Minimalistic multichip DIY LED build

[Floyd R Turbo]

Using two 180w multichips is even less work and looks more realistic.

One of the benefits of multichips over 2 or 3w chips is the lens quality. Large glass dome lenses provide even coverage, while cheap acrylic lenses used for "pegboard" arrays have hot spots in the centre with a fast drop off a few inches off centre. large dome lenses provide natural shimmer without disco effect and weird purple and yellow shadows. You can mount multichips higher up from the water surface to keep them free of salt creep and splashed.

LED arrays create a series of light beams that don't properly mix colours. You end up with a surplus of one spectrum in one area, and a deficit in another. You don't get the "beam me up Scotty" effect with multichips. Combining multiple colours in one chip is the best option. Using one blue, then one white fixture etc. always looks spotty and doesn't provide a homogenous mix of spectra even at high suspension levels.

Part of the aesthetic is personal preference. I've never liked the look of T5 lighting and find it to look clinical and artificial. Multichips replicate the look of metal halide and if done right, the sun. In addition to the quality light, it looks cleaner with one or two small pendents rather than a massive array looming over the tank.

The lower the wattage of each chip, the more efficient. A series of 1w lights is better than 3w so multichips are more efficient than arrays. Changing the multichip or lens is a quick and painless process, and the cost isn't even as much as a metal halide bulb.

I agree with Lassef, higher colour temperature is the way to go. The 7500 Cree chips cause cyanobacteria and slime algae and look too yellow. Corals tend to bleach or brown out due to excessive growth of zooxanthellae masking brighter pigments. You end up using a lot more blue chips to drown out the green light so efficiency is diminished. With high kelvin white chips, you can run a 50:50 white:blue ratio, while 7500K whites require 25:75, thus lowering lumen/watt and PAR/watt.

Thanks Mr. Wilson, your description of the "beam me up scotty" effect and color separation is exactly what I wish to avoid.

I get the advantage of the larger multichips for simplicity, etc, but I have a 66 x 21 tank so it's long and narrow, and I feel that even with 30" of headroom I don't think I would get a decent spread, plus I really want to PWM dim then and I can't do that with a DIY driver on the 100/252W chips.

Here is my original post of the tank I wish to put these on for reference

http://www.reefcentral.com/forums/showpost.php?p=20502130&postcount=1332

also I'm not sure if I have to worry about getting a hotspot or what not.

With a mixture of 20W chips on individual heat sinks, I can replace or rearrange rather easily, and I can control them all to my heart's content. So that is the thing for me.

Next, I will have to figure out if I want to do lenses or not, and if I will be clustering them. I see the "scotty" effect coming into play maybe a little if I don't cluster them (versus an even on-center spacing)

 

[tomservo]

See I'm not trying to compare apples to oranges. I'm saying you have a 100 watt 20kK chip that's fine, now slap together 100 watts worth of crees of various chips (RB, B, CW, WW, etc) whatever the equivalent of all those little LEDs thrown together on one chip is, and compare apples to apples.

I'm not saying one is vastly better than the other by any means, however not knowing if that's the case does bug me.

I get what you're saying, and the crees are more efficient than the epileds/epistar chips. All you need to do is compare the efficiency of the blues (since whites are blues with phosphors). Crees are 450-550mW/W and the chips ac-rc uses the datasheet tops out at 340mW/W, which is actually a huge difference. I put in an inquiry with him to see what power bin his chips are, and if he can get better bins.

Given the size of Cree emitters, if a person could get a custom MCPCB made, it's possible to make a cree multichip of similar size to the chinese ones. I did look around to do this but it seems custom MCPCBs are very expensive, the one place I found started at like $110. Given the density you'd also need these MCPCBs made out of copper, which I think was an option.

It turns out that bare Cree emitters are actually quite cheap.. like $1.98 each for blues, or $1.10 in qty 1k.

There's even a 410-420 nm LED that is in the same form factor package as the Cree stuff.

 

[sfsuphysics]

I get the advantage of the larger multichips for simplicity, etc, but I have a 66 x 21 tank so it's long and narrow, and I feel that even with 30" of headroom I don't think I would get a decent spread, plus I really want to PWM dim then and I can't do that with a DIY driver on the 100/252W chips.

