DIY LEDs - The write-up

[MosMike]

I'm actually replacing a 8X8 heatsink USA heatsink with one that has twice as many fins that are an inch taller (and I'm using a fan).

It should be noted, that this swap is suitable only for active cooling. Heatsinks for natural convection (=passive) must have large gaps between fins. BTW, the best type for natural convection is the pin heatsink, like this one:

http://www.aavidthermalloy.com/technical/papers/pdfs/elliptical.pdf

 

[der_wille_zur_macht]

This means that such losses will be higher in an XM-L @>1000mA vs. XP-G of the same cost.

I follow you until this statement. What data do you have that supports XM-L running hotter/less efficient than XP-G under given conditions?

 

[kcress]

For example, @350mA with a die temp of 35C the loss is 3%, @700mA with a die temp of 75C the loss due to overheating is 12%. This means that such losses will be higher in an XM-L @>1000mA vs. XP-G of the same cost.

How can you state vastly different currents and their efficiencies as 'temperature' efficiencies? Apple and oranges.

 

[TheFishMan65]

MosMike,
There is another possibility open to us. Drill a few holes to allow upward air flow. In most of our cases the fins/pins are pointed upward and the internal pins have trouble getting air flow. Imagine if a hole is drilled in the middle. This lets air flow up through the hottest part.

Go back here and look at the last picture. You can see how the heatsink I found has slot cut in it periodically (just luck that it matches a star size).

Do you have any documentation or opinion on whether this is better or worse?

 

[jpccusa]

The disadvantage of fans is they cost money to run and can be noisy. Long term is it cheaper to get a heatsink or run a fan. I don't know.

I can't get around not using a fan due to the space I have available in the canopy. I found really cheap ($3 ea. shipped) 120mm fans on eBay claimed to make less than 17 db of noise.

IMHO do not use screen. You will not get good thermal transfer from the star. Now if you could only punch holes away from the start that may help. A large sheet will not have much flow in the center. Do you have a break? Maybe you could make your won U channel?

Thanks for the input on the screen. I'll use sheets instead then.

I'm planning on having 2 panels, 12" x 25" each, with 2 fans side by side horizontally in the middle blowing air downward (cold air would enter through far sides of panels).

 

[Cuervo]

Another option would be to add a few heatsinks to the U channel. Either get a few computer CPU heatsinks and attach them to the U Channel with thermal paste or take one of those giant heatsinks and cut into pieces that would fit between the LED's.

 

[MosMike]

I follow you until this statement. What data do you have that supports XM-L running hotter/less efficient than XP-G under given conditions?

Very simple:
XM-L T6 - $7.90 280lm @700mA 35lm/$1
XP-G R5 - $4.90 139lm @350mA 28lm/$1

The 7 lumen per dollar difference can be "destroyed" by the fact that XM-L die is running hotter at a higher current. Actually an active cooling of the LEDs can increase the lumen output by 5-10%.

 

[MosMike]

Do you have any documentation or opinion on whether this is better or worse?

This is better. In reality extruded heatsinks work better when the heatsink is perpendicular to the ground and the fins generate convection currents.

 

[TheFishMan65]

MosMike, who says the XM-L will be running hotter? I admit they will be running at 2 watts (.7ma * 2.9 volts) and the XP-G at one watt (350ms * 3 volts). Are you just assuming they will be running hotter or did I miss something?

 

[MosMike]

MosMike, who says the XM-L will be running hotter? I admit they will be running at 2 watts (.7ma * 2.9 volts) and the XP-G at one watt (350ms * 3 volts). Are you just assuming they will be running hotter or did I miss something?

That's correct. The XM-L will have to dissipate twice the power in the same small footprint. It should be noted, that the lumen output of Cree LEDs is specified with a die temperature of only 20C !

 

[Belgian Anthias]

Xp-g

Xp-g

Very simple:
XM-L T6 - $7.90 280lm @700mA 35lm/$1
XP-G R5 - $4.90 139lm @350mA 28lm/$1

The 7 lumen per dollar difference can be "destroyed" by the fact that XM-L die is running hotter at a higher current. Actually an active cooling of the LEDs can increase the lumen output by 5-10%.

XP-G can run at max 1500mA 490 lumen = 100lm/$1

 

[der_wille_zur_macht]

Very simple:
XM-L T6 - $7.90 280lm @700mA 35lm/$1
XP-G R5 - $4.90 139lm @350mA 28lm/$1

The 7 lumen per dollar difference can be "destroyed" by the fact that XM-L die is running hotter at a higher current.

