Electronics with passion: amplifiers and power supplies, FET's and vacuum tubes..

High Current LDO Linear Voltage Regulator – the sequel

[Читать на русском]

The previous article about a nearly ideal LDO regulator gained quite some interest amongst readers. Meanwhile I managed to build two "barrel" multi-purpose power supplies based on this LDO circuit.

Working Prototype

Casing was always the biggest doubt in my DIY projects. Here I resorted to nice and cheap stainless-steel cans from Ikea with covers cut from a 6mm thick MDF. Perfect solution... to be stuck behind the desk.

120W barrel power supply


Why bother?

Some say I am (almost) crazy: these days there are all varieties of inexpensive power bricks available on the market, which can provide virtually any imaginable output. Even worse: these two barrels that I am writing about did cost me quite some money (mainly for transformers) and time to build. They replaced some 7 switching power supplies - those I already had and simply put into a storage box in a basement chamber.

And I can tell you why: it's my battle against electromagnetic smog in our living space. Guess where is my micro-wave oven? Right, it's been donated to the cleaning team. My conscience still punishes me for that donation - I'd better have teared it apart for parts: what a nice high voltage 1kW transformer it had 🙂

The topic of electromagnetic trash in the air worth a full-blown thesis work. I'm thinking of returning to EMI issues on these pages one day...

Quick casing solution: heat-sink with fan fits nicely
The pictures are "click-able" for higher resolution.


Components side: LDO + Mains FUSE
Soldering side (Teflon insulated + Kynar wires)


Schematic Variations

There are endless variations possible around the LDO schematic layout I am posting here, all based on a particular application requirements. Let's review here just few simplest mods around the same schematic.

In all the schematic snippets below I kept the elements numbering from my first "MOSFET+tl431 LDO" publication.


Dual Transformer's secondary + Soft Start

Here's the simple mod - adding a soft-start circuitry: just one additional resistor R9.

Most efficient raw power - dual secondary windings

Sample selection of components:

  • VD1, VD2 = Schottky rectifiers 8A 40V 
  • VD5-8 = small 0.5A 200V diode bridge
  • C1 = 15kuF 25V
  • C2, C3 = 47uF 25V
  • C4 = 1000uF 35V
  • R9 = 1kOhm
  • C6 = 0.1uF ceramic

Note the increased value of C4. Together with R9 it makes sure that the pull-up voltage "V++" raises slowly at startup. Since the regulator's output could not go higher that the potential on the C4' positive pin minus the MOSFET's threshold - the output voltage will also raise slowly at the power-up.

Single Transformer's secondary + Soft Start

Here the rectifier together with the voltage doubler seem to clog the schematic. Nevertheless in case of a single secondary the voltage doubler remains as simple as before: a little diode bridge and 3 caps.
Voltage Doubler works fine with single secondary too
Shall the system have another supply with higher voltage that shares the same "ground" connection - it would be a good candidate to replace the doubler altogether. The current needed at the "V++" connection is just few milliamperes and should not burden other rectifier and possibly regulator.


No Current Limiter

Let's do without the current limiter for now. This circuit can provide really a negligible dropout voltage under heavy loads. I could not find an industrial LDO regulator with such characteristics 🙂
TL431+MOSFET=great LDO regulator

For parts values please see below.

Attention: proper fusing is the must! I would suggest installing a fuse rated twice the nominal load current right after the transformer's secondary. The fuse must be of a "slow" type (with letter "T" - time). I would highly discourage you relying on the fuse installed at the transformer's primary in case the transformer provides several secondaries for different loads. The reason is simple: while one secondary is about to smoke - the full load might still be below the rated max provided other loads went off.

LDO with the Current Limiter

Here is the equivalent of the originally published schematic. That one was simply re-drown in an easier-to-read manner, I hope.

LDO regulator with Current Limiter

Sample selection of components:

  • R1, R6 = 2.2kOhm
  • R2, R3 = 470Ohm
  • R4 = 0.22Ohm 3W
  • R5 = 12kOhm
  • R7 = 2.2kOhm multi-turn
  • C5 = 10nF ceramic
  • VT1 = IRFZ40
  • VT2 = 2N2222
  • VD9 = 1N5244B (14V Zener)

Test it!

Here is my tried and true TESTING LOAD. It helped me countless times in testing and tuning audio power amplifiers. Now it helped also giving hard times to the linear LDO voltage regulator rated at 12.6V 2A. The current limit was set to some 2.5A.

