Hi. I'm new here, though I've been using HO-fluorescent and MH-powered Algae Scrubbers for a long time (ever since I read the first edition of Dr. Adey's book a long, long time ago). I'm interested in making the switch to a compact LED design. I'm a bit stuck on some of the engineering aspects of designing an LED system, so I thought I'd ask for your advice on this topic. Specifically, I'm getting stuck on the thermal power dissipation calculations. Unfortunately, the suppliers of pre-mounted LED on the metal-core PCBs don't publish heat transfer ratings for their products, which means that there's no way to approach the heatsink problem mathematically. I'd like to avoid the approach of just taking a stab in the dark at mounting LED on a randomly chosen heatsink and hoping for a good outcome. Optimally, I'd like to know the specs for heat transfer of the MCPCB-mounted Luxeon Rebel ES, like the ones that you can buy from Steve's. My goal is to crunch some numbers to determine how many LEDs operating at full power can safely be mounted on one of several possible heatsink choices, while ensuring that the finished product will maintain an operating temperature low enough to ensure that Tj on the LED is still in the high efficiency portion of the Thermal Pad Temperature vs. Luminous Output characteristic. In the absence of clear-cut engineering data, it might be helpful to know anecdotal experiences, like: how many LED you were able to mount onto a heatsink; how much power they were dissipating; what the C/W rating was for the heatsink; and what was the observed operating temperature on the heatsink, using something like an IR thermometer. Hopefully, obtaining some solid data like this would help me to approach the problem from an engineering perspective rather than just taking a shot in the dark. TIA.
Hey Bob, welcome! I'm always glad to see reefkeepers that have used scrubbers for a lot longer than I have. I met a few of those at MACNA in Miami last year and we had great conversations. Hope I/we can answer your questions! The guys who make the MakersLED heat sinks live 30 minutes from me and they would probably have some of the better technical insight about heat transmissivity and such. But I can give it my best shot from the anecdotal perspective. On my Rev 1 unit, I used an oversized heat sink because it was cost effective to also use it as a light blocker, but what I have found is that it does not even get that warm to the touch when running 7 LEDs at 700mA. That was a 7.5" x 6" one (roughly) from HeatSinkUSA, with the fins oriented vertically. I've shot that with a IR therm and rarely do I see over 100F on the back of the sink behind the LED position. I switched to the MakersLED heat sink for the next 3 versions and that one had the heat sink oriented horizontally, so obviously a bit less dissipation efficiency, but even without a fan on them, they never got much warmer than the Rev 1. I have had a few customers say that they did feel like they were a bit too warm so they put fans on them. This was mainly the case then they had to run them for long hours in a closed cabinet. Any airflow around the cabinet significantly reduces the temp on the heat sink, even just a clip-on circulating fan. For the Rev 3, I had the MCPCB made which spreads the heat out very well before it gets to the heat sink, and this allowed me to remove the need for the fan pretty much for every instance. On top of that, the board has jumpers on it that allow you to run the entire array at 350mA and in many cases I found that this is actually a pretty good level of light for many users, perfect for the maturing stage, and then you can just lengthen the photoperiod to extend the daily cycle until that isn't enough, then you can start flipping jumpers and reducing the photoperiod to get a nice balance. If I had to take a rough guess at a guideline for 700mA 3W 660nm Deep Red LEDs, I would say that (in conjunction with the LED spacing guideline) that an array of these at 2" on center, which is the "high intensity" configuration, would need your standard heat sink with 1" or so fins at least 1" around all sides. So if you have a 3x3 array, which would be 4" x 4", you would want minimum of a 6" x 6" heat sink with the fins oriented vertically. Keep in mind that most people who run that large of a heat sink mounted horizontally over a tank would probably do so with a tighter array of LEDs running a higher current, and still wouldn't necessarily need fans, especially if the fins are a bit longer. The 660nm Deep Reds don't run as hot as your DT LEDs simply because they have a lower max drive current of 700mA. Running them at 500-600mA further reduces their heat output while reducing the intensity proportionately less, so keep that in mind too. I hope that at least puts you on the right path. If you want I can get in touch with the Makers guys on their specs. This is actually the one I used for the Rev 1, cut to 6.5" long http://www.heatsinkusa.com/8-460/
I would like to chime in and note that the base of the profile Turbo uses is a profile with a nice thick base which helps the heat flow from the LEDs into the fins for good heat dissipation. When the profile has a thin base you run the risk of restricting the flow of heat from the base to the fins....basically the fins will feel cold to the touch and they aren't very efficient anymore.
