High Frequency SMPS

[cupoftea]

Hi,
We are doing an SMPS for a person at the moment….unconnected with that, but necessary as an add-on for his job, is that we must also produce a “high frequency SMPS” (10MHz +) for driving LEDs.

I apologise in advance, but the spec for the HF SMPS is very open, ….all we have to do is in some way demonstrate the cost and efficiency and size advantages of a high frequency SMPS….by making some kind of HF SMPS that drives LEDs.

Anyway, I take it offline LED drivers are out of the question at these frequencies?....since ferrite is impossible due to core losses, and an air core transformer, with 3500VAC isolation is going to be spewing RF all over the place?

As discussed, the spec is very open…..but the customer has mentioned that he likes the following advantages of HF SMPS…..

1…..Use of PCB printed coils instead of “normal” inductors.
2…The use of much smaller energy storage components, ie inductors and capacitors.
3….The use of GaN transistors, for high switching frequency.
4…Customer thinks some of his products don’t need a radiated emissions EMC test pass anyway…..so if there are extra RF emissions, then “so what”.
5…customer thinks there can be cost savings. (eg due to not needing “bought” inductors., but just PCB inductors)

..Anyway, I am thinking that skin effect issues could well make any solution bigger and more expensive? Also, the only capacitors possible are ceramics or “air dielectric” ones, but even ceramics have SRF of eg 5MHz for a typical 1uF, 0805, X7R.

I have read multiple web articles, but all are gushingly over-worshipping of 10MHz plus SMPS’s, without detailing any of the drawbacks.
I am suspicious that there are pretty good reasons why 10MHz+ SMPS can never realistically be cheaper than todays 50-1MHz SMPS?
Also, there is the situation of ferrite-core-less SMPS’s suffering issues of coupling if they are placed near conductive objects?
Also, there is the issue of the tolerance on ceramic capacitors which make up resonant SMPS elements?
Also, many modern RF Amplifiers use switching techniques anyway (Class E?)…they are almost an SMPS in their nature, and these are not known for their cheapness….so why is it that their application in SMPS is going to be cheap? (cheaper than what exists now)
Also, a 10MHz+ SMPS must surely be resonant, and these inevitably have high circulating currents....and if a wide vin and/or vout range is required, then the circulating current tends to be higher still.
Also, resonant SMPS tend to have higher no_load losses, (due to circulating currents still flowing at no_load)
Also, as discussed, the new High frequency radio transmitters, are essentially SMPS (of a kind) anyway...and they are neither cheap, nor small, so surely this technology for SMPS is wayward and far-out?...then again, if all you need to do is stuff some current through some LEDs....

So can you think of a good example HF LED driver SMPS that we can put forward for this?

 

[KlausST]

Hi,

SMPS can be anything.
* AC to DC or DC to DC
* 5V to 3.3V or 300V to 5V or anything else
* Constant current, constant voltage
* from milliwatts to kilowatts...

If this is what you mean with "the spec is open" .. then good luck with the project and your research.

Klaus

 

[cupoftea]

As discussed, research is very interesting for this, because there are very very few articles actually detailing any lab produced examples whatsoever.

Please may i then put it like this?.....your boss asks you to do a 11-15Vin, 8W out LED driver.....
*..Using only printed inductors or transformers.
*..No isolation required.
*..LED string is 10 LEDs in series.
*..White LEDs
*..Switching frequency should be >10MHz

 

[danadakk]

Seems to be an interesting paper on HF SMPS :

https://backend.orbit.dtu.dk/ws/portalfiles/portal/171273244/08664603.pdf

Note Digikey only shows controllers up to 750 Khz.....


Regards, Dana.

 

[cupoftea]

"A 10 MHz GaNFET Based Isolated High Step-Down DC-DC Converter: Design and Magnetics Investigation "

Thanks Dana,

That paper you kindly show looks like the best thing out there on HF SMPS design. (here it is again)....

"A 10 MHz GaNFET Based Isolated High Step-Down DC-DC Converter: Design and Magnetics Investigation "

https://backend.orbit.dtu.dk/ws/portalfiles/portal/171273244/08664603.pdf

Ayk, its good to know where your bridges are before you cross them…..so that you can actually cross them.

The paper details a 20W, 200-300v to 0-28v buck, at 10MHz.

I think even the authors would confess that what they have produced is at best “uncompetitive”, and more realistically, a very poor offering indeed.…..showing that HF SMPS is not yet feasible……what the paper produces is neither cheaper, smaller, or more efficient, than a “standard” ~100kHz solution.

Here are the problems of the 10MHz converter of the kindly provided paper…….
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
*..efficiency is only 85.3% maximum.

*…inductor & transformer ferrite pieces are custom made and made of fair-rite 67, which may be in short supply compared to good old 3F3, N87, 3C96, etc etc.

*…the shown design has a fixed frequency of 10MHz, and cannot handle any variation of vin and vout outside of the fixed range.

