EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: daqq on May 10, 2018, 07:05:13 am
-
Hi guys,
I've been researching transformers recently, doing simulations and stuff, and found that leakage inductance is a big nuisance. I was wondering, would potting the whole transformer in an epoxy (or similar material) that would be mixed with ferrite dust (finely ground up ferrite of a similar material to the main core) lower the leakage? It should in theory envelop the core and make something like an "outside core" around the windings.
Thanks,
David
-
There are two types of magnetic "leakage": one is causing an external stray field and the other is parasitic inductance that is effectively in series with the output (or input if one wants). In both aspects the usual toroidal transformers are relatively good.
Epoxy filled with Ferrite powder will not have magnetic properties not that different from air: with so much non magnetic material in between the permeability will be rather low - likely well below 2. For powder cores to get a good µ it takes really close packed particles and even than µ hardly reaches 100.
For a good magnetic shielding to the outside use a steel case. Having some space between the transformer and the shield also helps to avoid to couple local fields to the case.
-
Leakage inductance comes mainly from the space *between* the primary and the secondary. Filling this space with a higher permeability material will increase the leakage inductance. In a well made toroidal transformer, leakage inductance is often so low as to not have any practical implication apart from decreasing the available short-circuit current, which can be an advantage in some situations. What is the problem you are trying to solve and in which way is the leakage inductance a nuisance?
-
Kleinstein: Thanks for the information! So, no, that would not decrease the inductance.
Wolfram: Makes sense.
What is the problem you are trying to solve and in which way is the leakage inductance a nuisance?
I'm just starting with a special half bridge converter design (Input: 48VDC -> Output: 300VDC to 400VDC, adjustable, 250W). Not exactly a DC DC converter, rather a cap charger - there's a 480uF capacitor bank that needs to be charged to the target voltage and then needs to partially discharged (a 2us, 600Amp pulse, repetition rate 500 pps).
I'm trying to simulate something using LTSpice, but I don't know what kind of leakage inductance (coupling coefficient for LTSpice) to expect. I'm not at the moment trying to solve any real problem, yet, just looking at what's possible.
-
For some realistic numbers on toroidal transformer leakage inductance, here's a set of examples of different winding techniques and the resulting leakage inductance: http://richieburnett.co.uk/temp/gdt/gdt1.html (http://richieburnett.co.uk/temp/gdt/gdt1.html)
-
I think you are decreasing eminations of the field but you are providing a lower parallel path for the magnetic flux so you are lowering the impedance of the parallel path ordinarilyoccupied by air so its like a voltage divider?
-
If you want some help with making a realistic simulation, post a screenshot of your LTSpice schematic and ideally also the .asc file. Realistic k for a toroidal transformer on a MnZn ferrite core is in the range of 0.995 for one with windings spaced far apart to 0.9998 for one with windings on top of each other.
-
No, that would increase it, if slightly. Tightly wrapped copper tape (leaving a slit so as not to short the transformer) would decrease it, again if only slightly.
If you don't have low enough leakage, you need more layers in parallel, more interleave between primary and secondary.
You may also have too long wires. Inductance is directly proportional to length. Using a thicker core and fewer turns, allows less wire length. Though this also reduces winding area, so it tends to be more applicable to signal than power applications.
What frequency and maximum current are you operating at?
Tim
-
For minimum winding leakage inductance a torpid with bifiliar windings could be be best. However, that really works best for a 1:1 turns ratio, while you will need something around 8:1.
Therefore your best next step might be to interleaved the crap out of the windings (it’s a technical term). Given the high turns ratio, Teslacoil is probably on to something by suggesting parallel primary windings.
Alternatives:
- use ER core with foil windings for the primary and round windings for secondary
- try a planar magnetics solution (this forms the transformer windings in the PCB)
And of course keep a close eye on all the connections outside of the windings: transformer leads, PCB traces, long-legged components. Every little bit of loop area adds up to more inductance (which looks like transformer leakage inductance).
Also, how many units might you need? It can help you to choose between technologies.
Finally: there are DC DC topologies which can actually make use of transformer leakage, e.g. LLC (also full bridge LLC), which could be useful. Have you considered these?
-
I did something similar in the past, but input voltage was 12v, so turns ratio was much bigger. The primary winding had to be 5-turns center-tapped. I ended up paralleling 16 or so strands of wire while trying to spread the 5 turns uniformly around the core... performance wasn’t stellar under heavy hard switching.
There was an alternative technique to winding primaries of push-pull toroidal stepup transformers on a russian forum; basically, it goes like this: let’s say you want 2x5 turns- make multiple bifilar 2x5 turns sections with single strands of wire until the entire circumference of the core is filled up with sections of 2x5 turns. Then phase all such sections in parallel. The idea is to have the equivalent of one massive 2x5 layer evenly distributed around the core. This technique is cumbersome, to say the least.
In the final version of the device i resorted to using E core transformer with Cu foil primaries and interleaved secondaries... and a regenerative clamp for good measure.
