Boost Converter Efficiency Improvements

It started off with a typical medical confusion between correlation and causation - high cholesterol levels were found to be correlated with coronary heart disease in the 1950's.

This lead the researchers to look for ways of lowering blood cholesterol levels, which eventually pointed them at the HMG-CoA reductase enzyme.

They had to wait until the 1970's when a Japanese biochemist ran across a natural product with such an effect by accident when he was looking for a new antibiotic.

There was then systematic research on why it worked, and why it was too toxic to use in humans, and people started looking for structurally similar compounds with a better balance of effectiveness versus toxicity.

This seems to have been more of a suck-it-and-see procedure than a systematic search based on any kind of insight into what caused the toxicity.

Lipitor is fully synthetic, but when it was synthesised we didn't know what we do now about the fine structure of the surface of the HMG-CoA reductase enzyme, so it's synthesis predates entirely rational drug design.

My doctors are now giving me simvastatin, rather than Lipitor, so they aren't impressed by that particular product of big pharma.

The wikipedia article credits a Chinese university with the plan to use what we now know about the structure of HMG-CoA reductase to find a better drug - not a great advertisement for big pharma.

--
Bill Sloman, Sydney

My aunt's husband did a bit of that - he once took a flock of sheep to Mara linga, presumably just before the 1956 UK atom bomb test took place there. He died a few years ago (2009) at 90-odd. I doubt if the sheep did as well. He also worked on human metabolism and nutrition.

--
Bill Sloman, Sydney

See? That's what I meant with the "oh-oh" effect. Yuo have a totally unexpected outcome, and often it has undesirable side effects that then must be worked on.

That is how a lot of research is conducted. Has to, because we don't know all the stuff to make everything a mathematical slam dunk. This is also why a lot of the early aviation pioneers died.

Yeah, right, almost 30 years after an invention it is easy to say "not impressed". Fact is, Pfizer has enabled many people to live a longer life and most of all have a higher quality of life at an advanced age.

Quote "Atorvastatin was first synthesized in 1985 by Bruce Roth of Parke-Davis Warner-Lambert Company (now Pfizer)".

--
Regards, Joerg 

http://www.analogconsultants.com/

That arbitrary compounds sometimes are physiologically active isn't "totall y unexpected". That's why the guy was browsing through that particular stac k of natural products in the first place.

It's not all that scientific - keeping records of what did what is basic to science, but you have to publish the results in some sort of structured do cument to qualify as doing science.

That's what Pfizer claims. People who take statins do live longer, but it's more likely mostly due to their doctors keeping an eye on them and schedul ing them for stents and by-pass operations than to the statins on their own .

My doctor brother tells me that in 30% of cases, the first clinical sign of heart disease is sudden death. Getting people heart-disease-conscious allo ws earlier intervention, which can help.

Along with a lot of other compounds that didn't do as well. That wasn't sci ence, it was prospecting. And granting the dubious reliability of the few p ublished results of clinical trials - most of those results don't get publi shed - big pharma doesn't do science.

--
Bill Sloman, Sydney

Here're a few results from my recent experimentation:

I've been able to increase the frequency to 80kHz using the method discussed earlier. I have tested several different inductors on my original PCB that was giving me 79.4% efficiency at 4W output.

Efficiency COILTRONICS HC9-470-R 47uH 3.65A 77.0% COILCRAFT DO5022P-473MLD 47uH 2.6A 77.5% EPCOS B82477P4473M000 47uH 2.8A 77.5% COILCRAFT MSS1210-104KEB 100uH 3.4A 79.4% WURTH 7447709101 100uH 2.8A 79.6% COILCRAFT DO5040H-473MLD 47uH 3.7A 80.3% COILCRAFT MSS1260T-104MLB 100uH 2.1A 81.7% COILCRAFT DO5040H-104MLD 100uH 2.8A 82.5%

I did another PCB layout using an MCP1407 FET driver, a lower gate charge FET, the DO5040H-104MLD inductors and 1uF ceramic capacitors replacing the electrolytics.

The FET driver gives a 23ns rise time and a 22ns fall time at the gate, a huge improvement.

The efficiency is now 85.5%, A 6% improvement.

I also did another PCB using a tapped inductor as suggested by John Larkin, an SRF1280 47uH. This version got to 85.8% efficiency and has about half the number of parts and takes less board space.

The single inductor has around a 22C temp rise. The FET, now a DPAK, has around a 10C rise but some of that may be being conducted back from the inductor.

The only downside so far is slightly higher RF emissions.

If there're no major EMC problems, I think the winner for my application is obvious.

Thanks everyone!

Dave.

That indicates significant differences in core loss. It does not surprise me much that Coilcraft and Wuerth score better.

That's almost all in the layout.

Great. Considering the extreme step-up ratio and the low output power I think 85% is quite a respectable efficiency. The 22C rise on the inductor is still a bit high for my taste but that might be as good as it gets with catalog parts.

--
Regards, Joerg 

http://www.analogconsultants.com/

Choose a bigger catalog part with higher peak current limit. It will have lower RDC and higher saturation limit. That will reduce copper losses and core losses since you will be working at a lower peak flux.

Regards

Klaus

If you'd been up to winding your own transformer and probably the inductor as well, a Baxandall class-D oscillator with an 11:1 step-up ratio would ha ve given you your 200V from 12V.

