Seeking Low Voltage (3-12 V Input Voltage) Boost Converter Passive Zero-Current Fixed Frequency Switching Ideas/Papers

Use of 2 Mhz (or greater) boost converters are growing within the automotive industry. An important road block limiting their use at higher power levels (I realize 'higher power' is a subjective description) is the boost transistor switching loss (particularly the turn-on loss). For this thread, the input voltage of interest is in the 3-12 V range. I'd like to focus on 12 V output for purposes of this discussion.

My constraints are:
--> 3-12V input
--> ~12 V output
--> 25W output typical
--> Schottky boost diode (important, as there are no reverse recovery charge to contend with, rather a anode-cathode parasitic capacitance which behaves differently in soft-switching)
--> 2 MHz fixed frequency operating frequency. I can't use variable frequency solutions.
--> Soft-switching MUST be passive in nature (keeps cost down, makes reuse of standard PWM boost ICs)
--> Wish to focus on minimizing the boost transistor turn-on loss

There are a ton of papers on passive soft-switching. Rudy Severns gave a superb seminar on the topic some years ago. The problem I have identified with all passive schemes (so far, could be that I messed up as well) is that they are intended for non-Schottky diodes or do not work at low input/output automotive voltages (voltage required to reset the soft-switching inductors/capacitors at 2 MHz is too low).

Question: I'm looking for examples and/or public domain papers which provide practical implementation ideas and guidelines for a passive fixed-frequency boost transistor snubber based on above constraints.

Just off the top of my head: would a tiny saturable turn-on inductor with a catch winding work?

If its an off the shelf component (e.g. Coilcraft) and cost effective, absolutely. I'm all ears.

The reason for my fixed frequency constraint is because of my role at On Semi. I'm the Apps Engineer responsible for the NCV898031 (2 MHz boost controller) and would like to find a practical and cost effective method to increase the power operating range. I have found turn-on losses to be higher than turn-off losses. If I encounter a soft-switching snubber that addresses both, even better.

I haven't yet had time to explore the possibility of using the PWM signal to also drive a high-side P-channel for recirculating the energy (that a patent troll wouldn't come back to haunt me).

The saturable snubber just saturates at some small percentage of current, but it's inductance is large enough for a sufficient time to allow the transistor voltage to fall before it's current rises (it effectively creates a delay in the collector current rise, but voltage is already zero when it starts to rise). Snubber reset losses should be very small (you could reset with a zener diode), but at 2Mhz, it may be better to put a catch winding on the core, and hard-wire that to the DC bus, or the output bus.

This snubber does not mitigate diode recovery current spike (just turn on loss), but since you are using a Schottky, it shouldn't matter.

(An important road block limiting their use at higher power levels (I realize 'higher power' is a subjective description) is the boost transistor switching loss (particularly the turn-on loss).
What is the MOSFET you are using? Your input is 3-to12 and output is 12, are you using a bootstrap for your drive voltage?. Also I noticed that your IC's drive voltage is 6V. I suspect this could be the problem.

Jay, I've considered saturable inductors (still doing it in my spare time), but recirculating the stored energy has been my issue. I can't introduce a 'DC' bath between VIN and ground via a zener or resistor (standby leakage current is introduced). Can you point to a schematic somewhere? Is this an off-the-shelf part kind of solution? Another 'constraint' is that in an implementable solution, I must use off-the-shelf part that can be ordered by On Semi folks who would eventually assemble a new reference design for customer demo boards.

Romuald, VDRV = 6V is not an issue. I diode-or the input and output to power the VIN pin (45V capable). The boost transistor is the NVTFS5820NL.

Saturable reactors at this frequency will be quite lossy. It's not a common approach for this and other reasons.

Also, as frequency rises, zero voltage techniques usually dominate. Something else to think about.

DCM interleaved boost circuits are becoming popular as well, but to keep them fixed frequency is a different variation on the theme.

All this soft switching topic is nothing new. As you mentioned Rudy Severns, he wrote some fascinating historical articles on soft switching that go back 100 years or so.

Alain,
You state that the turn-on losses dominate the turn-off losses. If that is really so* then try using a smaller value inductor for the boost converter so that the current slope has ramped down to a lower magnitude at the turn-on edge. Then the turn-on loss will be smaller; but at the penalty of higher current at turn-off, and therefore higher turn-off loss. However, you can find the optimum point where total switch losses are minimum.

