In my Problem Circuit, I'm using a Bourns SRP1270 (), which is rated for 50A saturation current, and I'm not getting even close to that
-- I'm going up to about 25A at absolute most, yet even at 17A I'm seeing excessive power consumption.
I don't see anything about core losses on the data sheet -- is there any way to tell (other than just by knowing ahead of time) what the losses are? If there was some alternate part I could slap in there it would be lovely. If not, I'd at least like to know if there's some way of knowing what the losses will be so that I can simulate them or calculate them.
If the answer is "change to brand XYZ" -- I'm good with that, just let me know the brand!
Switching frequency is about 100kHz.
--
Tim Wescott
Wescott Design Services
http://www.wescottdesign.com
Powder cores tend to be terrible at ripple. Preferably no more than 20%. For high ripple (DCM?) you need a low loss mix, usually ferrite.
Powder composite cores can be quite good these days (e.g., Coilcraft XFLs usually have excellent specs), but I guess the present example isn't one of 'em.
Speaking of whom -- if you'd like a brand to hint at, Coilcraft typically has better prices, not to mention useful data, on most of their product line. They sample and sell direct from their website:
Iron powder cores are terrible, not hard to burn the paint off of. Permalloy powder is good but expensive. The "KoolMu" formulations are almost as good as permalloy, but cheaper.
Coilcraft is great.
--
John Larkin Highland Technology, Inc
picosecond timing precision measurement
jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com
You're pulsing it, so 17A is the max current during the pulse. (you're missing some transients? I know nothing of magnetics, I did pick up my SMPS book the other day, I'm reading the magnetics section. (maybe the blind pig will find an acorn. :^)
If you could work out what the core material was, and what it's magnetic pr operties are, LTSpice will let you run the John Chan model to get hysterisi s losses in the core, but only for an inductor.
The shorted loop in a moderately conductive core is much easier to model, b ut I've no idea how yo find out how conducive your core is, short of disman tling the inductor and looking at the core on it own. Just dropping the bar e core into a coil and seeing how much difference it makes to the coil loss es and inductance might be interesting, but I've never done it.
Note that that's traditional iron powder toroids -- Micrometals designations, one or two digits, old school names like "carbonyl iron" and "hydrogen reduced".
The cheapest and highest permeability of which are #26 (yellow-white) and #52 (green-blue, about half losses), both of which are only slightly better inductors than they are resistors. Typical Q factor of 10-20, so they're only useable at very low ripple fractions, under 10%.
The lower mu (10 to 30) varieties aren't bad -- I've used #18 and #2 cores in snubbers and DCM boost converters before, though not very powerful ones.
The alloy powder cores are better still. Which as far as I know, are still rather old technology, and still not well suited to DCM flyback converters, which still need ferrite, or something as good.
As for modern powder-composite (molded and pressed) cores, they must be related to these, but I haven't seen any descriptions of how the formulations differ, or what the core parameters are (when pressed into a more familiar form).
There's no freaking way they're based on #26 or the like -- some of them run cool, in the low MHz, with modest ripple fractions (30%?). Often performing even better than ferrite, and in smaller packages too.
But there are high and low loss materials. The Bourns part in question must be the shitty kind. I've personally had SRP1770's melt off the board -- THUNK -- when subject to rated current at AC (100s kHz).
1uH at 100kHz ... yikes ... that would drive an average ferrite to where you can fry eggs on there. You need more inductance. A lot more, depending on the expected voltage swing across the inductor.
Well, a 3757 running at 100kHz with a 1uH inductor is not a happy situation. Voltage levels above 10V assumed, since I think Tim mentioned something like that. Higher voltages across the inductor will make it worse.
Tim, with the layout being done this is a rock and a hard spot situation. See which inductor that fits the footprint will provide the highest inductance yet still tolerate the worst case peak currents expected. Then crank up the frequency, way up.
Peak. The I^2R losses are swamped by something else -- probably core losses.
Today's experiment is going to be taking four coils in a series-parallel arrangement to get back to the same inductance, and seeing what sort of loss I end up with.
I ordered half a dozen inductors from DigiKey -- basically I bought the top five or six by saturation current, making sure to get a broad choice of brands. I'm going to try them all.
--
Tim Wescott
Wescott Design Services
http://www.wescottdesign.com
Layout isn't a strong issue -- it's on a PC board, but that's just for evaluating the circuit. I'm OK with letting the coil grow considerably if it means better efficiency.
--
Tim Wescott
Wescott Design Services
http://www.wescottdesign.com
So you would recommend that rather than stringing four coils in series- parallel and giving things a whirl, I should be making a string of (at least) four in series? My resistive losses would go up, but as an experiment to see what happens to core losses it'd probably be educational.
At least in simulation the chip seemed to be cranky with continuous conduction -- is this a fault of the chip, or is it just me not working hard enough to understand the feedback-control problem I'm addressing?
--
Tim Wescott
Wescott Design Services
http://www.wescottdesign.com
Joerg, a silly question. Why do you recommend a higher frequency? I've been reading about magnetics.. and they (R.W. Erickson, "Fundamentals of Power Electronics") say that the core losses increase as something like the square of the frequency? They were talking about mostly eddy current losses.
Work with whatever you've got on hand that still won't saturate or at least not much. If saturating you may have to go series-parallel. I don't know what your max voltage swing across the inductor is but I'd start out in the tens of uH.
You'd have to work the compensation more if CCM. But you can't possibly be even close to CCM with a 1uH inductor.
The LT3757 is a nice chip except for the sluggish FB recovery in the non-A version and the cost. AFAIR I may be the reason why there is an A-version.
I have used that chip in the most unorthodox fashions and it always delivered. Often with the FB grounded, "brick on the accelerator" and regulated via the current sense. Before the A version came out I sometimes went in via the compensation pin Vc.
It is based on experience. The last time I almost got burned by an inductor that was trying to unsolder itself it was also a boost converter with a LT3757. Increasing the inductance to the max the footprint could take helped but I was still getting almost 70C. After doubling the frequency of operation that dropped almost to 60C.
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.