long ago, and it is definitely workable.
The trick is to use the maximum possible conduction time for the required voltage conversion ratio.
In your case 20v to 150v conversion would allow 88% on/12% off.

The advantage of that is the peak to average , and the RMS current can be lower.
And its the RMS conduction current that dissipates all those conduction loss watts.
Going well over 50% conduction is definitely advantageous, where the boost ratio is high.

The increased Rds on, of high voltage mosfets is still a problem. but its manageable if you use multiple devices.

I would be very wary of using a microprocessor to directly drive any switching power supply. The problem is that it simply cannot react fast enough under fault conditions. And stabilising the feedback loop in software is bound to be much more difficult without making it very slow to react.

While your software is busy thinking about the next instruction, the current could be spiking upward at 100 Amps per microsecond through your switching mosfet if the core ever saturates.

Much safer to use hardware that can shut things down in a few hundreds of nanoseconds if things ever go horribly wrong.

 

[samuelr]

warpspeed, you are right using MCU is not wise with regards to safety features that a dedicated might be able to handle with a great degree of reliablity. However, I am havnig some difficulty to find a boost that is able to do 150V boost without a flyback configuration do you have any recommendation with regards to the regulator?

 

[Warpspeed]

Almost any PWM control chip that can produce a single output 0 to 100% duty cycle would be suitable.
I don't understand the question.

 

[mtwieg]

warpspeed, you are right using MCU is not wise with regards to safety features that a dedicated might be able to handle with a great degree of reliablity. However, I am havnig some difficulty to find a boost that is able to do 150V boost without a flyback configuration do you have any recommendation with regards to the regulator?

You're not going to find a monolithic boost converter IC with integrated FET, you're going to use a general purpose controller which can drive external FETs, probably with external gate drivers suited to the GaN FETs. The EPC2034 or EPC2010C should work well for you. A SiC schottky diode would help too, or you could use a GaN FET with its gate and source shorted as the diode.

 

[schmitt trigger]

150 watt is a little on the high side of power rating for a DCM flyback.

I'm with TREEZ on this one, use a full bridge driving a transformer.

 

[samuelr]

Yeah found that its not possible to get >90% eff. with this configuration. I will do 75V and see if I can just put them in series. It seems 75V has some solution with 97% eff. from Texas Instrument

 

[Easy peasy]

A push pull configuration driving a tapped inductor gives very good efficiency for 20 -> 150V

 

[Warpspeed]

The problem with forward converters is the magnetising current. Trying to reduce that to one or two percent is going to be very difficult, which you will need to do to reach 95% efficiency.

Boost and flyback type converters side step the whole magnetising current issue, but in return creates some new problems which mostly revolve around having very high peak currents.

 

[Easy peasy]

To be fair, every converter (isolated) has magnetising current - the forward is no worse or better in this respect than any other topology.
We build 24 / 24V isolating converters at 400W at 96.5% efficiency and we run the largest magnetising current compatible with a reasonably low core loss, the issue is really switching losses (turning off high current) and conduction losses, and of course using FET's to do the rectification, again keeping switching and conduction losses as low as practicable, this is a non trivial exercise and usually requires decades of power electronics expertise to arrive at robust solutions - especially if you also want to pass commercial EMC standards..!

 

[treez]

We build 24 / 24V isolating converters at 400W at 96.5% efficiency and we run the largest magnetising current compatible with a reasonably low core loss,

This is interesting, I am a fan of putting small gaps in forward or bridge smps transformers, so that I know more accurately what is the magnetising current....the gap results obviously in more magnetising current...is this to what you allure?...ie you deliberately gap, and as you say, to run more magnetising current?

The following threads turned into debates over gappng of bridge/forward smps transformers...

Post #8 of thread 341980 below shows Sunnyskyguy's showing that if you don't gap, then you have a gap anyway, an undefined one, with a core with widely tolerant permeability..(unless you expensively lap the cores)

https://www.edaboard.com/threads/340530/

https://www.edaboard.com/threads/341980/

 

[Easy peasy]

Nope, no gap = lowest mag current, we don't gap (unless flyback) we just have good control.

 

[treez]

Thankyou, of course, then, regarding Sunnyskyguys reference sited above, ..if you don't gap, then do you lap? (ie get the mating surfaces of the ferrites lapped?

 

[Easy peasy]

there's no need to lap cores in a typical commercial design