Boost Converter Efficiency Improvements

So 4W, not a whole lot. But 12V to 200V is quite ambitious.

At 4W that means your total losses have to be well under 800mW. Very feasible but now you have to look in every nook and cranny.

250mW peak loss just from the DC resistance, too much. You need inductors with lower DCR. 40C mean they are either tiny or you also have RF losses in the core. So you may need a lower loss core as well.

Yikes! 170nsec is an eternity for a switcher designer. Either drive the driving transistors with more gusto or, better yet, us a real gate driver chip. TC or MC series, for example. And gate resistors are only for wimps :-)

At 40W that's ok. But 100uH doesn't sound like much for a 200V supply at

40kHz. Also, try not to drop this below 40kHz because it could drive animals nuts. They'll hear the magnetostrictive buzz.

Sure? Schottkies are generally The output electrolytic (4u7 @ 400V) gets to around 40C as well. I

40C in the cap is not good at all. Get a better one or hang several 1uF in parallel, and also make sure there is a 0.1uF/250V ceramic in parallel. 12V -> 200V is a strrrrretch but can be done. If you want >80% in such a low power case you've got to look at all the stuff meantioned above. No stone can be left unturned.

--
Regards, Joerg 

http://www.analogconsultants.com/

Wall wart flyback transformers are available. Run such transformer in backward direction.

Vladimir Vassilevsky DSP and Mixed Signal Designs

Got a schematic?

That's a high boost ratio for what is (I think) a 2:1 inductor stepup ratio. A lot of energy is going to go into charging and discharging capacitances.

Have you considered a C-W type multiplier? Diode drops won't cost much efficiency at 200 volts out, and you could get the basic boost ratio down by some factor.

I like to make boost converters using a 1:1 dual-winding inductor to double the p-p swing.

+12-------/////--+--/////--------|>|------ HV | nfet (diode or C-W thing) | gnd

--

John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
http://www.highlandtechnology.com 

Precision electronic instrumentation 
Picosecond-resolution Digital Delay and Pulse generators 
Custom laser drivers and controllers 
Photonics and fiberoptic TTL data links 
VME thermocouple, LVDT, synchro   acquisition and simulation

Your on-time is "long" and your off-time where the inductor delivers

200V is short.

You mayvbe need to measure if the diodes "shorts" for some time at the end of off-time. If so, then considerably energy might be lost during output capacitor discharge.

Use Silicon-carbide diode (?) if you increase your frequency considerably:

CSD01060A, No Recovery Time > 500mA (Io). Can they be bought with smaller current rating?:

You could try to decouple with 100nF and 10nF?

Mayby you should consider another topology?:

• Cuk-converter

• transformer/flyback

?uk converter:

home.netvigator.com: Cuk converter:

Quote: "...The Cuk converter is a new SMPS topology at this moment. It include higher efficiency, low input and output current ripple, minimal RFI, small size and weight..."

Use an isolated Cuk-converter:

-

September 1, 1994 Cuk converter yields 90% efficiency Peter Demchenko, VN, Vilnius, Lithuania

Ccuk=6,8nF ?? PS: Low power.

/Glenn

Yup. Roughly 1.25usec and the 170nsec are grossly eating into that, dissipating stored energy.

I'd probably get a CCFL transformer and use that instead.

[...]

--
Regards, Joerg 

http://www.analogconsultants.com/

It might be a little higher in the end, 220V @ 25mA. I've tried this on the existing design and I still get the same 80% efficiency so I'm happy there's enough wiggle room for later changes.

And there're two of them... The 2nd one is only around 1A peak though.

They're a Coilcraft MSS1210-104KEB, 12mm square and 10mm high and shielded. It was the lowest DCR I could easily find in a smallish SMD.

It looks like the next step up would be a leaded toroid at around 80mR or an SMD toroid at 32mR but an enormous 25mm diameter, and around twice the price.

I'll get some to experiment with and see how much difference they make, I can solder them to the existing pads to test.

To be honest, I wasn't really considering RF losses. I didn't think they'd be significant at 40kHz.

I'm driving the gate with an emitter follower plus a few other parts for level shifting with, of course, no gate resistor :) I tried tweaking a few resistor values to speed up the transition, I shaved off a few mA but nothing really significant. The current you save improving the efficiency of the FET is lost in the driver.

I do have a 15V zener on the gate but that should be > The two diodes Schottkys, so they should be fast enough

They're Vishay VSSB420S, 200V 4A (Cj of 120pF). Over-specced and expensive. I was thinking of just using an ES1J, much cheaper and a Cj of 8pF.

