LF: LED Build Critique

Now that we're only about a month away from starting sprouts indoors for our garden this year I built a grow light to see if I could give our sprouts a head start since sunlight near the windows only goes so far. The already made grow lights I've seen are either stupid expensive or are from a very questionable source, so I took up building one.

I still feel pretty newbish on LED arrays so I'm asking if anyone would be kind enough to share their opinion on this build. I've already tested these LED's on a mini hydroponics setup and they are working well for plant growth. If it turns out well this year with the sprouts I may build another light so we can have more of the shorter cycle plants going.

Essentially I'm asking -
How could I have done this better? (without a significant cost increase)
How could I have done this cheaper?

Parts list:
-LED's : 3W 50pcs Lot 3W Full Spectrum 400nm 840nm LED Grow Chip 45mil Bridgelux No PCB | eBay
$0.40 a piece x 32

-Drivers : 10W High Power Driver Supply 85 265 V Constant Current LED Light Chip Lamp 3W 100W | eBay
They were $2 a piece but are sold out now, I may need another source.

- 12' x 12' x 1/8" aluminum plate
About $18

- Fan from DTR
About $2.50

- Arctic alumina

- A salvaged heat sink, power cord, and wire

Total cost was a bit over $55 with salvaged parts.

I think it would be neater with series-parallel array of stings 8x LEDs in series + a 0.5 ohm resistor for balancing, for a total of 4 such strings in parallel. But you need to find a constant current power supply that can deliver up to approx 30V, and 3A.

I don't think driving 700mA chinese LEDs at 900mA is a good idea for lifespan.

Those drivers are NOT 0.98 power factor. There aren't enough components for that feature.

Fancy name-brand adhesive isn't necessary. I've found a $0.80 tube of chinese thermal adhesive works just fine, especially for really low powers like this.

I wouldn't feel comfortable with a mains device that close to an aluminium plate. The adhesive and surface contaminants will degrade with the high voltage over time. Might be fine. Time will tell.

Using a single higher power driver would have been a better option than several small ones.

I've built an LED array similar to this before by rectifying/filtering mains, choosing a number of LEDs to drop about 160-165 volts, and an LM317 to drop the extra few volts. Junk-drawer full bridge rectifier, junk-drawer filter caps, junk-drawer NTC for input surge to caps, junk-drawer LM317. And of course a fuse in case something goes tits-up. The cost was zero and the efficiency ends up being ~95%.

FireMyLaser said:

I think it would be neater with series-parallel array of stings 8x LEDs in series + a 0.5 ohm resistor for balancing, for a total of 4 such strings in parallel. But you need to find a constant current power supply that can deliver up to approx 30V, and 3A.

Click to expand...

How well does this balance the current?

Cyparagon said:

I don't think driving 700mA chinese LEDs at 900mA is a good idea for lifespan.

Those drivers are NOT 0.98 power factor. There aren't enough components for that feature.

Fancy name-brand adhesive isn't necessary. I've found a $0.80 tube of chinese thermal adhesive works just fine, especially for really low powers like this.

I wouldn't feel comfortable with a mains device that close to an aluminium plate. The adhesive and surface contaminants will degrade with the high voltage over time. Might be fine. Time will tell.

Using a single higher power driver would have been a better option than several small ones.

I've built an LED array similar to this before by rectifying/filtering mains, choosing a number of LEDs to drop about 160-165 volts, and an LM317 to drop the extra few volts. Junk-drawer full bridge rectifier, junk-drawer filter caps, junk-drawer NTC for input surge to caps, junk-drawer LM317. And of course a fuse in case something goes tits-up. The cost was zero and the efficiency ends up being ~95%.

Click to expand...

That makes sense now, I already had an problem with the "mains device" issue. While testing the light about 6 hours in a couple circuits were shorting out and I fixed the issue by increasing the silicone pad thickness the drivers are on.

Can I get a link to the $0.80 thermal adhesive?

