cheap ebay DIY LED

[larrylwill]

naoki: I have a Thermal camera I can get better readings. Which I will when I get it hooked up. Right now I do not have a driver so I will run it with a lambda power supply and make some measurements. You have done some great work measuring everything. At the price they can be bought at it should be rather easy to make a light stick to replace a florescent bulb fixture. I have a light meter and need to measure how much light I get at about 1ft above the pots and compare the 2. I do know from experience that the Leds do look brighter than incandescence bulb of the same luminescence rating.

[larrylwill]

naoki: I got the led hooked up today and all I can say is I still can't see. That sucker is bright. I hope to start making measurements tomorrow. I will measure the light output and compair it to a 75W (14W) pot light Led I bought for the house. I will also compare angles. I think my meter only measures Lux though. One advantage of the screw in Par Led lights is you don't need a power supply. I think the best way is to series 4 together and build a full wave rectifier circuit to run them from the ac line. I have to do some research on how the do it on the Par Leds, I do not see a transformer in mine, but I hate to break open one to see, but I might. Also I cant find your email can you pm me again with it? thanks Larry

[larrylwill]

Here are some Thermal Pictures.
The led strip is some exposed 5050 leds I have under the cabinet light, which we have used for a year. The cross hairs is the highest temp.

[naoki]

Larry, the camera is pretty cool. What are the differences between the photos? Did you run the cooling fan at different speed, and measured the temp?

Since you have fancy power supply, it would be cool to check how the light output changes with different current (i.e. to see how much efficiency gain you'll get by running it at the lower current).

So your other "Par" led is one of these household spot/flood light which screws into the normal light bulb socket, right? Are you planning to modify these to use full-rectifier circuit (I don't know what it is)? There are videos of how to open these LED bulbs (to see inside, or to increase the lumen output). You can put the bulbs in the oven to warm up the glue.

I went ahead and ordered a bigger one (Cree CXA3070, middle-bin, 3000K). It's much pricier ($40 + $12 for 50W ebay driver), but I wanted to see how the cheap ebay stuff compare with one of the most efficient COB. I'm going to check the output at 30W (to compare with the ebay 30W) and 50W (which I'll eventually run at). CXA3070 is the 2nd biggest CXA series, with max. of 117W. But it still uses 36V-class forward voltage. The biggest CXA3570 is more efficient but, it requires driver with higher voltage. Some people calculated that the efficiency of CXA3070 could be above 40% when it's running at 20-40W. I don't quite understand these calculation, but it sounds pretty efficient (I'm assuming 40% of electricity is converted to light energy, and 60% is lost in heat).

I just sent you an email, check your email now.

[larrylwill]

I did not use a fan, thats just the heat sink pictures taken at different times. The power supply I used is a standard Lamda 30-60V constant voltage. I did not measure the current. In fact the supply will deliver about 25-60V. I will run the supply at different voltages and measure the lux at 10" and also the current. I did buy a 2nd 50W Led with driver off ebay for $15.52 shipped. (auction).
I also did some lux measurements. Please have a look and pick it apart.

I only have a lux meter.
Outside in the shade I got approx 4000 lux
at the led I got approx 4000 lux at 10" dead center.

conversion to watts = 4000 lux
Enter illuminance in lux: 4000
Enter luminous efficacy in lumens per watt: lm/W: 60 for led lamp
Enter surface area: Ft sq: 1
Or enter spherical radius: 0.28209479177
Power result in watts: 61.93536

There is really no way to convert lux to lumens directly.
I know that a 4' T8 bulb puts out approx 2800 lumens
I also measured approx 4000 lux at 10" from 2 T8 4ft tubes in a fixture with tinfoil reflector.

Based on these readings a 50W led puts out about the same light as 2 4f t8 florescent bulbs.
Although the bulbs are 4' long and the led is a spot 5" out of either side only rates as 2000 lux.

It seems that florescent bulbs at $7.00 per 2 in a $12 fixture = $18.00 are about the same price but more efficient than 1 50 W led and cost approx the same. I have not measured the ac current yet. I will.

My setup is 2 - 2 bulb T8 4ft fixtures with one 2700k and one 5000k bulb in each. They are about 10" above the top of the pot, These are only on the Oncs and Cats. This year I have had 4 bloom. A sherry baby, a odium twinkle and max tenuifolia. The window is facing NW and I get light all day but only direct light around 4pm and its filtered throgh a tinted window. Approx 2000 - 3000 lux, in the morning and afternoon, 1000 - 2000 lux. I run the lights from 5am until 3pm. The more I look the more it looks like the florescent are a better solution.

