Connect with us

Want to build a bipolar (+,-) 24V Power supply....

Discussion in 'Electronic Design' started by Motty, Oct 21, 2003.

Scroll to continue with content
  1. Motty

    Motty Guest

    I would like to build a +,- 24V power supply for a project I am
    working on. Having not been taught practical design during my
    education, I need some pointers. I assume I need to use a center
    tapped transformer that supplies about 30 volts from 120 (or is it 60V
    since I will need +,- 30V?). Then a full bridge rectifier and some
    applicable filter caps. I would then input the +30 volts to a 24V
    regulator and its circuitry (heat sinked) and the negative 30V to a
    negative 24V regulator and its circuitry(heat sinked). But I need
    tips on where to hook the bridge up to. Do the end taps of the
    transformer connect directly to the bridge (with the center tap being
    ground)? I would really like any pointers/schematics that anyone can
    provide. I need to understand how to do this as well. Using a bunch
    of nine volts is OK for prototyping but not really practical
    (financially or spacially) for this project.

    Also, is there a way to have both the positive and negative outputs in
    one output "jack" or do you have to split them up? I can't remember
    seeing a three node connector. Basically the output of this supply
    goes into a +,- 15V regulator circuit and a +,- 5V regulator circuit.
    These mini power supplies power the rest of the cicuit. (This is
    possibly the most needlessly comlicated guitar pedal in the world, but
    that is another story). I really do need all these voltages...I
    promise! I know I could probably make just a +,-15V power supply, but
    I've already designed the cicuit with the +,- 15 Volts in it so I want
    the 24V. Nothing like having a project within a project!

    Oh yeah, I was thinking 1A max on the be on the safe side
    and for future endeavors.

    Sorry...long post...thanks in advance!
  2. The bridges will have 4 terminals. Two are marked
    "~" for the analog inputs and one is marked "+" and the
    other "-" for the rectified outputs. Connect one bridge
    "~" to the end tap and the other "~" to the center tap.
    Do the same with the second bridge but use the other end
    tap. You now have the center tap common to both bridges.
    You may tie the "-" to both bridges together - this is your
    "DC ground". You then have +V at one bridge "+" terminal
    and -V at the other bridge "+" terminal. I'd split the outs.


    Russell Powell

    Artisan Components
    SR. FAE - U.S. Central

  3. Byron A Jeff

    Byron A Jeff Guest

    Here's a pointer to a pretty good overview:
    Yes on the center tapped. Yes on the 60V.
    Just like any other. Note that the center tap is your ground for your DC
    What amperage are you planning on pulling? That determines if you need heat
    sinking or not.
    As I said the CT of the transformer is your ground. You use the + output of
    the bridge as your +Vin and the - output of the bridge as your -Vin.
    See the above link. I found the link in the epanorama section on dual supplies:

    But the original was gone. So the link is to a google cached copy.
    Think disk drive connector where each of the lines are isolated from the
    other. It's probably a bad idea to think of using an inline stereo jack
    for example.
    No problem. You need the headroom with audio.
    So you'll be dropping 6W or so. You'll only need a light heatsink.

  4. NO!!

    You only need one bridge rectifier. Connect the two ends of the
    transformer secondary to the two "~" terminals of the bridge. The
    center tap of the transformer will be the common or ground point of
    the two supplies.
  5. Jim Meyer

    Jim Meyer Guest

    You really shouldn't add the information about where you work or
    what you do until you get someone to check your posts for accuracy.

  6. I read in that Peter Bennett
    >) about 'Want to build a bipolar (+,-) 24V Power supply....',
    Indeed. If you do what Russell suggests, you get two *positive*
    supplies; you can't get -V out of a rectifier + terminal. And if you
    short - of one rectifier to + of the other, you short-circuit the
    transformer through two of the bridge diodes.
  7. N. Thornton

    N. Thornton Guest


    Others have explained how to do it with a 30-0-30 transformer. Now
    here's how to use a 30v transformer (1x 30v only):

    connect one TF tag to 0v.
    connect the other T tag to 2x diodes, one each way round. One diode
    will give you +V and the other -V.

    Now, point of information, you dont want a 30v or a 30-0-30v TF to get
    24v regulated out. Why? Your 30v TF winding will give you about 45v,
    so the setup will roast the regulator, waste power, and require an
    excesively large transformer & box.

    I dont know what 24v reg youre using, but lets say it wants anything
    upto 3v Vdrop. So your reg input needs to be 27v.

    Now lets allow for 11% mains Vdrop: now you need 30v dc to your regs.

