Help with modification of power supply for LED lighting

Discussion in 'The Projects Forum' started by Bhante, Dec 15, 2009.

  1. Bhante

    Thread Starter New Member

    Dec 15, 2009
    3
    0
    I need help with modifying the power supply for some LED room lighting.

    I am trying to replace some 12V halogen spotlights with 12V LED spotlights, but am having trouble with the power supply. The original Power supply is incompatible with LED. The LED lamps take about 300 mA each, and in total there are 12 of them. I am using a 5A 12.4V switched mode power supply. When I connected the power supply to just one lamp it lights beautifully without flickering, and there does not seem to be any problem with a minimum current requirement with this power supply. However with all lamps connected, the lamps flicker (all in synchrony); it seems the flickering gets worse and at lower frequency, the more lamps are connected. It looks like there is a feedback oscillation problem somewhere between the power supply and the lamps.

    Under some conditions the flickering vanishes: eg if there is one 50W halogen lamp in the circuit; also if the lights are turned on shortly AFTER the 50W lamp was connected and then disconnected - but naturally the advantage disappears the second time the lights are turned on.

    All this makes me sure that a trivial modification to the power supply should be able to correct the problem - such as a resister or capacitor. I tried first a 1.69K Ohm resistor and then a 240 Ohm resistor connected acros the 12V output - I expected that to help, but neither worked. Maybe some kind of capacitor across the output is what I need? A capacitor should short-circuit the AC component in the oscillation.

    If I connect a multimeter across the 240 Ohm resistor with all 12 LED lamps connected and no halogen (and therefore with the lamps flickering), surprisingly I find only 9.3 V coming from the power supply (with only one lamp directly connected it was 12.4V). AC current across the 240 Ohm resistor is 0. However if I replace one of the lamps with the 50W halogen all the lamps burn very brightly (all LEDs and the halogen) without flickering, and there is 12.4V DC and 26V AC across the 240 Ohm resistor!!!

    Now I am confused!!!

    Can anyone help me please?
     
    Last edited: Dec 15, 2009
  2. Bernard

    AAC Fanatic!

    Aug 7, 2008
    4,173
    397
    Might try 100 to 500 uF cap across 12V supply.
     
  3. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Your switching supply needs some minimum load in order to regulate properly.

    Obviously, 240 Ohms is not enough of a load.
    A 12v 50W Halogen bulb would have a current of about 50W/12v=4.17 Amperes, thus a resistance of about 2.86 Ohms.

    However, you might try 20 Ohms.
    Get a pack of these:
    http://www.radioshack.com/product/i...1&filterName=Type&filterValue=Power+resistors
    Two resistors, 10W each, 10 Ohms.

    Connect them in series across the supply.

    At 12v, 20 Ohms will cause a total power dissipation of 7.2W in the resistors, or 3.6W each. This is comfortably about 1/3 the power rating of the resistors; you always want to use a resistor rated for twice the average expected power dissipation.

    If it doesn't fix the problem entirely, it should at least improve it considerably. You may need to add another pair of resistors in series across the supply.
     
  4. Bhante

    Thread Starter New Member

    Dec 15, 2009
    3
    0
    As I stated in my original post the power supply connected directly to a single LED lamp works perfectly, so it doesn't seem to be a simple question of minimum load. Rather, it seems to be a question of oscillations caused by some obscure instability in the system.
     
  5. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    OK.
    You have two proposed solutions; add capacitance, and add more load.

    Why don't you try them out, and report back your results?
     
  6. nfhiggs

    New Member

    Dec 15, 2009
    3
    0
    LEDs are a non-linear load, while Halogens are a very linear load. The porblem is likely caused by inadequate filtering on the Switching Supply's output. Switching supplies put out DC pulses that are smoothed out using LC filtering. If this supply is designed for use with a very linear load that a Halogen light presents, then filtering of the output is not very critical and may even be left out entirely. I would put a 12DB/octave LC filter on the output with a very low cutoff frequency - maybe a 2200 Uf non-polarized cap in parallel with the load and a 1.5 mH inductor (with a power rating of a couple hundred watts at least) in series with the load and placed before the Cap. That may filter out the current surges caused by the non-linear load and stop the oscillations.
     
  7. Bhante

    Thread Starter New Member

    Dec 15, 2009
    3
    0
    @nfhiggs:

    Many thanks for your post, which looks like a very promising solution. The problem is that I have very limited access to components in the vicinity (other than by post - very slow especially at this time of year!). I was planning to try and get to Conrad tomorow, and have had a look in their catalogue. Non-polarised capacitor will mean non-electrolytic, which rules out most of the contenders. Other than the electrolytic there is a metalised polyester film rated 63V DC/ 40V AC.

    The inductor is more difficult. When you say rated at a couple of hundred watts I assume that means for transients, not continuous? Watts or milliwatts? The following are the ones I could find (all 1.5mH) with specified resonance frequency:

    100MHz 64800 Ohm 0.08A
    10MHz 2.4 Ohm 0.7A
    1.2MHz 5.1 Ohm 220mA
    1.25MHz 21.6 Ohm 100mA

    Doesn't look very promising, does it? If it is in series with the load, it must be able to take 5A passing through.

    If I just touch one halogen lamp to one of the sockets for a fraction of a second (200 to 300 ms) that is enough to stabilise the output and stop oscillations. Any ideas why? It suggests the main problem is with transients on startup only. What about putting the coil of a relay in parallel with the load? Maybe I don't even need to use the relay contacts - when the relay powers up it needs energy to switch the contacts and maybe that switching time might be enough to stabilise the power supply until the LEDs have switched on properly. Do you think that might work, and if so, how big would the relay have to be to have a reasonable chance? Would a small reed relay work, or would it have to be a heavy duty relay?

    If the relay on its own does not work, maybe I could connect the LED load through the relay contacts so that the LED's only come on after the relay; how long is the power supply likely to take to come up to power? I have just this moment checked what happens when I turn on the power supply first and then connect to the LED circuit - no oscillation (that was almost certainly what I must have done before starting the installation, which is why I thought the power supply would work). Leaving the power supply on is not an option in this installation, but maybe it suggests something with a relay might work.

    Regarding the previous suggestions (capacitors vs. resistors) I tried a few possibilities with the few components I could lay my hands on, and none of them worked. To be fair, I couldn't reproduce any of the suggested components exactly: for the capacitors I tried 10 uF, 47 uF, and 1000 uF, all electrolytic. The resistor proposal I haven't been able to check adequately to date. However it is certainly not a question of the minimum load of the power supply.

    What about just putting a zener diode across the outputs? Would that be enough to remove the transients when starting up? There is in fact an unoccupied position marked on the circuit board for a zener diode across the output, but then there are about 18 unoccupied component positions on the circuit board so whether that is for a completely different power supply I don't know. The power supply is nominally 12V but actually 12.4V. What about a 13V zener diode? 12V would take constant power, whereas the 13V would only take spikes. If it worked, it would be by far the simplest solution ...

    The Conrad catalogue has zeners in 13, 15 or 18V and in 500mW or 1.3W. Should I try 13V 1.3W?

    Regards, Bhante
     
Loading...