Maximum voltage/current for a single SST-90

[SmurfTacular]

I saw this thread, and it gave me motiveation to do the same or something very similar. I want to build the farthest throwing LED plug-in spotlight, with two or more aspheric lens's.
But I want to use an SST-90 instead of an SST-50. And I want to know what is the max you can drive an SST-90 when its mounted on a massive CPU heatsink with a fan. Something similar to this.

Do you think an SST-90 could push 50 watts? Also, since it is not going into a flashlight, is there a voltage regulator thats adjustable and can do 50 watts? I was looking at this voltage regulator, but at 5 volts, how many amps would that draw?

 

[Th232]

Don't have that much experience with pushing the SST-90, but since (as you know) the issue is getting the heat away from the die, why not use the CST-90 instead? Junction to heatsink thermal resistance is nominally 0.9 deg C/W, much lower than the SST-90 (note that the front page value of 0.64 for the SST-90 is junction to case, mounted to a star the junction to heatsink thermal resistance is nominally 2.15 deg C/W).

Same die and dome, but much better thermal conductivity, specs allow it to be driven up to 13.5A.

Or you could also use the domeless variety (CBT-90) since you're going for absolute throw.

 

[Al Combs]

The regulator you picked has a fixed 5 volt output. Also if I'm reading it right, the minimum order quantity is 110. There is a variable output model in the Murata line, but it also requires external low ESR caps to stabilize ripple. If you liked EasyB's spotlight, the regulator line he picked requires no external caps. There is a 16 amp big brother version of the one he chose. It's a little more money (excluding the 110 thing), but they have it in stock at the moment.

I think Th232 is right about using the CST-90 instead of the SST-90. According to the spec's, a CST-90 N bin @ 13.5 amps puts out (3,408-4,089 lumens) vs (2,724-3,269 lumens) for an SST-90 N bin @ 9 amps. You might have to wait a while since Avnet only has the CST-90 in the 'L' bin (2,385-2,896 lumens @ 13.5 amps) at the moment. Avnet does have the top R bin CBT-90. But without auxiliary optics for the secondary fluorescence, the R bin puts out much less light at 13.5 amps (2,344-2,813 lumens) than the CST-90. Post #31 of that thread links to the site making the mirrors they use. Saabluster had a picture in post #168 of that thread with and without a reflection collar.

When they become available, the CST-90 could be thermal epoxied directly onto the CNPS9500. Or is that a 9700? Speaking of Zalman, you're not allowed to hotlink images like that. I think if you just say heat pipe CPU cooler, people will get the idea. BTW, I think that is an excellent idea. That was the first thing I thought of when I saw EasyB's spotlight post. With a CST-90, you'd still have a place to solder the wires without worrying about a short. Also the CST-90 has a thermistor to monitor the heat buildup. This is a good thread about using a voltage regulator instead of a constant current regulator. Pepko's post #19 in that thread mentions how he hooked the thermistor to the trim pin of the regulator. I guess he used trial and error to set that up. Perhaps you could PM him and ask.

As far as voltage, here is a jtr1962 post from the White LED lumen thread where he tested an early SST-90. There are 4 LED's in that post. He only goes up to 11.75 amps, but that should give you a good idea. It's only ball park of course. Since every LED is different, you're going to have to measure the one you're using. If you don't have a DMM that goes over 10 amps, here a 100 amp shunt for 24 bucks. This shunt is only 1 milliohm, so you can count on the results being almost the same when you remove it from the circuit. A good clamp meter would be better if you have or can borrow one. Another way would be to borrow a second regular DMM to measure the voltage at the LED directly while you measure the current with a second meter. It's not that the number isn't real, but the internal resistance of the ammeter's shunt is going to cause the voltage to the LED to rise when it's removed from the circuit. All you have to do then is turn down the voltage regulator until the LED meter reads the same as it did when you were measuring current.

Good luck and send pic's...:wave:

 

[ma_sha1]

Here is a SST-90 @ 15 Amp :devil:

 

[SmurfTacular]

why not use the CST-90 instead? Junction to heatsink thermal resistance is nominally 0.9 deg C/W, much lower than the SST-90 (note that the front page value of 0.64 for the SST-90 is junction to case, mounted to a star the junction to heatsink thermal resistance is nominally 2.15 deg C/W).