Also FYI, I used heatsinkusa's 6" low profile and it works great as well. I have 5 Phillips 660's running at 700mA and the profile is 6" long. Heatsinkusa does provide an estimate of the heat dissipation of their profiles. It in C/W/3"
Thanks for the replies. I have to admit, I'm a little more focused on the math aspects of this stuff than the average guy, because I'm thinking about scaling-up my build to a larger size filter to replace an ATF that's using HID lighting. So please forgive me if it seems like I'm overly concerned about numbers. I'm thinking about building something that has a large number of LEDs, and I'm trying to avoid the problem of overspending or underspending on heatsink material. I'm trying to collect as accurate a set of numbers as I can so that I don't end up building the same unit several times through successive approximations. It's really hard to get a complete set of numbers that I can hang my hat on, because the guys who sell star-mounted LEDs aren't publishing their product's heat transfer characteristics, and it's hard to generate accurate data without it. Also, whenever I read forum discussions on the subject, it seems that there's always at least one or two important pieces of data that get left out of the conversation. For example, everyone usually mentions how many LEDs are on a chain, and how much current the chain is being driven with. But it's also important to know the composition of the LEDs on the chain, and whether they are all the same or if any of them are different (like having a blue LED or two on the chain). This is important because diodes that have different values of Vf will dissipate different amounts of energy when they're placed in series on the same string. You know what I mean ... each Deep Red LXM3-PD01 that has a Vf of 2.4V @ 700mA will dissipate ~1.68W, where any other LED in the chain would probably have a different Vf, and a different power dissipation. A Royal Blue 3W LED that's running on the same string at 700mA might have a Vf around 3.0, so it might dissipate 2.1W into the heatsink while the Deep Red would only dissipate about 1.68W. Of course, these minor differences don't amount to much when you're only putting 12W or so into a heatsink, as almost any heatsink would have the ability to accomodate that much power. But the differences can become very significant as you scale-up the number of heat sources. What I'm really trying to determine is which heatsinks you've tried using, what their C/W ratings are, how many watts your LEDs are putting into the heatsinks, and what the temperature of the heatsink is relative to ambient air when it is being driven. Essentially, I'm trying to work backwards to determine junction temperature, to assure that the LEDs are never running hot, which would decrease their output and shorten their life. Turbo, I'd like to ask some questions about your Rev 1 unit. You mentioned that it had 7LEDs running at 700mA. Was that string comprised solely of the 3W Deep Red Luxeon LXM3-PD01-0300, or did it have some other LEDs on it? Knowing that would help to work out the answer of how many watts were going into the heatsink. Knowing the C/W rating for that heatsink would help, as would knowing the number of degrees above ambient temperature that you recorded with your IR thermometer. by any chance, is this the heatsink that you used? http://www.heatsinkusa.com/7-280-wide-extruded-aluminum-heatsink/
Headed to the sack here shortly, but I can answer one of the questions easily. The Rev 1 used 6x 660nm in a 2x4 array, 2" o.c. with a blue dead center, so in and "I" shape I later change this to splitting the blues into 2 and running them in parallel within the series like this I do the same now, except on the LED board (old pic warning) The Vf across the terminals of the parallel pair of blues is still 3.0V compared to 2.2-2.4V across each of the reds, but they essentially acts a current divider, although probably not a perfect one as while each L blue does come from the same batch and are likely to be very closely electronically matched, they are probably not perfectly matched, but IMO that is not a major concern. If they were cheap chinese LEDs I would be more worried. If you have some highly technical questions, you might try e-mailing Steve with Steve's LEDs. It might take him a while to get back to you, but from my exchanges with him, he really knows his stuff.
