# Some thermal testing and analysis

In these high power mods we do, thermal management is often a limiting factor. Understanding how and which characteristics of a flashlight are desirable for good thermal management can be confusing. I have done some testing and analysis that has helped my understanding of the subject so I thought I would write something up to share.

Ultimately, it is the LED junction temperature that we want to manage. 150C is the quoted maximum junction temperature for several Cree emitters. That is very hot, much hotter than the outsides of flashlights tend to get. But because of the finite thermal conductivity of flashlight parts, the junction temperature is hotter than the body of the flashlight. Cree also quotes the thermal resistance of emitter packages. The XHP-50 thermal resistance is 1.2 (degrees C)/W. This is the thermal resistance between the junction temperature and the solder point at the bottom of the package, where the package is soldered to the MCPCB. The temperature difference between the junction and the solder point is proportional to the thermal power flowing between them. So for a high driven XHP-50, 6A(7.3V) is the input power. LEDs are typically 40-60% efficient, so 40-60% of the input power leaves as photons and don’t contribute to heat. Let’s say it is 50% efficient, so 21.9W of heat is produced, which means there is 21.9W(1.2C/W)=26.3C temperature difference between the junction and solder point.

Thermal resistance:
This brings us to the first thermal characteristic of a flashlight; the thermal resistance between the MCPCB and the surface of the flashlight where the heat fins are. This resistance should of course be minimized to maximize heat transfer to the outside of the flashlight. I measured this thermal resistance for two flashlights, the Eagle Eye X6, and the Ultrafire F13 with integrated shelf. To do this I powered an LED in the flashlight heads directly using a current regulated power supply. Without the driver in the way, I could measure the temperature of the bottom of the shelf directly below the MCPCB, using an IR thermometer. The shelf temperature should be very close to the temperature at the solder point.

I used a Latticebright LED driven at 4A, so the thermal power was approximately 9W, assuming this LED is only 40% efficient.

With the Eagle Eye X6, there was approximately a 2.8 degree C (5F) temperature difference between the shelf and the hottest part of the head where the heat fins are. This temperature difference was approximately constant as the temperature rose from 43C to 65C (~110F to 150F). So the flashlight thermal resistance is approximately 2.8C/9W=0.31C/W.

The Ultrafire F13 has an integrated shelf like the X6, but it is very thin. There was approximately a 6.7C (12F) temperature difference between the shelf and the hottest part of the head. Again, the temperature difference was approximately constant as the flashlight heated up. Thermal resistance is approximately 6.7C/9W=0.73C/W.

If we add the LED package and flashlight thermal resistances, we can know the total temperature difference between the LED junction and the flashlight head at different LED powers. Using the XHP-50 driven at 6A as an example, the total temperature difference would be 33C for the X6 and 42.3C for the F13.

So we can estimate the junction temperature if we know the head temperature, using this relation. At 140F (60C), a flashlight is hard to hold for more than a second and the junction temperature in the example above would be 93C for the X6 and 102.3C for the F13. Well below the max junction temperature of 150C. This is assuming the shelf temperature is very close to the LED package solder point temperature. Maybe someone knows the thermal resistance of the MCPCB/thermal paste/shelf joint?

Heatsinking:
Now we consider the heatsinking ability of flashlights, the ability to transfer heat to the air. I measured this characteristic for the same two flashlights. I powered the same latticebright emitter at a constant 3.2A. The electrical power was 3.6V(3.2A)=11.5W. The thermal power was then approximately 6.9W if the emitter was 40% efficient. The flashlights were mounted horizontally with the body tube installed, but no tailcap (so the power wires could reach the emitter). It was about 82F (27.8C) in the room.

In still air, the X6 head stabilized near 174F (78.9C) after 26 minutes. The F13 head stabilized near 155.5F (68.6C) after 26 minutes.

With a fan on that provided a slight airflow, the X6 head temperature stabilized at 151.5F (66.4C) and the F13 head stabilized at 136.5F (58.1C).

Now, to predict the head temperature:
Thermal power produced=(alpha)(T_h-T_a)
T_h is the head temperature and T_a is the ambient air temperature.
In equilibrium, the thermal power produced by the emitter is equal to the thermal power dissipated by the flashlight body surface. And this dissipated power is proportional to the temperature difference between the head and the air temperature. The (alpha) will be different for different flashlights and airflow conditions.

