Core losses in SLR converter

[tomizett]

Hi All,
I've been working on a DC-DC converter to drive a portable amplifier. Input is 12V nominal (battery) and output is +/-20 at about 1.5A max.

The topology is series-load-resonant (SLR), and I elected to use an external inductor (rather than the leakage of the transformer) so that it was easier to play with the vaule. The transformer is re-purposed from a computer power supply, so is a fairly low-leakage type, with a ratio of about 1:5:5 (centre-tapped secondary).

What has surprised me is the amount of core loss in the inductor - it gets uncomfortably hot to the touch fairly quickly at 1A of output. I've used 4 turns on a pair of CS270075 ("Sendust") cores, for an inductance of about 3uH. At 1A output we should have about 15A peak at a frequency of about 50kHz (converter is running continuous-conduction, above resonance). I saw similar losses with CM270060 (MPP) and MS-130060-2 cores.

These cores where simply ones I had lying around. Do you think they are suitable, or would there be better choices for the application?
I chose the SLR topology because it should have lower losses than a hard-switching design. Is inductor loss usually a significant contributor in these converters? Presumably the same counts if transformer leakage is used?

With hindsignt, I'm not so sure it's the best choice of topology given the low voltage and high current - I'm pretty sure that conduction losses dominate over switching losses, which are what the SLR is supposed to minimise.

To be honest, I probably won't bother changing the design now... it does what it needs to. But I'd be interested to hear opinions on topology choice and inductor selection.

Datasheets:
**broken link removed**
**broken link removed**

 

[ronsimpson]

I do not have enough information but;
Is the heat from copper loss or core loss? It is a little hard to know.
Can you post schematic?
Pictures of the transformer?
Wire size?

 

[tomizett]

Sorry if the info is a bit sketchy - the design is not very scientific.
Some of the heat is definately copper loss. I've currently got a "temporary" winding of PVC insulated wire, I think 1 or 1.5sq.mm, which I intend to replace later with a tidier milti-filer winding of magnet wire.
However, with that few turns it's easy to get hold of just the core and it certainly feels like it's heating on its own - not being heated by the wire.

I don't have a schematic drawn up at the moment, but the power circuit is basically just this, with the addition of a turn-off snubber capacitor between the midpoint of the half-bridge and supply:

There are zillions of grades of ferrites (etc) out there, all optimised for different things. I'm just wondering whether the one I've picked seems like the right general class of material?

I did try to calculate the theoretical losses (just out of interest) before I posted, but didn't get anywhere. I'm too tired to work it out at the moment.

(edit for typo - grades, not grates)

 

[ronsimpson]

What is voltage time on the transformer?
12 volts, 3uS, 3 turns ??

 

[tcmtech]

Did you modify the stock windings and if so did you have to split the core to get at the winding bobbin and if so by chance did you lose any spacer sheet or like item that went between the two core halves?

If yes to the first tow and maybe to the last then that might very well be your problem. I've played with repurposing old SMPS transformers many times and during the core spliting more than once I misplaced that spacer which lead to horrible HF operation characteristics for it.

Also, on a small power supply as what your working with, you may want consider redesigning it to work as a simple free running push pull oscillator. That design works surprisingly well for lower wattage power supplies like what you are dealing with.

 

[tomizett]

Thanks for the interest.
Voltage time across the LC tank would be about 6V x 20us. 6V because the tank sees half the supply voltage (it's a half bridge, and the transformer returns to half the supply voltage) and 20us is one 1/2 cycle at 50kHz.
The transformer would see about 4V for the same period of time - that's the 20V output refected back throug the 5:1 ratio.

However, it's not the tranformer that's at issue here - it's the external inductor. I don't think there's anything "unexpected" happening (I've no reason to believe the losses are higher than theory would predict), I'm just surprised that they seem to be dominating losses in the conveter.

TCM; good call on the suggestion of a missing spacer. In this case though, I've re-conficured the windings but not actually disassembled and re-wound anything. And, as I've mentioned above, the transformer itself is fine.
I'm sure you're right that a self-oscilating hard switcher would work. It was a bit of a learning excercise, although the controller has got a bit more complicated than I was hoping for.

Here's a better sketch of the way things are laid out - it's electrically equivalent to the one I "borrowed" from google above, but might make things clearer. L1 is where the losses are occuring.

 

[tcmtech]

TCM; good call on the suggestion of a missing spacer. In this case though, I've re-conficured the windings but not actually disassembled and re-wound anything. And, as I've mentioned above, the transformer itself is fine.
I'm sure you're right that a self-oscilating hard switcher would work. It was a bit of a learning excercise, although the controller has got a bit more complicated than I was hoping for.

I see. Overly complicated for learning purposes. Been there done that too many times to count!

I think you new concept should work just fine for what you need once you get the switching parameters figured out.

One thing though. At low voltages the half switched H-brides like you have drawn up tend to not be all that great at times in the low working voltages and mid power ranges due to the characteristics of the capacitors needed on the right side to make them work properly can in rare cases send them into odd harmonic oscillations and other unwanted actions if the ESR or any other primary HF characteristic of the two don't match properly or if the switching frequency just happens to hit where a harmonic can work it's way up from and star robbing power off they system.

