DiY LEDs - How to Power Them

Mellodrama

Well-Known Member
Supra!

Your latest creation, the one with a mix of COB's and stars. Are 3 drivers running this thing? One for each of the COB's and one for the star string?
 

SupraSPL

Well-Known Member
You guessed it, the red/blue string is running at 650mA on the cheapo $6 driver and the CXA3070s are running on the cheapo $10 880mA drivers. Total dissipation on the heatsink ended up 86W.
 

SupraSPL

Well-Known Member
Ouch. I see my mistakes. Cree has a minimum efficiency for a certain generation of tech before the part is scrapped, so if you pick the bottom bin from two different parts they'll have the same efficiency. I understand the thermal and mass production advantages of COB but what is it that makes it more efficient when Tj is constant?

Don't think I'm a COB hater! I just recall you or Flux talking about how the XML-2 is a bit more efficient but not worth the expense or effort. I couldn't find the post then naively did the calculations and ran with it.
It is all good I love when people question and shake things up. I always learn from it and it helps me find mistakes I have made. I would not be surprised if I had said that about XML2 on the Mr Flux CXA analysis thread or somewhere before I realized my own mistake. If I recall (my memory is terrible), I made a quick study of the CXA1820 and wrote them off. I also thought the Vero was better than the CXA3050 until I ran the numbers of the CXA3070 and realized that is the best growing LED so far (IMO).


Regarding thermal resistance, I am suggesting that since the CXA package handles each watt of heat heat better than the XML2, that at a given drive current the CXA should be at a lower Tj than the XML2 (assuming they both had the same amount of heatsink/W and the same ambient temp.) That decrease in Tj should offset or more than offset the slightly higher current droop.


For example, when Cree upgraded the XPE to the XPE2 ceramic, the XPE red gained 10C for each dissipation watt while the XPE2 red gained 5C for each dissipation watt.
 

SupraSPL

Well-Known Member
I have been doing some testing to try and determine the maximum current I can run before a Tj of 50C is literally impossible, due to the ambient temp in a grow and thermal resistance. In other words there is a point where no matter how good your cooling is, the Tj will rise above 50C. So far it seems like that limit could be about .9A. Luckily I just got a great deal on an Amprobe that has a thermocouple so I can confirm my other test.

So if Tj is rising above 50C, the current droop teams up with the temp droop to reduce photon output to the point that there is no economical gain to run them harder (cheaper to just buy more and run them soft). If this turns out to be true, I will limit my CXAs drive current to about 1A.

DSC07069.jpg
 

Abiqua

Well-Known Member
Just got a chance to test this wall wort style 2A 9V switching adapter. It is $6 shipped, powers at least 8 .3A fans quietly and 75% efficient across its entire range (68% with 1 fan).
Supra, I was wondering if you could answer a few question, as I am fugging around with fans tonight too.....

#1 How are you calculating P-in [power in]
#2 I hooked my multimeter up a to a non-powered circuit 2nite and was showing .47 volts with power just from the multimeter, is this a short issue, do I need to check continuity?

#3 I am confused about calculating efficiency, even though you probably have gone over it twice....:) I am running a 9v 600ma wall wart and running about 600mA of fans on them. Closest I got was 10.15 volts and that was after switching out 120mm for another......but like I said, even before the circuit was powered I had .47 volts reading on the multi, so that raises a red flag for me, but I can't seem to search for the right terms!!!!

But....I did run a 5v 1a phone charger I had as well, and with about 500mA load on it, the multi showed 4.96 volts, however again, before being powered the circuit did show .17 volts. This is on my 20v switch by the way....

Maybe you have a suggestion?

View attachment 3030208View attachment 3030209
 

Observe & Report

Well-Known Member
Regarding thermal resistance, I am suggesting that since the CXA package handles each watt of heat heat better than the XML2, that at a given drive current the CXA should be at a lower Tj than the XML2 (assuming they both had the same amount of heatsink/W and the same ambient temp.) That decrease in Tj should offset or more than offset the slightly higher current droop.
In the Vero sheets they say they test them with a 10ms pulse at 25C. I assume they do that so they can measure them without thermal solution biasing the results. Cree probably does the same, no? Without a long enough pulse to heat up the junction, the COB still wins, why?
 