This is the very definition of apples to oranges! Compare the numbers when the currents are the same. Otherwise we could contrive things to make just about any LED look better than any other LED.

Plus, IMHO, it's fairly useless to use only upfront purchase cost when computing cost efficiency. You only pay upfront cost once. If you keep the unit 2 years vs 10 years, upfront cost is a very different proportion of the total cost of ownership.

 

[MosMike]

This is the very definition of apples to oranges! Compare the numbers when the currents are the same.

Then the XM-L will deliver 160lm @350mA and it will give 20lm/$1
It's like using a Ferrari as golf cart - not very interesting.

Nethertheless, XM-L is the way to go. What can be the RB & XML mix if both are run @700mA ?

I found a nice online supplier of U channels: http://www.onlinemetals.com/merchant.cfm?pid=6992&step=4&showunits=inches&id=297&top_cat=60

 

[chris023]

Okay, I need some help please, sorry if this sounds like a bunch of long winded babble. I think I have read/researched sooo much that I think I go stupider. Just when I think I understand, I get confused again.

I want to LEDify my 60x24x24 SPS reef from T5s. My main issue is the Driver(s) and understanding what is needed to power the LEDs.
I did make a 12 LED dimmable module for my 20 gallon frag tank (picture below), which I used a RapidLED kit that came with the ELN-60-48P and I trusted what I was advised that this driver will handle the 12 LEDs. Thanks again to all you who helped me out with that first rig (you probably know who you are =). I now think I understand how to determine how many LEDs a driver can run based on the formula called out so many times in this and other threads:

driver voltage / forward voltage of LED = number of LEDs you can add to driver.

In the case of my previous mentioned frag module : 48 (volts of driver) / 4 (forward voltage – not sure if this is actually right, heard that ~3.5 is more like the avg) = 12 (number of LEDs). I run these LEDs at 800ma which is well under the max 1.3 amps this driver is capable of.

So I want to run approximately (not 100% determined until I understand more on drivers) 96 LEDs for my tank (60x24x24 sps). Based on another guideline (on these threads) of: Length X Width / Number of LEDs = 15. 15 being the target number for SPS. I think I want to set these up in series based on the complexity/dangers of running in series/parallel may be too risky for my electrical skill set (if that isn’t already obvious). I will probably configure them on U-Channel lengths as heat sinks, with say 18 LEDs 3 inches apart for 54 inches. Will probably need to make 6 of these ‘rails’, actually totaling 108 LEDs. Again the configuration is all just a possible option at this point in the process. I want to run over almost the whole length and width of the tank as I have pretty much got full grown colonies covering most of the tank (see picture below). Please provide any feedback on this aspect of this build.

The big question is what driver(s) should I use to run as many LEDs together as I can??? Definitely don’t want to do more than say 3 or 4 drivers total as I believe (based on reading threads) that will become counter-productive and possibly dangerous? I thought I could just bump up to a driver that could handle more than 12 LEDs (HLG series?), but I am having trouble understanding how? It seems like I need more volts, but the max volts on any of these drivers is 54, correct? So those drivers would only power maybe 16 per string, correct? Seems like the watts on these drivers are the major increasing factor, but I’m not sure that wattage is the concern here (although I could be very wrong).

I probably should just stop and let you all set me straight. Thanks in advance for your help. :beer:

pic of my 12 LED frag module:

pic of my current tank that I would like to LEDify

 

[kcress]

Okay, I need some help please, sorry if this sounds like a bunch of long winded babble. I think I have read/researched sooo much that I think I go stupider. Just when I think I understand, I get confused again.

I want to LEDify my 60x24x24 SPS reef from T5s. My main issue is the Driver(s) and understanding what is needed to power the LEDs.
I did make a 12 LED dimmable module for my 20 gallon frag tank (picture below), which I used a RapidLED kit that came with the ELN-60-48P and I trusted what I was advised that this driver will handle the 12 LEDs. Thanks again to all you who helped me out with that first rig (you probably know who you are =). I now think I understand how to determine how many LEDs a driver can run based on the formula called out so many times in this and other threads:

driver voltage / forward voltage of LED = number of LEDs you can add to driver.

In the case of my previous mentioned frag module : 48 (volts of driver) / 4 (forward voltage "“ not sure if this is actually right, heard that ~3.5 is more like the avg) = 12 (number of LEDs). I run these LEDs at 800ma which is well under the max 1.3 amps this driver is capable of.