8/4/2 Ohm   30/60 W   test load


Further developments of this idea

  1. External shutdown + soft start;
  2. Thermally controlled fan;
  3. Thermal shutdown / fuse;
  4. DIY kits for tube amps upgrade and beyond;
  5. Programmable supply...

Stay tuned - subscribe to my blog! 😉


I do enjoy sharing ideas and helping people. But to become even more efficient I need to know what my readers are actually looking for.

  • Did you find this interesting but want to know more?
  • You know how to do things better?
  • Do you have a question or need help?

Nothing will change without YOU taking an action! Please do a small step right now: leave a comment or drop me an e-mail (see link in the profile or mailto: my family name AT gmail.com).

Sincerely yours,
- Serge Patrushin.


  1. Abdullah Kahraman

    "Did you find this interesting but want to know more?" This suits me. Wanna see more about it, especially external shutdown+soft start. But why do you need a thermally controlled fan? Does it get hot at 2A?I subscribed to your blog. Great blog!

  2. Serge Patrushin

    Thanks Abdullah,your comments are very encouraging. That's what helps me to keep this blog going ;)The design I described here was squeezed to fit into that funny canister. Thus I attached a small CPU fan to the MOSFET. It's Ok (below 50C) while feeding some 1.2A load and the fan is off. At 2A it's still Ok, but gets too hot (~~80C). Since I planned to plug more stuff in it – I'll have to spin the fan, at least slowly (I hope).

  3. significato nomi,significato dei nomi

    I’m really impressed together with your writing talents as neatly as with the layout to your blog. Is this a paid topic or did you modify it yourself? Either way keep up the nice high quality writing, it’s uncommon to see a nice blog like this one nowadays..

    • serge

      Thanks for the good words 🙂
      That’s my humble personal blog, nothing more. Sure I do have some hopes to monetize it somehow in the future…

  4. Andy

    Interesting, but how do you choose the component values for a different voltage range, and a different MOSFET?

    I’d be interested in making at least a 4A linear regulator to provided 6V DC from a 2 cell lipo (maximum 8.4V unloaded, minimum, say, 6.6V before it starts to suffer).

    How small are we talking for the gate bias current? Would a watch battery be enough, or does it definitely need a voltage doubler or step up (boost) converter?

  5. serge

    Hi Andy,
    please accept my apologies for late/irregular replies – I am on the road these days…

    Honestly I’ve not considered battery operation while publishing this. However why not giving it a thought?

    The gate current here is negligible since we are operating in analog mode and not recharging the gate cap at high speeds. The only consumer of the V++ current will be TL431 (remember to feed it with at least 1mA) plus some more milliamps to make it swinging over all possible input voltage range, all dictated by R1.

    Since you are considering battery running at high currents I would assume the device will not run for many hours in a raw. Then indeed let’s try using a lithium coin to add 3 Volts for the gate (V++) and also use the “L” type MOSFET to make sure it can conduct high currents happily having just 3V at its gate.

    • Andy

      Hi Serge,

      Thanks. There’s no rush. How do I choose the resistor values for different input and output voltages? Do you have formulas?

      I guess C5 will have the same value, and I assume there’ll be no need for a zener diode when the input is only going to be 8.4V max.

      Does having the current limiter mean that the output voltage will drop more with a higher load (due to R4 adding to the output impedance), or will it adjust automatically? Have you done any analysis or testing of the output voltage vs output current?

      I want to use it to drive servos in a moderately large model (i.e. a BEC), so the load will vary considerably, and quite quickly. Will it be able to keep up with those changes? E.g, if the load switched quickly between, say 0.5A and 4A several times a second, how much variation would there be in the output of the regulator, and how low is the output voltage likely to fall, if it’s designed to output 6V? I’d need to be able to guarantee a minimum 5.5V at the design output current (preferably 4A, though 3A might be okay).

      • serge

        Hi Andy!

        First and foremost I would definitely consider using a switching power supply for your application.

        Regardless the route you choose I am happy to respond to all your requests – and I believe this will be beneficial to anyone reading this post.

        C5 is there to reduce the loop gain at high frequencies hence killing the possible oscillations. It does have sort of a negative impact on the regulator reaction speeds, but nothing to worry in the applications employing motors and such. On top of that the output filter caps (omitted on a simplified diagram) must take care of high speed transients.