If you had the space (and technical know how) you might be able to get away with shining in at a shallow angle, and then rotate overhead to simulate peak sun, then end back at a shallow angle... However the engineering involved in making that happen might be a bit cost restrictive

I have an 8' x 3' tank that has been collecting dust waiting to be cleaned up in my garage for way too long, that has me contemplating lighting solutions, and this is one thing that did enter my mind... again though making it work is another matter

I get what you're saying, and the crees are more efficient than the epileds/epistar chips. All you need to do is compare the efficiency of the blues (since whites are blues with phosphors). Crees are 450-550mW/W and the chips ac-rc uses the datasheet tops out at 340mW/W, which is actually a huge difference. I put in an inquiry with him to see what power bin his chips are, and if he can get better bins.

Given the size of Cree emitters, if a person could get a custom MCPCB made, it's possible to make a cree multichip of similar size to the chinese ones. I did look around to do this but it seems custom MCPCBs are very expensive, the one place I found started at like $110. Given the density you'd also need these MCPCBs made out of copper, which I think was an option.

I'm sure if you cram 30 or so emitters within a tiny area like the other multichip you'll also tank in efficiency simply because of thermal issues of that many emitters being so close to each other. While yeah you can handle the heat issues of the macroscopic view, the spacing between each might contribute to a lot of heat issues.

 

[tomservo]

sfsuphysics: I don't see why there would be thermal issues with a good cooling solution if the MCPCB is copper. Modern processors have at the very least similar heat density, 150w on a 38x38 heat spreader? And that's coming from a die underneath that is much smaller, for example the bloomfield quad cores are 130w thermal power from a 263mm^2 die. That's about 16.5x16mm.

Floyd: 30 inches of headroom is tons. You could use 60 degree optics with that much space and get good coverage from 3 emitters. I say 3 b/c that results in almost square areas, 22x21. You could always pick up a 20w and play with it, see what you think.

 

[ronreef]

Here are some close-ups of mounting the 50W 16000K multichip using Arctic Silver thermal adhesive. The thermal adhesive requires about 8 hours cure time. Note the standoffs between the aluminum angle and the gooseneck bracket. This prevents the bracket from interfering with installation of the fan.

https://picasaweb.google.com/103048...&authkey=Gv1sRgCPihk6v6o6PkGQ&feat=directlink

Ron

 

[BeanAnimal]

you need to do is compare the efficiency of the blues (since whites are blues with phosphors). Crees are 450-550mW/W and the chips ac-rc uses the datasheet tops out at 340mW/W, which is actually a huge difference.

Be aware that the you can't just translate "blue" efficiency to "white" efficiency, as the phosphors themselves differ from company to company and greatly in efficiency!

That is, the cheaper white LEDs likely have lower efficiency phosphors, meaning that they suffer even more of a loss compared to a good (Cree) blue and good (Cree) phosphor.

In other words, cheap phosphors compound the inefficiency of an already cheap inefficient led.

 

[tomservo]

BeanAnimal: That is certainly true that phosphors differ! However, highly efficient phosphors can also mean excess green/yellow spectra, and you can definitely see this in Cree spectrum graphs. As a baseline, emitter efficiency is the best place to start when considering this factor, as a low efficiency emitter will always have poor output! When considering higher kelvin emitters I think phosphor efficiency is less of a factor, as there is simply a lot less phosphor present.

High CRI emitters always have lower efficiency as well, with top ends at 75-90 lumens/watt. Likely because lumens are weighted so heavily towards green/yellow, and high CRI means less green and more red. Doesn't mean that there's less actual light output, although it's impossible to know this without some numbers in a sensible format (like mW). So low lumen/watt efficiency doesn't necessarily mean it's a lousy emitter, could simply mean it's lousy at making green light.

 

[BeanAnimal]

Certainly.... The point I was trying to make is that (likely) when all things are considered (for our purposes) the cheap China leds are really not in the same ballpark as the currently (no pun) shipping emitters from the big 3. Add to that (not a point I made) the exactness of binning (or utter lack of) and the China leds are a huge gamble (if not clearly inferior). That said, if the cost is right and the output meets the needs of the end user, who am I to argue

 

[bhazard451]

As someone who has seen multiple Chinese leds from multiple bins, I can say firsthand that they vary greatly. One of the worst in output I've seen to date were epistar 14k leds. One epistar 3w "14k" compared to my Cree XT-E CW at 1.3mah is about 1/6th of the light output. I'm not kidding, it's that bad.

The cpu heatsink requirement for a multichip is also a co ncern. That is some serious heat if the fan were to fail, and you would never know in an enclosed fixture.

The best mid point for a build between multichip or 3w to me, is closely placed, 3w led clusters running at close to max current. I have 48leds of 180 watts of Crees and Luxeons running in a 12"x12" case with heatsink. The output is truly impressive over my 34 gallon.

