No announcement yet.

My PSU idea

  • Filter
  • Time
  • Show
Clear All
new posts

  • My PSU idea

    Hi, I'm new to the forum but not electronics. I have been experimenting with DC-DC computer PSUs for a while.

    The first one I made was a (shock, horror) linear regulator based, using a 7805 with 2N3055 pass transistor, for 5 volts / 15 amps, and an LM350 with 2N3055 for 3.3v at 15a. The 12v rail was straight through, so I guess you could call it "95W + 12v rail".

    Anyway, I have an idea for a reasonably efficient switching PSU that should really work.

    My thought is to build a transformer that would have probably about a 4:5 ratio of input to output, so that 12v AC (switched from 12v DC) in would result in 15VAC out which would rectify to just over 21V DC.

    Consequenty, if the input was only 7V (while cranking for example), the output would be about 12.3v DC.

    I would use PWM to keep this at exactly 12 volts. This means that around ten amps would have to pass through the transformer for a decent rating, which will require reasonable low gauge wire but is certainly possible at a reasonably high frequency.

    I might have got the ratios a bit wrong (switching DC makes same voltage AC or lower) but you get the picture.

    As for the other rails, I would simply take the original 7 - 15 volts or so and pulse width modulate it down to 5v and 3.3v, no transformer needed.

    Are there any fatal flaws, or should I start working? Also, what size wire and how many turns are recommended?

    I plan to use the TL494 chip for regulation and switching, and power mosfets to do the actual switching.

    Thanks in advance guys. If I can pull it off, I plan to make a fully functional 286W cont (343W peak) DC-DC PSU (12V/10A (12A peak) + 5V/20A (24A peak) + 3.3V/20 (24A peak). I would design these as a fully featured ATX power supply and sell it for around $50 US. If you think that's too high/low then say so please.

    George Dewar

  • #2
    It sounds very interesting to me. I'll certainly be watching this for developments.
    Ford Focus MP3 : Blog Updated 29 January 2009!
    Car PC Status: Complete - Undergoing Software Redevelopment


    • #3
      Excellent. Out of curiosity, why do I have 0 posts instead of 1..... Now I suppose I will have 1 post, after making my second post.....

      Anyway, I think I'll start some planning etc, the worst that can happen is I'll waste a small amount of money.


      • #4
        no offence, but lots of people have tried and gave up. It is possible but not and easy thing that you propose. I have a cct based on a sg3525 chip, but have been too busy / lazy to make the board.
        It's quite a lot of work, I wish you good luck.
        Never let the truth get in the way of a good story


        • #5
          Yeah, it does sound like it would have been done and popular if it were very easy, but it won't really hurt me to try, so I think I will have a go .

          I can easily experiment, but I think I'll start with trying 3 turns per volt.

          So 36 turns for 12v rail,
          15 turns for 5v
          10 turns for 3.3v

          And 21 turns for primary, with a fair few wires paralled. It might work .


          • #6
            have a look at this cct using the sg3525. it might give you some good info.
            Linky If you can't read it properly, let me know and I will send you a better copy.
            Never let the truth get in the way of a good story


            • #7
              Nice meddler!

              That's almost exactly what I was planning on doing, although I have decided to use the transformer for every rail now, to make things easier AND better.


              • #8
                yeah, that was just a sample of the circuit. I have the completed circuit on bits of paper in a drawer in my desk. I also have a circuit for a shutdown controller and the software as well to do
                Never let the truth get in the way of a good story


                • #9
                  Hang on, does that mean you plan to build it too? I'm guessing it does. That's cool, will you sell it, out of curiosity?

                  Now if I have full load (286W), at minimum voltage (7 volts) it will draw just over 40 amps through the primary, so I will have to very carefully design the primary side of the circuit, and use several paralled windings in the primary.

                  Not too much problem, as many computer PSUs put out about that much on the 5v rail, and at 12v it will only draw 24 amps.

                  The absolute peak load is 343W, which at 12 volts means a (momentary) draw of 28.5 amps (fine).

                  So the maximum continuous current through the primary of the transformer is 24 amps, which can be increased momentarily (under 30 seconds) to 28.5 amps for any reason, and a huge 40 amps for a few seconds while cranking hard on a cold day with a flattish battery or a massive diesel engine (ie a crank that drops the battery to 7v).

                  That's not going to be too hard to pull off in theory, but we'll see .


                  • #10
                    I might get round to doing it one day. I am in no rush to do it at the moment because I am about to start renovating my house.
                    I haven't checked your maths, but it sounds about right for the current draw. The 12V rail is the hardest part of the PSU so that should be your first concern. The rest could be quite basic if you wanted using linear regs. (it's nasty but it would work). There are many options to get the desired result.
                    if you want to discuss the more intricate design of my supply, shoot me a pm and we can talk further.
                    Never let the truth get in the way of a good story


                    • #11
                      This sounds interesting...keep us posted on your progress.


                      • #12
                        The point of this is to make a highly efficient fully switching power supply capable of supplying power on par with a regular 240v (120v for you guys) ATX power supply, while smootly handling the conditions in a car.

                        So no linear regulators. Either switching through the same xformer as the 12v rail, or using PWM. Preferrably the first.

                        Thanks for the support Meddler, I might send you a PM soon .

                        I'll buy some stuff from DSE in the next few days and start testing some things.


                        • #13
                          Another thing I was planning on was integrating the shutdown controller etc into it. Saves a lot of hassle, just need to connect it to your car battery and ACC line if you want shutdown functionality.


                          • #14
                            Originally posted by George00
                            So no linear regulators. Either switching through the same xformer as the 12v rail, or using PWM. Preferrably the first.
                            I wasn't saying this is the way do the other voltages, I was just using it as an example to show that the 12V is the hardest to achieve while cranking.

                            Originally posted by George00
                            Thanks for the support Meddler, I might send you a PM soon .
                            Never let the truth get in the way of a good story


                            • #15
                              It's going well.

                              I did a kind of test with parts I already had lying around, and got it to produce the voltage, although much lower voltage than normal as I didn't have a correctly wound transformer just lying around.

                              But I got it switching correctly etc.

                              I'll keep you guys updated - next step is to wind a transformer.

                              BTW, I'm using IRFZ48 transistors, they're great for the purpose. They'll pass 50 amps continuous at 100 degrees C (212 farenheit, Americans, and forgive my spelling...). And I can get them for $0.65 each from the USA - even cheaper in bulk.

                              I'm trying to decide whether to do push-pull with a centre tap, or use a capacitor voltage divider and switch either side of that. I can't experiment properly with either until I wind it though, as 1.6 volts is useless for load testing....