    Diode Vdrop allow 1.5v per diode, and lets say youre using a single
    winding TF, so only one diode drop. Now you need a peak TF V_out of
    31.5/1.414 = 22.3v. So you need a 24v transformer.

    And that will give you more than 11% mains sag allowance, if you work
    back thru the numbers, and use more accurate values for the reg and d

    Regards, NT
  8. Fred Bloggs

    Fred Bloggs Guest

    You will need a fairly big transformer and the voltages are high enough
    that discrete regulator components are most likely required. If
    everything is purchased new, you are looking at $50 just for the major
    components: transformer (>50VA) , bridge rectifier, filter
    electrolytics, high voltage (>40V) rated regulator power elements in
    TO-3 etc; not including hardware such as enclosure, heat sinks, power
    cord, panel jacks, and circuit board. All of the circuits in the
    "epanorama" are wimpy reject hobbyist imitations and not real power
    supplies. You should not need 24V @ 1A to supply unregulated voltage
    into the existing +/-15V supplies; 18V with reasonable ripple at maximum
    load should do it- this will make a BIG reduction in the scope of the
    project. For example, an optimally sized transformer will now be
    something like Vrms= (18Volts/0.8+2)/0.85/1.414*2= 40VAC,CT @ 2A; add in
    a FW bridge at 4A @100PIV, and 2x 50WVDC 10,000uF filter caps, fuse
    primary at 1A.
  9. I read in that N. Thornton <>
    Indeed; I've been using this 'double half-wave rectifier' to get + and -
    supplies from AC-output wall warts for a couple of years. Provided the
    load currents of both supplies are nearly equal, the classic problem of
    the half-wave rectifier, that the DC load current causes asymmetrical
    saturation of the transformer core, doesn't happen.
    This and what followed is ALL GOOD STUFF. 'Hobbiests' please note.
    Hobbyists as well. (;-)
  10. Motty

    Motty Guest

    I am actually downgrading my requirements because of the size
    restraints I have been encountering. I don't want a 4 pound TF here.
    Now I am designing for a +,- 15V regulated output at 1.5A max...using
    LM317 regulator and LM337 (I the top of my head) negative
    regulator. These regulators are rated at 1.5A. Do I need 2A at the
    TF or can I use something smaller? Can I use something smaller that
    the regulators? My circuit should pull no MORE THAN 1.0 amp at any
    time. I want to build the supply a little more heartier for future
    projects that may require more. Also the regulators have an input of
    3V<(Vin-Vout)<40V so its a floating reg that is happy as long as the
    input-output voltage diff is no more than 40V. That I can handle.
    But I do not want to waste power and use a heat sink the size of a
    small dog.

    So lets say I want 19V on the input to the regulators. Add in the 11%
    mains sag (oh,yeah...what is this?) and the diode drop of 1.5 and I
    get: peak TF voltage of - ~22.6V. So I divide this by the RMS
    conversion factor to get a TF output of: about 16 volts. Does this
    mean I only need a 16V transformer to get +,- 15 regulated voltage?
    This is what doesn't make sense. If you could please explain futher
    or somehow point me to a schematic I would greatly appreciate it.
    Feel free to email me directly if you want to.

    Thanks so much for the help.

    PS How do you know what fuse to use at the input to the primary? Is
    it the max current of the TF?
  11. Motty

    Motty Guest

    Here is what I can make of it:

    One tap on the TF goes to ground. The other tap goes through two
    diodes...pointing in diff directions. Each diode then is tied in
    series to a large cap (is 4700uF good?). These caps are tied to
    gound. That way you get +,- Input voltage out of one TF. Is this the
    correct approach? Does wiring it this way divide the current? Is
    half going to the + signal and the other half going to the -? I
    simulated this and got the correct output. The simulation, of
    cousrse, doesn't take into transients and noise on the line etc, but
    it appeared to do what I wanted.

    Let me know.
  12. Motty

    Motty Guest

    I did a SPICE simulation of a circuit with an 18Vrms input (23.95Vpeak
    DC). I used 200V 3.0A diodes and 2 10000uF caps in parallel with each
    diode/resistor combo. Also I used 15 ohm loads on the circuit. Is
    there a good way to determine this resistance? I figure that line
    resistance is about all I will have since I am going directly into
    regulators. I don't know of their input resistance and it is not
    specified on their data sheet but the input is going into a transistor
    array on their schematic of the regulator, so I would figure it to be
    low impedance. Anyway, with these values the voltages worked out
    correctly. The ripple voltage ended up being about 5.6% of the peak
    dc voltage. Is this acceptable? Also with these values, the surge
    currents got pretty high. The negative cycle had about 170A while the
    positive had about 90A. I don't know why they are so different???
    The peak currents are about 30-32A. Are the diodes going to need heat
    sinks or not? It seems that they may be dissipating alot of heat.
    Any problems with what you see? Is there a flaw somewhere? The way I
    see it, this is a good approach to what I want/need. The full bridge
    rectifier would cut the surge currents and capacitor values down but
    at the cost of a much larger transformer. Comments welcome...actually
  13. Yes, LM337.
    Okay, so use a 1A transformer. Actually, use a 2A transformer, because
    you're going with a double half-wave rectifier, so you need 1A for the +
    side and 1A for the - side.