-
This is what is required. The underlying fundamental is this: consider if the secondary is short-circuited. It acts like a solid piece of metal, a ground plane. This makes the primary a transmission line over that ground. Transmission lines have characteristic impedance and length, and those two factors become low-frequency-equivalent inductance. This inductance is the leakage.
Thus, you have exactly two options:
1. Use more pairs in parallel, or wider conductors (like foil, wide-facing not edge-facing by the way), to get the impedance down. Standard transmission line geometry and formulas apply!
2. Use shorter windings and fewer turns, to get less length = less leakage. This requires a fatter core, so as to require fewer turns (at somewhat more length per turn) while meeting Bmax. A different core shape may be desirable (pot core > toroid). Alternately, choose a different frequency and core material, like nanocrystalline at 50kHz and 0.6T. You can only do so much this way, though.
If you do not have tightly interleaved windings, then the leakage is more complex, because each bank has self-inductance which adds to the leakage inductance. If you wind three layers primary then three layers secondary, you will have far more leakage (uh... about 3 times?) than for a single layer of the same number of turns and wire length.
Note that leakage is not always your enemy: it's just that most SMPS applications are for lower voltages and higher currents, such that the transmission line impedance needs to be low, tens of ohms or less. For high voltage applications, a high impedance is desirable, which leads to bank-wound structures like you see on CCFL drivers, and special-case designs like used inside FBTs. :)
Tim
-
Sorry for the late reply guys.
Thanks for the excellent suggestions!
1. Use more pairs in parallel, or wider conductors (like foil, wide-facing not edge-facing by the way), to get the impedance down. Standard transmission line geometry and formulas apply!
Interesting suggestion, but seems reasonable. But won't there be some kind of tolerance problem? Say, on one winding I'll have 6.2 turns, on the other 5.8, won't the difference be seen as some kind of conflict?
-
Sorry for the late reply guys.
Thanks for the excellent suggestions!
1. Use more pairs in parallel, or wider conductors (like foil, wide-facing not edge-facing by the way), to get the impedance down. Standard transmission line geometry and formulas apply!
Interesting suggestion, but seems reasonable. But won't there be some kind of tolerance problem? Say, on one winding I'll have 6.2 turns, on the other 5.8, won't the difference be seen as some kind of conflict?
Where are these fractional turns coming from? 1 loop around the core = 1 turn.
You can get apparrent non-integer ratios when measuring a real transformer, due to coupling factor (which is a real number, unconstrained by turns) -- but this parameter will be equal between different pairs of windings, so that no unbalanced currents circulate between them. Even if it did, it's precisely what you aim to improve by using more parallel windings, so that the unbalanced current arises from shorting out leakage. :) (Eddy current in a shield winding serve this purpose.)
Tim
-
I'm just starting with a special half bridge converter design (Input: 48VDC -> Output: 300VDC to 400VDC, adjustable, 250W). Not exactly a DC DC converter, rather a cap charger - there's a 480uF capacitor bank that needs to be charged to the target voltage and then needs to partially discharged (a 2us, 600Amp pulse, repetition rate 500 pps).
I'm trying to simulate something using LTSpice, but I don't know what kind of leakage inductance (coupling coefficient for LTSpice) to expect. I'm not at the moment trying to solve any real problem, yet, just looking at what's possible.
Might want to look at modern two stage automotive ignition designs since those also have an intermediate voltage in the 300-400V range but the input is 12V. Have you considered using a resonant switching design that uses leakage inductance to its advantage?
-
Kleinstein: Thanks for the information! So, no, that would not decrease the inductance.
Wolfram: Makes sense.
What is the problem you are trying to solve and in which way is the leakage inductance a nuisance?
I'm just starting with a special half bridge converter design (Input: 48VDC -> Output: 300VDC to 400VDC, adjustable, 250W). Not exactly a DC DC converter, rather a cap charger - there's a 480uF capacitor bank that needs to be charged to the target voltage and then needs to partially discharged (a 2us, 600Amp pulse, repetition rate 500 pps).
I'm trying to simulate something using LTSpice, but I don't know what kind of leakage inductance (coupling coefficient for LTSpice) to expect. I'm not at the moment trying to solve any real problem, yet, just looking at what's possible.
What about a dual-switch (primary side) flyback architecture to charge the capacitor?
The primary dual-switch can completely recover the energy from the primary leakage into the primary side bulk capacitor, and a variable frequency controller can deliver constant power to charge the capacitor at the maximum rate. Somewhat like a "Joule Thief", but with reasonable efficiency and a somewhat more sophisticated controller circuit: Turn on the switches until the primary current reaches the limit, then let the transformer fully discharge into the capacitor. You'd also want a secondary controller to limit the output voltage when the capacitor is fully charged. Most commercial photo flash capacitor chargers were built this way, often with self-oscillationg single transistor circuits. With the older ones the variable frequency is clearly audible.
Otherwise, I believe it's been already mentioned here: the have a minimum stray field from a toroidal transformer, you want to place the windings in a way that they always go completely around the core (once or multiple times). So the turns of the windings are evenly distributed along the core.