For the X6, alpha is 0.135W/C in still air and 0.178W/C in slight air flow. For the F13, alpha is 0.169W/C in still air and 0.229W/C in slight air flow.

Using the relation above, one can estimate the head temperature for different emitter powers (after the temperature is left to stabilize). Then with this head temperature along with the flashlight and LED package thermal resistances, one can estimate the LED junction temperature. However, this relation is not useful in practice for over powered lights because these lights are almost never left to equilibrate; even with the slight breeze, the F13 with XHP50 at 6A would equilibrate at 95C above ambient temperature.

Hopefully this might help someone understand or approximate the thermal characteristics of hosts.

If people are interested in this here are some other links that might explain some things better than I did:

http://www.digikey.com/en/articles/techzone/2011/jun/calculating-led-junction-temperature-in-lighting-applications

http://www.cree.com/~/media/Files/Cree/LED%20Components%20and%20Modules/XLamp/XLamp%20Application%20Notes/XLampThermalManagement.pdf

Note that in these links the thermal resistance of the heat sink is discussed. This is the temperature of the heat sink above the ambient temperature when it is dissipating a certain power. This is a steady state measurement so it is valid only when the temperature of the flashlight has stabilized after a long time. For general solid state lighting this is relevant, but like I said for our overpowered lights we turn them off before equilibrium is reached because the equilibrium temperature is much too high so they are always in a transient state.

The thermal resistance between the shelf and the head that I measured above is a more relevant and useful value as it can be approximately used in the transient state we use our flashlights in to approximate the junction temperature when we know the head temperature.

By the way, I estimated the thermal resistance of the MCPCB/shelf joint. If the thermal paste has a thermal conductivity of 5 W/m*K and is 0.075mm thick under a 20mm star, the thermal resistance is about 0.05 C/W, much lower than the package and head thermal resistances.

Very cool post. I guess those 2 lights are somewhat popular so this should be good info to a number of people.

One thing I would have liked to have seen though is a comparison between mcpcb and dtp stars.

But great work, thanks for sharing.

Whether the star is DTP or not certainly has a large impact on the performance of the emitter, but I don’t think it would affect the measurements that I performed. At a given power, whether the thermal path is direct or not would affect the temperature of the emitter package, but the thermal power flowing down through the MCPCB and into the shelf would be the same in either case, so my measurements would be unaffected by whether the star is DTP or not.

For fun I estimated the thermal resistance of a non-DTP board (using equation 8 in the Cree pdf I linked above). I looked up the dielectric materials used and they are things like acrylic, epoxy, polyurethane… which have thermal conductivities in the range of 0.25 W/m*K. If the dielectric layer is 0.075mm thick and the thermal pad is 2mm x 3mm, the thermal resistance of the dielectric layer is 50 C/W, which is huge.

Great work.

From my understanding a dtp would have a cooler led and higher overall head temperature. Rendering any data you have gathered here useless to anyone using dtp stars. I could be wrong but just my 2c

Like I said, I think the measurements I did are intrinsic to the hosts and describe thermal resistances of the flashlight heads. The thermal power flowing through the MCPCB would be the same whether it is DTP or not.

Your understanding is qualitatively correct. A DTP would result in better transfer from the LED to the MCPCB, and this keeps the LED cooler. Right after the light is turned on, this would indeed cause more heat to enter the flashlight head. But the LED package itself is such a small thermal mass that it would heat up and reach equilibrium quickly (probably in a couple seconds), and at this point the heat flow to the MCPCB would be equal for the DTP and non-DTP. For the non-DTP star, the LED would be hotter in order to flow the same heat through the less thermally conductive non-DTP mount. But the same thermal power will flow for a given LED power; this is a consequence of the conservation of energy. OK, this is not completely true; for large currents the non-DTP LED will become very hot and the efficiency will drop, meaning more thermal power will flow with the non-DTP MCPCP.

Very cool (hot?) measurements, thanks for doing this. It confirms my long-term observation that flashlights with well mounted DTP-boards can be left to become burning hot with no apparent damage to the led. It is really nice to see some data on this now

I’ve asked before, how is it we don’t have little heat-dissipating modules/heads/collars/fins at the driver/LED level.

There’s one nice illustration here, from Ronin:

Or take something like this heatsink,

with added threads into the center on the pin side to take the battery tube, and a bezel over the emitter on the other side of the heatsink

Or this

http://e3d-online.com/v5-3.00mm-Direct-Heatsink

It’s not critical, yet — but we are heat-cycling our batteries rather dramatically with the hotrod flashlights.
That can’t be good, over the longer term.