I do have a ceramic on there, a 1206 100nF.

I just measured a 3.2ohn ESR on one of those caps. That won't be helping any.

Thanks for you help, I've got a few things to try.

Dave.

On Tue, 04 Jun 2013 16:57:51 -0700, Joerg wrote:

It's not that bad, the duty cycle is around 40% for 200V. Take a look at the attached schematic. I didn't model the gate driver but the pulse source rise/fall times are approximately correct.

Version 4 SHEET 1 880 696 WIRE -32 128 -224 128 WIRE 288 128 32 128 WIRE -560 192 -720 192 WIRE -432 192 -560 192 WIRE -240 192 -352 192 WIRE -224 192 -224 128 WIRE -224 192 -240 192 WIRE -144 192 -224 192 WIRE 16 192 -80 192 WIRE 144 192 16 192 WIRE 288 192 288 128 WIRE 288 192 224 192 WIRE 336 192 288 192 WIRE 480 192 400 192 WIRE 640 192 480 192 WIRE -240 272 -240 192 WIRE 16 288 16 192 WIRE 480 288 480 192 WIRE -720 304 -720 192 WIRE -400 352 -448 352 WIRE -288 352 -320 352 WIRE 640 352 640 192 WIRE -560 368 -560 192 WIRE -448 416 -448 352 WIRE -720 592 -720 384 WIRE -560 592 -560 432 WIRE -560 592 -720 592 WIRE -448 592 -448 496 WIRE -448 592 -560 592 WIRE -240 592 -240 368 WIRE -240 592 -448 592 WIRE 16 592 16 352 WIRE 16 592 -240 592 WIRE 480 592 480 352 WIRE 480 592 16 592 WIRE 640 592 640 432 WIRE 640 592 480 592 WIRE -720 656 -720 592 FLAG -720 656 0 SYMBOL ind -448 208 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L1 SYMATTR Value 100µ SYMATTR SpiceLine Ipk=3.4 Rser=0.13 SYMBOL diode -144 208 R270 WINDOW 0 32 32 VTop 2 WINDOW 3 0 32 VBottom 2 SYMATTR InstName D1 SYMATTR Value ES1D SYMBOL cap 0 288 R0 SYMATTR InstName C1 SYMATTR Value 4µ7 SYMATTR SpiceLine Rser=3.2 SYMBOL res -304 336 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R1 SYMATTR Value 1m SYMBOL voltage -720 288 R0 WINDOW 123 0 0 Left 2 WINDOW 39 24 124 Left 2 SYMATTR SpiceLine Rser=0.01 SYMATTR InstName V1 SYMATTR Value 12 SYMBOL voltage -448 400 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V2 SYMATTR Value PULSE(0 12 0 100n 200n 8u5 25u) SYMBOL cap -576 368 R0 SYMATTR InstName C2 SYMATTR Value 1000µ SYMATTR SpiceLine Rser=0.02 SYMBOL nmos -288 272 R0 SYMATTR InstName M1 SYMATTR Value IPP110N20N3 SYMBOL ind 128 208 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L2 SYMATTR Value 100µ SYMATTR SpiceLine Ipk=3.4 Rser=0.13 SYMBOL diode 336 208 R270 WINDOW 0 32 32 VTop 2 WINDOW 3 0 32 VBottom 2 SYMATTR InstName D2 SYMATTR Value ES1D SYMBOL cap 464 288 R0 SYMATTR InstName C3 SYMATTR Value 4µ7 SYMATTR SpiceLine Rser=3.2 SYMBOL res 624 336 R0 SYMATTR InstName R2 SYMATTR Value 10k SYMBOL cap 32 112 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C4 SYMATTR Value 1µ TEXT -704 680 Left 2 !.tran 0 100m 0

Dave.

[...]

Yeah, I see your predicament. Another option is to increase the frequency but for that you'd need an external oscillator outside your uC. Or see if you can trick the uC by using a timer and let the loop alter the CC register on the fly.

But there have to be some. Because the Coilcraft lists 2.5A for a 20C rise. Your average current is maybe a 1/5th of that, yet you still see a

20C rise. Since you are in the UK and the weather wasn't that great lately in Europe I assume ambient was 20C or less :-)

If that is the case then the FET may be oversized. If you use one with a very fat gate capacitance that can result in a penalty. With a 20:1 step-up ratio there is always a penalty. The Miller feedback is a real issue.