So I want to run approximately (not 100% determined until I understand more on drivers) 96 LEDs for my tank (60x24x24 sps). Based on another guideline (on these threads) of: Length X Width / Number of LEDs = 15. 15 being the target number for SPS. I think I want to set these up in series based on the complexity/dangers of running in series/parallel may be too risky for my electrical skill set (if that isn't already obvious). I will probably configure them on U-Channel lengths as heat sinks, with say 18 LEDs 3 inches apart for 54 inches. Will probably need to make 6 of these "˜rails', actually totaling 108 LEDs. Again the configuration is all just a possible option at this point in the process. I want to run over almost the whole length and width of the tank as I have pretty much got full grown colonies covering most of the tank (see picture below). Please provide any feedback on this aspect of this build.

The big question is what driver(s) should I use to run as many LEDs together as I can??? Definitely don't want to do more than say 3 or 4 drivers total as I believe (based on reading threads) that will become counter-productive and possibly dangerous? I thought I could just bump up to a driver that could handle more than 12 LEDs (HLG series?), but I am having trouble understanding how? It seems like I need more volts, but the max volts on any of these drivers is 54, correct? So those drivers would only power maybe 16 per string, correct? Seems like the watts on these drivers are the major increasing factor, but I'm not sure that wattage is the concern here (although I could be very wrong).

I probably should just stop and let you all set me straight. Thanks in advance for your help. :beer:

First off, for being "stoopid" you seem to have a pretty good bead on things..

You only go a little off the rails towards the end.

Once you have an estimate of how many LEDs you want to use you have to figure out what driver(s) to use and then work backwards to find exactly the number of LEDs you can have. From there you return to how to lay out reasonably that number of LEDs.

IMO don't bother with 54V drivers unless the numbers shake out really specially that way.

You should go with parallel strings....

Looking for about 96 LEDs.

3.3 x 0.8 x 96 = 250W

If you are up for mixing the colors then this can easily be handled with a single driver. If you are going to screw with the color balance then you will need two drivers. Which drivers can't be determined until you answer that question.

 

[mseepman]

I have to say that I understand where the fear of parallel builds comes in. Today I received 180 Cree XP-G and XP-E LED's and figuring everything out has my head spinning.

 

[kcress]

Take a breath..

Sit down and do it like you'd eat an elephant.

Do it one bite at a time...

 

[chris023]

First off, for being "stoopid" you seem to have a pretty good bead on things..

You only go a little off the rails towards the end.

Once you have an estimate of how many LEDs you want to use you have to figure out what driver(s) to use and then work backwards to find exactly the number of LEDs you can have. From there you return to how to lay out reasonably that number of LEDs.

IMO don't bother with 54V drivers unless the numbers shake out really specially that way.

You should go with parallel strings....

Looking for about 96 LEDs.

3.3 x 0.8 x 96 = 250W

If you are up for mixing the colors then this can easily be handled with a single driver. If you are going to screw with the color balance then you will need two drivers. Which drivers can't be determined until you answer that question.

Just to ensure I am on the same page...what is the 0.8 in your above numbers...current? (3.3 = forward voltage, 96 = total LEDs, 0.8 =?)

I think I have a good idea of the number of LEDs (give or take a few). Lets say I want to run 6 u-channels with 18 LEDs per u-channel = 108 LEDs. Colors will be mixed per u-channel.
I will be running these at .7 amps...so 3.3 x .7 x 108 = 249.48
so the HLG-320-?
wouldn't each string need ~60 volts in this case (18 LEDs per string x 3.3 = 59.4? Is that correct?)

I think I get the concept of parallel strings based on page 200 (and other threads), although the consensus seemed to be not to go with parallel. Maybe that is just for smaller setups. I just don't fully understand the wiring.
Other concerns with parallel I have as a electrical newbie: each strings total forward voltage needs to be basically the same; you need the correct resistors and fuses on each string, etc.
I'm pretty sure I can overcome those concerns if parallel is the way to go...

 

[chris023]

I have to say that I understand where the fear of parallel builds comes in. Today I received 180 Cree XP-G and XP-E LED's and figuring everything out has my head spinning.

wow! spinning is turning into an understatement :spin1:
what size tank are you putting them on and how are you wiring/driving them?

 

[kcress]

:furious::furious::furious::furious:

I spent 20 minutes on this answer and it vaporized.

So will give the budget repeat.

LED drivers are categorized by wattage. You need to come up with a ball park wattage to know where in the families to look.

Watts = 3.3V x 0.8A x the total number of LEDs.

Once this is known you look at the 60, 100, 120, 150, 200, 250, 320 watt models.

The parallel isn't real hard.

Here's the basic setup:

The resistors are all the same value so you don't have to think about 'different values'.

You'll use them to shift around a few LEDs to get the strings matched current wise.