* ( I would think that turn off loss is inherently greater than turn on, even for the inductor value you have been using, because of the delta i slope. Maybe the real culprit is the gate driver, and it needs to have higher drive current capability at turn-on polarity?)

Lots of Rudy's snubber ideas I looked at seemed practical for off-line topologies, but didn't appear to work at low voltage automotive battery voltage conditions.

Saturable inductors (eg Toshiba 'spike killer' type cores) at first glance appeared to be a possible approach. When I discussed this with an automotive customer, it was pointed out to me that they couldn't use Toshiba cores because they would not support the AEC-Q200 automotive requirements. Sigh!

Randy, the gate drive NCV898031 gate drive capability is adequate (800 mA typical). The problem I have observed at 2 MHz is the added turn-on loss contribution from the boost Schottky capacitance. This isn't to say that I wouldn't benefit from a turn-off lossless snubber as well or instead. So yes, operating near DCM would be beneficial, but there is always tradeoffs (output current ripple increases --> higher filtering stress on the output capacitors).

Rehashing, I have been surprised to see how little appears to be published for low voltage boost soft-switching applications. Lots of higher voltage articles that depend on the minority carrier recombination of high voltage PN diodes, but when it comes to low voltage Schottkys...

Down at low voltages, the switches are extremely low on resistance, so heading towards higher current stress vs switching stress may be the optimum.

Some rough calculations:
assume 500pF of Schottky cap plus drain-source cap. Discharge that cap from 13V, 2 million times per second -> V^2xC/2=85mW Not much loss.
As Ray pointed out, 25W / 3V = 8.3A Assume 10mOhm Rds and .77 duty On-time; Conduction loss = 8.3x8.3x0.01x0.77=530mW

Ray, wouldn't higher current stress also result in costlier boost capacitor filtering solutions? Its give and take but something I could (and probably should) revisit.

If ZVS is the way to go instead of ZCS (I now regret mentioning this), then that's the way I may very well be going if I can figure out a cost effective, simple circuit solution that automotive power supply designers would consider.

I appreciate everyone's comments, but would like to keep this discussion thread focused on practical soft switching methods for low voltage boost topology. My switching losses are dominant in my design (runs a lot cooler for a similar NCV887100 IC running at 170 kHz) and my analysis and measurements support that the 2 MHz switching losses are limiting useful power operating range.

Ran some quick calcs: right now your hard switched turn on loss is around 0.5W or so...sound riight? Do you know what your losses are?

If you designed a saturable turn on snubber with 500nH of unsaturated inductance that saturated at 160mA, your reset losses would be about 15-30 mW, and swich turn on loss would be about zero. This is great, but it comes at a price.

Your IC max duty is 90%, so your clamp voltage (zener reset voltage) would have to be about 2V, which would increase your turn off losses. So, you would have to look at it big picture: is the total switching loss reduced with the snubber (turn on + reset+ turn off)?

As far as an off the shelf component, I don't see why it would be hard to just find a small core (like a toroid) that would do the job. The reset network is just a zener and a diode connected in parallel to the inductor (no DC path to ground this way). A resisitive reset usually causes way too much voltage stress at turn off.

There might be other losses that I am not aware of with this approach (like core losses) you may need to consider at such a high freq.

Early attempts at PFC implemented an auxiliary switch with small series inductor to deal with the diode recovery issue with ZCS. The problem was that the losses in the small inductor were just about the same as the switching loss, Better recovery diodes gave another solution and the circuit fell out of use. It might be applicable at the higher frequency and lower voltage of your application.

It is quite common that the input filter of the PFC is bigger than the circuit itself. At one time we designed a 7th-order Chebychev filter to minimize the size of this.

Use of saturable inductors is of interest, but wouldn't one be forced to use Amorphous Alloy cores to keep 2 MHz losses to a managable level? I once considered using a ferrite material but the hysteresis losses were considerable (that was decades ago, but don't think ferrites have improved enough to make them viable)

Still haven't had the chance to look at the references you mentioned Ray. Hope to get to it today.

Yes, you would need some special material to keep core losses low. No sense in getting rid of turn on loss, but then causing more loss!