Making sense out of this "alphabet soup" of numeric specifications is difficult. I've been looking over the specs that are mentioned on the Steves LED website, and comparing them to Phillips datasheet DS68 and I can't reconcile the numbers. I'm wondering if some of the specs that are listed on Steve's site may reflect a misunderstanding of the data sheet. For example, this page: http://shop.stevesleds.com/Philips-Luxeon-ES-ROYAL-BLUE-3-Watt-LEDs-Luxeon-ES-Royal-Blue.htm lists the claimed specs for the LXML-PR02-1100 LED that Steve is selling. The DS68 data sheet states in Table 1, Page 3 that the Minimum Radiometric Power for the LXML PR02-1100 @ 700mA is 1100mW, and that the Typical Radiometric Power is 1120mW: In Fig. 12 on Page 13, Philips shows a plot of Relative Flux vs. Forward Current for the PR02 series of LED, showing that the relative radiant flux is standardized to a value of 1.0 at 700mW (red lines). The table shows how the relative flux will change as a function of forward current. At 1000mA the relative flux function yields a Y-value of 1.35 (green lines). Multiplying 1.35 by the reference value of 1100mW @ 700mA gives us a minimum radiometric power value of 1.35 * 1100 = 1485mW @ 1000mA. (It's important to recognize that the Y-axis in Fig. 12 lists the relative flux compared to the standardized value at 700mA. Notice that the axis isn't labelled in watts. It's labelled as a dimensionless number, because the values are ratios and not actual milliwatt values.) If you read Steve's page for this LED he mentions: Now I have to admit, my day job is not in LED engineering, so I'm confused by this. It would seem that either my interpretation of the data is just off, or that Steve's presentation of the data is off. I'm not sure which, but the discrepancies that I've noticed like these make the job of designing a system rather difficult. When the merchant's numbers don't line up with the factory's numbers, it gets me to start scratching my head in frustration. Unfortunately, these kinds of discrepancies keep popping up as I'm doing my research, which makes the whole process of designing a system to replace my current lighting system a real challenge! Thanks for your help.
Steve got back to me pretty quick: Comment from Steve: That math is not accurate. The Luxeon ES LEDs, such as the Royal Blue, is around 53% efficient (datasheet DS68, Table 1, far right column), so although it draws 3W of power, the total dissipated heat is less than 1.41W each, and even less than that for the red. The total wattage draw does not equate to dissipated heat because more than half of the energy is being turned into light with the Royal Blue LEDs. Comment from Steve: He is looking at the charts correctly, but he made in error because he is did not reference the BIN code, which for the LEDs we sell, have a higher than standard output of the Royal Blue LED. The current flux (radiometric power) BIN "N" (Datasheet DS68 Table 9) Steve's LEDs currently has in stock has a radiometric power output maximum of 1,300mw max. We have found that Luxeon typically label BINs their LEDs on the lower side, likely due to liability issues, which means their output is always towards the higher end, as with any quality product. We have verified this with the help of the Chemistry Department supervised doctoral candidate students using a radiospectrometer. [What is a radiospectrometer doing in a Chemistry Department you ask? - They use it to obtain light intensity and calibrate those light sources to be used as references when measuring the efficiency and efficacy of solar cells produced by the students.] Using the correct BIN information, the math should read 1,300mw X 1.35 = 1,755mw total output, which is precisely what is indicated on our product page. The thermal conductivity for all of our MCPCB stars and boards (including yours) are 1.6 W/m.K (watts/meter*Kelvin) which is evaluated in terms of Fourier's Law - I'm having college physics and thermodynamics flashbacks just thinking about that. Additional Comments from Steve: The calculation of heatsinks depends on more than a dozen variables to accurately calculate thermal dissipation of a heatsink - ambient temp, heatsink material, heatsink mass, heatsink fin mass, convective airflow, radiative properties of the heatsink finish, ambient humidity, ambient airflow, forced airflow, linear windspeed, turbulence from air source, parasitic turbulence from heatsink finish, turbulence from nearby heatsink fins, thermal conductivity of PCB, thermal conductivity of PCB thermal paste, reflectance of light from the splash shield he is using, which will partially reflect light back onto the heatsink, etc, etc, etc just to name a few. They have 2 part college courses devoted to heatsink selection. To solve the issue - the most elegant solution is to simply estimate a heatsink's capability and err on the higher side and select a heatsink that is more than capable of performing the task at hand. ...Basically what I took from that was that it is extremely complicated, and if you want to get really detailed about it, you'll never end up building it!! I would keep it simple. Come up with the layout of LEDs and the drive currents, and then shoot it off to a heat sink person, they will likely be able to give you an answer off the top of their head - saving you all the math.