Those would definitely be a step in the right direction. 170nsec turn-off time are really painful here, much more so than any turn-on time. Even if you used your transistor scheme, if you could have a snappier path to zero that would help. Or maybe if you could even pull negative, that helps a ton.

40% duty cycle? That has me confused, normally it would be much more extreme.

But you should also be able to do that with one of then timers in there, hoping it has one. The off state current would also be very low. Your loop would have to set values in a CC register.

Yikes. Those are monsters, way too big, too much capacitance:

I was thinking of just using an ES1J, much cheaper and a Cj

Mucho mejor. No Schottky but at 35nsec I think they could be snappy enough.

Ouch. But electrolytics in that capacitance and voltage range will always be around that value. I never use electrolytics on switchers if I can help it. And of course no tantalums either. TDK has 2.2uF/250V in

2220 size, but they are over a buck a pop. This type is usually still the better deal:

You are welcome. Getting the FET to turn off in well under 100nsec will go a long way here.

--
Regards, Joerg 

http://www.analogconsultants.com/

On Wed, 05 Jun 2013 01:44:22 +0100, David Jordan wrote:

This is a tapped-inductor boost. It shows an improbable 96.4% efficiency at 217 volts out. I added a few parts for observability.

There are lots of nice surface-mount dual-winding inductors around.

Version 4 SHEET 1 880 696 WIRE -672 192 -720 192 WIRE -560 192 -592 192 WIRE -432 192 -560 192 WIRE -240 192 -352 192 WIRE -176 192 -240 192 WIRE -64 192 -176 192 WIRE 96 192 16 192 WIRE 144 192 96 192 WIRE 288 192 208 192 WIRE 368 192 288 192 WIRE 448 192 368 192 WIRE 544 192 448 192 WIRE 656 192 624 192 WIRE 688 192 656 192 WIRE -240 272 -240 192 WIRE 288 288 288 192 WIRE -720 304 -720 192 WIRE -400 352 -448 352 WIRE -304 352 -320 352 WIRE -288 352 -304 352 WIRE 448 352 448 192 WIRE 688 352 688 192 WIRE -560 368 -560 192 WIRE -448 416 -448 352 WIRE -720 592 -720 384 WIRE -560 592 -560 432 WIRE -560 592 -720 592 WIRE -448 592 -448 496 WIRE -448 592 -560 592 WIRE -240 592 -240 368 WIRE -240 592 -448 592 WIRE 288 592 288 352 WIRE 288 592 -240 592 WIRE 448 592 448 432 WIRE 448 592 288 592 WIRE 688 592 688 416 WIRE 688 592 448 592 WIRE -720 656 -720 592 FLAG -720 656 0 FLAG -304 352 G FLAG -176 192 D FLAG 368 192 OUT FLAG 96 192 BOOST FLAG 656 192 FILT SYMBOL ind2 -448 208 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L1 SYMATTR Value 100µ SYMATTR SpiceLine Ipk=3.4 Rser=0.13 SYMATTR Type ind SYMBOL res -304 336 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R1 SYMATTR Value 20 SYMBOL voltage -720 288 R0 WINDOW 123 0 0 Left 2 WINDOW 39 24 124 Left 2 SYMATTR SpiceLine Rser=0.01 SYMATTR InstName V1 SYMATTR Value 12 SYMBOL voltage -448 400 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V2 SYMATTR Value PULSE(0 12 0 100n 200n 12u 25u) SYMBOL cap -576 368 R0 SYMATTR InstName C2 SYMATTR Value 1000µ SYMATTR SpiceLine Rser=0.02 SYMBOL nmos -288 272 R0 SYMATTR InstName M1 SYMATTR Value IPP110N20N3 SYMBOL ind2 -80 208 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L2 SYMATTR Value 100µ SYMATTR SpiceLine Ipk=3.4 Rser=0.13 SYMATTR Type ind SYMBOL diode 144 208 R270 WINDOW 0 32 32 VTop 2 WINDOW 3 0 32 VBottom 2 SYMATTR InstName D2 SYMATTR Value ES1D SYMBOL cap 272 288 R0 SYMATTR InstName C3 SYMATTR Value 1µ SYMATTR SpiceLine Rser=3.2 SYMBOL res 432 336 R0 SYMATTR InstName R2 SYMATTR Value 10k SYMBOL ind -688 208 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L3 SYMATTR Value 50µ SYMBOL res 640 176 R90 WINDOW 0 71 60 VBottom 2 WINDOW 3 83 54 VTop 2 SYMATTR InstName R3 SYMATTR Value 10K SYMBOL cap 672 352 R0 SYMATTR InstName C1 SYMATTR Value 100n TEXT 72 352 Left 2 !.tran 0 30m 0 TEXT -216 112 Left 2 !K L1 L2 1 TEXT 16 424 Left 2 ;JL's Boost June 4, 2013