        Provided input voltages V+ and V++ never exceed MOSFET’s Vgs max – the zener can be omitted.

        The current limiter will drop some 0.6 Volts at max load current. That is when VT2 opens and starts limiting the MOSFET’s conductivity.

        Certainly the regulator will keep its output voltage stable as long as the V+ – I*R4 is higher than required Vout. Thus the drop on the current limiter resistor should be taken into account only when there is not sufficient input voltage. Please note that R4 does not actually add up to the output impedance provided the regulator works within the designed limits. Because the voltage feed-back is taken directly from the output.

        The regulator must be capable to cope with the loads you outlined without problems. Just count those 0.6 Volts dropped on R4 for the current limiter, should you use it at all. For the huge load changes one may consider adding some capacitance at the output too.

  6. Andy

    Thanks, Serge.

    But how do you choose the resistor values for a particular set of input and output voltages, and to set the limiting current for the current limiter?

    The idea behind using a linear regulator is to avoid EMI from a switch mode power supply interfering with the receiver in the model. It’s still tempting to use a voltage doubler, which means some switching noise, but hopefully only a negligible amount because it would only be providing a few mA.

    • Serge

      Andy, I got the idea: linear regulator indeed means a silent one :).

      R4 is easy to digress: R4 = 0.6 / I
      where I is the max current you want to limit at. Note that the limiter will not act sharply. Thus that’s rather a safety feature without much precision.
      R4 power dissipation is about PmaxR4 = 0.5 / R4.

      BTW there were some more talks about voltages etc in the comments to the first LDO post: http://myelectrons.com/mosfet-tl431-ldo-linear-voltage-regulator

    • serge

      In order to calculate values for the divider that sets the output voltage (R5, R6+R7) one just needs to make sure that the desired output voltage will generate 2.5V at the TL431 REF pin.

    • serge

      One more idea: why not using a charge pump to get few mA for V++? That would produce very little noise, probably less than anything with coils.

  7. Andy

    Okay. Thanks, Serge.

    What sort of characteristics should I be looking for when i choose the MOSFET, other than the current rating? E.g. how important is the gate threshold voltage, and should I avoid one with a really high current capability to avoid it acting too much like a switch?

    • serge

      Andy, thanks for your questions! I hope our discussion may help others too. And it definitely helps me to sharpen my writings.

      While choosing MOSFET I’d pick one with say ten times higher current rating than the regulator’s max. One can easily use much bigger device too, the only concern in that case may be its input capacitance. That is easily alleviated by increasing the control current, if really needed.

      There is no worry that a big MOSFET may start acting as a switch in this particular topology – the source follower. Such effect indeed might be observed in a topology with high loop gain and improperly engineered feedback, but even there the bigger device would only mean higher capacitance and will not add gain just because of its higher current capabilities.

      In case you can get enough V++ voltage to manage conventional 4V Vgth MOSFET (something V++ >= 7v + V+) – I would refrain from using “L” type (low Vgth ones) just as a matter of reliability.

      Should we be limited to V++ = 3V + V+ or so (as we considered using lithium “button” battery for that purpose earlier) – then a low gate threshold type “Logic Level” MOSFETs remain the only viable choice.

  8. Andy

    Thanks, Serge.

    One option would be a small capacitor only charge pump (e.g. with an ICL7660), as you suggested, though I wonder if the inevitable ripple would find its way to the output?

    I still quite like the coin cell idea, or possibly the third cell of a 3 cell lipo – though it would mean the battery would need a lot of balancing after use, as the third cell would get almost no use, while the other two got a lot of use. For either of those two options, a logic level FET would have to be used.

    A friend of mine asked if it could be made to work from two battery packs, in case one failed. I guess there could be two V+ inputs, fed through pretty sizeable diodes (or several in parallel), but maybe it would be better to have two FET ouputs in parallel? Would that cause any problems? I guess the voltage reference part could potentially be shared.

    I’m drifting off topic a bit with that last part, I know. I’ll order some parts soon and have a go at the single V+ version first.

    • serge

      Hi Andy,
      regarding two batteries – at the first glance I would NOT use two FETs in parallel because the structure diodes would make a weak of failing battery charging from the regulator’s output. Instead I would scavenge a big good Schottky pair from a PC power supply and use that one to form V+ for this regulator from two batteries.