If you have an O-scope (and preferably a spectrum analyzer too) use it to check your waveforms and pay attention to any noticeable harmonics blips on anything you may see. If they are noticeable there is a fair chance that one of those frequencys could ramp up and destabilize things under just the right load conditions.

Its been a long time since I played with that type of half switch H-bridge design but I do recall that they had the rare tendency to do odd stuff like burn up the core or windings if some higher up harmonic ran away with itself because it would dump excess power into a frequency that was way too high for the core or coils to work with and they would just cook themselves for no obvious reasons.

I cant say thats why your present design was overheating but it is a odd (dumb luck coincidence) bug that could explain it. Just a guess at this point but worth your thinking about when all the obvious stuff checks out yet the thing wont work right.

 

[dr pepper]

The core could be warming up excluding copper loss from the ferrite being operated above its design frequency or its being saturated.
The former is difficult to know with salvaged items, the latter is simple enough, are you operating the care above 0.2T?
I've not had much success with resonant designs, it'll probably be even trickier with salvaged bits.

 

[tcmtech]

I've not had much success with resonant designs, it'll probably be even trickier with salvaged bits.

That's sort of where the simple self oscillating circuits have advantages. they tend to self tune to their most efficient frequency regardless of the load right up until they pass the overload limit and quit in one way or another.

Also just doing a simple self oscillating test circuit with one to find the likely frequency band that the transformer works is not a bad thing to start with. I've used that approach before and it tends to be fairly accurate.

Just start with no C in the LC tank part of the circuit and see what frequency things top out at and at what current draw and then start adding capacitance until the amp draw hits its lowest value. Whatever frequency that works out to be is likely at or near the effective frequency bandwidth that that particular core can handle and the point you want to design your actual control circuit to to be near.

Beyond that consider that the frequency that some cores will work in on a full alternating current push pull driven mode will be off from what they originally ran at as PWM type SMPS's.

Its those odd little quirks that I found are the key frustration factors in repurposing old SMPS transformers. They tend to be way fussier about how they are used than the common LF iron core power transformer we are most familiar with, but once you know how to make them talk and tell you what they do like, designing around repurposed mystery ones is not really that bad. It just takes a few extra steps to nail down where you start at.

 

[dr pepper]

That is very much my style.
Might fiddle around with that next time I need something.
Iron cores can be hard work too.

 

[tomizett]

Thanks for the input folks.

Overly complicated for learning purposes.

seems to be the defining characteristic of a lot of my projects...

When I get then chance I'll re-visit the loss calculations and see if they seem to ring true with what I'm seeing in real life. Everything seems to be doing what I expect in terms of the waveforms I'm seeing, and I'm pretty sure that I'm not saturating (off the top of my head I think saturation current was calculated about 80A). It's a good incentive to get to grips with the magnetics equations again - I've mastered them in the past, but every time I grasp them I seem to forget them again just as quickly.

I suspect that this is simply a poor choice of topology for a low voltage converter - I'm sure that the core losses are far outweighing any gains from reduced switching losses.

There is one successful aspect though: This is actually a replacement for a flyback converter I originally designed for the same project... I had to abandon that due to it's horrendous EMC problems. The new design is far better in that respect, at least.
I'll keep you posted and put up a schematic etc as it progresses.

 

[tcmtech]

I suspect that this is simply a poor choice of topology for a low voltage converter - I'm sure that the core losses are far outweighing any gains from reduced switching losses.

One thing that has scratching my head is your 1:5 turn ratio for what in realistic working terms is a ~3:5 step up requirement. It feels unreasonably lopsided for the work it need to do given a simple 1:2 push pull type converter based around a 3:6 turns design could easily put out 20+ volts at load.

It's been a while since I took a SMPS transformer apart but from what I recall they used to be built around something like a ~ 1/2:1 turns per volt primary to secondary ratio which for your design I would be tempted to experiment with a rewound transformer that has maybe 4 - 6 turns primary and ~15 - 20 secondary and see what sort performance it gives you.

As I stated before. HF transformers are not hard to work with but you have find out what they want before you can effectively use one and just dong a bunch of calculations on a random guess work approach tends to not tell you the right info.

Number of turns and frequency band in their given mode of operation are critical numbers you cant just throw random guesses at and expect them to work. They will tell you what you need to know if you ask the right way but you have to build a test circuit to get them to do it.

 

[tomizett]

Yea, 1:5 is probably rather higher than I would have chosen, but it was what I had available. While I have re-wound SMPS transfomers before, I've found it a bit of a fiddle (chiefly, dissassembling the original without breaking anything) so I wasn't going to if I didn't have to.
On the other hand, 1:5 has actually worked out better than I'd expected. It gives plenty of "headroom" for the regulation to work in - the power source is a 12V lead-acid, so could concievably drop to 10.5V or so under heavy load. Admitteddly, the regulation is actually pretty poor as it stands, but I'm sure that it could be tightened up (at least in theory).

The transformer I'm using (from an ATX power supply) would originally have been configured as 10:1, to produce 5V from about 170V... which is even moe lopsided! I suppose they must have needed that ratio to maintain regulation under worst-case line/load conditions.
Actually, I'm quite surprised I've not saturated it, given that I'm driving at probaly 1/4 of it's normal operating frequency - but it's showing no signs if it.