Same die and dome, but much better thermal conductivity, specs allow it to be driven up to 13.5A.

Or you could also use the domeless variety (CBT-90) since you're going for absolute throw.

If the CST-90 is better, than why doesn't everyone use the CST-90 instead of the SST-90?

Also, dont domeless LED's emmit less lumens than an LED with a dome? I even made a thread asking that here.

The regulator you picked has a fixed 5 volt output. Also if I'm reading it right, the minimum order quantity is 110. There is a variable output model in the Murata line, but it also requires external low ESR caps to stabilize ripple. If you liked EasyB's spotlight, the regulator line he picked requires no external caps. There is a 16 amp big brother version of the one he chose. It's a little more money (excluding the 110 thing), but they have it in stock at the moment.

I think Th232 is right about using the CST-90 instead of the SST-90. According to the spec's, a CST-90 N bin @ 13.5 amps puts out (3,408-4,089 lumens) vs (2,724-3,269 lumens) for an SST-90 N bin @ 9 amps. You might have to wait a while since Avnet only has the CST-90 in the 'L' bin (2,385-2,896 lumens @ 13.5 amps) at the moment. Avnet does have the top R bin CBT-90. But without auxiliary optics for the secondary fluorescence, the R bin puts out much less light at 13.5 amps (2,344-2,813 lumens) than the CST-90. Post #31 of that thread links to the site making the mirrors they use. Saabluster had a picture in post #168 of that thread with and without a reflection collar.

When they become available, the CST-90 could be thermal epoxied directly onto the CNPS9500. Or is that a 9700? Speaking of Zalman, you're not allowed to hotlink images like that. I think if you just say heat pipe CPU cooler, people will get the idea. BTW, I think that is an excellent idea. That was the first thing I thought of when I saw EasyB's spotlight post. With a CST-90, you'd still have a place to solder the wires without worrying about a short. Also the CST-90 has a thermistor to monitor the heat buildup. This is a good thread about using a voltage regulator instead of a constant current regulator. Pepko's post #19 in that thread mentions how he hooked the thermistor to the trim pin of the regulator. I guess he used trial and error to set that up. Perhaps you could PM him and ask.

As far as voltage, here is a jtr1962 post from the White LED lumen thread where he tested an early SST-90. There are 4 LED's in that post. He only goes up to 11.75 amps, but that should give you a good idea. It's only ball park of course. Since every LED is different, you're going to have to measure the one you're using. If you don't have a DMM that goes over 10 amps, here a 100 amp shunt for 24 bucks. This shunt is only 1 milliohm, so you can count on the results being almost the same when you remove it from the circuit. A good clamp meter would be better if you have or can borrow one. Another way would be to borrow a second regular DMM to measure the voltage at the LED directly while you measure the current with a second meter. It's not that the number isn't real, but the internal resistance of the ammeter's shunt is going to cause the voltage to the LED to rise when it's removed from the circuit. All you have to do then is turn down the voltage regulator until the LED meter reads the same as it did when you were measuring current.

Good luck and send pic's...:wave:

OK, so I already have the heatsink. Now you suggesting that I change the LED I use. I want to make heat transfering as maximum as posible. Would using a star board deminish the transfer? I think using the star would be alot easier. Tthe surface of the heatsink is not anodized, and as you said, it could short out.

And I just purchased a DMM from Harbor Freight for $20 that can measure 20 amps and also measure temperature.

 

[IMSabbel]

Do you think an SST-90 could push 50 watts? Also, since it is not going into a flashlight, is there a voltage regulator thats adjustable and can do 50 watts? I was looking at this voltage regulator, but at 5 volts, how many amps would that draw?

In the while lumen testing thread, an SST-90 was far from the end at the maximum current of 12.5 A or so the testing PSU could deliver.
And the cooling setup was worse than one of these CPU HSF assemblys.

50W should be possible with one, I am sure.