I'm with Turbo. The cost of an over sized heat sink may be an extra $10-$20, whereas, the cost of a hand full of LED's is in the same range.... Cost in oversizing that sucker = cost of replacing burnt diodes
Thank you, Turbo, for taking the time to contact Steve on my behalf. And thank you to Steve for the detailed replies. Your helpfulness is sincerely appreciated. That is precisely why my math was not "right" -- it was "intentionally wrong." I deliberately chose the first iteration of calculations to imply a worst-case scenario, where the heatsink has to be capable of being able to dissipate all of the power being applied to the circuit. It's an established principle of conservative design. My math was "wrong" because I'm already doing what Steve suggested. I do appreciate the reference to Table 9 on DS68. That fully explains what I needed to know about the maximum radiometric flux rating. That, and the thermal conductivity value of 1.6 W/m-K get me a lot closer to the answer. Note: it would also help to know the thermal contact surface area of the star boards. If that information is posted on the site, I managed not to see it. Solving equations is not a problem once you have the coefficients to plug-in. Thanks again.
In an effort to clarify things a little: All of those variables pertain to the thermal interface between the heatsink and air, but they don't have anything to do with the thermal interface between the heatsink and the thermal pad/LED assembly, which is the question that I asked about. I'd just like to clear up that distinction, in order to dispel the impression that there are so many variables to take into account that the problem is un-solveable, or that there is no precise answer to the question. The answer is indeed solvable, and all of those extraneous variables don't have anything to do with the problem at hand -- that being the thermal interface between the MCPCB stars and the heatsink material. Now that I've had a chance to look over the data sheets, I've tried to get a little closer to the theta-j value that everyone wants to know when they're selecting a heatsink. Looking at the MCPCB design parameters table in Application Brief 32, it looks like Steve's value for dielectric thermal conductivity of 1.6 W/m-K falls closer to the "typical" ratings for MCPCB expoxies than it does to the "high conductivity" epoxy ratings. (ie: it looks like Steve's quoted value of 1.6 is a lot closer to the Table 1 value of 0.8 than the value of 4.0.) Solving the pair of simultaneous equations to interpolate the data, my best guess is that the MCPCB stars that Steve is using for thermal pads probably have a Junction Thermal Resistance value in the neighborhood of 8.44 K/W. If you're a heatsink guy, I think that theta-j is the only number that you really want to know. I just thought I'd post this in case anyone else comes along looking for this exact figure. Hopefully these calculations will save that person a lot of time and jumping through hoops. If there are any mistakes in my calculations, please let me know. thanks again.