*…The design of the resonant L and C needs use of a calibrated Agilent 4294A Impedance Analyzer. (many thousands of dollars)

*..A expensive Maxwell 3D simulator is needed to develop the magnetics (it’s a finite element modelling simulator and is very expensive).

*..As page 6 shows, foil windings are necessary…these are going to be custom made , at great expense.

*…As the paper confesses……QUOTE>> There is limited data available on the design of high frequency (HF) and very high frequency (VHF) power magnetics. Power magnetics have high flux drive. For most of the materials, large signal loss data are not available at above a few MHz. <<UNQUOTE

…..So its going to be very difficult, in any HF SMPS design, to select the magnetic material…..becuase there is no data available for showing its characteristics…………and there is no data regarding the tolerance of its characteristics….and the temperature drift etc…so any design of any HF SMPS is going to be a “hope it works” thing, with a likely high field failure rate.



*..The paper is honest enough to confess that high frequency magnetics materials are simply not available at realistic costs for operation at 10MHz plus…….

QUOTE>> With the emergence of GaN and SiC devices, there has been a significant advancement in semiconductor device switching speed, but magnetics has become a primary limitation constraining miniaturization. By increasing the switching frequency of the converter, the absolute value of capacitance and inductance can be reduced but the actual size reduction at very high frequencies depends on the allowable loss power density. Appropriate core material and winding structure have to be selected for these high frequencies to reduce the loss and realize the achievable miniaturization. Emerging thin-film magnetic materials are a good choice for frequencies greater than 10 MHz. These materials are typically alloys with Fe, Co and Ni. But these are not commercially available at economical costs<<UNQUOTE

*..The paper is honest enough to confess that the only way they could drive the primary and secondary FETs at 10MHz was to use a costly Virtex-5 FPGA development board !!.....An FPGA needed just to act as the switching frequency oscillator….this alone makes the entire thing totally un-commercially-viable.

*… The paper is honest enough to confess that the only way they could provide the high side drive was to use batteries!..again, this alone makes the entire thing totally un-commercially-viable.

*.. The paper (pg 9) is honest enough to confess that the expensive simulator that they used to calculate the magnetics parameters was very innacurate compared to what they actually measured with the precision impedance analyser…and that this was due to the under-estimate of parasitic winding capacitance.

*.. The paper (pg 10) is honest enough to confess that the only way that they could get a GaN FET with low enough gate capacitances was to use a 650V part , which has a high Rdson of 220milliohm….and with all the circulating currents you get in resonant converters, this is going to hit efficiency.

*..The paper is honest enough (pg 10) to confess that they need a biopolar gate drive supply in order to ensure proper switching of the GaN FETs at 10MHz.

*…The paper is honest enough to confess that they can only manage output power variation by doing dead time adjustment…this comes with extra complexity and cost….not only that, but with high dead time, the cct current RMS values go up.

*..the 5v max gate drive voltage of the GaN FETs could be breached by ringing in the gate drive…because at this high switching frequency, damping in the gate drive wont be permitted…..so we will likely see dead GaN FETs at some point.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
So this paper details an SMPS which provides for a specification...and does it in a diaboilcally bad fashion.....even a junior EE could do it massively better using the standard approach of today with switching frequencies around say 100kHz.

Every other paper one reads about HF SMPS examples is just as diabolically bad as this......has anyone even heard of any even remotely sensible SMPS being done at 10MHz plus?

 

[danadakk]

@cupoftea

So whats happed in 5 years since the paper was published ? What have the majors done in this
area of development .....

This by the way is not my area of work, so I love the time you took to comment. I learned
a few things.


Regards, Dana.

 

[cupoftea]

Thanks Dana , and of course i do not criticise yourself....its just that there do not exist any realistic example papers on 10MHz+ SMPS...all the ones that you and i find, are diabolical, because its just not possible with currently commonly available parts.

It was partly the fact that an FPGA had to be designed to drive it, aswell as the other stuff..

In the last 5 years..nothing has changed...all the critique of the post above still applies to 10MHz+ SMPS....

Ayk, Any RF engineer will tell us that those switching amplifiers (eg Class E) for RF transmitters cant be used like for SMPS applications.....they are way too inefficient.

 

[cupoftea]

What is the highest frequency for a sucessful SMPS?, say DCDC 24vin to 5 vout at say 20w, or similar.
Is it 1MHz, 2MHz, 3MHz, 5MHz?....at what frequency does it start getting ridiculously bad? (like in the above paper which is like all of them)....i'd say above 2MHz, but what are your thoughts?

 

[danadakk]

I would go to the mainstream vendor sites like TI, ST, Analog Devices and
see what they are promoting to make that decision.


Regards, Dana.

 

[mtwieg]

High switching frequency is not, by itself, a figure of merit. Going higher than a couple MHz doesn't have any benefit if you're just making a DC-DC converter. If you need isolation then even 1MHz is likely excessive.

But there are novel applications where extreme switching speed (and control by a fancy FPGA) is necessary, for example envelope tracking power supplies: