A serious proposal for serious Luxeon Moders

I am amazed with the amount of innovative work with Luxeons that is being done by the some of the more serious moders here. I am pleased that so many take the time to share what they are doing. I do experience disappointment, however, when I see someone who obviously put much thought and effort into a creative mod but their own analysis of their work is in the vein of "its brighter than my XXX and is not so hot that I have to drop it".
I want to encourage the *serious* moders to instrument their mods. With only a tiny hole for very fine wires [that can be later plugged if desired] it is fairly easy to measure actual LED voltage, current, and heat sink temperature. From these data, actual junction temperature is easily calculated.
How to do it:
Voltage: obvious
Current: I recommend using a 0.1 ohm sense resistor in the LED circuit. This is low enough to have negligible influence on the circuit yet will provide 1mA resolution when measured on the 200mV scale of the typical 3 1/2 digit DMM [100mV = 1A]. The price of low-ohm sense resistors has dropped in the past year. At Digikey you can get 0.10 ohm, 1%, 0805 SMT package ones of 10/$3.44. On a Luxeon Star, it is easy to cut one of the power traces with a Dremal type tool and then bridge the cut trace with the sense resistor.
Temperature: National Semiconductor make a number of three terminal temperature sensing ICs. These include the LM20, LM34, LM35, LM45, LM50, LM60, and LM61. For our uses, the LM61 is the best bet. It can be had for about $1/each at Digikey.
See link
http://www.national.com/pf/LM/LM61.html
It is available in both SOT-23 and TO-92 packages and has a low enough operating voltage to power directly from the voltage applied to the LED. Current consumption is a negligible 100uA. I strongly recommend the SOT-23 package because it thermally couples much better to the surface it is mounted on than the TO-92. In the case of the Luxeon Star, the V+ and GRD pins of the SOT-23 will span the (+) and (-) pads of the Luxeon thus providing a good measure of the PCB temperature.

So what can you do with the data obtained?
First, and foremost, you can impress your CRF buddies.
Here is an example of data usage: You have instrumented you new XYZ mod as recommended and obtain the following data; Vled=3.31V, Iled=380mA, temp=98C. You can calculate power, P=(3.31V)x(.38A)=1.26W. Now, from the datasheet, the junction to PCB thermal resistance is 17C/W so now you can calculate the actual junction temperature as 98C + (1.26W)x(17C/W) = 98+21.4=119.4C.
OK, so what if you don't think that your mods are in the same league as, say, dat2zip? You can play too. Suppose your mod is simply N-cells driving a Luxeon through a dropping resistor? With 3 tiny wires you can bring out the LED voltage, the battery voltage, and the drop across the dropping resistor. With this information, you can measure/calculate actual LED voltage, current and power. Example, you are using a 2 ohm dropping resistor. You measure Vled=3.2V and Vres=.58V. To get LED current, use ohms law I=E/R =.58V/2ohm=.29A. Now you can calculate LED power = (3.2V)x(.29A)= 0.93W. With your instrumented light, you can actually measure what happens to LED current as the batteries run down instead of guessing. You will also gain an appreciation of just how much light Luxeons can put out at very low current levels. You can also step forward and answer the [seemingly endless] questions that appear on the board "how much current will a Luxeon draw on 3 C cells?" You can say... I don't know about your particular sample but mine draws xxx under the following conditions...

Doug S,

I like your thinking, but, in reality the data would be basically useless. Luxeons vary way too much for such scientific analysis. All the data goes out the door when the next Luxeon produces twice/half as much light at the same current/voltage/temp/lux. It's just the nature of the beast. Current technology for high power LEDs just isn't that consistent.

I had the same thoughts initially, but soon realized that it just wasn't worth the effort without control of the manufacturing process. So, I leave such testing to LumiLeds where they can make a difference more than me.........

I think I understand what you're saying Doug, but try and understand that I build my mods only as a hobby. If I were to step up and start producing my mods for sale, then it might be a worthwhile endeavor.

On the other hand, I DO test my mods for current consumption and publish my results along with other specifications for what I have built. I do not want to "drill a hole", no matter how tiny in any of my mods, just to gather a bit of extra information. Such information tends to vary wildly due to the components used for these mods, and would probably not be all that valuable like Lambda says.

Originally posted by lambda:
Doug S,

I like your thinking, but, in reality the data would be basically useless. Luxeons vary way too much for such scientific analysis. All the data goes out the door when the next Luxeon produces twice/half as much light at the same current/voltage/temp/lux. It's just the nature of the beast. Current technology for high power LEDs just isn't that consistent.

I had the same thoughts initially, but soon realized that it just wasn't worth the effort without control of the manufacturing process. So, I leave such testing to LumiLeds where they can make a difference more than me.........

Click to expand...

<font size="2" face="Verdana, Arial">Lambda, I am in complete agreement that there is great variability in the Luxeon product. I do not agree with the judgement that the data is useless. This is especially so for the temperature data. I think that having the ability to judge the effectiveness of various heatsinking arrangements is very useful. The heat that must be dissapated for each watt of input power only varies a few percent from the very best to very worst samples. I think that some people upon knowing the actual junction temperatures they are attaining with a particular arrangement would factor that information into decisions about what power levels to drive at.

Doug S,

Have to agree on the temp data, as I do use that when making mods. Only I just point my TPI IR thermometer at it read the temp that way. In most applications, when bonded with Artic Silver, the Luxeon never gets more than 5F over the heatsink temp. Very high power mods may see bigger differences.

Originally posted by lambda:
Doug S,

Have to agree on the temp data, as I do use that when making mods. Only I just point my TPI IR thermometer at it read the temp that way. In most applications, when bonded with Artic Silver, the Luxeon never gets more than 5F over the heatsink temp. Very high power mods may see bigger differences.

Click to expand...

<font size="2" face="Verdana, Arial">Lambda, yes, those IR thermometers are wonderful. What you are doing seems like it would work well for designs where the outside of the assembled light is very closely thermally coupled to the Luxeon PCB as appears to be the case in some of you clever designs.
For many mods however, the measured temp of the heatsink as operated on the bench will be quite different from the temperature as operated in the assembled mod. Most mods when assembled do not offer a direct view of the heatsink [or better, the Luxeon PCB] for the IR thermometer to read. An IR thermometer cannot read the temp through a seemly clear lens [it will mostly read the temp of the lens itself] since most materials that are clear to light of largely opaque to the IR wavelengths being measured.
From the designs you have posted it is apparent that you are one of the moders that has a keen appreciation of thermal issues. On average, however, I think that thermal issues are underappreciated by many moders here at CPF.
As an electrical engineer who has designed circuitry [such as switching regulators] I can remember occasions where I went to great lengths to squeeze an extra few percent of efficiency out of some circuit. Sometimes this level of anal retentiveness doesn't make sense if the same gain in the desired end product [light in our case] can be had more easily from tweeking the system elsewhere. A 10C reduction in junction temp can deliver that few extra percent light with no increase in power consumption.

Doug S,

I agree with what is being discussed here and I too have gone out and bought a IR thermometer.

I am very aware of heat and die temperatures.

It is this one specific issue I have that concerns me releasing the higher power BADBOY boards since the total combined power being dissipated can exceed the LED specification.

All my mods are rated pretty conservative to this point. Even my joking about my first badboy needing to be waved around is still on the conservative side since my hand is not callous since I work in an office and I burn around 150F which is 65C and add 10 or so for the LED difference leaves the LED at 75-85C.

This is far below the LED rated specification.

This is and will be a very serious issue when the 5W comes out.

The heat that must be dissapated for each watt of input power only varies a few percent from the very best to very worst samples. I think that some people upon knowing the actual junction temperatures they are attaining with a particular arrangement would factor that information into decisions about what power levels to drive at.

Click to expand...

<font size="2" face="Verdana, Arial">I beg to differ here. Remember that the Luxeon spread in VF is very broad. At a constant current of 350mA the power extremes would be:

2.79V * .35A = .977W
3.99V * .35A = 1.396W

This like Lambda says varies way to much.

Constant current does not equal same brightness. Luxeons vary in a way that if two were driven at the same current say 0.35A it is probable that neither would produce the same light output.

The Luxeon is a fickled beast.

-WayneY

Originally posted by dat2zip:
Doug S,

Constant current does not equal same brightness. Luxeons vary in a way that if two were driven at the same current say 0.35A it is probable that neither would produce the same light output.

The Luxeon is a fickled beast.

-WayneY

Click to expand...

<font size="2" face="Verdana, Arial">Thanks for that Wayne. I've been waiting to hear your results.

I was thinking of using a 3-way switch to run a 3 AA at 2.2 ohm or direct drive. I would start off using the 2.2 ohm setting, when batteries die down (or I needed short run brighter light) I could then switch to direct drive.

Can someone tell me approximately how efficient a 2.2 ohm resistor would be on a 3 cell 1w Luxeon? I am not an EE but could test it myself if someone would show me the way.

Tim

With the IR thermometer measurements, the emissivity has to be accounted for; the less expensive units are calibrated for a fixed emissivity in the range of .95 or so- does the unit you have allow you to set the value to provide an accurate reading?

Originally posted by dat2zip:
Doug S,

</font><blockquote><font size="1" face="Verdana, Arial">quote:</font><hr /><font size="2" face="Verdana, Arial">
The heat that must be dissapated for each watt of input power only varies a few percent from the very best to very worst samples. I think that some people upon knowing the actual junction temperatures they are attaining with a particular arrangement would factor that information into decisions about what power levels to drive at.

Click to expand...

<font size="2" face="Verdana, Arial">I beg to differ here. Remember that the Luxeon spread in VF is very broad. At a constant current of 350mA the power extremes would be:

2.79V * .35A = .977W
3.99V * .35A = 1.396W

-WayneY</font><hr /></blockquote><font size="2" face="Verdana, Arial">Wayne, I stand by my statement. My statement addresses the heat to be dissapated per watt of input power, not per unit of input current. In all cases, over 90% of the input power results in heat to be dissapated. Your reminder that for a constant current, power varys considerably due to the variable Vf is why it is worthwhile to instrument our mods for both current and Vf if we really want to see what is going on.

Originally posted by dat2zip:
Doug S,

I agree with what is being discussed here and I too have gone out and bought a IR thermometer.

I am very aware of heat and die temperatures.

All my mods are rated pretty conservative to this point. Even my joking about my first badboy needing to be waved around is still on the conservative side since my hand is not callous since I work in an office and I burn around 150F which is 65C and add 10 or so for the LED difference leaves the LED at 75-85C.

This is far below the LED rated specification.
-WayneY

Click to expand...

<font size="2" face="Verdana, Arial">Wayne, I don't in any way wish to appear to be casting aspersions on your work. I am very much in awe of your abilities to design and cram such good stuff into such small packages. That said, I find it improbable that at the power levels that your Badboys run at you have achieved as low as a 10C deltaT from the Mag housing *to the PCB* the luxeon emitter is mounted on. Even if it is only 10C, allowing for the additional 17C/W rise from there to the junction, it may not be as "far below rated specification" as you think. You are exactly the type of serious modder that I would love to see instrument their mods with a temp sensor mounted on the board in immediate proximity to the emitter slug.

Noted. That's a good idea. If the temp IC was installed to the emitter board.

One thing that would be better would be to regulate or monitor the temp with a control or feedback loop mechanism.

My long term outlook might be the addition for the 2AA mag 14 mm board is a three board stackup with one board being the microprocessor.

In this way you could monitor it and run it overdriven for short periods and not exceed the rated temperature specifications.

On the other hand running it for extended periods would allow the die to heat up since the heat sink would rise in temperature too. As it approached some upper limit trip, the micro could thottle back the power to the Luxeon automatically.

That's really what I would prefer, ts to guarantee total safe operating under all conditions.

It's a project on the horizon since there has been discussions of PWM circuits.

I think that I will be able to get back to near full time R&D in developing new products or prototying new circuity next month. I think my current goal is to move the two converter boards off to outside contractors so that I get back finished products which will free me to do more design work.

I like the idea of putting the SOT23 right on the emitter board and I might even steal your idea with the changes I'm currently updating if you don't mind.

I'm just as frustrated as many others that getting uniform consistant light from LEDs in general is a more laborious process requiring tweeking instead of being designed in.

The more I think about, the IC version makes perfectly good sound advice.

It should in almost all cases give much better and accurate readings of the LED base temperature. Assuming the LED is mounted with good thermal grease it should provide solid repeatable measurement.

-WayneY

Originally posted by dat2zip:
Noted. That's a good idea. If the temp IC was installed to the emitter board.

One thing that would be better would be to regulate or monitor the temp with a control or feedback loop mechanism.

My long term outlook might be the addition for the 2AA mag 14 mm board is a three board stackup with one board being the microprocessor.

In this way you could monitor it and run it overdriven for short periods and not exceed the rated temperature specifications.

On the other hand running it for extended periods would allow the die to heat up since the heat sink would rise in temperature too. As it approached some upper limit trip, the micro could thottle back the power to the Luxeon automatically.

That's really what I would prefer, ts to guarantee total safe operating under all conditions.
-WayneY

Click to expand...

<font size="2" face="Verdana, Arial">I think the idea of closed loop control using temperature is a good one! The ability to rely on the thermal capacity of the heatsink for short periods of high current drive has appeal. BTW, you can't beat aluminium in this role as it has 2.5 times the specific heat of copper. I recently pushed the numbers around and found that the temp rise in aluminum is almost exactly 1C per joule/g assuming no loss to the environment. So, for example, if we had a 30g aluminum heatsink and dumped 5W into it for 60 seconds (5W X 60sec =300J) we would have a 10C rise assuming no loss to the environment.
For actually implementing the control, I think that a microprocessor would be overkill. For example, with your particular choice of IC in the Badboy, control could be done via input to the current control pin 4. Upstream would be an op amp w/Vref and a temp sensor IC with voltage output.

Originally posted by dat2zip:

I like the idea of putting the SOT23 right on the emitter board and I might even steal your idea with the changes I'm currently updating if you don't mind.

The more I think about, the IC version makes perfectly good sound advice.

It should in almost all cases give much better and accurate readings of the LED base temperature. Assuming the LED is mounted with good thermal grease it should provide solid repeatable measurement.

-WayneY

Click to expand...

<font size="2" face="Verdana, Arial">Hey, no *stealing* is involved. I see the purpose of CPF to be able to exchange info and ideas.

BTW, for me, the SOT-23s are about the limit of size for hand soldering multileaded components. Your skills are obviously at a much higher level since you are working with 0.5mm pitch stuff. I'm sure that board real estate is at a premium on your boards. Temperature sensors do come in smaller packages, the LM20 comes in SC-70 and smaller. With the LM20, however, you need a calculator to convert the output voltage to a temperature, with the LM61 its easy to do in your head.

No problem, you foot the bill and send the instruments and I'll measure whatever you like. Until that time ,however, I will only be capable of the "compare/hold if you dare" measurements that you mention.
It's a reasonable desire tho and I don't blame you for requesting it.

Doug,

Great ideas. What I plan on doing with the uC is to implement a soft switch mechanism. Toggling the power would cycle through power modes from Low, Med, and High. Two models could be available, one that starts at high power and one that starts at low.

I'm thinking of using the Atmel Attiny15l 8 pin uC. I've used this line before and like the family line.

In this way you get a long lasting flashlight that in a pinch can deliver a large punch of light if needed. For most usages the low would be more than sufficient.

I forgot when I mentioned *stealing* to add

As for auto regulating badboy direct. I've got a sketch and have found the appropriate size NTC SMT part that should work. Haven't crunched the numbers through and determined what the trip point would be.

That's another possibility since it allows full power until it warms up and then slowely throttles back as it crosses over the thermal trip level.

I believe that my current badboy 14mm is already layout restricted from adding some of these features.

I have plans on making a 17mm board and will explore some of these options for a real full 5W driver since full 5W will probably need thermal management of some form.

-WayneY

Originally posted by dat2zip:
Doug,

As for auto regulating badboy direct. I've got a sketch and have found the appropriate size NTC SMT part that should work. Haven't crunched the numbers through and determined what the trip point would be.

-WayneY

Click to expand...

<font size="2" face="Verdana, Arial">Wayne, by NTC I am assuming that you are talking about a Negative tempco thermistor device. If so, I would suppose that you are applying it as part of a voltage divider attached to pin 4 of the IC you are using in your badboy. If so, this is definitely a simpler solution than what I had suggested earlier and one that you might even be able to get on your 14mm board since the sensing device would be wherever the Luxeon is mounted. I seem to remember seeing some thermister devices offered somewhere that had large transitions in narrow temperature bands. You actually bought them for specified temperatures. I think that they were PTC but that doesn't matter, you just put them in the other leg of the voltage divider network. I note that the datasheet for the IC you are using in the Badboy doesn't give a bias current specification for pin 4 That would be useful. Another approach using the voltage output type sensors like the LM61 and similar, would be to act on the IC Voltage feedback input. This would eliminate the need for a separate opamp and reference.

Took a quick look at the family that National has and I must admit I haven't looked at Temperature ICs for quite some time. I'm familiar with the LM35 which is pretty old stuff now.

I now see both positive and negative temperature output gain.

Pretty cheap price wise too.

I think I could incorporate this pretty easy on my next round which would be pretty nice. Too bad I just FTP'd the files last night. I still could hack up one of the current ones to test the design tho.

That would ease my conscience about over exceeding the specifications of the LED.

Just thinking about it now it's too bad I didn't keep the IC and LED on the same board. Could kinda let the IC overtemp shutdown be the crude temperature regulator. If the LED got both the board and IC hot the IC would shutdown. Pretty crude tho. Would have to do some good testing to verify where this in fact would shutdown.

Pin 4 specifications is very vague I agree. Even the reference usage in the data sheet is kinda hooky.

-WayneY

Originally posted by dat2zip:
Just thinking about it now it's too bad I didn't keep the IC and LED on the same board. Could kinda let the IC overtemp shutdown be the crude temperature regulator. If the LED got both the board and IC hot the IC would shutdown-WayneY

Click to expand...

<font size="2" face="Verdana, Arial"> Do that and I would never again believe you when you say conservative
Besides, I don't remember any mention on the datasheet of overtemp shutdown for that IC.

Let me see if I can summarize:

Use a 0.1 ohm in series to measure current since most DMM affect the measurement and can't be trusted.

preferably measure the temperature as well since that determines the safe operating region for both the LED and ICs.

Besides, I don't remember any mention on the datasheet of overtemp shutdown for that IC.

Click to expand...

<font size="2" face="Verdana, Arial">Well dang, I better retract my foot out of my mouth. Just checked the datasheet and can't find it. Must have been a different IC. Never mind...

-WayneY

Originally posted by dat2zip:
Let me see if I can summarize:

Use a 0.1 ohm in series to measure current since most DMM affect the measurement and can't be trusted.

preferably measure the temperature as well since that determines the safe operating region for both the LED and ICs.
-WayneY

Click to expand...

<font size="2" face="Verdana, Arial">I guess the only thing that I would add to your succinct summary is to get the temp sensor as thermally close to the slug as possible. That way, we can reasonably use the published Luxeon thermal resistance from board to junction when calculating Tj using the power we calculated with our measured I and V. When the emitter is mounted on an aluminum slab as is the case with the "Star" configuration, the backside of the board under the slug would be optimum. It gets more difficult when the emitter is mounted on regular double sided PCB material. In this case, the backside under the slug is not that thermally close. In this case, mounting on the frontside as close to the slug as possible would probably be best. If necessary to mount on backside a possible solution would be to have a circle of thermal vias through the board surrounding the slug similar to the picture on your website of the little circle of etched ovals that you have as a guide to centering the emitter on the current Badboy board. The sensor would mount inside the circle. Yeah, I know you sent the files for fab, just offered for future reference. If you are interested in IC temperature, I would use a separate temp sensor that is thermally close to the IC.