Thermal paste/grease preparation

[SmurfTacular]

I've been modding Maglites since I registered on this forums. And so far its been a blast. I've probably modded around 10 Maglites with P7's (and counting).

Anyway, I was wondering how you guys prep your Maglites heatsink to be coated with thermal paste. First of all, the stuff gets everywhere. I didn't know that when I first started working with it. I ended up getting thermal paste on everything; my clothes, my hands, everything I touched, the P7 dome etc. I have eventually learned to cope with thermal paste and avoid its unwanted stick-to-everything characteristic. Now before I insert the heatsink (H22A) into the Mag body, I put painters tape over the LED and rim of the heatsink, I also put tape all over the grated grip portion of the Maglite. Because the last few times Thermal paste stained the grated portion of the Maglite, literally! I couldn't get it out. I even tried nail polish remover and that barely did anything. I eventually used Acetone and that worked out pretty well.

how do you guys deal with and prep for using thermal paste? Also, what brand do you use?

 

[olephart]

Been there and done that. Still cleaning up. The hard core apply it from gallon cans with a 4" brush. Actually, ya don't need much or yer doing something wrong.

It all starts with the fit and finish of the heat sink and the body of the light. What you want is polished surfaces with a very tight fit. In that case, ya hardly need any paste. Nothing better than full metal contact for heat transfer.

The paste is used once you achieve the polished tight fit. It fills the microscopic holes that you can't see. It's not as good as metal to metal contact, but, it's better than air. Once ya get to gaps over .001", yer not accomplishing anything with a wad of thermal grease. It's not magic, just a little better than nothing. Won't cure a bad fit.

So, the answer is to work on the fit and polish and use a minimum amount of grease. Hard to imagine having so much that the LED needs to be taped up.

As to brand, I read the reviews where they test it on computer heat sinks. I haven't checked in a year, or so. Artic Silver and Zalman's seemed to be the best back then. I like the Zalman cuz it's in a finger nail polish type bottle and it's easy to paint on a thin coat.

 

[SmurfTacular]

Thanks for the advice, I just bought this

http://cgi.ebay.com/Zalman-ZM-STG1-...ultDomain_0&hash=item23077dacfa#ht_1372wt_911

 

[Justin Case]

The large contact surface area between your Mag heat sink and Mag body will most likely dominate over any of the other relevant variables. Thus, to a large extent, any thermal compound will work fine and you can withstand fairly large bond line thicknesses well above the desired "as thin as possible" thickness.

See this post for an example calculation. In the example, the contact surface area is for an LED turbo tower for a SureFire TurboHead. A Mag heat sink will have far more surface area. Yet, even for the turbo tower case, you can have a huge bond line thickness of 5 mils and still get a very low thermal resistance.

I doubt that there is any need to paint the entire surface of anything that you need to heat sink. That's just too much thermal compound. You aren't going to be able to paint a layer only 1 mil thick, are you? Thus, you are just going to slop on a lot of excess that is going to get squeezed all over the place.

Put some dots or lines of compound over the surface of interest. Interface that surface to the other surface of interest. Apply some pressure and some shear movement to spread the compound, e.g., rotate the Mag sink within the Mag tube. Inspect the amount of compound coverage and apply more dots/lines of compound as needed. Use the shearing movement to apply an thin, even coat to your surfaces, not some gross applicator brush.

 

[SmurfTacular]

The large contact surface area between your Mag heat sink and Mag body will most likely dominate over any of the other relevant variables. Thus, to a large extent, any thermal compound will work fine and you can withstand fairly large bond line thicknesses well above the desired "as thin as possible" thickness.

See this post for an example calculation. In the example, the contact surface area is for an LED turbo tower for a SureFire TurboHead. A Mag heat sink will have far more surface area. Yet, even for the turbo tower case, you can have a huge bond line thickness of 5 mils and still get a very low thermal resistance.

I doubt that there is any need to paint the entire surface of anything that you need to heat sink. That's just too much thermal compound. You aren't going to be able to paint a layer only 1 mil thick, are you? Thus, you are just going to slop on a lot of excess that is going to get squeezed all over the place.

Put some dots or lines of compound over the surface of interest. Interface that surface to the other surface of interest. Apply some pressure and some shear movement to spread the compound, e.g., rotate the Mag sink within the Mag tube. Inspect the amount of compound coverage and apply more dots/lines of compound as needed. Use the shearing movement to apply an thin, even coat to your surfaces, not some gross applicator brush.

Thanks for the really useful info. What about the anodization on the inside of the Maglite? Should that be removed for better heat absorption?