If it's any indicator of diode life, which I don't know maybe you'll have an feel for it. I build this light using the same LED's (pictures attached below). It's using a salvaged 12V 1.5A supply with the fan drawing .2A. It has a good 500 hours on it and is still going but then again that being only 1% of it's expected life maybe that's not surprising, I haven't exactly been recording the light output to check for any degrading.

BTW ty for the tips, my next build I'll try to go for the array build you recommended and get 95%. Happen to have a wiring diagram to link?

TheDukeAnumber1 said:

How well does this balance the current?

Click to expand...

nearly perfectly. If the LEDs are from the same batch/manufacturer and on the same heat sink, the current will be already mostly balanced without resistors. Those 100W chinese LED things for example are just a 10S10P array of 1W LEDs with no balancing resistors.

TheDukeAnumber1 said:

Can I get a link to the $0.80 thermal adhesive?

Click to expand...

It's $0.70 shipped I guess. And of course cheaper if you buy in bulk (I just searched for thermal adhesive). Hot Thermal Compounds Conductive Heatsink Adhesive Industrial Glue for PC GPU IC | eBay

TheDukeAnumber1 said:

I build this light...

Click to expand...

Driven by what? At what current?

TheDukeAnumber1 said:

Happen to have a wiring diagram to link?

Click to expand...

No, but I threw this together in paint just now.

Most of the Chinese 100W LEDs are not working properly. I highly recommend using balancing resistors, especially with china LEDs, and test each one at low current (1mA-ish) before using them. Unless you want Murthy to f*** you.
Having that many bare LEDs in series on an metal plate seems dangerous, perhaps at least put each LED on a star pcb for more isolation.

FireMyLaser said:

Most of the Chinese 100W LEDs are not working properly.

Click to expand...

Yes, but that's because they often put shitty dies in there with non-matching characteristics and partial-shorts. The ones that work properly... work properly.

FireMyLaser said:

I highly recommend using balancing resistors

Click to expand...

Where possible, ditto.

FireMyLaser said:

Unless you want Murthy to f*** you.

Click to expand...

It's Murphy, FYI.

FireMyLaser said:

Having that many bare LEDs in series on an metal plate seems dangerous...

Click to expand...

Except they're not bare. The thermal plate is isolated from the electrodes. To what voltage rating, I'm not sure. I'll have to look that up or test it later.

Cyparagon said:

Driven by what? At what current?

No, but I threw this together in paint just now.

Click to expand...

I'm glad you asked, I have the power supply directly connected to the LED's and it turns out that they aren't getting nearly as much current as I had initially thought, only .25A, and ty very much for that diagram.

Edit: Also if you'd be willing to test these LED's, I'd gladly send you what I have leftover.

FireMyLaser said:

Most of the Chinese 100W LEDs are not working properly. I highly recommend using balancing resistors, especially with china LEDs, and test each one at low current (1mA-ish) before using them. Unless you want Murthy to f*** you.
Having that many bare LEDs in series on an metal plate seems dangerous, perhaps at least put each LED on a star pcb for more isolation.

Click to expand...

I tested each LED before this build, I've learned about Murphy the hard way in the past. I prefer the star pcb's but personally I think they are harder to solder to post attachment to the heat sink than the bare LED's, unless I'm making a mistake doing it that way.

Cyparagon said:

The cost was zero and the efficiency ends up being ~95%.

Click to expand...

So I built that circuit and my LM317 "seems to me" to heat up very quickly. How did you figure the efficiency? My circuit is slightly different from your sketch. My cap is 270uF @ 150V, and I used 35 LED's since the FV was about 3.4V, and I had my LM317 set to 300ma.

My efficiency was based on the series string running at about 165V 700mA with the total input being ~172V 700mA, since the LM317 is only dropping about 5-6V on average, and rectifier dropping 1-1.5V on average. 165/172 is about 95-96% efficiency.

You're forgetting rectified voltage is a lot higher than RMS voltage. 150V isn't high enough. That's gonna blow in a matter of minutes if not sooner. Rectified mains is 160-180VDC. Use a 200V+ cap. Based on the formula here, you're going to have 9V p-p ripple with that low value. So the cap will peak at maybe 170V but drop to maybe 160V before it is charged again by the next cycle. This will be filtered out by the regulator of course, but that also means it has to drop 10V more at some points than others - more heat. You can add more capacitance to reduce the ripple, but then you have to deal with high inrush current. This is where a NTC or soft-start resistor with its shorting relay would be handy.

