Power supply help needed.

[matt832]

I'm in the process of replacing my under-cabinet lighting with LED. I have 6 fluorescent tubes of various sizes that I'm replacing. The largest is a 64" T6 tube. I'm replacing that with 16 3W leds.

Here are the spec's.

Specifications: ​

LED Emitter: 3W ​

- Output Lumens: 180-210 Lumens
- DC Forward Voltage (VF) : 3.6-3.8Vdc
- DC Forward Currect (IF) : 700mA
- Color Temp: 2700~3200K (Warm White)

I'm making a simple power supply using a 50V center tapped transformer, bridge rectifier with a 2200uf filter capacitor. I'm connecting that to a Lm317t regulator operated in current limiting mode with a 2 ohm resistor. I want about 600ma current so using the calculation with the LM317's internal voltage of 1.2 v, P=1.2/2 = 600ma. I need a voltage of 16*3.6=57v i opted to use the 25v half of the xformer so I could use the other half to power other lights, perhaps that is a wrong assumption.

It's based on this circuit I found at another site:

The issue I'm having is the regulator is running quite hot. I'm only driving 8 of the LEDs with it rather than the 16. I'm wondering if I'd be better off using the full 50v feed from the transformer and running all 16 LEDs from it.

I'm open to any other circuit designs that are easy to make at home. Trying to avoid buying a driver from China.

 

[mercrazy]

did you measure voltage going to lm317? sounds like excess voltage being burned off as heat. need to bring up LED string voltage close to input voltage-overhead. if you can hold your finger on the 317, it's ok.

 

[Steve K]

Have you considered increasing the heatsinking? The LM317 is pretty hard to kill... if you don't provide enough heatsinking, it'll limit the current passing through it.

 

[matt832]

I had a small clip on heat sink, depending how and where I mount it I can go much bigger. I'll have to play around with the voltage, odds are the best option is to load it with a comparable load and see what the voltage actually is, open no load is not a realistic number. Since I'm doing a lot of LEDs I may have to shift some of the units to another strip if I go across the 50V terminals on the transformer, which may make more sense in the long run.

 

[DIWdiver]

+1
+1

Using the circuit you posted you won't get 63 VAC at the output, you'll get more like 32 VDC. Oh, maybe the capacitor is rated for 63 V and you knew the voltage would be only 32V. Sorry. Probably the cap is DC though.

However, the exact voltage will depend on your line voltage, which varies based on many factors including location, time of day, day of week, and weather (if you and your neighbors are all running the air conditioners, voltage drops). It will also depend on whether the transformer is rated for an input of 110V, 115V, or 120V. Probably 115 since that's what's in your post, but just wanted to be sure.

Also, with 2200 uF and 700 mA, you'd expect about 2.5V of ripple on your power supply. If you fall out of regulation, this will introduce a bit of 120 Hz flicker in the light output.

With a string of 8 LEDs, you'd expect 28.8 - 30.4V at rated current (700 mA). That will give you a nominal overhead of 1.6-3.2V. The LM317 requires over 2V of overhead plus the voltage on the programming resistor, plus the ripple voltage, so you will likely find that the circuit does not stay in regulation very well.

In any case, you have very little excess power being dissipated in the LM317 and the circuit should operate with pretty good efficiency. But the power dissipation in the LM317 isn't trivial. I'd guess you'll typically operate with 2-4V overhead and 500-700 mA, the lower overhead and lower current both occurring when your supply voltage is low and you are dropping out of regulation. That would create a power dissipation of 1.0-2.8W.

Assuming you have the TO-220 package, it can dissipate the 1W without heatsinking. 2.8W however requires heatsinking. Not a lot, just some aluminum or copper to spread out the heat and give it some more surface area to dissipate from.

At max operating temp, which you would likely reach with 1-1.5W in the TO-220 package with no heatsinking, the part will be too hot to touch more than briefly. The temperature at which it starts dropping the output is substantially higher than this. The thermal limit should not be counted on to protect the part for long term operation. It's great for testing and development, but for normal operation you should stay well below the thermal limit temperature.

