Boost Converter Efficiency Improvements

On 06/24/2013 11:59 AM, JW wrote:

All the ground transients are contained inside the cups. Ripple on the supply and ground traces is in the low nanoamps.

I used to try pretty hard to avoid SMPSes, but at this point they're pretty much a necessary evil. Toroids are a big help.

I can't find the exact circuit at the moment, but it was something like this with an extra cap multiplier stage. IIRC I decided to leave the second stage out, since the ripple current was in the low nanoamps anyway. The heater was a bunch of 0603 resistors on a flex, indium-soldered to a thermal shield made of anodized, annealed aluminum foil for fast response.

Not the most efficient thing in the world, but interesting in some respects. The FB loop may need a bit of recompensating--I gave up after about 15 ms of simulated time.

Cheers

Phil Hobbs

Version 4 SHEET 1 1892 2928 WIRE 64 -160 -304 -160 WIRE 176 -160 144 -160 WIRE 240 -160 176 -160 WIRE 496 -160 240 -160 WIRE -304 -128 -304 -160 WIRE 496 -128 496 -160 WIRE 176 -64 176 -96 WIRE 240 -64 240 -160 WIRE 576 -32 560 -32 WIRE 688 -32 656 -32 WIRE -304 -16 -304 -48 WIRE 496 -16 496 -48 WIRE 496 -16 432 -16 WIRE 560 32 560 -32 WIRE 576 32 560 32 WIRE 688 32 688 -32 WIRE 688 32 656 32 WIRE 736 32 688 32 WIRE 736 48 736 32 WIRE 64 80 -80 80 WIRE 240 80 240 16 WIRE 240 80 64 80 WIRE -80 112 -80 80 WIRE 64 128 64 80 WIRE 64 144 64 128 WIRE 304 144 240 144 WIRE 368 144 304 144 WIRE 496 144 496 80 WIRE 496 144 368 144 WIRE 240 160 240 144 WIRE 736 160 736 112 WIRE 496 176 496 144 WIRE 560 176 496 176 WIRE 656 176 656 32 WIRE 656 176 640 176 WIRE -80 208 -80 176 WIRE 304 208 304 144 WIRE 496 208 496 176 WIRE 496 288 496 272 WIRE 368 304 368 144 WIRE 432 304 432 -16 WIRE 64 352 64 192 WIRE 240 352 240 304 WIRE 240 352 64 352 WIRE 304 352 304 288 WIRE 304 352 240 352 WIRE 304 368 304 352 WIRE -512 384 -544 384 WIRE -416 384 -432 384 WIRE -352 384 -416 384 WIRE -208 384 -352 384 WIRE -208 400 -208 384 WIRE 432 400 432 384 WIRE 496 400 432 400 WIRE 592 400 560 400 WIRE -240 416 -304 416 WIRE 16 416 -128 416 WIRE 112 416 16 416 WIRE 432 416 432 400 WIRE -544 432 -544 384 WIRE -128 432 -128 416 WIRE -128 432 -176 432 WIRE 592 432 592 400 WIRE -240 448 -272 448 WIRE 112 448 112 416 WIRE 128 448 112 448 WIRE 256 448 208 448 WIRE -352 480 -352 464 WIRE -304 480 -304 416 WIRE -304 480 -352 480 WIRE -352 496 -352 480 WIRE -352 496 -416 496 WIRE -208 496 -208 464 WIRE 304 496 304 464 WIRE -416 512 -416 496 WIRE -352 512 -352 496 WIRE -272 528 -272 448 WIRE 368 528 368 368 WIRE 368 528 -272 528 WIRE 432 528 432 496 WIRE 432 528 368 528 WIRE -544 544 -544 512 WIRE 432 544 432 528 WIRE -416 608 -416 576 WIRE -352 608 -352 592 WIRE -352 608 -416 608 WIRE 16 608 16 416 WIRE -352 624 -352 608 WIRE 432 656 432 624 WIRE 16 752 16 672 FLAG -304 -16 0 FLAG -304 -160 +48 FLAG 304 496 0 FLAG 496 288 0 FLAG 432 656 0 FLAG -544 544 0 FLAG -208 496 0 FLAG -352 624 0 FLAG 736 160 0 FLAG 16 752 0 FLAG -80 208 0 FLAG -416 496 set FLAG 432 -16 out FLAG 176 -64 