 

[Justin Case]

Thermal conductivity of aluminum oxide is far higher than any commercial thermal paste you will use -- probably 2X to 10X. The anodizing thickness could be a few mils thick, which won't matter since you have a large contact area. I think that if you remove the anodizing, you'll just create a larger bond line and have to fill that extra gap with thermal paste that has poorer thermal properties than the layer you removed.

 

[olephart]

I always thought a good fit and polish was important on any light. So, it's good to know that mags have enough area to tolerate sloppy fits and rough surfaces. Always nice to have excess capacity.

I notice threads that measure lumen reduction after a brief period. This is usually attributed to heat build up. Do you think the heat sink could be improved or is the body unable to shed the heat produced?

 

[Justin Case]

The figure of merit is the junction to solder point thermal resistance. For the Cree XR-E, as an example, this thermal resistance is 8 C/W. If you drive the LED with say 4W, then you are looking at a min temp rise of about 30C over the solder point temp.

You can then estimate the lumens drop from the curve of relative luminous flux vs junction temp, which is often found in the LED's datasheet.

For the XR-E, if you start at 25C, then the junction temp increases to about 55C, giving a relative luminous flux of about 0.95 (5% drop).

If you want to eliminate the drop, you'd probably have to chill the heat sink.

 

[fyrstormer]

Despite the joke further up, you can actually buy thermal paste by the gallon, but if you're using it right it'll take you months to use up a small syringe -- years even, if you're not constantly building stuff that needs it.

I use Arctic Silver 5, and the way I use it safely is I assemble the heatsink first, then squirt a drop of the thermal paste onto the mating surface of the heatsink, then I either smear it a little with a tiny plastic spatula or I just install whatever will be producing the heat and I let it smush the paste around. If you have a layer of the stuff thicker than a sheet of paper on-average, there is something very wrong with your installation or the way the parts fit together, and you need to fix that before final assembly.

I only ever touch the stuff with my fingers if absolutely necessary, and only with the tip of my pinky finger, and I clean my finger off with turpentine immediately afterwards.

If you're using the thermal paste to produce a better interface between the inside of the Maglite head and the heatsink, then put the stuff inside the Maglite head and then install the heatsink -- don't put it on the heatsink first, because that will just smear it all along the inside of the Maglite head while you're assembling everything.

 

[SmurfTacular]

Yeah, thats usually what happens when I insert the heatsink; it gets everywhere.

 

[olephart]

If you want to eliminate the drop, you'd probably have to chill the heat sink.

I wasn't asking about a tiny, and unavoidable drop. I was referring to P7's loosing 1 to 200 lumens (or more) after 1 to 3 minutes. Others don't appear to loose much, at all. Seems more about the heat sink or surface area. I was looking for some insight that might improve such a build.

 

[Justin Case]

Since you didn't specify any particular light, LED, or other details, don't blame me for an answer that doesn't address your question.

In any case, I am not referring to a "tiny" drop. Increasing the junction temp by 30C over RT results in a relative luminousity decrease of 5%. That example was only for one Cree XR-E driven at about 4W, and assuming that the solder point temperature was 25C, giving a junction temp of 55C and thus giving the estimated 5% drop.

If the solder point temperature is hotter than 25C, which is probably going to be the case in real life, the lumens drop will be greater.

Do the math yourself if you don't like my answer. I gave you all of the info you need. It's a series resistance problem. Add up the thermal resistances that are in series and calculate the temp rise back at the LED junction, and thus estimate the lumens drop. You can look up or figure out for yourself what the thermal resistance might be from the solder point to ambient for whatever flashlight floats your boat.

I can tell you right off that I've measured over 160F at the turbo tower heat sink for an MC-E tower driven at a nominal 613mA per core. The junction to solder point thermal resistance for the MC-E is given as 3 C/W in the MC-E datasheet. If we assume that the thermocouple temperature of the AW tower for the 2S2P MC-E/SOB1227/KT1 configuration is equivalent to the max solder point temperature (not strictly correct since the thermocouple temp is actually Ths, not Tsp, and thus there is the thermal resistance at the slug-tower interface), then the steady state junction temperature is approx 71C + 3C/W * 8.3W = 96C. Based on the relative flux vs junction temperature graph in the MC-E datasheet, output could go down to about 80% compared to the initial output when the LED starts at room temp (25C).

At ~600mA drive per core, the estimated emitter lumens from a K-bin MC-E is 1.55*370 ~570 lumens. A 20% decrease gives a lumens drop of about 114 lumens.

Does this >100 lumen drop meet your interests?

 

[olephart]

Just trying to understand something. Thanks for trying. Look at the first entry on the chart in post#204. http://candlepowerforums.com/vb/showthread.php?t=260659&page=5

No doubt the LED is getting hot and output is reduced by over 1/2 - all per the calculations you provided. Seems like it would be cooler if more heat could be removed.

