Connect with us

eddy currents in SMPS xfrm

Discussion in 'Electronic Design' started by Adam. Seychell, Mar 6, 2005.

Scroll to continue with content
  1. What would cause copper heating in an unloaded transformer constructed
    the following way.

    core: EF20 ferrite (20 x 20mm E core)
    primary: 8 turns, of 0.3mm wire x 4 strands.
    secondary: 135 turns, of 0.2mm wire
    input: 12V 98% duty square wave 200kHz.
    topology: push pull

    When I have only the primary winding the FET+transformer dissipation is
    around 300mW. As expected, the heating feels mostly from the core
    material and is acceptable. However when I add the secondary winding the
    transformer gets very hot as it dissipates a couple of watts. The power
    consumption rises with frequency, reaching 4W at 350kHz.
    There is no significant improvement between the order the primary and
    secondary windings are laid.

    What exacly is causing this loss ? Is it the transformer's distributive
    capacitance of the secondary winding causing loading at high frequencies ?
    Do I need a bigger E core just to combat this effect , even though the
    specified power output will remain relativly small ?
  2. Adam. Seychell wrote...

    Reduce the number of turns by two, three or four times.
  3. James Meyer

    James Meyer Guest

    But keep the turns ratio the same?

    I have a fuzzy recollection of transformer construction details where
    the fringe-ing magnetic field inside the winding area will cause copper losses.
    In most transformers, the field is almost totally confined to the core
    structure. But things like air gaps can result in parts of the field going
    places where it shouldn't and parts of the winding act just like a shorted turn.

  4. I read in that Adam. Seychell
    Did you measure the secondary output voltage? If it's over 200 V, the
    secondary is resonating with its own self-capacitance. Note that the
    actual resonance frequency may be below 200 kHz or above.
  5. Terry Given

    Terry Given Guest

    8T x 4 strands * 0.3mm = 32*0.3mm = 9.6mm, should fit on one layer.

    132 * 0.2mm = 26.4mm, at least 2 layers (probably three or four)

    Proximity effect is what is killing you. At 350kHz, 20C the skin depth
    in Cu is d = 66mm/sqrt(350kHz) = 0.11mm. Because the primary winding is
    only one layer, proximity effect does bugger all, ie the effect of
    having Tcu >> d is negligible (no more than 12% greater power loss).

    Very different story with the secondary though. Assuming the primary and
    secondary are *not* interleaved (primary then secondary, either order) then:

    If dia = 0.2mm then area = pi*r^2 = 0.03mm^2. The equivalent rectangular
    wire is 0.18mm x 0.18mm, so h = 0.18mm. h/d = 0.18mm/0.11mm = 1.61.

    [all formulae/charts use equivalent rectangular cross-sectional wire]

    Looking at Snelling Fig. 11.14, assuming 4 layers then the ac-dc
    resistance ratio Fr = 10, so the *actual* AC resistance is 10 times
    higher than the DC resistance. So the losses will be correspondingly
    higher, likewise temperature rise.

    (4W - 0.3W)/10 = 370mW, which is roughly the primary losses, and no
    doubt about what you expected. So you probably have 4 layers on the
    secondary. More layers makes this a *lot* worse.

    OTOH if you interleave the windings (1/2 secondary, primary, 1/2
    secondary) then there is a line of symmetry thru the centre of the
    primary winding, and we only need to consider one half of the windings -
    the primary effectively becomes 1/2 a layer (we still dont care,see
    above) and the secondary becomes 2 effective layers.

    for 2 layers with h/d = 1.61 Fr = 3, so the AC resistance is 3 times the
    DC resistance. Simply splitting the secondary into 2 separate halves,
    with the primary in the middle, has reduced the secondary copper losses
    by a factor of *THREE*

    Ideally, Fr = 1.33 for wire (1.5 for foil), so the optimal wire diameter
    is h = 1.33*d = 0.15mm, Acu = 0.0225mm^2, optimal diameter = 0.17mm

    Or, if you dont want to split the layers, use much smaller wire for the
    secondary - 0.1mm dia wire has 4x the DC resistance of 0.2mm wire, *but*
    h = 0.008mm^2 so h = 0.088mm. Then h/d = 0.81, so Fr = 2. The DC
    resistance quadrupled, but the total resistance has gone down by
    1-(4*2)/10 = 20% lower than in the original case.

    You can clearly see that splitting the secondary into two halves is a
    *LOT* more effective than simply reducing wire size.

    I have seen transformers that catch fire because of this effect. One in
    particular had 12 layers of 0.6mm thick Cu foil for the primary,
    sandwiched between two halves of the secondary winding. It set the
    UL94V-0 bobbins on fire. I reduced the foil thickness from 0.6mm to
    0.1mm (amidst hoots of laughter from the techs, who though I was an
    idiot). Fr before was about 100, afterwards it was about 1.5. So
    although the DC resistance got 6 times higher, the overall resistance
    was (6*1.5)/(1*100) = 11 times *lower* than before. The measured 400C
    temperature rise dropped to a nice cool 35C. And the techs stopped laughing.

    A good reference is:
    "Soft Ferrites" E.C. Snelling, 2nd ed., Butterworths, ISBN 0-408-02760-6

    another is:
    "Switchmode Power Supply Handbook" K. Billings, McGraw-Hill ISBN

    and the Unitrode magnetic design app notes, available free from (somewhere....I have paper copies)

  6. I read in that Terry Given <>
    The OP said 'unloaded transformer'. Your stuff is good, but appears to
    be irrelevant.
  7. Interwinding Capacitance in the Secondary!

    Instead of winding it like you probably did:


    You Need to do i like this:


    All windings the same way.

    The Alterneative is a stack of Disc Windings, series Connected. There are
    coil formers for that.

    The real easy alternative is to get a CCFT Transformer, where everything is
    already done the right way. You do not say what "relatively small" exactly
    is. And what it is for - If you want DC output, things become easier.
  8. Terry Given

    Terry Given Guest

Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day