The core loss problems for snubbers which I've seen are due to a misunderstanding of the frequency characteristics of core loss. The basic equation for core loss given by several magnetic manufacturers is: P=v*K0*F^(Kf)*B^(Kb). Good data sheets supply equations or tables for Kf and Kb for various ranges of flux and frequency.
This equation is typically for a continuous sine wave with a slight modification for square waves. For a narrow pulse however, which is what a snubber gets, the effective frequency on the core is much higher than the switching frequency. A duty cycle factor must be included in this equation. An IEEE paper by W. Roshen in 2007 goes over this in detail but, in summary the equation can be simplified by substituting (1/2Tp) for F and then multiplying by the duty cycle Tp/Tsw, where Tp is the pulse time in which the high delta flux is applied to the inductor. I've seen several engineers use the basic switching frequency for F in this equation which has led to a very incorrect result. One client I helped had a small inductor power calculated at 0.8W when actually it was about 13 watts and the core was cracking because it got too hot. It's not only special materials as Jay mentioned but, also the correct equations applied.

Just peeking in,.... magamps are too lossy to consider. Soft switch defines a resonant period that will cost you duty-cycle or max Fsw rate.

This project should simply use 2 NextFets (I've used up to 2MHz) or EPC E-mode GaN
as Boost switch and synchronous rectifier. This is low voltage, so it's easy to drive.
This could be put into a tiny module at this power level.
Is 2MHz arbitrary? Did anyone plot losses versus frequency? What if you got 98% efficiency at 1MHz but only 94% at 2MHz??
I would also interleave boost sections to provide lower ripple current and more "on" time charging output caps. Parts are still tiny at this power level.

Out of curiosity, did you ever work at TDI?

Thanks to all for your feedback. I do appreciate it.

Response, though interesting, are getting off topic. My question is very narrow in scope.

I'm the Apps Engineer responsible for the NCV898031. This is a simple and inexpensive SOIC8 2 MHz controller. Our end customers are highly cost sensitive (automotive). Cooling is 'inexpensive' compared to added circuitry. A limiting issue I face with 2 MHz in boost topology is the switching losses from the MOSFET and am looking for simple passive solutions to increase practical maximum power implementation (yes, this is subjective).

Much of what has been published are more specific to higher voltages having reverse recovery current that can be taken advantage of to reset the snubber. I've noticed that Schottky diodes don't perform so well in these same circuits, and that 2 MHz result in switching results in impractical snubber values.

So far, I'm contemplating using a separate inductor to redirect the some of the switching losses (still a loss though) away from the transistor. Don't know when I'll have time to try anything though.

Our customers manage losses with PCB thermal planes (i.e. no external heatsinks) using u8FL or S08FL type MOSFET and Schottky packages (e.g. NVTFS5811NL and MBR440MFS).

Vin can be down to 2-3 V. Automotive Start/Stop applicaiton battery issues (customer reverse battery protection diodes included in this figure).

Vout varies with customers/applications. I would be quite pleased to manage the MOSFET losses to deliver, for example, 12V@2A from a 3V input. Ambient temperature can reach 105C (so PCB and junction temperature does get hot).

So there you have it. Sync converters (not feasible with the NCV898031), mag amps, active solutions are 'out'. My 'constraint box' is quite limited.

A couple comments:
From your description of your fairly narrow "constraint box," why do you even care about efficiency? Reliability is another issue, but if you have 60% efficiency, does your customer even care?
I would think your customer is cost driven. You have to keep in mind the EMC implications. Remember, automotive customers need to survive high voltage spikes and relatively low emissions. A 2 MHz solution might make the EMI filter smaller, but still not insignificant (as Ray mentioned). I've seen that ugly problem before--where the EMI filter is larger than the converter! Your exotic inductive materials might be in the filter(s).

Heat may not be the problem--given your constraints--but your choice of solution may lead to untenable EMI.
Don't forget to look at all of the implications of your approach, not just switching losses.
I would consider some of the newer material MOSFETs for both switch and diodes. That may be contrary to your cost driver, but it allows operation at 2 MHz. I'd give up on the simple, cheap Schottkey diode if you can afford the cost.

Approach works at present power level.

EMI not an issue. Tested and passes with margin.

Boosting the power to a higher level is my only issue because of FET losses.

Cost is critical. Some efficiency penalty is acceptable. Of course, 60% efficiency is an exaggeration and would be unmanageable.

I fully realize that my 'box' may be unachievable, but I won't say 'quit' so easily.