3 multichips within the same enclosure, would be a major heat concern, and I don't think it would even be possible safely. The output may not even be as strong, and would lack the 420nm actinics and other colors I used.

 

[BeanAnimal]

To that end, if you look at the Orphec pendant, it is is HUGE... I mean HUGE as in larger than a full size coffee can, closer to a gallon of milk.

Why? I suspect a massive passive heatsink.

 

[bhazard451]

Using two 180w multichips is even less work and looks more realistic.

One of the benefits of multichips over 2 or 3w chips is the lens quality. Large glass dome lenses provide even coverage, while cheap acrylic lenses used for "pegboard" arrays have hot spots in the centre with a fast drop off a few inches off centre. large dome lenses provide natural shimmer without disco effect and weird purple and yellow shadows. You can mount multichips higher up from the water surface to keep them free of salt creep and splashed.

LED arrays create a series of light beams that don't properly mix colours. You end up with a surplus of one spectrum in one area, and a deficit in another. You don't get the "beam me up Scotty" effect with multichips. Combining multiple colours in one chip is the best option. Using one blue, then one white fixture etc. always looks spotty and doesn't provide a homogenous mix of spectra even at high suspension levels.

Part of the aesthetic is personal preference. I've never liked the look of T5 lighting and find it to look clinical and artificial. Multichips replicate the look of metal halide and if done right, the sun. In addition to the quality light, it looks cleaner with one or two small pendents rather than a massive array looming over the tank.

The lower the wattage of each chip, the more efficient. A series of 1w lights is better than 3w so multichips are more efficient than arrays. Changing the multichip or lens is a quick and painless process, and the cost isn't even as much as a metal halide bulb.

I agree with Lassef, higher colour temperature is the way to go. The 7500 Cree chips cause cyanobacteria and slime algae and look too yellow. Corals tend to bleach or brown out due to excessive growth of zooxanthellae masking brighter pigments. You end up using a lot more blue chips to drown out the green light so efficiency is diminished. With high kelvin white chips, you can run a 50:50 white:blue ratio, while 7500K whites require 25:75, thus lowering lumen/watt and PAR/watt.

I disagree with the "large glass dome" comment. Since it isn't focusing the light specifically like a 90 or 60 degree optic would, it is essentially doing the same thing a led with a "plastic" dome would do with the led angle, nothing. I've never had a "purple or yellow shadow" with my setup either.

It is tough to avoid a blue/white shimmer using optics however. With that said, hanging a multichip higher over an open tank without focusing the light will end up shooting a lot of that light all over the room instead of the tank itself. The led angle is still 110-120 degrees, and hanging that 10" or more over the water will light everything in the room.

I agree with using a multicolor multichip, but unless that multichip can also include warmer whites or actinic violet in it, it is much less flexible and less visually appealing. I've run a Cool White/Royal Blue 1:1 only fixture for a few months, and several colors end up lacking in fluorescence. Many coral pieces become dull and lifeless looking. After introducing warmer whites along with CWs and 2:1 royal blue, violets, reds, and some green, my coral have never looked better.

Nutrients grow cyano. My lights do not and have not. When said nutrients appear in the tank, the lights will probably grow the cyano too.

Chinese "High Kelvin" leds also differ vastly from bins they have. One 3w "14k" led that was supposedly the same as another company's 3w "14k" led was half as bright and completely different in color. Using Crees from WW 3.5k to CW does look yellow on their own, which is why the blues are needed.

According to you, my tank should look like urine. Does this look yellow to you? I would challenge any multichip against this, and I have full Apex control of the color for each separate channel.

 

[ronreef]

According to you, my tank should look like urine. Does this look yellow to you?

I can't tell. Too many variables... the color accuracy of your camera... exposure setting... aperature setting... ISO setting.... monitor color accuracy... cone sensitivity of the eyes.... Unless someone were to see your tank up close and personal, gauging color accuracy from an internet photo is a mute point.

 

[bhazard451]

I can't tell. Too many variables... the color accuracy of your camera... exposure setting... aperature setting... ISO setting.... monitor color accuracy... cone sensitivity of the eyes.... Unless someone were to see your tank up close and personal, gauging color accuracy from an internet photo is a mute point.

Now that's just nitpicking. I could just turn the white channel on to give more perspective, but why bother.. since I won't be able to bring out the DSLR and give every single camera setting for you.

The point is, until there are more and better multichip options available, the same or better output can be created with 3w leds in tight clusters, without needing exotic cooling options covering each chip. Yes, I consider a Hyper 212 exotic considering the size of the cooler needed over the size of the chip.