 

[Al Combs]

If the CST-90 is better, than why doesn't everyone use the CST-90 instead of the SST-90?

Also, dont domeless LED's emmit less lumens than an LED with a dome? I even made a thread asking that here.



OK, so I already have the heatsink. Now you suggesting that I change the LED I use. I want to make heat transfering as maximum as posible. Would using a star board deminish the transfer? I think using the star would be alot easier. Tthe surface of the heatsink is not anodized, and as you said, it could short out.

And I just purchased a DMM from Harbor Freight for $20 that can measure 20 amps and also measure temperature.

I'm not sure if you saw it or not based on your response to TH232, but the absence of the Lambertian dome makes a big difference in an LED's lumen output.

I didn't realize you had already bought the LED. Or did you? On page 12 of the SST-90 doc's, Luminus lists the junction to heatsink thermal resistance as 2.15°C/W. That's using a clad MCPCB aluminum star. If you try to push 50 watts through it, the junction will be 107.5°C above the heat sink temperature. With an absolute junction temp max of 150°C, your heatsink temp can never go above 42.5°C or 108.5°F. That's only luke warm.

Now this is where the CST-90's base being made of copper comes into play. It's junction to heatsink TR ratio is 0.92°C/W. So 50 watt with that LED will only leave the junction 46°C warmer than the heatsink. The heatsink could be hot enough to boil water and the LED will still be OK.

So whats wrong with an SST-90 only running at 9 amps. It will still be unbelievably bright.:cool:

 

[Al Combs]

And I just purchased a DMM from Harbor Freight for $20 that can measure 20 amps and also measure temperature.

How do you plan on compensating for the fact that if you adjust the output of your voltage regulator until your meter reads 9 amps or 13.5 amps, when you take the meter out of the circuit, the current will rise? Also there is the fact that as the LED's get hot, its forward voltage drops. If you looked at the voltage regulator thread I linked to, KillingTime said he abandoned his runtime test after 30 minutes because the drop in the LED's Vf triggered the 10 amp hiccup mode of his voltage regulator. His voltage regulator had been adjusted to 5 amps to drive an SST-50. A thermistor mounted next to the LED can help compensate for that. As far as an SST-90 and the aluminum star, I recently built a light using an idea I got from LED Zeppelin. It's surprisingly easy to reflow an LED onto a piece of copper using nothing but a stick lighter and some solder paste. It will get your SST-90 closer to the thermal resistance characteristics of a CST-90.:)

 

[MikeAusC]

Constant-current with temperature-limiting

Converting a Constant-Voltage supply to Constant-Current with Temperature-Limiting doesn't take a lot of components, provided the supply has a voltage-trim input.

Place a current sense resistor in the LEDs output, feed to an Opamp, then to the Trim input. A thermistor near the LED fed to another Opamp in the same package can reduce current once a maximum temperature is exceeded.

When RS finally deliver the LDO10C converter (which they said was in stock in Australia) I'll start building and testing and publish the details.

 

[Th232]

If the CST-90 is better, than why doesn't everyone use the CST-90 instead of the SST-90?

Also, dont domeless LED's emmit less lumens than an LED with a dome? I even made a thread asking that here.

If you check the datasheet, the CST-90 isn't used so much because it's so much bigger, 42 mm x 28 mm. That's longer than a D cell's diameter! And because the die isn't in the centre of the package, the light will need to be bigger still (I think a circle about 70 mm is required). If you want to get fancy, there's also a spot on the board designed for a thermistor, which might be useful given how hard you're pushing it. Also see Mike's response about using a thermistor to drop the current once the temperature is exceeded.

As for the domeless part, two things happen. Less lumens are emitted from the package, and the apparent die size decreases. We get more of a decrease in apparent die size than we get a drop in output, so the result is a greater surface brightness. When you put this through a lens (or several) you end up with a smaller collimated beam that is more intense (as Saabluster noted in your thread), i.e. greater throw at the expense of output in lumens.

As per your first post, you want maximum throw, hence my recommendation of the CBT-90. If you want greater efficiency, then the CST would be the way to go.