40-60% efficience is false.
25-30% at manufacture specified currents, and much lower at currents that blf users like to drive.
Host and other external parts resistance calculations are wrong. You can use this method for selfcontained systems only.
I like this idea - making measurement in real hosts (I can add one more suggestion-make it with real cell). All this test on several kgs of copper or aluminium powered by 20amps current supply gives users lots of information, but this methods are good for science and not good for simple users. Conditions are far away from real life and most can make false conclutions.

I will try to make something but it is very expencive. Lux-rc fx-30 and fx-60 can disperse 8w and 16w (led power) and I can suppose that more half of their price is mechanical parts and machining. EDC triple or quad can take 30w from 1x18650 for 15-20 minutes. All wants more power, but nobody wants to carry heatsink that will make holding light turned on for 20 minutes able.

Yes, you are probably right about this. Lowering the assumed efficiency would mean there was actually more thermal power flowing through the host and would lower the thermal resistances values that I derived above.

I don’t know what you mean by this.

Technically, the idea of thermal resistance is only valid in steady state systems, which a flashlight heating up is not. But as I noted, the temperature differences between the shelf and the head were approximately constant as the host heated up, so I think these measurements can approximately be used to infer the shelf temperature (and junction temperature) if you know the temperature of the head and the thermal power flowing from the LED.

Very useful analysis!

I’ll just add that while 150C may be the maximum junction temperature; according to Cree’s LM80 testing, you should keep the junction temperature below 100C if you don’t want the LED to wear out prematurely. Ideally, keep it below 85C for a long life-span. Above 100C, and you’ll never get the rated ~50,000 hours to 70% original brightness. You might get about 5,000 hours. Still a long time, though.

So, it looks like from your analysis, if you assume a 1.5C/W to 2.0C/W total thermal resistance, then a typical 1,000 lumen light which produces 5W of heat (and 5W of light), the junction point should only be about 7.5C - 10C warmer than the head of the light?

My guess would have been quite a bit more than that. Is the built-in thermal protection that some lights use, only for the purpose of not burning the user?

You can use your method if both parts (or both surfaces on one part) do not work as heatsink or this heat trasfert to air is very low.

No. It produces 8w of heat, 3w of light. And 10-15C is difference inside led, difference between led die and led thermal pad.
Yes, built-in thermal protection protects your hands only. Big lightening systems work for years, but component are too hot to contact them.

Yes, approximately, unless I missed something in my analysis. Note that the XHP50 junction to solder point resistance is 1.2C/W, but the XPL thermal resistance is 2.2C/W

I guess so, and as you noted, to increase the life of the LED and other components in the light.

I don’t quite understand. Is this so the thermal power flow is well defined?

With a flashlight in air, only when the temperature profile has stabilized can you precisely measure the thermal resistance of the head. Before it is stabilized one cannot precisely define what the value of the heat flow through the head is, because energy is going into heating the head. When the temperature profile is stabilized you know that the heat flow through the head is equal to the thermal power produced by the LED.

Regardless of whether the measurement I did measured the true thermal resistance of the head (I think it is close, though), it can be used empirically to infer the temperature of the shelf if you know the head temperature.

The calculations depands on what do you want to count.
If you want to know temperature difference between two parts or two places of one part, you need to exclude heat transfert to air.
Heat transfert to air difference from lots of extra factors, it need to be calculated separately.

You can use heat transfer to air to set experimental conditions. For example, with a flashlight in air that has stabilized in temperature, the thermal power transferred to the air is equal to the thermal power produced by the LED. So then you know the heat flow through the head, and if you measure the temperature difference between the shelf and head, you can calculate the thermal resistance.

However, this is assuming that all the heat produced by the LED is transferred through the head. In my measurement, the shelf was exposed to air and so some heat was transferred directly to the air instead of flowing through the head. So you are correct that this would cause an error, but the surface area of the exposed shelf is quite small so the error would be small. (only the bottom of the shelf was exposed. The reflector and lens were in place during the measurements)

This was another thought, a while back, along the same lines (it’s hard to find heatsinks that don’t have nasty sharp edges)

The earlier picture, the shiny metal thing with big fins and a little central hole, is part of a 3D printer.
Lots of heat-dissipation hardware out there, just not much for flashlights.