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation 
Picosecond-resolution Digital Delay and Pulse generators 
Custom timing and laser controllers 
Photonics and fiberoptic TTL data links 
VME  analog, thermocouple, LVDT, synchro, tachometer 
Multichannel arbitrary waveform generators

at 217

96% would be high, but the classic circuit for the job - Peter Baxandall's class-D oscillator - usually gets pretty close

Peter Baxandall described his parallel- and series-resonant class-D oscillators in 1959 in a paper in the Proceedings of the (British) Institute of Electrical Engineers (P.J.Baxandall "Transistor sine-wave LC oscillators, some general considerations and new developments" Proc. I.E.E. vol. 106, part B, supplement 16 pages 748-58, May 1959). It's hard to get hold of, so my web-page discussing the circuit

scillator1.htm

provides a link to a pdf copy of the paper.

He is reputed to have developed the circuit for driving high turns- ratio step-up transformers ? which tend to have low self-resonant frequencies ? being used to generate high output voltages from battery- level inputs.

The circuit is probably best known from Jim Williams? series of application notes for Linear Technology, on high frequency inverters for driving cold cathode back-lights used in laptop computers (application notes AN45, AN49, AN51, AN55, AN61, AN65). Jim Williams describes the inverter as a current-driven Royer inverter, referring back to the non-resonant inverter described by Bright, Pittman and George H. Royer in 1954 in a paper ?Transistors as on-off switches in saturable core circuits? in Electrical Manufacturing. AN65 does include a reference to Peter Baxandall, but to his 1960 paper ?Transistor Sine-Wave LC Oscillators? in the British Journal of the IEEE paper number 2978E which is cited in a discussion of root-mean- square power measurements.

As Jim Williams points out, this circuit can offer good efficiency ?

90% - when driving difficult loads. Even when lightly loaded, it produces a slightly distorted sine wave, with an appreciable odd harmonic content, about one percent of the third harmonic, and decreasing proportions of the higher odd harmonics.

--
Bill Sloman, Sydney

Interesting but for a DCDC step up I can't see that this inverter is the classic approach. All automotive HID ballasts which are using an internal conversion:

5V-32V -> 30VDC-500VDC -> inverter square wave to bulb are using flyback transformers.

If in case you need just 5W at fixed 200VDC you can do that with a very small transformer core already like RM6. In case you want to have it as a compact type you may use a planar transformer and hang it into cut outs of the PCB itself, etc.

Because you may not like or are used to develop your own transformers for such applications and want also an SMD type, it makes sense to start to google for such step up transformer and base the design on it.

If the project is high volume and you have no luck finding existing type, I am sure you can get full design support for instance from EPCOS and they will supply you transformer samples based on the specs.

I have also some small SMD bobbins in mind like from SUMIDA EFD15

But like the other Joerg already mentioned you may just work on the details of the existing design for some improvements here and there......

Joerg

--

Nucon Gesellschaft bürgerlichen Rechts Geschäftsführer: Joerg  
Niggemeyer & Gert G.Niggemeyer  Email: Joerg.Niggemeyer@nucon.de WEB:  
http://www.nucon.de

It's the approach that copes most elegantly with the relatively high interwinding capacitances and low self-resonant frequencies of high turns-ratio transformers.

The Baxandall circuit uses two cores - one for an inductor and the other for the transformer. This makes it expensive in very high-volume applications, where it can be worth bank-winding a single inductor to get low inter-winding capacitance.

So nominate one.

EPCOS makes cores and formers, and their application notes are the best in the business, but I've never heard that they wind coils for anybody - coil winding is pretty much a cottage industry.

For development work I've mostly wound my own - when I've been able to get hold of a minimal coil-winding machine with a built-in turns counter.

Voltage step-up applications do tend to be idiosyncratic. Jim Williams' series of application notes are aimed at a single application - high frequency inverters for driving cold cathode back- lights used in laptop computers. The curious thing is that he latched onto the Baxandall architecture - probably from a U.K. informant - and ran with it.

The Baxandall approach is one of the existing designs, and definitely worht looking at.