The peak voltage across the switching MOSFETs would have been pi/2 times 12 V - 19V - so you could have used the opposite ends of teh primnary widings to drive the MOSFET gates.

You would have got a bit over 90% efficiency,and the peak primary current w ould have been around 400mA. Your inductor would have to be wound to cope w ith that current. If you did use MOSFET switches, the oscillator probably w ouldn't squeg, no matter how big you made your filter inductance.

RF emissions wouldn't be too bad - the Baxandall oscillator does generate o dd harmonics, but at relatively low levels. The MOSFETs would be switching on and off at relatively low current levels - way below the peak 400mA circ ulating in the primary. The rectifiers required to get the 200V DC out of t he secondary would be a bit fiercer, but a ferrite bead between them and th e high voltage reservoir capacitor would probably take out any fast edges.

You obviously aren't going to build one, but ti probably would have been a nicer solution, if you had access to a coil-winding machine with a turns co unting dial.

--
Bill Sloman, Sydney

[...]

I think David already has the biggest one on there that will fit.

--
Regards, Joerg 

http://www.analogconsultants.com/

There is a theoretical disadvantage to the dual-winding inductor thing: the windings are identical so the copper in the boost winding is too thick relative to the driven winding. So the copper use is not quite optimized. That's a small price to pay for using a cheap stock part.

What is the actual price differential between the cheap stock part, and an more nearly optimal one that you might have gone to the trouble of designin g and getting wound?

My impression was that most of the cost of small ferrite-cored transformers was in the copper and the ferrite - the labour required to wind the copper onto a former wasn't all that expensive, and automating that part of the j ob didn't usually save enough to pay for - usually - unnecessarily generous copper and core in the off-the-shelf part.

--
Bill Sloman, Sydney

Oh, the other issue is that it will probably need snubbing, which costs efficiency. A single-inductor continuous-mode boost converter doesn't need a snubber.

--

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

I do actually have quite a bit of space now. I've got rid of the big electrolytics and there're fewer parts than before.

I wanted to try this topology quickly as an experiment and the SRF1280 seems to be the largest dual inductor available from Farnell.

I'm just doing a proper PCB layout now, that should help reduce the RF noise. I can get rid of that big heatsink pour on the drain tab for a start.

The pinout of those inductors leads to a realy nice layout; DC on top left, output to diode on top right, FET pulling down the bottom two pins.

Do you think a "better" doide than the 35ns ES1J would help any? I can get something like a Bourns CD214A-F1400 which is 25ns for a similar price.

With the added PWM resolution I can lower the inductance, which will lower the DC resistance.

I wonder how high I can go before the comparator in the PIC craps out...

Dave.

Farnell is often kind of limited in selection.

That is always nice. But mind the "air gap" in the form of this anthracite-colored epoxy up top. That's where magnetic field galore spews out. You can test it this way: Take a 2nd SRF1280 and solder an LED across one winding. Then run the converter under load and lower that

2nd SRF1280 inverted onto the one in the circuit. But slowly, like the docking procedure at the space station, else the LED my go poof.

Since you are operating in DCM every nsec helps. But if you run the sim again with a smaller Schottky it becomes apparent that only around 10mW can be gained. In your case 0.25%. As my old boss used to say, that's not much to write home about.

Those aren't exactly rockets but I believe they can do under 500nsec.

--
Regards, Joerg 

http://www.analogconsultants.com/

Yes, I saw that mentioned somewhere else in the thread. I went for a

250V FET, thinking it would be able to cope without any snubbing. The highest voltage I'm seeing is only 135V though. And then only for 10ns, right after switch off, before it settles back to 110V or so.

I'll post a trace in ABSE, under this subject heading.

Dave.

Le Tue, 18 Jun 2013 20:58:21 +0100, David Jordan a écrit:

You don't need any snubber for a two inductor boost too. At least the one you use with 2 rectifying stages. That is with a properly wired circuit (low parasitic inductances).

How did you connect your scope probe? The 25/30V overshoot might be a measurement artefact due to your probe

5-6" ground lead (check the resonnant frequency which probably is circa 100MHz)

--
Thanks, 
Fred.

Of course, if you get you own coils wound, you can put them into RM cores, which keep the magnetic field closely confined within the core outline. Toroids are even better, but harder to get wound.

--
Bill Sloman, Sydney

But one has to mind core saturation. If it saturates in a switcher that is usually accompanied by a loud band, flying debris, and occasionally spattering molten solder :-)

The best thing to do when the EMI from the core is a concern is to plop a Laird shield can over the whole thing. Laird is even in David's home country.

--
Regards, Joerg 

http://www.analogconsultants.com/

Of course one has to pay attention to core saturation, power dissipation in the windings (bearing in mind that the Curie temperature for ferrites is ap preciably lower than - say - the maximum junction temperatures for power t ransistors

It's stuff like that that makes electronic design interesting, and keeps yo u well paid. Iy kept me well paid too, when I was younger ...

The best thing to do when EMI from the core is a concern is to design it ri ght in the first place. Added extra shielding can be necessary if you have extraordinarily flux-sensitive components in the immediate vicinity, but it 's usually cheaper to make sure the flux sources are further away and point ed in a different direction.

--
Bill Sloman, Sydney