0 FLAG 592 432 0 FLAG -416 448 0 SYMBOL voltage -304 -144 R0 WINDOW 123 0 0 Left 2 WINDOW 39 24 144 Left 2 WINDOW 3 24 44 Left 2 SYMATTR Value 48 SYMATTR InstName V1 SYMBOL Misc\\2N7002 256 368 R0 SYMATTR InstName U1 SYMBOL ind 288 192 R0 SYMATTR InstName L1 SYMATTR Value 68u SYMATTR SpiceLine Ipk=320m Rser=1.3 Rpar=100k Cpar=2p SYMBOL cap 480 208 R0 SYMATTR InstName C1 SYMATTR Value 330n SYMBOL res 480 -144 R0 SYMATTR InstName R1 SYMATTR Value 500 SYMBOL res 224 432 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R2 SYMATTR Value 200 SYMBOL res 416 400 R0 SYMATTR InstName R3 SYMATTR Value 500k SYMBOL res 416 528 R0 SYMATTR InstName R4 SYMATTR Value 80k SYMBOL voltage -544 416 R0 WINDOW 123 0 0 Left 2 WINDOW 39 24 144 Left 2 SYMATTR InstName V2 SYMATTR Value 5v SYMBOL res -368 368 R0 SYMATTR InstName R5 SYMATTR Value 1k SYMBOL res -368 496 R0 SYMATTR InstName R6 SYMATTR Value {vset/(5-vset)*1k} SYMBOL pnp 560 80 R180 SYMATTR InstName Q1 SYMATTR Value BC857B SYMBOL polcap 720 48 R0 WINDOW 3 24 64 Left 2 SYMATTR Value 4.7u SYMATTR InstName C2 SYMATTR Description Capacitor SYMATTR Type cap SYMATTR SpiceLine V=63 Irms=170m Rser=0.05 Lser=0 SYMBOL res 224 208 R0 WINDOW 0 -49 55 Left 2 WINDOW 3 -118 111 Left 2 SYMATTR InstName R8 SYMATTR Value {Rsnub} SYMBOL cap 224 160 R0 WINDOW 0 -42 8 Left 2 WINDOW 3 -54 -38 Left 2 SYMATTR InstName C3 SYMATTR Value {Csnub} SYMBOL schottky 32 672 R180 WINDOW 0 24 72 Left 2 WINDOW 3 24 0 Left 2 SYMATTR InstName D2 SYMATTR Value BAT54 SYMATTR Description Diode SYMATTR Type diode SYMBOL Comparators\\LT1720 -208 368 R0 SYMATTR InstName U2 SYMBOL diode 80 192 R180 WINDOW 0 24 72 Left 2 WINDOW 3 24 0 Left 2 SYMATTR InstName D1 SYMATTR Value BAV70 SYMBOL cap 352 304 R0 SYMATTR InstName C4 SYMATTR Value 10n SYMBOL cap -432 512 R0 SYMATTR InstName C5 SYMATTR Value 1 SYMBOL cap -64 176 R180 WINDOW 0 24 64 Left 2 WINDOW 3 24 8 Left 2 SYMATTR InstName C6 SYMATTR Value 4.7u SYMBOL res 224 -80 R0 SYMATTR InstName R9 SYMATTR Value 250 SYMBOL res 560 48 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 0 56 VBottom 2 SYMATTR InstName R10 SYMATTR Value 25 SYMBOL res 656 160 R90 WINDOW 0 -16 76 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R11 SYMATTR Value 2k SYMBOL cap 192 -96 R180 WINDOW 0 31 2 Left 2 WINDOW 3 75 -34 Left 2 SYMATTR InstName C7 SYMATTR Value 4.7u SYMBOL res 160 -176 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 -39 -4 VTop 2 SYMATTR InstName R12 SYMATTR Value 2 SYMBOL ind 560 -16 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L2 SYMATTR Value 100n SYMBOL res 416 288 R0 SYMATTR InstName R7 SYMATTR Value 500k SYMBOL cap 560 384 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C8 SYMATTR Value 1n SYMBOL res -528 400 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 0 56 VBottom 2 SYMATTR InstName R13 SYMATTR Value 10 SYMBOL cap -432 384 R0 SYMATTR InstName C9 SYMATTR Value 4.7u TEXT -224 896 Left 2 !.tran 0 100m 0 50n TEXT 504 728 Left 2 !.param vset=1.5 TEXT 496 792 Left 2 !;.step param vset list 1m.1 0.2 0.4 0.8 1.5 2.5 3.2