--
Bill Sloman, Sydney

When using a PCB planar transformer you don't suffer on those interwinding capacities.

On our own ballasts we do 2 turns on the primary and 10 on the secondary. On a ELP22 core set gapped with 100um. Runs with output max 75W avg 200DC of input 10DC can generate up to 600V with a transformer ratio of just 5 :-O

Because it's not a sinus driven transformer you don't need that high transformer ratios, the switch off speed gives you the high volts.

Because you have only low volt you must reach some amps to get the energy stored into the core - you won't do that with the transistors of what have been available the last century.

Actually they don't wind for anybody, that's what I said and they are only to considered if you need their quality otherwise it's too expensive.

But they had or still have an EHP16-SMD Xenon transformer

3/15/8 L 1-2 1.4uH Rdc 11/105/205 mOhms H 10mm 23mm gullwing distance width 16mm

(I have of that type of transformer EHP19 samples available

4:24.15 primary inductance 5uH 9mR/210mR/165mR)

In case you need just some transformers and you know what type you can try to go to some Chinese companies and pay some bucks to get ready wound samples for little money only.

The ferrite powder EPCOS or Ferrox is using is from CHina anyhow. Custom design of transformers is not expensive there.

I like todo the copper windings in the PCBs, this is very precise and no one makes a mistake or crosses a wire on another and is covered by this yellow tape......

If I am winding my owns I have an more than 50 yers old drilling machine from the US with a mechanic gear to make very slow turns and having a very hi force available for bending thick wires arroung small bobbins :-) yes the clock counter is a must have.......

That's dam right but the software control of the PWM generation is in this critical designs very important too and gives you nowadays the option to get rid of most analogue PWM ICs, their shunts and even worse their "nice" compensation networks, because we have now inexpensive micros in our hands which get by software the safe operating areas and the reg speed simply coded.

Those type of converters are driving an AC high voltage load directly. It makes no sense to use for Cold cathode: DCDC plus 50Khz inverter. The inverters can be designed and operated by none or primitive PWM generators.

Yes it is very good to have a look on old circuit design, because it was not yet the time of AN cookbook design finding the right IC or component over the internet. Because this base principle was already working well with old semis for the CCL application the type of transformer and the principle is still working very well and even better with modern switches.

However if you need a DCDC converter from input 12V and 0V-200V also controlable on a shortened output, minimum, no load, etc. or for fast charging up flash cap, I prefer to you use the flyback. If it must be a transformer for the 200VDC here it is to be fined out and it depends on price and if it's available ;-)

rgds

Joerg

--

Nucon Gesellschaft bürgerlichen Rechts Geschäftsführer: Joerg  
Niggemeyer & Gert G.Niggemeyer  Email: Joerg.Niggemeyer@nucon.de WEB:  
http://www.nucon.de

I don't believe the 96% at all, I would be very happy if it got to 90% in reality though.

Tapped inductors are not something I've looked at before... I thought they were a bit boutique. Something like a Bourns SRF1280 type seems to be easily available though and I won't need to 1uF between the two inductors any more.

Interesting idea, thanks. I've got a couple of those inductors on order.

Dave.

[...]

Oh c'mon, Bill. There is a plethora of COTS flyback transformers available. Such as this example for a buck fifty:

Fast forward to the 21st century. There's lots of companies who will gladly wind prototypes and production volumens for you. Last time for me was a few months ago when we needed a really odd form factor.

[...]

--
Regards, Joerg 

http://www.analogconsultants.com/

It appears from your LTSpice circuit that the big culprit is C1 with about a half watt of loss. Maybe concentrate on that one first.

217

Coilcraft makes duals, too. They are very much mainstream these days and I use them in switchers all the time.

In the simulation one should set the coupling coefficient to around

0.98. That's where it realistically is for most of them. Unfortunately SPICE does not easily take core losses into account. When simulating loop behavior set it back to 1 for the initial runs because that saves a lot of CPU runtime.

If you are ordering stuff anyway, why not add a few mains flybacks, CCFL transformers and PoE flybacks? Those are cheap.

--
Regards, Joerg 

http://www.analogconsultants.com/

Or use JL's boost circuit.

217

If you build a regular buck or boost switcher with one winding, you get free AC out of the other winding. For floating supplies, negative supplies, whatever.

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation 
Picosecond-resolution Digital Delay and Pulse generators 
Custom timing and laser controllers 
Photonics and fiberoptic TTL data links 
VME  analog, thermocouple, LVDT, synchro, tachometer 
Multichannel arbitrary waveform generators