      Certainly we could imagine a more efficient solution for paralleling batteries with virtually no loss and perfect control, but that would mean quite a few components extra.

  9. Andy

    Thanks. I had a feeling there was a reason two FETs shouldn’t be used in parallel (unless it’s with the gates in parallel too, I guess), but I couldn’t think what it was.

    I see there are Schottky pairs that can have a maximum Vf of about 0.5V or so, for quite a few amps, and they’re not expensive. Personally, I think it would be better to use one battery so there’s another 0.5V before the input voltage gets critical, giving more time for a low voltage alarm to do its work. My friend seems determined to use two battery packs though.

  10. Cristian

    Good article, keep going!
    How about the wide band noise, and flicker noise on this circuit?
    How can you improve it?

    Best regards

    • Cristian

      Another question came into my mind, it’s possible scale this circuit to an output voltage of say…200V ??
      I think the main problem would be protect the TL431 with another linear but high voltage pass device right, line another N MOSFET.

      Best regards

      • serge

        Thank you, Cristian, for your comments and questions!

        Putting tl431 in a unity gain on AC provides the lowest possible noise here (just take a bigger C5). Though we should not expect this to be the most silent regulator in the world. Anyway it’s adequate for most solutions, and to my experience it’s better that what lm78XY or even lm317 can do.
        Should you need an extra-quiet regulator – I would go for a two-transistor shunt, exactly as I did for my recent DAC project (still to be published).

        Sometimes it’s amazing how things fall together: another friend of mine is requesting me to create a HV version; and I will need one very soon. Actually I have a sketch along the lines you mentioned, but I am not 100% happy with its security, thus I would not publish that one as is. Stay tuned! 😉

  11. Ben

    Good afternoon! Built your LDO circuit above as part of a power supply I’m building for a tube transmitter. I did find one interesting thing – the choice of output capacitor can cause issues. At first, I used a 0.47uF polypropylene capacitor and noted that it would not regulate properly. When I changed it out for a 10uF tantalum, it worked perfectly. I regret that I did not do more troubleshooting than that – so I am not sure if there were oscillations or something else that was causing the regulation to be poor.

    thanks much and 73,
    ben, kd5byb

    • serge

      Hi Ben, appreciate your input! I am very glad this humble site is helping good people 😉
      Though I would not claim this regulator being unconditionally stable – I have not observed any irregularities with this topology during my experiments. In any case your solution of having a good 10uF cap at the exit should be a good guidance for anyone who would use this LDO.
      Take care!
      – Serge.

  12. kevin

    I really really like your regulator design , simple ,clean and very effective .
    I just finished designing one using multiple series mosfets and it started out nice and simple
    then when I started adding all the input protections required it turned into a turd :(.
    This design is a keeper .Thank you ,and best of luck with your blog .

    • serge

      Hi Kevin, thank you for the feedback!
      By the way, if you’d ever need a second opinion on some schematic – I really enjoy finding weak (and strong!) points in designs. It has been sort of sports between me and friends 🙂 Thus do not hesitate dropping me a note at: patrushin {AT} gmail {d0t} com
      Take good care!
      – Serge.

  13. Pingback: стабилизатор напряжения на мосфете | MyElectrons.ru

  14. BillC

    Serge, thanks for your circuit and extensive explanation!

    I’m not much of an electronics guy (E=IR is as far as my grasp goes), but I think your circuit will solve a problem for me. I have built a nominal 24V VRLA battery/solar-powered gate opening system, but it has a problem. The supply voltage swing beween full sun charging (max. 31 V) and multi-day no-sun discharge of its VRLA batteries (min. 22 V) is more than the gate controller can tolerate. The controller seems to “train” based purely on the gate open/close duration (I had hoped it was smarter … using Watt-minutes consumed or some such). Not surprisingly, the motors run faster at 30V than at 22V and the controller gets confused when the gates hit their stops too soon or too late compared to its training.

    I want to use your circuit’s concept to limit the voltage swings seen by the controller to 22-25V, which I think will be narrow enough to keep the controller happy. A series MOSFET regulator is ideal because of its minimal voltage drop, but of course I need a V++ supply well above my 22V minimum VRLA voltage. I’ve seen a similar circuit that uses an LM555 timer chip to create a voltage doubler for this purpose (http://electronic-circuits-diagrams.com/psimages/powersuppliesckt3.shtml), but its discussion of circuit behavior is not as complete as yours. Do you think this approach to generating my V++ supply will create instability or other problems?