I haven't seen too many other DIY builds use this type of setup, but if you break it down, it would be essentially the same as 4 50w multichips in a foot long fixture. There would be no way to install the drivers into this fixture, and also have proper cooling for the leds. It's a very clean build which doesn't get hot, that could not be replicated with multichips.

To each his own though. I could see the benefits of multichips in a large build with some form of supplementation like T5s. Using T5's negates one of the reasons I switched to leds in the first place though. Bulb changes.

 

[mr.wilson]

See I'm not trying to compare apples to oranges. I'm saying you have a 100 watt 20kK chip that's fine, now slap together 100 watts worth of crees of various chips (RB, B, CW, WW, etc) whatever the equivalent of all those little LEDs thrown together on one chip is, and compare apples to apples.

I'm not saying one is vastly better than the other by any means, however not knowing if that's the case does bug me.

Cree spent a fortune developing a green chip for task lighting. The spectrum of their white chip is nothing like what one would use with T5 or metal halide, so why use a desk lamp on your reef?

Cree white chips have an impressive lumen/watt rating, but it's drastically lower with the blue and royal blue chips that you need to compensate for the green light. We aren't concerned with lumens, we need adequate PAR and the right spectrum for coral pigmentation.

It is hard to achieve reds, blues and purples using Cree white chips and you grow more nuisance algae. The energy used to generate green light is a waste of resources, therefore Cree whites are inefficient for reef lighting, but great for flashlights and desk lamps.

 

[mr.wilson]

Thanks Mr. Wilson, your description of the "beam me up scotty" effect and color separation is exactly what I wish to avoid.

I get the advantage of the larger multichips for simplicity, etc, but I have a 66 x 21 tank so it's long and narrow, and I feel that even with 30" of headroom I don't think I would get a decent spread, plus I really want to PWM dim then and I can't do that with a DIY driver on the 100/252W chips.

Here is my original post of the tank I wish to put these on for reference

http://www.reefcentral.com/forums/showpost.php?p=20502130&postcount=1332

also I'm not sure if I have to worry about getting a hotspot or what not.

With a mixture of 20W chips on individual heat sinks, I can replace or rearrange rather easily, and I can control them all to my heart's content. So that is the thing for me.

Next, I will have to figure out if I want to do lenses or not, and if I will be clustering them. I see the "scotty" effect coming into play maybe a little if I don't cluster them (versus an even on-center spacing)

PWM dimming may interfere with the photon acceptors that "flash on and off". Unfortunately there is little in the way of research on the subject. Some believe that you can time the light flashing on and off in synch with the photon receptors in order to save energy, but it's questionable how to achieve this.

0-10v analog dimming makes the native blue chip react differently with the phosphor coating, thus rendering a bluer light. This will compensate somewhat for low kelvin chips but dimming needs to be significant.

A 90˚ dome lens will cover 21" x 21" if you lower it enough (12" from surface?). That means three multichips to cover your tank. The wattage of the chips hinges mostly on the water depth. If the tank is <24" deep, then 50w chips may be sufficient, but 100w leaves you room to dim them down.

As far as generic (budget) drivers go, it's a false economy. I have replaced quite a few generic drivers in the last few years, but I have never had problems with Mean Well or Inventronics drivers. If you buy discount drivers, buy spares to have on hand so there is no down time. There are some reliable generic drivers, its just takes two years of use to learn which ones

A series of 20w multichips will work, it's just not how I would do it FWIW.

 

[mr.wilson]

I get what you're saying, and the crees are more efficient than the epileds/epistar chips. All you need to do is compare the efficiency of the blues (since whites are blues with phosphors). Crees are 450-550mW/W and the chips ac-rc uses the datasheet tops out at 340mW/W, which is actually a huge difference. I put in an inquiry with him to see what power bin his chips are, and if he can get better bins.

Given the size of Cree emitters, if a person could get a custom MCPCB made, it's possible to make a cree multichip of similar size to the chinese ones. I did look around to do this but it seems custom MCPCBs are very expensive, the one place I found started at like $110. Given the density you'd also need these MCPCBs made out of copper, which I think was an option.

It turns out that bare Cree emitters are actually quite cheap.. like $1.98 each for blues, or $1.10 in qty 1k.

There's even a 410-420 nm LED that is in the same form factor package as the Cree stuff.

When you are looking at technology that can replace a 250w metal halide with 100w of LED, the lumen/watt difference between premium brands is moot.

