Optimizing LEDs / the dilemma of tall enclosures / tank build log (possibly pic heavy)

Calvin, it's great to talk about DIY-LED! The heatsink you found seems to be pretty good (I don't know how to calculate or read the heatsink specs., though). It seems to be pretty quiet, 16.5dbA @ 1m, but I'm not sure if 24dbA of Arctic Alpine 11 Plus is measured at a closer range? I'm not familiar with the bearing system (Vapo Bearing System), but is seems to be equivalent to Fluid Dynamics Bearing of Arctic (according to this page).

If you want something cheap for the swivel, hanger strap might be a possibility. You can make a shape like|_|, and mount the heat sink to the open end (and screw the strap to the enclosure). It is flexible, so you can adjust the direction by bending the strap. If you want to be more fancy, how about mini-ball heads? something like this, but you could find it cheaper in eBay. Square heatsinks are probably easier to attach to the swivel, but I'm sure you can do it with the round one, too. For example, there is probably enough space at the LED mounting surface, where you can make holes to attach the hanger strap or make L-shaped stay from aluminum bar, and attach it to the ball head.

I got one of each:
BXRC-50C10K0-L-04
BXRC-40E10K0-L-03/url]
and
2x Molex PicoEzMate (optional connector)
The links are to Canadian digikey, and I noticed that the price is much higher than in the US.

I wouldn't use BXRC-50E10K0-L-04 (CRI 80) because of the lower total output.

I also ended up getting the Philips XF-3535L strip light to play with.

Are you going with 50W class driver (around 1.5A) for Vero 29? I'd like to drive it at 80W eventually (efficiency drop at a higher current seems to be smaller with Vero 29 than CXA3070), but I haven't found a cheap driver (around 2.1A), so I'm sticking with cheap 1.5A driver for now.

With regard to lens vs reflector, here is some info:

http://www.ledsmagazine.com/article...es-vie-for-usage-in-ssl-systems-magazine.html
http://www.bridgelux.com/wp-content...l-Consideration-for-Bridgelux-Vero-Arrays.pdf

I'm not knowledeable enough about this comparison. I thought lens could provide more homogenenous light, but lower efficiency (I may be wrong about this). But heat might get trapped inside with lens. At the very end of p.7 of the 2nd document, it says that since COB has large emitting surface, optimized lens system become too large. So reflectors are more commonly used for COB LEDs.

 

Calvin, thanks for the link to the thread (and youtube video). Some stuff is over my head, but I'll learn about it more. I think I read somewhere that if you have a constant voltage AC-DC driver and constant current DC-DCdriver, efficiency suffers a little bit (compared to the direct AC-DC constant current driver). So I haven't considered that route, but I don't remember where I saw this (and I might be wrong).

The Mean Well LPF-90D-42 seems to be nice with dimmable feature and high efficiency (90.5%). But it is pricey. The cheap driver I'm talking about is something like this ($12+1.50 shipping). This one is tested by a knowledgeable person, and around 89% efficiency (and it has good specs). I have one similar to what you linked (one at 30W=1050mA and another at 50W=1.5A). I measured the efficiency, and it was also pretty high (89.6% and 89.3%). But both of them were giving lower current than rated (864mA instead of 1050mA and 1.142A instead of 1.5A). I tested another waterproof 50W, 1.5A eBay driver (YM-361500A), which was giving a higher current 1.427A and also a high 89.6% efficiency. So this is pretty nice, but I couldn't find it again when I looked for it last time. One thing I worry about these cheap drivers are the quality of electrolytic capacitors (do they last for the life of LED?). With the waterproofed version, I won't be able to replace the capacitors. But I've been using the waterproofed version since I don't have to make a case or something for the naked driver.

We can do the rough calculation by considering two scenarios:
1. Vero 29 + $40 driver to get 80W
vs
2. Vero 29 + $13.5 driver to get 50W
If I use the US price, option 1 costs $70 to get 80W ($0.875 per watt). option 2 costs $43.5 to get 50W (or if you double it you get 100W for $87; $0.87 per watt). The cost-performance is pretty similar, but option 2 might be a slightly better deal. Also if you drive Vero 29 at 1.5A, you get a slightly higher efficiency (and lower heat production in theory) than that at 2.1A. But MeanWell driver may or may not last longer. This calculation shows that it probably doesn't matter too much whether you go with the cheap driver or MeanWell driver. It depends on how much total wattage you need. If I need around 150W, I would probably go with 3x 50W instead of 2x 80W to make the coverage more even.

At one point, I went through Mean Well data sheet and looked up the possible drivers for Vero29 at 1.8-2.3A level. Here is the list (I went through quickly, so there may be some mistakes).