    The regulator rating is a maximum, not a minimum. It says "you're not going
    to get more than 1.5A through this no matter what you do." It's perfectly
    fine to not use the maximum. It's like driving your car slower than its
    maximum speed.
    Mains voltage (in the US) could be anywhere from, say, 107V to 130V.
    (Actually I've seen as low as 95V, but that's extreme.) You need to be sure
    that your circuit will still provide adequate juice at the low end, and that
    it won't blow up at the high end.
    You should be aware that an unloaded transformer puts out a lot higher
    voltage than what it's rated for. 25% higher isn't unusual. Grab a 24V
    transformer, plug it in, and measure the secondary with your multimeter
    (with no load) - you'll probably see 30V. Bigger transformers are less
    prone to this than small ones, I think.
    In the 1980's National Semiconductor used to have a Voltage Regulator
    Application Manual, that had excellent chapters on power supply design,
    including worked examples, and all the details on how to figure heat sinks
    and so forth. Don't know if it's still available; maybe it's online now.
    (I have one, but I'm not selling...) Also, Art of Electronics (Horowitz &
    Hill) has a chapter on power supply design.

    Another good guide is experience: take apart some gear that's got comparable
    power requirements, and see how it works.
    No, because the transformer changes current along with voltage. It depends
    a bit on what you're trying to achieve; but as a first approximation,
    multiply secondary voltage * secondary current to get transformer VA rating
    (VA = volt-amps = sort of like watts but not quite). Now, divide VA by
    primary voltage (120V, in the US) and add 5% to get rated primary current.
    Now, double it to allow for some surge current; or use a slow-blow fuse.

    In another posting, you mentioned your simulation results:
    I'm guessing that in your simulation, you didn't take into account the DC
    resistance of the transformer, nor the ESR of the capacitors. 24V * 1.4 /
    170A = 0.2 ohms, and there's more resistance in the circuit than that.

    But with those big 10000uF caps, you will get surge currents that are pretty
    high, if not 170A. Diodes are rated for "average rectified forward current"
    and for peak current. The very-common 1N4004 is rated for 1A average, 30A
    peak; it would work okay in this application. Or you could go for a 3A
    diode, like 1N5404. Don't worry about heat-sinking: on average, those
    diodes are going to be dissipating 1A * 0.7v = 0.7W, which is not that much.

    You also asked:
    this acceptable?

    The regulator is specified with a certain amount of ripple rejection, listed
    in the datasheet - say, 60dB. How much ripple is acceptable at the output,
    in your application? It's never going to be zero, so you just have to
    decide what you can accept.

    Usually a bit of ripple is not a problem.
  14. N. Thornton

    N. Thornton Guest

    Hi John. How much of an issue is this assymetrical current? I've used
    these kind of supplies where the load is close to symmetrical, but
    what happens when rail loads get out of symmetry? I mean to what
    extent is it a problem for the transformer?

    Thanks, NT
  15. N. Thornton

    N. Thornton Guest

    from multiple contributors:

    'Want to build a bipolar (+,-) 24V Power supply....'

    If you mean load R, R= V/i. Modelling the load as a resistor isnt 100%
    accurate, but its close enough in this case.

    your main limiting R will be the transformer's windings. If the TF has
    5% regulation, that means going from no load to full load will change
    its V_out by 5%, hence you can work out its effective resistance.

    on the reg input, normally yes. if you mean on the output, probably

    yup, factor in transformer R, and dont overlook the limited R in the
    diodes too. You'll still get a fair surge tho, even if nothing like
    that high.
    no, not in real life, not on a 1A supply. They wont be peaking at 30A,
    and they only spend a small percentage of their time conducting.


    FW rect shouldnt make any diffrence to the TF size needed, it just
    halves your reservoir Caps.

    For 1.5A output you need a 1.5A transformer, or if you want 2x 1.5A
    outputs off of one TF winding, you'll need a 3A TF. There are minor
    technical complications to that, but we can skip them.

    for what? regulators? transformer?

