Building general purpose multi-ouput mains supplied PSU

Discussion in 'The Projects Forum' started by Gebus, Sep 10, 2016.

  1. Gebus

    Thread Starter New Member

    Sep 10, 2016
    3
    0
    Hey everyone - first time here, nice to meet you all!

    I want to build myself a PSU for general use, like powering MCU dev kits and other low voltage electronic projects. I want to power it from the UK main supply (230VAC rms), and I want to have say, 3 outputs:

    *0 to 20V adjustable, maybe 1A (20W max)
    *5V fixed, 0.5A max, 2.5W
    *3.3V fixed 0.5A max, 1.65W

    So, first thing I need to do is get a DC supply from the mains AC. This supply can be rough and and I plan to sub-regulate with SMPS to get nice clean outputs. I haven't decided on a full topology just yet as I don't know yet how to make sure I won't burn out my transformer, which is the primary purpose of my visit here. Looks as though it's a good idea to use an isolated transformer for safety reasons. I have found this part 173-9579 on RS, capable of 230VAC to 2*24VAC with 60VA capability. If I connect the two transformer outputs in series I can get 48VAC with my 60VA. Using a full-bridge rectifier with smoothing capacitor will get me around 33VDC.

    All good so far?

    Now for the first hurdle: How much DC current can I draw from this supply while staying within my 60VA limit? Preferably I'd like to stay at around 75%, 45VA, of this for reliability (is this sensible or overkill?). So after the initial inrush to fill the smoothing cap the current pulled from the transformer will vary with my load, which in turn I want to be variable! If I use a massive amount of smoothing capacitance the inrush risks burning the transformer out. Too little and the supply has more ripple, leading to bigger gulps of current from the transformer. I just don't know where to go here... how do I manage smoothing capacitance with adjustable load and not start a fire?

    Thanks in advance :)
    G
     
  2. AlbertHall

    Well-Known Member

    Jun 4, 2014
    1,929
    381
  3. ian field

    Distinguished Member

    Oct 27, 2012
    4,415
    783
    You'd probably be better off building around an ATX PSU - it has precisely regulated 3.3 & 5V outputs that will be most useful for logic work. But you have to load both those rails to get the full ticket current from the 12V output. You can probably get LDO linears for the lower "core" voltages. A real plus is; the metal box with all the dangerous voltages safely inside.

    The continuously variable PSU can get pretty complicated depending on what you want to do, and how.
     
  4. Gebus

    Thread Starter New Member

    Sep 10, 2016
    3
    0
    Thanks Albert, unfortunately this doesn't quite go far enough for me. One of the links has no explanation of the equations (I want to understand this) and the other assumes a fixed load current for cap value calculation.

    I'm thinking maybe I could fit multiple caps and switch them in/out with FETs depending on the load current? Might be problematic if the load current is transient though...
     
  5. Gebus

    Thread Starter New Member

    Sep 10, 2016
    3
    0
    Thanks Ian,

    By an ATX PSU you mean basically a computer PSU? I had thought of that, but I'm looking to learn something by this exercise, otherwise I could just buy one. I know designing one could get complicated, but that's the fun part.
     
  6. AlbertHall

    Well-Known Member

    Jun 4, 2014
    1,929
    381
    I have attached a more detailed discussion of the calculations. There is much information on this topic on WWW and you could search to find whatever information you want.

    There is no advantage in doing this. You fit a capacitor which is large enough to produce an acceptable ripple at the maximum current load current you are designing for.
     
  7. #12

    Expert

    Nov 30, 2010
    16,276
    6,788
    What country are you in?

    No, you don't want 48 volts to make a 20 volt supply.
    More like a dual 20V transformer.
    20 x 1.414 = 28.28V
    28.28 - 1.4 for the rectifiers, 2.2 volts for the regulator, 20 volts for the output = 4.68 v for the ripple.
    2 amps/radical 2 (4.68) 120 = 2516 uf
    Buy 2700 uf @ 35V

    Then forget about going to zero volts on the output. Nothing runs on zero volts.
    A stupid chip called LM317 will have a minimum output of 1.25 volts. Not a problem.

    Then forget about inrush current. These parts are designed to survive that.
     
  8. ian field

    Distinguished Member

    Oct 27, 2012
    4,415
    783
    For less than 1A per rail, the 78xx 3 terminal regulators still have their uses. Most appnotes show how to add an external pass transistor to regulate more current - but you'll end up with a big heatsink. You can use the same trick with a LM317 for the variable.

    Multiple rails can get untidy fast. A 7805 needs at least 8V input unless you use a LDO type.

    The salvage and surplus trade used to be awash with open frame 5 & 12V switchers, but they got harder to find.

    For a learning project, you'll be better off re purposing a ready made PSU - do the learning on projects that are running on a PSU you can have confidence in. Building your own mains switcher is very much expert level. Anything over 48V isn't a good idea - you can get telecoms PSU boxes at that voltage, designing buck regulators to run from that would keep you amused for a while. Car batteries aren't without their dangers - and you'd probably need to catch up with buck/boost topologies.
     
Loading...