The original question was, "Do you think the heat sink could be improved or is the body unable to shed the heat produced?" Another way of asking the question is would improvements to the heat sink increase the lumens, or is the surface area too small to do any better?

I am contemplating a tri-XP-G and see no point if all it will do is 400 lumens due to heat. I am trying to understand the size and type sink and amount of surface area necessary to produce 7 or 800 lumens for at least 15 minutes. Maybe this can't be done, I donno. That's why I'm asking.

 

[Justin Case]

Just trying to understand something. Thanks for trying. Look at the first entry on the chart in post#204. http://candlepowerforums.com/vb/showthread.php?t=260659&page=5

No doubt the LED is getting hot and output is reduced by over 1/2 - all per the calculations you provided. Seems like it would be cooler if more heat could be removed.

The original question was, "Do you think the heat sink could be improved or is the body unable to shed the heat produced?" Another way of asking the question is would improvements to the heat sink increase the lumens, or is the surface area too small to do any better?

I am contemplating a tri-XP-G and see no point if all it will do is 400 lumens due to heat. I am trying to understand the size and type sink and amount of surface area necessary to produce 7 or 800 lumens for at least 15 minutes. Maybe this can't be done, I donno. That's why I'm asking.

I doubt that heat affected the output within 30 sec of turn-on. To drop the output by 50%, the junction temp would have to be way over 150C, which just isn't going to happen that fast if the LEDs are well-sinked to something like a Mag sink. And if they weren't well-sinked, you'd see a tint shift quickly followed by LED death.

Most likely, the quick drop-off followed by constant output is due to the direct drive of the light and the discharge profile of the IMR26500 cell at 5.19A draw.

I've tested an IMR26500 on my hobby charger and Vbatt quickly sags from 4.2V to about 3.8V within about 30 sec. After 3 min, Vbatt is about 3.7V.

 

[olephart]

So, If I use a Fatman at 1.3A with a pair of 18650's in series and 3 x XP-G's in series, do I have a shot at maintaining decent lumens for a reasonable period of time?

 

[Justin Case]

Successful running depends on good thermal management. 3xXP-Gs is comparable to a Seoul P7 in terms of power draw. Certainly, that Mag mod has been routinely done.

The issue with a boost driver is the Vbatt:Vload ratio, which dictates the battery current that gets sent to the driver.

How low do you plan to drive your 2x18650 cells? 3.0V? 3.2V? 3.4V?

The recommendation for the Maxflex is to keep Ibatt < 2A. If we assume XP-G Vf at ~1A drive current to be 3.2V, Maxflex efficiency at 90%, and Vbatt=6.4V, then

Ibatt = (9.6V*1A)/(0.9*6.4V) = 1.7A.

Driver power dissipation = 9.6W/0.9 - 9.6W ~1.1W.

You should ask George if this level of power dissipation requires heat sinking the driver. My guess is that you can get away without sinking the driver, but heat sinking never hurts. If you sink the driver, I would try to sink it to the Mag tube, not the LED's heat sink. IMO, you don't want to add yet another heat source directly to the LED's heat sink that should be kept as cool as possible.

If you go down to 3.0V per cell, then Ibatt = 1.8A, so you are still within the recommended limit for the driver.

 

[olephart]

Thanks, that makes sense. I'll give it a try. I'm using a Fatman, not a Maxflex. The efficiency and current limits appear to be similar, so the math is good. The actual Vf on the XP-G's is 9.620 at 1300mA. Pretty close!

George sez the Fatman should be kept under 2.0A, but could be stretched to 2.25 with good heat sinking, etc.

I plan to keep the batteries above 3.6V and should not go over 2.0A. I'm using AW protected cells. Aren't they supposed to cut off in this range?

I think heat sinking the converter is a good idea and will do it as you suggest. I'll be making the head from scratch, but the dimensions should be similar to the components in a mag.

 

[Justin Case]

I meant Fatman, fingers kept typing Maxflex. Sorry for the confusion.

I don't know what the discharge current limit is for the AW protection circuit to kick in. I would think it is greater than 2A, since the present version of his 18650 cell has a 2600mAh capacity and the usual recommendation is a max discharge of 2C, or 5.2A.

 

[olephart]

I thought I read that there was low voltage protection that kicked in around 3.5 or 3.6V and kept it from over discharging. Maybe not. Hard to find actual specs on those batteries.

 

[Justin Case]

I thought I read that there was low voltage protection that kicked in around 3.5 or 3.6V and kept it from over discharging. Maybe not. Hard to find actual specs on those batteries.

The typical batt protection chips I've seen have an undervoltage limit more like 2.7V or so. If it were 3.5 or 3.6V, you'd waste a large percentage of the Li-ion cell's capacity.