This thread is turning into quite a laugh, as nobody else seems interested in such a technical subject, and it looks like I'm having a conversation all by myself. So unless anyone else is interested in this problem and posts a response, this will be my last post on the subject. I received a follow-up email originating from Jeff at StevesLEDs.com, related to the questions in this thread. Jeff emailed some data that said that the previously cited value of 1.6 W/m-K was a typo, and that the thermal conductivity for the thermal adhesive that they use is 0.84W/m.K. The revised number eliminates the need for estimations/interpolations on my part -- it lines up almost exactly with the reference value of 0.80 W/m-K in the left hand column of the table in AB32.pdf. In other words, after adding up the Junction Thermal Resistance for each component, it looks like the aggregate Junction Thermal Resistance for the LED pad assembly is going to be 10*K/W. IMO that number is actually pretty bad. But it is what it is. What's most interesting about this is that if you work backwards to calculate the Tj values for the LED in order to determine the actual efficiency / operating point for the diode, the numbers get pretty scary. You have to have a first-class heatsink with a very low C/W rating (like Bud is using), you have to limit the number of LED on it, and you have to keep it at room temperature, otherwise Tj for the LEDs is going to be sufficiently high that they're going to start to operate on the steep (inefficient) portion of the Relative light output vs. thermal pad temperature characteristic. (see tables 7 & 8, DS68) What this means is that you won't be getting 100% of the light output that you're expecting. The bad news is that those plots demonstrate that unless the thermal pad temperatures are kept low (ie: the intersection of the characteristics is at a pad temperature of 25*C, which I don't think is even attainable outside of a refrigerator) you suffer a significant amount of lost light output as the temperature of the LED rises, and you're not actually experiencing the high-efficiency lumen/watt output that the marketing data would suggest; this revelation of a 0.75E operating point gives me some serious doubts about how honest the LED industry has been in the way that it has been marketing LED lighting. Fortunately for us, the Deep Red and Royal Blue are the best behaved of the Rebel color LED, as their light output does not drop as much as the other colors when the termal pad temperature increases. But the take home point from all of this is that if you take the time to crunch all of the numbers, the light output of these LED isn't even close to what you might expect by reading the "specs" that are quoted by the LED marketers who suggest replacing your MH or fluorescent lights with LED. I won't go through actual number crunching, out of fear that I'd bore everyone. But here is a handy heatsink calculator that will help anyone who is interested in performing a quick check of the numbers: http://support.luxeonstar.com/custo...at-sink-i-need-includes-heat-sink-calculator- My take-home from all of this number crunching is that Bud has a really good ATF design, and his design maximizes the number of LED that you can fit on a board while maintaining decent operating efficiency. My unsolicited opinion is that Bud has a pretty good design, and that when compactness and wavelength-specific illumination outweigh the lost light output due to heat, the LED scrubber design really makes sense. Kudos, for a great design, Bud. My other take home from this is that for my large-scale application, LED just don't make sense. MH and VHO remain the best solutions for a large-scale application. In the end, I'm glad that I did the math, as it prevented me from making an expensive mistake. I'm still going to build an LED scrubber, but it's going to be sized appropriately (on the order of Bud's design) for a smaller tank. HTH
I am definitely following the thread and reading all you post, you just happened to catch me on the tail end of a 2 month crunch at work where I have literally been working 12-16 hour days, including weekends, so I haven't had any time to absorb anything very well. Now with the holidays upon us, the "work" has just shifted to another "resource"... I think this is very useful information that really hasn't been discussed to the level of detail you present here, so if you want to go down the road of detailing the calculations, by all means, please do so! I'm 100% sure that someone will find it extremely useful I for one would like to see the numbers that back up what you are saying w/r to light output at operating temperatures for various LEDs. I can't recall this really being discussed on any forum - but then again, I haven't really dove deep into LED threads on the big boards. As far an an aecdotal comparison, when you reference the issue with replacing MH/VHO with LED and the lowered quality of light output, I think that most people that do PAR measurements might take the other side of the fence - saying that their actual constructed fixture (which likely disregarded the intense calculations you have gone through, and was more of a "trial and error" type of process) output far more PAR than then MH fixtures, and the result was