EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: Artlav on November 25, 2016, 01:18:07 pm
-
I got a set of 7 big Bridgelux LEDs, rated at 38-40V forward voltage and 2.1A of drive current, and i want to use them all in one light.
The total forward voltage of the chain is quite close to the rectified 220V mains - 266-280V, so i wondered if there is a way to avoid wasting $600-ish on individual drivers for each of the LEDs, and make some sort of a constant current power supply to drive them directly from the mains.
The LEDs would be enclosed and up near the ceiling, so them being live is not an issue.
How would you go about making such a PSU?
I can think of three options.
-One is a linear regulator. The appeal is simplicity, and i have enough spare radiators to dump 100-ish W it would produce. The problem is that i'm not really sure how to design one of such a scale and what problems the scale can bring.
-Second is a reactive limiter. A capacitor or an inductor. Here i don't really have a clue - can you rectify and smooth the current after it? How large a capacitor (physically) would it take? Does it make sense on such a scale?
-Third one is a switching power supply. I had made constant voltage ones before, and i have made DC-DC constant current ones, but never a high-voltage DC constant current one.
So, the question is - which way is better for the given problem, and what to look for as far as it's implementation would go?
-
For awhile now I have been looking into making a light with 20 100w leds and the best power supply I came up with are 3 12v hp proliant 750w (62.5A) server power supplies in series. the problem with this is that your leds are rated for 38-40v but I think you change the voltage with a trim pot a little. Although you would be using less than 700w of 2250w they are cheap (25$ for 2) and if you only draw 700w the fans should be quite. You could also try the Chinese 12v switching supplies. The server psu's are great, I have melted 14awg steel wire with it. (They have over-current protection so I wasn't worried about frying them)
Sent from my iPhone using Tapatalk
-
I forgot to mention that they are 115v or 220v
Sent from my iPhone using Tapatalk
-
I did a one of these,
(http://seventransistorlabs.com/Images/LED_Light2.png)
Would be scaled up slightly for those.
Tim
-
For awhile now I have been looking into making a light with 20 100w leds and the best power supply I came up with are 3 12v hp proliant 750w (62.5A) server power supplies in series.
Hm, that is actually a viable idea - i already ran these LEDs from 12S LiPo cells with an LD1084 based linear regulator to drop the few extra volts.
A couple of 500W 48V PSUs are reasonably cheap, and would work with the existing controllers...
-
I did a one of these,
(http://seventransistorlabs.com/Images/LED_Light2.png)
Would be scaled up slightly for those.
Tim
Hmmm. 7 transistors used. Coincidence? (-:
-
Definitely possible to series connect & drive off rectified mains. A suitable (chunky!) series resistor would work just fine if the temperature rise is OK (but no dimming). Remember to add a current limiting fuse and a thermal cutout! However, you would then need to treat all LEDs (and cabling, and connectors) as live & dangerous (which might be no trouble if you're doing a one-box design).
You'll need to think about flicker too. If you use rectified mains & no bulk capacitor, the lights will flash at 100Hz (or 120Hz). This can make for terrible video footage and will even make fast-moving objects look strange to the naked eye.
It's also definitely possible to design a buck-mode PWM current regulator to limit the LED current from a 350V DC bus, but that's a project you should only embark on if you don't mind blowing up some stuff. You'll need a current control chip plus separate MOSFET, diode, filter inductor, EMC filtering etc. It's quite a lot of work!
A couple of 500W 48V PSUs are reasonably cheap, and would work with the existing controllers...
This would result in an easier system design. If you're only going to build one, I'd go this way.
PS: holy #### that's a lot of light!
-
Hmmm. 7 transistors used. Coincidence? (-:
Sure, why not? ;D
You'll need to think about flicker too. If you use rectified mains & no bulk capacitor, the lights will flash at 100Hz (or 120Hz). This can make for terrible video footage and will even make fast-moving objects look strange to the naked eye.
Yeah, for which it's nice to have filtering, or a constant current driver. But then you need rectification and filtering, which ruins your PFC.
Actually, you won't have very good PFC to begin with, especially if you go for high efficiency -- unfiltered mains plus a high voltage stack of LEDs means it'll only draw current towards the peaks (when V(mains) > Vf(string)).
So that might be another incentive towards active circuitry as well.
It's also definitely possible to design a buck-mode PWM current regulator to limit the LED current from a 350V DC bus, but that's a project you should only embark on if you don't mind blowing up some stuff. You'll need a current control chip plus separate MOSFET, diode, filter inductor, EMC filtering etc. It's quite a lot of work!
Well, you don't need a chip. :P But it does take some thought to pull off a good design.
Not shown in my schematic: CLC filter on the input and output. (The LEDs are up on a separate panel, fed by a cable, so the driver is inline like a "cord bump" device.) I also added shields around it, for additional RF shielding, and so it's touch safe (because, hey, lots of exposed metal, this was a copper-clad prototype!). And a fuse and switch. The attenuation was confirmed by conducted test, but not by radiated.
Tim