CEN-75-42: 31.5-42V, 1.8A, 75.6W, 90%, $40.30
CLG-100-48: 36-48V, 2A, 96W, 88.5%, $48.10
HLG-80H-42: 25.2-42V, 1.95A (1.17-1.95A), 81.9W, 91%, $47.30
HLG-80H-48: 28.8-48V, 1.7A (1.02-1.7A), 81.6W, 91%, $47.30
HLG-100H-42: 21-42V, 2.28A (1.4-2.28A), 95.76W, 93%, $56
LPF-90-42: 25.2-42V, 2.15A, 90.3W, 91%, $42
LPF-90-48: 28.8-48V, 1.88A, 90.24W, 91%, $42
LPF-90D-42: 25.2-42V, 2.14A, 90.3W, 90.5%, $44.60
NPF-90-42: 25.2-42V, 2.15A, 90.3W, 90.5%, $42 (also 90D version)
PLC-100-48: 36-48V, 2A (1.5-2A), 96W, 88.5% $40.6
HLN-80H-42: 25.2-42V, 1.95A (1.17-1.95A), 81.9W, 91%, $41
HLP-80H-42: 25.2-42V, 1.95A (1.56-1.95A), 81.9W, 90% $34.10 (need special terminal connector)

First column: the model number.
2nd column: Voltage range where constant current operation is possible.
3rd column: Rated Current (for some of them I noted the adjustable range)
4th column: Rated Power (not so relevant)
5th column: Typical efficiency
6th column: typical price from Octopart (not the cheapest, though)

LPF-90D-42 (or 90 without D for a bit higher efficiency but no dimming) seems to be a good one from this compilation. HLG seems to have a bit higher efficiency, but it costs a bit more.

When I compared the cost of Vero 29 and 18, Vero 29 was cheaper (per watt after considering the driver and shipping cost), but it may be different if you check the Canadian price. There are also some cheap efficient 1050mA eBay driver, too. But driving multiple Vero 18 in serial with HLG-120H-c1050 (US$64) could simplify the wiring and not a bad price if you are going to drive 5x Vero 18.

 

I'm sorry Naoki but I can't tell you more about it for now until all production details around it is settled. But it is a 24 V-system.

I will tell you more and show the finished result later =)

 

Cool, I'll be looking forward to seeing the results of your product!

 

After reading the reef thread a bit more, I still don't know the advantage of building the driver from scratch in that reef thread (I'd like to know if someone know the reason). But I looked at the data sheet of similar DC/DC constant current step-down driver: http://www.meanwell.com/webapp/product/search.aspx?prod=ldd-h The efficiency could be up to 97% at full load for this DC/DC conversion. But efficiency decreases with the lower load. Also I looked at a few AC/DC constant voltage power supply. The efficiency isn't any higher than AC/DC constant current drivers (and efficiency usually goes down with lower load). So it seems that you get double hit in efficiency loss with that route.

 

Hmm, you are right, some of ebay 30W drivers used to be around 1050mA, but it is not easy to find it now.

How about going with the cheap one you linked in msg. #22? It looks very similar to what I tested; it was rated at 1500mA, but I measured it to be actually 1142mA. The cheap 1.5A ebay driver (the one I linked) is widely recommended because it really gives 1.4A or so. But in your case, inaccurate driver may be good! If the driver actually gives a higher current, it isn't too bad (a bit lower efficiency); Vero 18 (4000K, CRI80) goes from 129lm/W (1.05A) to 121lm/W (1.4A). Or you can go with cheap 900mA driver (and get efficiency higher than 129lm/W).

Alternatively, you could get 3 of the ebay drivers I linked and 3x Vero 18 (($13.5 for V18 + $13.5 for ebay driver) x 3 = $81) to get 120W total. This will give the similar amount of light as 4x Vero 18 at 1050mA. When we drive Vero 29 at 1.4A, we know that we get higher than 129lm/W (maybe mid 130lm/W?), but 121lm/W isn't bad at all.

You can connect all of the AC input wires with wire nuts into a single AC outlet. But I guess that this isn't what you mean by simplifying! The flip side is that if a driver breaks, you only need to replace a single driver. With the big driver, the replacement is expensive. HLG-120H-C1050 seems to have amazingly high efficiency (94%). HLG-120H-C1400 can be used to drive 3x vero 18 at 1.4A, too.

 

Calvin, check this out:
http://www.ebay.com/itm/Inventronic...t=BI_Control_Systems_PLCs&hash=item1e955b7187
I didn't check the spec in details (I have to go home now, my digikey order should be waiting in the mailbox), but 1050mA x 4 channel with 19-38V. $20+9 shipping. Inventronics is a decent maker, I believe. So you need only 1 of this to power 4x Vero 18 (not serial connection).

Data sheet for this driver is here. Up to 90% efficiency. Pretty good.

 

Actually 87% efficiency @120V, which is pretty good. I couldn't edit the above post because of 2 hour edit limit (which is inconvenient).