    If you mean transformer, you could use an underrated one and heatsink
    it or fan cool it, but I'd leave that till you're more familiar with
    the ins and outs of PSU design. At the mo I'd stick with the maker's
    ratings and avoid fire.

    So IIUC you need to allow for 3v regulator V_drop.

    I dont think its a floater, unless I'm misunderstanding here.

    Yup. Your reg actually only needs 3v, but that other 1v will cover
    ripple nicely.

    Mains sag means that the mains isn't always exactly nominal voltage,
    it can drop a bit, so we allow for 10% drop. I used 11% before because
    it made the numbers nice n easy.

    16v is the ac voltage that the TF produces, ie the rms value. 16v ac
    has a V_peak of 23v-ish, and the reservoir cap will be charged upto
    peak V every time, less whatever V drops the circuit has.

    Half amp antisurge type. And do use a mains fuse, not a car one.

    Ideally one would rate it that way, but transformers take a big
    current surge at switch on, and a fuse under half an amp isnt reliable
    in this situation, so half amp it is.

    sounds good then.

    yup, basically.

    they'll mostly get swallowed by the reservoir cap, and what remains
    will be filtered out by the regulator. A 0.1uF cap in series with 1
    ohm R wired across the TF secondary would be a good idea too.

    no, We're talking stock small regs dropping 4v at 1.5A, handling max
    25v. No problem.

    a) you cant run 2x FW bridges off a CT TF and excpect + and - rails
    out of it. It wont even work.
    b) you havent done your calculations
    c) you dont need a 4A bridge for a 1.5A supply
    d) you dont need 50v caps on a 23v line.

    Power lead: almost anyone has some suitable 3 core lying about.
    Circuit board I wouldn't bother with for such a small project. Its
    just not worth it. Tie the big caps, bolt the regs to the case, and
    string the rest between them. A PCB would just be more work for no

    Finally, this info is intended for people who can build a power supply
    safely. If you dont understand the safety issues, or wont follow them,
    dont do the project.

    Regards, NT
  16. I read in that N. Thornton <>
    It can be a bit naughty with a toroid, because they saturate with very
    little residual d.c. (no vestigial air-gap, as there is even with
    interleaved laminated cores). With laminated transformers, I don't think
    it's an issue if the loads are nominally equal, e.g. op-amp supplies. If
    there is a difference between the load currents, and you don't know the
    winding details of the transformer, only its physical dimensions, you
    can use V = Bac.A.w.n and Bdc = [mu].I.n/l to estimate the induction
  17. I read in that Walter Harley
    about 'Want to build a bipolar (+,-) 24V Power supply....', on Thu, 23
    Oct 2003:
    Note that this IS a first approximation, although many commercial
    products (not safety-tested) use a much higher fuse rating than this
    'first approximation' gives. I've seen 20 VA transformers fused at 1
    A/230 V and 2A/120V. Don't worry - the fuse is quite safe. The
    transformer will burn up long before the fuse goes.(;-)

    The proper way to determine the fuse rating is to measure the maximum
    half-cycle r.m.s inrush current. You need to do several (e.g. 25)
    switch-on-run-switch off operations to find the maximum. You get
    different results according to when, in terms of the supply cycle, you
    switched on, AND when you switched off. If you can't measure it, divide
    the supply voltage by the **d.c.** resistance of the primary winding:
    the inrush current can't exceed that. Then you look at fuse
    characteristics - curves of pre-arcing time against current. Even with
    double-log graph paper, these curves are highly non-linear. You choose a
    fuse that will accept your maximum inrush current for the duration of
    the half-cycle, with a reasonable margin for variations.

    Then you do 25 to 100 switch on - run - switch off tests to check that
    your fuse holds up. If paranoid, repeat with the next lower fuse rating
    to check that it *doesn't* hold up.....
  18. Rich Grise

    Rich Grise Guest

    Has anyone ever actually connected two bridges like this
    to a CT transformer? It seems almost intuitive that it
    will blow something up, but I'm too lazy to actually
    trace it out.

    I've always seen bipolar supplies built with only one
    bridge, hooked up in the usual way, then the CT is 0V,
    and the + of the bridge is +, and the - of the bridge
    is -. Basically two back-to-back full-wave CT supplies.

    Anybody want to sacrifice a bridge or two and maybe
    a transformer testing that two-bridge klooge? ;-)

  19. I read in that Rich Grise <>
    wrote (in <P22mb.3900$>) about
    'Want to build a bipolar (+,-) 24V Power supply....', on Fri, 24 Oct
    Yes people have tried it. That's why I already had the explanation of
    its problems written up and posted two days ago.
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