 

[matt832]

Ah yes, it's a relearning curve for me... It's been too many years since I did any work in this area and thanks to the posts here and elsewhere stuff is coming back to me. I forgot about the 2v needed to keep in regulation. It looks like the limit may be 12 LEDS for the 50v output on the transformer. I will have to explore whether to get a different transformer or split up the lights on differing circuits. I have 3 more I'm building with 10 LEDs each on them - all powered on together so I may have some options. If I were to split the total number of LEDs in this setup (46) into 4 branches it might work with a 50v transformer. 46x3.7=170/4=43v leaving some overhead for the LM317 and the resistor.

That translates into 11.5 LEDs so rounding up to 12 per bank still allowable for voltage drop on a 50v AC source. I have a number of CPU coolers laying around so I may mount the LM317s to several of them somewhat remotely. Thanks for the input, it's very helpful. Anything else I'm missing?

 

[SemiMan]

Ah yes, it's a relearning curve for me... It's been too many years since I did any work in this area and thanks to the posts here and elsewhere stuff is coming back to me. I forgot about the 2v needed to keep in regulation. It looks like the limit may be 12 LEDS for the 50v output on the transformer. I will have to explore whether to get a different transformer or split up the lights on differing circuits. I have 3 more I'm building with 10 LEDs each on them - all powered on together so I may have some options. If I were to split the total number of LEDs in this setup (46) into 4 branches it might work with a 50v transformer. 46x3.7=170/4=43v leaving some overhead for the LM317 and the resistor.

That translates into 11.5 LEDs so rounding up to 12 per bank still allowable for voltage drop on a 50v AC source. I have a number of CPU coolers laying around so I may mount the LM317s to several of them somewhat remotely. Thanks for the input, it's very helpful. Anything else I'm missing?

Keep in mind that peak output of the transformer and the output post the bridge rectifier is 1.414*VAC - 2 diode drops. You are looking at almost 70V off a 50V transformer. There will be other voltage drops in the transformer and level dependent on what the rated current is of the transformer and whether it is saturating.

You are avoid Chinese drivers, but the LEDs you describe sound Chinese. Most modern LEDs have much lower Vf than 3.6-3.8V at 700mA ... and certainly no LED have that tight of voltage binning. Vf could be a lot lower than 3.8 contributing further to your LM317 burning up.

If you are having fun with the circuits, great, but if not, it could be cheaper to buy an off the shelf driver and then you can get dimming and other features.


Semiman

 

[mds82]

As another option, you can get some a pretty good pre-made one. Recom makes some good ones, such as the RACD-60 http://www.recom-power.com/pdf/Lightline/RACD60.pdf .

Also, 16 LED's for under cabinet sounds like a lot. According to the calculation that 3300 Lumen. If you want to reduce the forward voltage you could use different LED's that are rated around 2.9 or 3v instead, which would help.

 

[matt832]

The 16 LEDs are replacing a 64T6 florescent tube that is rated at about 2800 lumen's. And yes, there is a bit of "fun" to the circuit, sort of rediscovering my younger years when I was doing a lot of electronics. It's been a few decades and it's amazing what you forget and how doing things like this starts to jog your memory.

Another part of the reason for doing this is to avoid replacing the tubes every 3 or 4 years, plus the cost of the 64T6 bulb has risen to $25+ locally so for the cost of the tube I can build a replacement that should last more than 5 years. That coupled with some energy savings as the existing lights are using old standard ballasts that are 20+ years old and are potential energy hogs.

 

[matt832]

Well i opted to just use the 25V tap on the transformer. Only capacitors available locally are rated at 50V DC. That given the regulated output was running 40.7 volts no load. If I divide that by 3.7v per LEd I get 11.02 so I rounded up to 12. I'm running at 560ma and 3.3 volts on the LEDs. The LM317 is running very cool, only slightly warm to the touch with a simple heat sink. The LEDs are also running significantly cooler than their counterparts with the constant current supply I purchased, not sure why.

Now all I have to do is work on getting the remaining 6 LEDs powered up and build the other 3 light bars. My design calls for 10 LEDs per unit, but if I power them off the same transformer I might just opt for 12, a little more light won't hurt. I'll have to see if the load is too much for the transformer though. BTW, I have the 12 light unit running on my kilowatt meter and it's showing an operating cost of .06/day, so the cost to run them will be negligible since I only use them a few hours a day.