I get about 6500 lux at the center (4 bulbs, 2 fixtures), at the pot top and it drops off to approx 3000 lux at the 4 edges of the shelf. The biggest drawback is the spread of light, drops off rapidly from center, much greater with the Leds.

If you can find flaws in my thinking please do.
Larry

More info:
At start up I measured at 8"
Voltage Current Lux
30 .41A 500
32 .66A 900

After 1 min
30 .43A 600
32 .69A 1000
33.34 .92A 1200
I could only get 33.34v out of this power supply.
As the led warms the current and Lux go up. I do not know the ideal Temp. I turned off at approx 1.5 min.
I only have 1 meter. The voltage and current are spot on but the lux varies a little due to I have to hold the meter at approx 8"

---------- Post added at 05:26 PM ---------- Previous post was at 04:43 PM ----------

More info. I found that: The vast majority of orchids grown commercially for sale require between 800 and 3000 foot candles of light to thrive and bloom.
We can convert lux to FC.
1000 lux = 92.9 FC approx 93 fc. So 800 FC is approx 8611 lux and 3000 fc is 32291.
However they do not need all the light at the same time, so I have to find a how many fc at what period of time chart.


To grow and bloom properly : Low End High End
• Low light orchids need 1,000 fc 1,500 fc
• Moderate light orchids need 1500 fc 3,000 fc
• High light orchids need 3,000 fc 4,500 fc.

Low (1000–2000 fc) = 2 hr filtered sunlight, or 14 hr under lights (20 cm distance)
Medium (2000–3000 fc) = 4 hr sunlight, or 16 hr under lights (15 cm distance)
High = (3000–4000 fc) = 6 hr sunlight, or 16 hours under lights (8 cm distance)

Illumination Conversion - FREE Unit Converter

[naoki]

Thanks for the measurement, Larry! It's quite interesting.

Originally Posted by larrylwill

conversion to watts = 4000 lux
Enter illuminance in lux: 4000
Enter luminous efficacy in lumens per watt: lm/W: 60 for led lamp
Enter surface area: Ft sq: 1
Or enter spherical radius: 0.28209479177
Power result in watts: 61.93536

There is really no way to convert lux to lumens directly.
I know that a 4' T8 bulb puts out approx 2800 lumens
I also measured approx 4000 lux at 10" from 2 T8 4ft tubes in a fixture with tinfoil reflector.

Based on these readings a 50W led puts out about the same light as 2 4f t8 florescent bulbs.
Although the bulbs are 4' long and the led is a spot 5" out of either side only rates as 2000 lux.

I don't quite understand the calculation above, but the 50W LED probably have similar lumens (not sure about PAR) as 2x 32W T8. So LED "should" be slightly more efficient (but we can't easily measure lumen output as you mentioned). There are also a couple complication for the direct comparison. I agree that fluorescent bulbs have more uniform coverage (and lower leaves are more likely to receive light). It can be good or bad depends on what you are trying to do, but for me, it is a strength of T8/T5. Also LED is more directional, and fluorescent fixture needs reasonable reflectors (T5HO generally have decent reflectors, though). COB LED is super easy to DIY, but it is a point light source. To get uniform (and penetrating) light, you still need to use multiple emitters (at lower wattage). One advantage of LED may be the reduced heat, though. My grow tent gets overheated with fluorescent light in the summer, so I have to use some LEDs in addition to T5HO.

Originally Posted by larrylwill

More info:
At start up I measured at 8"
Voltage Current Lux
30 .41A 500
32 .66A 900

Since LED efficiency goes lower with heat (you can see it from your measurement), let's look at the start up measurement. I think you actually put fc instead of lux for your table (it can't be that low; 500lux=46.5fc).
12.3W 500fc 40.7fc/W (yours)
21.12W 900fc 42.6fc/W (yours)
26.85W 1170fc 43.6fc/W (mine, I measured mine at 12", but converted to 8" distance with inverse square law)
This results seems to be a bit weird. First, LED is supposed to run more efficiently when you are using lower current, but it is opposite from your measurement. Second, mine is just 30W module, and yours is 50W, so I expected that yours would be much more efficient.

Originally Posted by larrylwill

I found that: The vast majority of orchids grown commercially for sale require between 800 and 3000 foot candles of light to thrive and bloom.
We can convert lux to FC.
1000 lux = 92.9 FC approx 93 fc. So 800 FC is approx 8611 lux and 3000 fc is 32291.
However they do not need all the light at the same time, so I have to find a how many fc at what period of time chart.