You need more LEDs... enough that the LM317 is only dropping a few volts. Your number will be different than mine depending on your nominal mains voltage and the LEDs you used. You can get a ballpark figure, but fine-tuning for best efficiency will be trial and error due to variations mentioned above. Based on 3.4V, you may want to start with 46, and add more one-by-one if the LM317 is still dropping too much voltage.

That is a fairly dangerous approach.

The problem is that it sounds good on paper, untill you start looking at some variables that can change in the real world:

Line voltage is a nominal 120 volt sine, but could vary 5% from that or so. That gives a rectified voltage of 162 to 176 volts. The string of leds could also change voltage drop, to a lower forward voltage when they get hotter. This effect is not huge, but will make a difference of a couple of volts along a string designed to take in the order of 150 volts.

If the variation in the led string voltage is 5 volts combined, the difference in voltage drop across the LM317 would vary by 19 volts between extremes (i.e. minimum mains with cold leds vs maximum mains with hot leds).

As it's minimum forward drop is 3 volts and the maximum is usually 30 volts, you would need to get this circuit -exactly- right in terms of the number of leds used. Adding it all up there is about 8 volts of playroom on a 120 volt system, which is pretty tight.

Obviously you could design in such that it just becomes unregulated when the line voltage drops below 120 volts giving you more playroom, but that would cause the light power to drop if that ever happened (and it's not -that- rare for mains voltage to be a bit under nominal).

Dangerous? Arguably, if you don't know what you're doing. I'll admit my rectifier blew up after the first week, but that's because I didn't account for the inrush. :undecided: I added a NTC to the design, replaced the fuse and rectifier, and it's been fine ever since. I've used this method to light my front entryway for about 3 years now, ~6 hours a night. So in an anecdotal sense, when done properly, it's very reliable.

It's not a project for someone with no mains voltage experience, of course.

Benm said:

the maximum is usually 30 volts

Click to expand...

40V is the maximum differential rating, actually. Furthermore, like most "maximum ratings", it doesn't mean instant death if you poke above it momentarily.

Benm said:

Line voltage is a nominal 120 volt sine, but could vary 5% from that or so.

Click to expand...

It may be different in the EU, but in my experience, mains voltage never rises above nominal (transients aside). It only drops (by maybe a volt) when loads are switched on elsewhere in the house. Mine is 122V. I've never seen it above this, and only seen it below this by about a volt.

Benm said:

The string of leds could also change voltage drop, to a lower forward voltage when they get hotter.

Click to expand...

And that's one reason why we heat sink LEDs. I don't remember what I measured for sure, but I think it was less than a couple volts difference when warmed up.

If you set it up so the regulator is dropping 5V on average, there's really no way the mains variation plus voltage drop difference is going to get above the regulator's rating of 40V. If mains drops to 110V, that only means a dimming of the setup, and everything is unharmed.

When I tested the my "original" circuit I had a string of 45 LED's setup and it consistently lit up 35 but when I tried 36+ they wouldn't lite up. (they did lite but extremely dim). Is there another error that could have caused this?


Unfortunately I ran out of time before my sprouts needed more light, but I did manage to find decently priced drivers that push .25A through (20)LED's. Thanks all for the comments on these lights. I removed half the drivers on the light in the OP and it's running good without issues, and I like that "thermal plaster" a lot more than arctic silver. Hopefully I'll get that circuit to work out for the next build.

TheDukeAnumber1 said:

When I tested the my "original" circuit I had a string of 45 LED's setup and it consistently lit up 35 but when I tried 36+ they wouldn't lite up.

Click to expand...

First rule of troubleshooting: check voltages.

Mains voltage, Cap voltage, Load voltage, Vf of one LED, and voltage on Rset would probably enough to tell the problem. Both at series of 35 and 36.