Floating an Aux supply for non-isolated SMPS

Discussion in 'Power Electronics' started by mchl_hemingway, Aug 12, 2016.

Thread Status:
Not open for further replies.
  1. mchl_hemingway

    Thread Starter New Member

    Mar 16, 2013
    7
    0
    Hi All,

    I was wondering if anyone could spell out my options, if any, in my latest venture. What I'm hoping to do is create a large (~2kw) switcher to drive a large number of high powered LEDs. Now I'm sure there will be plenty of people rushing to tell me this is a bad idea but for cost/size/complexity reasons I'm going for non-isolated. The basic scheme is going to be:

    Mains > EMI Filter > Bridge Rectifier > Boost PFC (400V, 100kHz) > Buck (~300v) > 30x n type buck (~160v)

    This final stage is 30 low current, asynchronously firing, current controlled n-type bucks. The question here is to do with ground referencing. There is a considerable amount of logic and n type mosfets,etc and need +12, +5, +3.3 levels. The hope was to use a PC PSU as an auxiliary supply to provide these levels and spare a lot of work(there's enough already).

    As I understand, the voltages mentioned above (400, 300v) are referenced to the -ve node of the bridge rectifier. This isn't ground earth referenced. Other than the assured death that awaits me, is there any reason why a PC PSU cant have the ground earth connection removed and have its outputs reference to the -ve node of the bridge. I appreciate this would all need some incredible insulation strategy and a lot of caution but generally speaking am I missing anything? Thanks
     
  2. AlbertHall

    Well-Known Member

    Jun 4, 2014
    1,959
    387
    They'll be making a film of this project. I've already got a name for it - Death Wish VI.
    I don't think there are any problems with this - except the obvious.
    We are not allowed to discuss mains powered LEDs, but if we stick to making a PC PSU lethal then that's probably OK :eek:

    PS: Have you made a will yet?
     
  3. crutschow

    Expert

    Mar 14, 2008
    13,035
    3,242
    I sense someone in the running for the Darwin award. :rolleyes:
     
  4. mchl_hemingway

    Thread Starter New Member

    Mar 16, 2013
    7
    0
    Well if anyone could suggest how to reduce the Darwin award to an honourable mention it would be appreciated. The development plan would include a 3KVA isolation transformer for probing on the scope. Probably have some sort of relay hooked up to bleed the caps. Everything would end up in a fully enclosed grounded case, fused, RCD.
     
  5. crutschow

    Expert

    Mar 14, 2008
    13,035
    3,242
    Certainly if you build this in spite of the warnings then, at a minimum, power it from a GFI outlet when testing.
    That way you should just get a nasty shock rather than a visit to the morgue if you touch the wrong node.
     
  6. tcmtech

    Well-Known Member

    Nov 4, 2013
    2,039
    1,667
    To me, this sounds like an unnecessarily and overly compllicated way of doing something simple. :(

    What do you expect to gain from it that just running 30 or so common line powered SMPS LED driver units could not accomplish for far less complexity and cost?

    PF correction for a load that size has no relevant gains to justify the added complexity and costs.

    Neither does boosting the line voltage to a higher DC voltage just to buck it down to mid-stage somewhat lower voltage then again to much lower DC voltage to ultimately run a bunch of LEDs either and all in a line driven unisolated multistage power circuit on top of that.

    Tell us what your power source is and what the LED's loads are and we can probably give you a far more practical and cheaper method to achieve the same equivalent function.
     
  7. mchl_hemingway

    Thread Starter New Member

    Mar 16, 2013
    7
    0
    Fair enough Tcm, ill give you a better idea what I'm trying to achieve. The hope was to have independent control of the currents going through the 30 strings. This was going to be achieved using a VHDL implementation on a Spartan 3 FPGA as well as a 32 channel multiplexer and a ADC (obviously other parts).

    The fpga side of things is something I know and seemed a reasonable option given the parallelism of the task. It would have a set of target currents and a module would implement an asynchronous 30 channel PWM to track the currents.

    I agree, I seem to have come to a silly conclusion on how this should work with the up and down buses. I figured I needed the PFC because there may be a number of these units running together so obviously a very large current spike. The 2nd stage I thought I needed to remove the the low frequency (100 Hz line) ripple coming out the PFC stage. I know some 100 Hz ripple isn't going to make much difference to LED strings but I thought it might make it difficult when sensing the current. I'd imagine there may be a be a few solutions to this but I just preferred a well regulated bus that didn't require £100+ of output caps for the holdup.

    hypothetically, if I was going to develop this in a robust way, with the FPGA control, and 1.5 - 2.2kW output into 30 strings, reasonable size, what would the scheme look like?
     
  8. tcmtech

    Well-Known Member

    Nov 4, 2013
    2,039
    1,667
    PF control doesn't always have to rely on raising the voltage to some arbitrary level then reducing it back down to another level. A properly designed simple SMPS with minimal capacitor filtering on its input DC rail and a good feedback control loop design followed by normal properly sized secondary side filtering can do the same thing.

    As for ripple on power to LED's they aren't that fussy about it and the same goes for your eyes. Odds are the display you are looking at now only uses a 24 - 30 FPS (24 - 30Hz) frame rate any you can't detect that even with your peripheral side vision which is way more sensitive to high-speed movements and strobing effects than your center of vison you are reading this with is.

    Same with dimming techniques. Variable current DC Vs PWM is not an issue so long as the PWM frequency is higher than our visual range can detect which for the most part any base frequency over several tens of Hertz is way beyond our eye's ability to detect even as rapid movements in your peripheral side vision and are still well below what any basic networked array, even a huge one, would have trouble dealing with.

    Given that I see no need to mess with the individual power supplies voltage and or current regulation to achieve LED dimming effects. PWM with the proper frequency and array control can easily handle that with minimal components.
     
  9. mchl_hemingway

    Thread Starter New Member

    Mar 16, 2013
    7
    0
    Ok, I think I see what your saying. Could you suggest a topology? would a simple buck converter do the job or would a non-inverting buck-boost be better to exploit the full ac cycle? the load needs about 160V. Could you make any suggestion about a feedback scheme/controller IC that might be suitable.
     
  10. ScottWang

    Moderator

    Aug 23, 2012
    4,855
    767
    The Transformer-less power supplies is not allowed now, so this thread will be locked.

    User Agreement.
    6. Restricted topics[/B]. The following topics are regularly raised however are considered "off-topic" at all times and will result in Your thread being closed without question:
      • Any kind of over-unity devices and systems
      • Devices designed to electrocute or shock another person
      • LEDs to mains
      • Phone jammers
      • Rail guns and high-energy projectile devices
      • Transformer-less power supplies
      If You come across a thread or post discussing any of the above topics please report the thread so We can close, and if necessary remove, the offending thread/posts.
     
Loading...
Thread Status:
Not open for further replies.