If 2000fc (peak intensity; noon in the greenhouse) is recommended for a plant, you can get similar amount of cumulative PAR by 2/3 of 2000fc (=1300fc) of constant florescent light with 14 hr/day. This is a rough approximation, but take a look at this article and here. The 2nd link will tell you the different amount of PAR for a given foot-candle (sun vs artificial light). But lots of plants seem to be ok with lower than these numbers under artificial light.

[larrylwill]

My chart was in Lux. I was on the X10 scale. I got the calculations on the net. They were not mine. Also the recommended requirements were from a Orchid web site.

All readings were taken with a 5 digit multimeter. Current was made with a series connection. I have no idea how accurate the Lux readings are, I do know that a very small movement above or below 8" could vary the readings by 100 lux. Also in direct sunlight the meter is over range.
I suppose I could check the lux accuracy with a light source. Since CP is more useful than Lux I can convert readings to CP.
I went back and redid the readings, I must have been drunk when I did it the first time. Also the reason the current and Lux are the same after 1 min is the power supply was crowbar'ed at 0.92A at 32.34 as the led got hotter the current went up untill the power supply crowbar'ed.

Here is a new chart.

At start up I measured at 8"
V Current Lux CP
30 .36A 6000 557
32 .66A 8000 743
33.85 .92A 12000 1115

After 1 min +/- due to cool down.
30 .43A 7000 659
32 .69A 11000 1022
33.34 .92A 12000 1115

---------- Post added at 01:11 AM ---------- Previous post was at 12:57 AM ----------

In the 90s I had a 55 gal coral reef aquarium. I spent about 7 years playing with the lights trying to grow corals, mostly against the advice of the pro's on the reef forum. I was successful with Murcury Vapor lights, 2 150W but the cri was terrible. I then tried HID 5500k 250w which looked great but expensive and hot. Then I settled on 6 T12 tubes 2 atinic 360nm, 2 cool white 5000k and 2 grow lights (pinkish color) This was the cheapest and had the best or good CRI. I was able to grow just about any coral I wanted.

[naoki]

Wow, there is no comparison between Cree vs ebay cheap LEDs. Cree CXA3070 is one of the cutting-edge white LEDs at this moment. Compared to ebay COB LEDs, CXA3070 is putting out about 60% more light per watt!!! Most of us are happy with whatever we have been using until we see the difference since we don't know what's out there.

OK, here is the punch line. CXA3070 is $40+$2.23 for holder, and 100W ebay LED is $8. So Cree is 5 times more expensive initially. But after running it 1.04 years, Cree becomes cheaper (I'm using $0.19497/kWh, which is probably higher than most of you). So it doesn't make sense to get the cheap LEDs. I started this thread with the cheap ebay LED. But now, I would recommend Cree CXA3070 over ebay LED. It's difficult to compare against florescent light (or red+blue type LED grow light). But I'm guessing that this CXA3070 is one of the most cost effective ways to grow with artificial lighting. The total costs is about $60 per 50W module of CXA3070 (pretty comparable initial cost to T5HO). You can drive it at a higher wattage (e.g. 100W), but I didn't want to push it to the maximum capacity.

Materials

1. LED:
   • Cree CXA3070-0000-000N-Z230F ($40 from digi-key) + LED Holder: TE Connectivity 2-2154857-2 ($2.23 from digi-key)
     vs
   • LED: ebay "100W" Warm White (similar to the one posted in the begining of this thread, but it has 100 diodes instead of 30 diodes for 30W). $8+free shipping from China
   
   CXA3070 (3000K) comes in 3 grades (bins) depending on the efficiency: Z4 (most efficient), Z2, Y4 (least efficient). Mine is the middle bin (Z2). The LED holder is a bit tedious to use (drill holes to the heatsink), but you don't have to solder/glue the LED.
   
   Here is the CXA3070 + holder:
   
   
   Comparison of the size:
   
   Clockwise from top left: Cree CXA-3070 (3000K, middle bin; Z2), ebay "100W" Warm White (around 3000K), ebay "100W" Cool White (around 6000K), ebay "30W" Warm White (around 3000K)
   
   
2. Drivers: I had 3 kinds of constant current drivers, all from ebay:
   
   From top to bottom: 50W (1.5A), 30W (1.05A), 20W (0.6A).
   
   
3. Heatsinks: whatever CPU heatsink with a fan.
   
   From Left to Right: Cree CXA3070, 3000K, ebay "100W" Warm White, ebay "100W" Daylight White, ebay "30W" Warm White.
   