3.4 3.5 3.6 3.8 TEXT 504 848 Left 2 !;.step param Rsnub 1100 1600 100 TEXT 504 680 Left 2 !.param Rsnub=6000 TEXT 864 680 Left 2 !.param Csnub=22p TEXT 496 760 Left 2 !;.step param Csnub list 180p 200p 220p 270p 330p 390p 470p TEXT -800 664 Left 2 !.MODEL BAV70 D \n + IS = 3.2E-009 \n + N = 1.85 \n + BV = 125.9 \n + IBV = 2.906E-007 \n + RS = 1.1 \n
  • CJO = 4.957E-013 \n + VJ = 0.54 \n + M = 0.03 \n + FC =
0.5 \n + TT = 0\n + EG = 1.1 \n + XTI = 3 TEXT 568 -104 Left 2 ;Heater (47 47k in parallel) TEXT -552 136 Left 2 ;Catch diode decoupling TEXT 504 904 Left 2 !.options plotwinsize=0
--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs
Loading thread data ...

You know, after much kicking and screaming, I now kinda like SMPS. Especially, when they are external, with the supply over there and the circuit here. It's pretty easy to filter out the high frequency stuff, and there's never the

60 Hz magnetic pickup problem I use to always get with internal linear supplies.

George H.

Reply to
George Herold

Wow! a 48V square wave at 1MHz, with 140mA sawtooth. I just looked at the first couple of hundred uS, but I am astounded. If you can keep the switching transients out of the ground(s), there may be hope for humanity after all.

I racall one of my HP spectrum analyzers had to resort to dithering the SMPS frequency to get the spurs low enough so they wouldn't show in the noise floor. I'm sure they spent a great deal of effort in shielding and circuitry before they finally gave up and accepted dithering. And you just waltzed in and did it! Amazing.

Reply to
JW

"Waltzed in" might be a bit cavalier--more like "clomped in wearing belt, suspenders, crash helmet, foul weather gear, and a portable air raid shelter". ;)

I relied on separate 4-layer boards in individual bomb-proof steel shield cans, toroidal coils with a compensation turn on the board to cancel the solenoidal part of the field, flat flex cable going through very small slots in the steel shield, a very tight layout, and lotsa bypasses. One of the nice things about low power supplies is that the layout can be really really tight.

Plus I haven't shown any pictures of the actual gizmo, because I haven't got any--the company ran out of dough, as I said. The prototype laser worked great, but didn't have all this down-hole stuff finished, more's the pity. As I said when I walked in on this thread, I'm just coming up to speed on SMPSes myself.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

What! No Ghostbusters to back you up?

Most of my HP gear is sealed in aluminum cans (better conductivity than steel), often double-shielded coax or hardline, feedthroughs all over the place, and they still ran into problems with switching transients.

At the beginning of transient analysis in your circuit, capacitor C6 in the Catch Diode Decoupling has a 120mA sawtooth with about 2nS fall time.