  15. BillC

    PS: I recognize my 24V system voltage exceeds the LM555 maximum. I was thinking of using a 16V Zener to connect its ground pin (1) to your circuit ground, giving the LM555 an operating supply range of 6-15V, based on my system supply of 22-31V. I presume this means the “doubler” could only generate ~ 5V gate-to-source at 22V input. Is this enough to keep the IRFZ40 fully on for my target current of ~2A? Looks to me like the data sheet says “yes” but I’m not sure.

  16. Risam

    Dear Serge,

    this is very nice aticle.

    Btw, today, i am plan to build a 50Amp @13,8V adjustable from 10 to 15V liniar power supply for HAM use. in my box, i have 10 pcs of VSPP20N60C3 MOSFET from infineon. Does i can use this mosfet in cascade or paralel (10x20Amp each = 200Amp total) for solution?

    Thankyou very much for your reply.

    • serge

      Dear Risam,
      welcome to MyElectrons! 🙂

      You’ve got a serious project, indeed.

      1) Please consider the power dissipated per each transistor carefully. 5A per transistor, depending on the input voltage and considering the spec you mentioned, may result in anything as from 25W and higher.

      2) Sure we have to put them in parallel. In order to assure an even current distribution between trahsistors please consider adding some resistance in series with each transistor’s source. To my mind 0.1 Ohm 3 W would fit there nicely.
      By the way you can simply tap R3 from one of those source-resistor connections in order to get a decent current limiter.

      3) Last but not least: a little tl341 will have to deal with some 24000pF of gate capacitance. Thinking of your challenge … I would put a mighty hot PNP emitter follower between tl431 and those ten gates.

      I wish you all the best! Please also share with us your achievements 😉

      • Risam

        Ah, Thankyou serge, thankyou very much for your quick reply.
        Yeah, this is my big project.

        This project began since i found a transformator, ex microwave oven. it is a big and heavy transformer once. Input 220V, and output is 2300V at 500mA for anoda magnetron. so its total powers about 1,2KVA :-)..

        And then, i think i can remove the secondary winding, and then re-wind a new winding to get only 15V at secondary winding. with the simple calculation, i can get about up to 60Amp at 20V (1,2KVA at 20V) with new winding (with 2 x 2,4mm square of email) 🙂

        and then this project was begin….

        i’ve send an email about raw schematic LDO to you. I feel very proudly if you check and give a comment about schematic diagram.

        Thankyou and best regards

        • serge

          Risam, very smart use of a micro-oven indeed! 😉
          Have you got your super-transformer ready?
          May I suggest you to perform initial testing of your setup using just transformer + rectifier bridge + capacitors? The goal would be to check how low the ripple voltage lets you driving the load.
          Hmmm… building a test load of such a scale would be quite a challenge on its own 😉

  17. Risam

    Due to my job (this week is very very busy…), i still not yet finishing to wind new winding on my transformator, but all of material to build have been completed. I hope, in the end of this week i have more time to make it.

    Okay… i’ll perform some test to transformer, rectifier, and capacitor first. i’ll sent to you the documentation of my project.

    There is one challenge that today i haven’t idea… How i can build the dummy load with capable to handling power about 750W, 15V @ 50 Amp? The biggest Resistor on the market around me is only 20W. 35 pcs… @ 10 Ohm in paralel.. oh no.. …. “lieur” in sundanese language 🙂

  18. Jeffrey Ohlson

    Dear Serge ,

    I am a retired electronic tech , and current hobbyist , and I questioned why your LDO regulator circuit had the adjustment pot at the bottom of the divider , until it occurred to me why . . . . If the adjustment pot hits a bad spot , or goes “open” at any time , the power supply’s output voltage goes to zero , instead of maximum voltage ! That is EXCELLENT , good sir ! You should mention that important feature in your circuit description . I want to say thank you for opening my eyes to a important consideration in the design of any power supply !

    • serge

      Thank you, Jeffrey, for kind words! And welcome to the site!
      Eah, pots can be spotty, unfortunately I could not recall from whom I learned that technique so many years ago..

  19. Seffin Arsyad

    Dear Serge,
    Great site, it’s very useful, thank you for your effort.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.