4. Wire: Cerrowire 18-2 Thermostat Wire (18 gauge, 210-1002BR from HomeDepot).
5. Thermal paste: Arctic Silver Ceramique 2

Measurement:
Explanation of the data columns:

• "fc.1ft": I placed a light meter 12" away from the emitter, aligned to the center of the emitter. It was measured after running it for a minute or more (to stabilize the temperature).
• "input.W": watt used by LED + driver. This was measured by kill-a-watt meter.
• "Vf": forward voltage, the voltage across the + and - or the LED (unit: V)
• "current.A": current going through the LED (unit: A)
• "LED.W": watt used by LED (excluding driver). This is Vf * current.A
• Drvr.Effic: Driver efficiency. (=LED.W/input.W)
• fc.per.inW: footcandle at 12" per given input watt (=fc.1ft/input.W)
• fc.per.LEDW: footcandle at 12" per given input watt (exclude power loss due to driver) (=fc.1ft/LED.W)

LED        Driver     fc.1ft input.W    Vf current.A LED.W Drvr.Effic fc.per.inW fc.per.LEDW
CXA3070    20W(0.6A)     520    19.0 33.58     0.469 15.75      0.829       27.4        33.0
CXA3070    30W(1.05A)    970    33.8 35.04     0.864 30.27      0.896       28.7        32.0
CXA3070    50W(1.5A)    1500    59.0 37.03     1.427 52.84      0.896       25.4        28.4
ebay-100W  20W(0.6A)     320    18.8 27.51     0.552 15.19      0.808       17.0        21.1
ebay-100W  30W(1.05A)    480    27.2 28.06     0.866 24.30      0.893       17.6        19.8
ebay-100W  50W(1.5A)     730    46.0 28.81     1.425 41.05      0.892       15.9        17.8
ebay-30W   30W(1.05A)    520    30.1 31.01     0.866 26.85      0.892       17.3        19.4

LED on the left, and lightmeter on thr right:


Results (graphical representation):


Notes:

• With regard to the driver efficiency, those cheap ebay drivers (30W and 50W) are pretty good (89% efficiency). But the 20W driver is not great (around 82%).
• When you look at fc.perLEDW, you notice that efficiency of Cree and ebay LED increases with lower current. I think some other people calculated that radiometric efficacy of this Cree can be >40% with 20W (0.6A) driver. In other words, 40% of electic enegy can be converted to light energy (the rest is wasted as heat).
• However, running Cree CXA3070 at lower current is a false economy. You can run 2x CXA3070 @50W, or 3x CXA3070 @30W to get the similar amount of light. But it takes 10.7 years to make the cost (initial + electricity) even (13h/day). So I'm going to run it with 50W driver
• As mentioned earlier, after 1.04 years (13h/day), Cree CXA3070 becomes cheaper than ebay LED.
• I was a bit surprised, that luminous efficiency of the 30W ebay LED isn't so different from 100W ebay LED.

These measurements are approximate. For example, the color spectra could be slightly different even though both Cree and ebay LEDs are warm white (around 3000K), and I used footcandle instead of PPF of PAR. Additionally, the light spread could be slightly different and 1-point measrement at the center may be misleading (I think the light spread looks similar).

[larrylwill]

Very good data. I bought 2 T8 electronic ballast 4 bulb fixtures with polished reflectors. I have several Cats and oncs in bloom right now. I also have some afternoon sun because I cut off some low branches, so Im happy with what I have.

[naoki]

This is a super easy version of DIY LED with high performance. Almost no tool required (wire insulation stripper does help). No need to solder nor drill, so much easier than the other DIY LED. Efficiency is pretty high (> 140 lumen/W).

Parts:

• LED: Philips Luxeon XF-3535L. ($19.37 from DigiKey)
• Driver: 900mA, 30W constant current driver from FastTech ($12.02)
• heatsink: Aluminum 3/4" C channel (about $13 from HomeDepot)
• 3-prong AC-plug
• 24-gauge wire
• Wire-Nut (something like this)
• Optional: Kapton tape (something like this, but you can get it cheaper via eBay)

Step-by-step assembly
The LEDs come as a flexible strip. Mine is 525mm long.



The backside of LED strip has adhesive already, so you peel the backing and attach it to the aluminum C channel (I cut the C channel to 22" length with a hacksaw) like a sticker.



This is the close-up of the diodes:



Each end of the LED strip has a connector. One is for + and the other is - (labeled). This connector (AVX 9176-400) is pretty simple to use. If you use 24 AWG (gauge) wires (either solid or stranded wire), all you have to do is push the wire in. You don't have to strip the insulation neither.



After I pushed in the wire, I used Kapton tape to cover it. It's unlikely to short it, but I happen to have Kapton tape.