bleaching out their corals. This was the early-on happenings with a lot of people, they cranked their lights up until things "looked the same" only to burn everything out because the light was actually too intense. Or so was the theory. Since then, with the advent of the higher quality out-of-the-box fixtures like AIs or Radions, Reefbreeders, etc, mixing in a lot of different colors in order to achieve "full spectrum", and incorporating good cooling fans, I think this is where you start to take the edge off the heating issue and you start to get better functioning fixtures. It's good to know that the DR and RB LEDs operate at a better efficiency level across various heat levels. Might this also have to do with lower maximum drive current, at least for DRs? Another thought, the output of the LEDs getting reduced due to heating might not have as serious of an impact on scrubber fixtures when compared to DT fixtures. As I have continued to develop the LED fixture/MCPCB design, what I've found out is that the "default" setting of the jumpers on my boards works quite well to cure the screens fast. That default setting is to have all the LEDs running at 350 mA via jumpers that put pairs of LEDs in parallel. I have an L4 on a 200g tank that also runs a skimmer, filter sock, and infrequently changed carbon, and I have found that I don't need to run that scrubber at the high-output level to get excellent growth and filtration - it does just fine at the low-current level. What may explain this is that the output of the LEDs at the 350mA level is still good, because the junction temp is kept lower. Upping the drive current to 700mA likely results in high light output, but how much of an increase is probably not linear - but I wouldn't expect it to be either. I just had not considered the junction temperature issue, but rather, just the "spec" efficiency issue. Of course, the game might be changing a bit with the new Cree LEDs that were just announced. I knew that something was coming from them back mid-year, I just didn't know what.
The beauty of the scrubber design, compared to DT lighting, is that the LED allow you to eliminate the energy that goes into parts of the spectrum that aren't particularly helpful for growing algae. The result is that an LED scrubber is highly efficient, in terms of the usefulness of the light that you generate. With DT lighting I think people are more likely to want "balanced" light output, which means that the DT lighting wouldn't be visually appealing if it were tailored to support absorption peaks, like an LED scrubber. I don't know about you, but I can't stand to look at a tank that's lit entirely at 450nM. One of the areas where I think that MH lighting still has a leg up on LED is in terms of directionality. LED without reflectors act a lot like omnidirectional radiators in a half-space environment, rather than as focused light sources. In that respect they're much like a MH bulb that's placed up against a horizontal surface in a half-space environment with no parabolic reflector. They provide a very wide light distribution. Looking at the Luxeon spec sheets (DS68, Figs 16, 17), the polar response for the LED is 50% output at 60* off-axis. In other words, the light from the LED spreads out to a total coverage angle of 120* before light falls off to half-intensity. The result is non-focused light, much like an MH bulb without a reflector. Failure to control light distribution can result in the possibility of having a lot of useful light spilling out of the tank. To make the most out of LED lighting, you either have to put the LED in a box that reflects the light back to the subject (like your scrubber boxes) or you need to use a reflector to focus the output, like you'd use with an MH light. Most of the LED fixtures I've seen aren't incorporating the little flashlight-type reflectors, which are ridiculously expensive for what they are, and which are required to make LED a viable option in a DT environment where you don't want too much light spillage into the room. LED that are allowed to radiate their light out 120 degrees without reflectors are going to cast a lot of light out of the tank, which amounts to ambient lighting and wasted energy. Getting to the point about operating efficiency, here's that chart I was talking about: To make the most out of this, we'll have to spend some time doing math to work backwards to the LED temperatures. That's the subject for another post.
does the board software shrink images after they are uploaded? in the previous post I tried to upload images that would be large enough to read the charts clearly, but it seems like they've been shrinked.
There is a hard ceiling that is server based. If it's less than 1 MB it should upload as that size. But I might have a setting off. I'll check into it.
I see that my images in Post 7 are sized the way I was expecting them to be sized, but the most recent images in Post 14 are small and hard to read. I'm trying to determine if there's a size limit (pixels or MB) that results in the images being shrunk. It's hard to demonstrate things with the charts when they get auto-sized to be small and illegible. I don't know if it's a browser problem at my end or a board problem. If you can shed any light on this <rimshot> I'd appreciate it.