You added 2pF parallel capacitance to the 68uH, L1. According to some specs on different style inductors, the SRF of a 68uH is from 10MHz to

15MHz. So 2pF is about right. Even if the fall time increases to 4nS, it still has harmonics well into the VHF range. The energy will happily flow along any unshielded wires to other parts of the circuit.

In addition, capacitor C1, 330nF on the other end of L1 has a 120mA p-p sawtooth at 1MHz. That's a lot of energy and would be extremely difficult to prevent it from going everywhere.

No doubt the current swing decreases when the voltage across the 47k load resistors reaches the final value, buut there is still plenty of 1MHz stuff floating around. I am baffled at how you can get that down to nanovolt levels, even with all the precautions you take.

There must be plenty of other companies drilling holes in the ground. Would your system be any use in fracking? If you could get some company interested in following up, you stand to make a fortune. If the original company has gone bankrupt, you might be able to get the rights for a song if you don't have them already. I'm sure everyone here would wish you the best of luck!

Reply to
JW

Oops - not sawtooth. Triangle waveform. 120mA p-p is a lot of energy!

Reply to
JW

Steel has the advantage of a large mu with lots of loss at high frequency. Plus it's cheeep.

Yeah, but my simulation will beat anybody's actual hardware. ;) (Maxim #1 of PowerPoint Engineering)

There's not much of anywhere for it to go, though. Faraday cages and all that.

I'm not sure that the nanovolt part (the two-stage cap multiplier) would have worked with everything in the same box. I mostly cared about the nanoamp part, because the heater blanket would completely surround all the sensitive sensor and control electronics, and I really wanted to avoid any inductive coupling.

Thanks. It's a pretty specialized business, doing downhole geophysical measurements, and really the only customers are the big oil services companies. (The start-up was funded by one of those, but my NDA prevents me saying which one, or exactly what the measurement was.)

One of the things about that business is that there's an enormous amount of engineering lore, learned by a mixture of (a) hard experience and (b) folks pulling stuff out of their lower abdomen and telling other folks. I have a lot of respect for the people in that business, which is a difficult one on very many levels, but it's really hard to know where real experience ends and the other stuff begins.

We were planning to use a thermoacoustic refrigerator to make an optical bench that would stay at around 25C even with a 175C ambient. Thermoacoustic fridges are incredibly tough--they're made entirely of metal, and the only moving part is the gas. However, the customer wouldn't have anything to do with the idea, apparently on account of some misadventures with Stirling fridges in the dim distant past.

Assuming that had worked, it could have been used for a huge variety of measurements. The two main issues with downhole stuff are, first, that it has to work at a 175C ambient, and second, that it has to survive being banged around really, really badly, every single day.

Thermoacoustic fridges would have fixed #1 more or less completely, and not made #2 any worse than it already was.

Maybe another time!

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

I've run into that recently- the golden formula failed unit consistency. 8-(

Best regards, Spehro Pefhany

--
"it's the network..."                          "The Journey is the reward" 
speff@interlog.com             Info for manufacturers: http://www.trexon.com 
 Click to see the full signature
Reply to
Spehro Pefhany

re.

No need now ...

I don't even need that - John Devereux did it for me.

.shtml

Sure, but it included the drawing that you were so mad keen to see.

As I've repeatedly said, you are confusing the Ayrton-Perry non-inductive r esistor with the Ayrton-Perry astatic inductor - two different components i nvented by the same two people. Why not go for broke and confuse both devic es with the Ayrton-Perry variable inductor, invented by the same fecund pai r?

That way you could make jokes about a variable non-inductive inductor ...

--
Bill Sloman, Sydney
Reply to
Bill Sloman

That example is not the Ayrton-Perry winding. It's the simple loop-back done on toroids to reduce an external field, nothing special about that.

[...]
--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

But Aryton and Perry were apparently the first to discover this "simple" tr ick, along with the astatic resistor winding also - separately - credited t o them, and people who know a bit about what electronics was being done in the UK around 1900 have labelled it as theirs.