Flexibility in phosphor coating, and subsequent colour temperature is more of an advantage.

Again, lumens are a poor measurement for reef lighting. Adding a ton of 555nm green light that humans see as "bright" will raise the lumen value, but it's counterproductive to our goal for illuminating a reef tank.

 

[bhazard451]

Cree spent a fortune developing a green chip for task lighting. The spectrum of their white chip is nothing like what one would use with T5 or metal halide, so why use a desk lamp on your reef?

Cree white chips have an impressive lumen/watt rating, but it's drastically lower with the blue and royal blue chips that you need to compensate for the green light. We aren't concerned with lumens, we need adequate PAR and the right spectrum for coral pigmentation.

It is hard to achieve reds, blues and purples using Cree white chips and you grow more nuisance algae. The energy used to generate green light is a waste of resources, therefore Cree whites are inefficient for reef lighting, but great for flashlights and desk lamps.

The same could be said for the Epistar multichips, although the efficiency is reduced even more so.

This can be corrected by introducing leds like the colored Luxeons, and with violet leds peaking near 420nm. Cree not having a very good Green/Cyan/Red/Violet option led me to use Luxeons for my color channel. I'm starting to doubt having the green in for aesthetic reasons myself, and will probably replace them with more violets soon. There is no multichip equivalent that I know of that would allow any of these options.

 

[ronreef]

Now that's just nitpicking. I could just turn the white channel on to give more perspective, but why bother.. since I won't be able to bring out the DSLR and give every single camera setting for you.

The point is, until there are more and better multichip options available, the same or better output can be created with 3w leds in tight clusters, without needing exotic cooling options covering each chip. Yes, I consider a Hyper 212 exotic considering the size of the cooler needed over the size of the chip.

I haven't seen too many other DIY builds use this type of setup, but if you break it down, it would be essentially the same as 4 50w multichips in a foot long fixture. There would be no way to install the drivers into this fixture, and also have proper cooling for the leds. It's a very clean build which doesn't get hot, that could not be replicated with multichips.

To each his own though. I could see the benefits of multichips in a large build with some form of supplementation like T5s. Using T5's negates one of the reasons I switched to leds in the first place though. Bulb changes.

Theres nothing nitpicky about accuracy.

 

[mr.wilson]

Certainly.... The point I was trying to make is that (likely) when all things are considered (for our purposes) the cheap China leds are really not in the same ballpark as the currently (no pun) shipping emitters from the big 3. Add to that (not a point I made) the exactness of binning (or utter lack of) and the China leds are a huge gamble (if not clearly inferior). That said, if the cost is right and the output meets the needs of the end user, who am I to argue

The most popular LED aquarium fixtures from the USA are made in China and they have proven to be less efficient than the so called "generic Chinese lights". Sanjay Joshi's testing proved that you get more PAR per watt, PAR per 24" x 24" area, and more peak PAR with Chinese offerings. As a matter of fact, the big US manufactures were pretty far down the list in efficiency despite the "award winning" Cree chips.

 

[tomservo]

The cpu heatsink requirement for a multichip is also a co ncern. That is some serious heat if the fan were to fail, and you would never know in an enclosed fixture.

Well if you're willing to spend the kind of money it takes to buy big plate heatsinks, you could get a NoFan CR95. I'm kind of amused since your fixture uses fans, though. If the failure of a single fan is an overriding concern, and you don't just use a temp probe + your apex to monitor the fixture, most heatsinks these days allow for dual fans.

This can be corrected by introducing leds like the colored Luxeons, and with violet leds peaking near 420nm. Cree not having a very good Green/Cyan/Red/Violet option led me to use Luxeons for my color channel. I'm starting to doubt having the green in for aesthetic reasons myself, and will probably replace them with more violets soon. There is no multichip equivalent that I know of that would allow any of these options.
Today 11:16 PM

Or you could use a phosphor combination that was suitable in the first place? It seems to me that adding a bunch of color channels is pointless unless you really need to fiddle with the color. MH/T5 guys usually find a bulb they like and stick with it, though some do seem to tinker.

I did a sort-of multichip build using a 4000k (85CRI) citizen emitter, these have a really good spectrum, stronger cyan than any other brand I've seen, and a strong red and the deep red output is respectable. I tried cyan/turquoise with it and it just wasn't needed, same with 660 red. It just didn't add anything except disco light shadows. I'm not saying that's the case with all setups. It seemed to me at the time that adding extra cost and complexity to fix the spectrum was not the ideal path.

mrwilson: That is very interesting news, I'll have to go check that out.