You connect the + connector of LED to + (red) wire of the driver, and - connector to - (usually white or black) wire of the driver. I used this cheap driver (LEF-30W). Note that the sticker is reversed. So the DC output is actually on the left side with red and white wires (even though the label says that the DC out is right side). These are connected to the LED connectors. The wires of the driver are short, so you need to add a couple feet of 24-gauge wires (with wire-nut or solder).



Then the other side of the driver has brown, blue, and green/yellow (yellow with green stripe) wires, which is AC input. You need to connect an AC plug. The round stick is the Ground (refer to the next photo for the AC outlet). Then there are two flat blades; the bigger one is Neutral and the smaller one is Hot (Live). So you want to connect:
Green/Tellow = ground (round stick)
Brown = Live/Hot (smaller flat blade)
Blue = Neutral (bigger flat blade)



That's all you have to do. Here is the assembled and finished LED. I use soldering, but you can use wire-nuts to connect wires.



A test run. It doesn't look bright in the photo, but it is pretty bright, so you might want to wear sunglasses.




The next photo is showing the beam angle. From 1' distance, it provides about 2' wide homogeneous light. The C-channel looks like a decent reflector. The Cheerio box is not an ordinary Cheerio box. It is modified, and it is actually a DIY spectrophotometer.



The aluminum C-channel, which is acting as a heatsink, does become quite warm, but it is not hot. A better heatsink would improve the efficiency, but the cost will go up quite a bit (LED is more efficient, when it is kept cooler).

Measurement
Driver:
AC input: 22.7W
Vf: 21.38V
DC current: 906mA
Driver efficiency: 85.3% (a bit low-end)

Light measurement at 1' away from the LED in the middle:
PPFD: 110 micromoles/m^2/s
footcandles with LX1330B: 655fc
footcandles with Gossen Ultra Pro: 420fc (the response curve of this meter is quite different, but I'm including it for the comparison with the measurements of my older posts).

The LED is about 2' long. Here is the values of 1x 24W 2' T5HO (actually consumes about 18.15W per bulb in 4 bulb fixture).

AgroBrite 6400K 24W (brand new)
PPFD: 58 micromoles/m^2/s

AgroBrite 6400K 24W (2 years of use)
PPFD: 33 micromoles/m^2/s

Odyssea Plant 24W (new, purplish "plant" light)
PPFD: 81 micromoles/m^2/s

Odyssea Plant 24W (after 3-5 months)
PPFD: 63 micromoles/m^2/s

With 4x bulb fixture with 4 bulbs in it (bulbs are not brand-new), you get
PPFD: 160 micromoles/m^2/s

So 2x XF-3535L (45W) can outperforms (about >30% more light than) 4x 24W T5HO (consumes 72.6W). The cost of this DIY LED is similar or cheaper than 4x 2' T5HO.

Comments:
There are several versions of XF-3535L. I chose 4000K version. This can do 147 lumen/W if you are driving it at 800mA. I couldn't find a cheap 800mA driver, so I'm slightly driving it harder (900mA). This means the efficiency (i.e. the amount of energy converted to light instead of heat) is a bit lower than 147lumen/W. With any LEDs, if you drive harder, the efficiency goes down. Warmer ones like 3000K may have slightly better light suitable for plants because it has more red light. The lumen/W goes down if you use warmer light, but lumen isn't a good measure for plants. From several measurements I have done with LED (with PPFD of PAR, not irrelevant fc or lux), the output relevant for plants (PAR) doesn't go down with warmer light. So I think using 4000K or something lower than that is optimum. Also XF-3535L comes in several CRI. Stick with the lowest CRI for the given corrected color temp (80 CRI instead of 85 in case of 4000K). Higher CRI ones have lower output for plants. There is also 4' version (and 900mA driver works well), so this would be a good candidate to replace 54W T5HO.

Note that the performance of this LED is impressively high. But the price is relatively high compared to COB LEDs like Cree CXA3070 or Bridgelux Vero series (I recently assembled Vero 29, and I'm very impressed by it, I'll try to post more about it when I get some time). It is difficult to compare the performance between COB and the linear XF-3535L, but I'm guessing that they are pretty similar to each other. So if you are for ultimate cost-performance ratio, go with COB. DIY COB-LED is pretty easy, but it is slightly more time-consuming than this strip light (e.g. drilling holes and tapping). There are also other high performing COBs from Citizen and Samsung, but I haven't found a good source to get a few samples. I also heard that improved LEDs are coming out pretty soon. So year of 2015 will be exciting for us artificial light growers!

Note that there are lots of other "strip-light"-style LEDs. Most others are not worth using for plants. The efficiency is really low (for most of them).