If you want to pose as an expert on the period, do try, but even a dilettan te like me will probably show you up embarrassingly fast.

It's not the simplest loop-back that minimises the external field, which is just a single turn in the plane of the toroid, as mentioned by Rayner and Kibble who go on to describe a slightly more complicated winding scheme whi ch they find even better than the Ayrton-Perry "bootlace" for some applicat ions - less inter-winding capacitance.

--
Bill Sloman, Sydney
Reply to
Bill Sloman

The bifilar trick is much older than 1900. Some reading material:

formatting link

And why do you think Ayrton and Perry did not publish this if they'd invented it? They did invent the non-inductive winding and published that, so there cannot be any excuse about a scarcity of ink or a paper shortage.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

and

almost a

any

uses

inductors, who

process

We may just go digital on this cap charger boost thing.

formatting link

It can do arbitrarily cute algorithms and takes a lot less parts than using a dedicated analog controller. The uP will likely be cheaper than an analog controller chip, too.

The fet source current sensing may not be necessary, but I think it simplifies the control algorithm. Might need a gate driver, but probably not.

--
John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
 Click to see the full signature
Reply to
John Larkin

process

formatting link

If it doesn't have to be fast and agile this can be done by uC. I just did 1/2 of one converter with a uC. Unfortunately that one also had scores of other jobs. It was a situation where an analog loop was simply too primitive and slow. Sure enough we ran out of horsepower and had to resort to a simpler algorithm. My experience is: If you think you need a uC of a certain size, take one at least two sizes bigger. Bigger is better, at least when it comes to fast PWM or SMPS jobs.

Cycle-by-cycle current sense is a royal pain with uC. They are too slow and even their built-in comparators are too sluggish. And forget about leading edge blanking so that lowpass has to do it all.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

process

formatting link

Right: we weren't planning on cycle-by-cycle control. But if we review the situation say, every 50 or 100 usec maybe, easy on a 50 MHz ARM or something, that's still faster than most analog feedback loops. It's unlikely anything will go bad wrong in 100 usec.

If we know the gate drive waveform and we know the averaged fet current, we know the peak current. We really know everything we need. There's just this little cultural gap between the analog designer and the C programmer. He's about 30 feet North of here.

Somebody could do some nice consulting business around digital power. Get a few uP eval kits, an open-source compiler, build a few breadboards, accumulate a library of simple apps and subroutines, set up a web site. The most expensive thing would be a Rigol oscilloscope.

--
John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
 Click to see the full signature
Reply to
John Larkin

I assume you mean .org, not .com?

Jon

Reply to
Jon Kirwan

Roight. Something like that.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

and

"continuous process

formatting link

Well, I want to adjust the PWM duty cycle to keep the cap charging approximately constant-power. One easy way is to servo the PWM off the average source current, which is a control loop of sorts. Then efficiency might be optimized by making the frequency a function of the load cap voltage. Something like that.

--
John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    
 Click to see the full signature
Reply to
John Larkin

approximately

current,

Or, since you have to monitor the cap voltage anyhow, and assuming a somewhat constant input voltage rail:

Provide a simple math function or LUT in the uC that changes the duty cycle to the voltage of the cap at any given time.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

approximately

current,

making

That would work: don't bother to measure fet current, just map cap voltage into optimum-efficiency PWM settings. The uP would fit into a 6 pin SOT23. There are still some advantages to measuring fet current too, and we'll probably wind up with a uP with more pins than six.

This digital power thing is interesting. If you look at the recent analog switcher controller chips, they are immensely complex (hard to understand exactly what they do) and need tons of external parts. The complexity creates quirks, too. I was going to use one LTC dual-phase switcher in a laser driver, to regulate current, but it has mandatory current foldback, so it won't start up.

--
John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    
 Click to see the full signature
Reply to
John Larkin

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.