SupraSPL

Well-Known Member
Abiqua good questions. To get a reasonably accurate input power reading, I use the Ensupra power meter (no affiliation). I have a kill a watt P4400 also but mine doesnt measure tenths of watts (from what I have seen the 4460 does not either) and the Ensupra does. Also the Ensupra is more accurate when dealing with small power loads. I have seen a power meter for sale at Walmart for $16 but I have not tested it. Belkin has a power meter that has .1W resolution (Belkin insight) but I have heard it does not do well with small loads. The Ensupra seems to have availability problems. Here is one I have not heard of that does have a .1W resolution and is $17 Weanas Plug Power Meter

On the DC side, you need a measurement for volts (wire in parallel and mind where you tap in to the circuit, it does matter) and also measure amps by wiring in series (tap in anywhere in the circuit will give the same reading). Multiply your amps and volts to get your dissipation power in Watts.

The final step is to divide dissipation power by input power and you get a percentage of efficiency. This works exactly the same way for fan power supplies and LED drivers.

If your voltmeter is measuring phantom voltage have no worries it is normal. However, many cheap voltmeters are inaccurate because high quality resistors must be used to get good accuracy and they are somewhat expensive. I highly recommend Amprobe as an affordable but accurate meter. You can see reviews of voltmeters on youtube check EEVblog videos.
 

SupraSPL

Well-Known Member
You are correct O&R, pulse helps avoid that problem. Thermal resistance of the XML2 is 2.5C/W and so far I have tried to estimate the thermal resistance of the CXA3070 I am getting about 1.2C/W. Hopefully my thermocouple can help verify that.
 

Abiqua

Well-Known Member
Abiqua good questions. To get a reasonably accurate input power reading, I use the Ensupra power meter (no affiliation). I have a kill a watt P4400 also but mine doesnt measure tenths of watts (from what I have seen the 4460 does not either) and the Ensupra does. Also the Ensupra is more accurate when dealing with small power loads. I have seen a power meter for sale at Walmart for $16 but I have not tested it. Belkin has a power meter that has .1W resolution (Belkin insight) but I have heard it does not do well with small loads. The Ensupra seems to have availability problems. Here is one I have not heard of that does have a .1W resolution and is $17 Weanas Plug Power Meter

On the DC side, you need a measurement for volts (wire in parallel and mind where you tap in to the circuit, it does matter) and also measure amps by wiring in series (tap in anywhere in the circuit will give the same reading). Multiply your amps and volts to get your dissipation power in Watts.

The final step is to divide dissipation power by input power and you get a percentage of efficiency. This works exactly the same way for fan power supplies and LED drivers.

If your voltmeter is measuring phantom voltage have no worries it is normal. However, many cheap voltmeters are inaccurate because high quality resistors must be used to get good accuracy and they are somewhat expensive. I highly recommend Amprobe as an affordable but accurate meter. You can see reviews of voltmeters on youtube check EEVblog videos.
Again thank you for the answers.....and of course for measuring efficiency a Kill A Watt is what I need, brain fart on my part.

The Amprobe is more than affordable and what sucks is that my old Solex stopped working, so I went out to Radio Hack and bought a small meter. Ended up paying $30 while the Amprobe is $40 on Amazon, ouch! Decisions decisions.

Have you tested that particular 9v wall wart @ half load, just to see what the voltage was? The reason I ask is, that I checked another set of fans I am using in my small 65w led cab.

Two fans @.25 amps being powered by a 12v 500ma wart. I am getting almost 14v's at the junction but I do have rheostats wired inline and the fans are only seeing about 9-10w's.

It just seems that wall warts aren't regulated very well especially below full load, which is to be expected, but I am even a little more dismayed.....:peace:
[again....this could be the meter, 8%+- tolerance is a bit much for these cheap meters....]
 

SupraSPL

Well-Known Member
Haha I have a similar radio shack meter, they forced me to buy it for an energy class (in which we learned very very little sadly). I am sure they overcharged us and the one I have does not measure amps. You are right the amprobe can be had very cheaply. I got my 520 for about $35 shipped.

Here are the results of the 9V 2A $6 switching wall wort:
1 fan - .252A - 9.2V - 68%
2 fans - .46A - 8.9V - 73%
4 fans - .846A - 8.74V - 75.4%
6 fans - 1.14A - 8.43V - 73.3%
8 fans - 1.425A - 8.31V - 74%
10 fans - it burst into flame - jk I don't have enough pigtails on hand for 10 fans

You are right on both counts. The wall worts are not regulated (and the vast vast majority of them are only 50% efficient).Unfortunately I had a hard time getting good readings with my radio shack voltmeter and I think it was mostly due to the leads. Now I can't find it to compare against the fluke and amprobe.
 

guod

Well-Known Member
You are correct O&R, pulse helps avoid that problem. Thermal resistance of the XML2 is 2.5C/W and so far I have tried to estimate the thermal resistance of the CXA3070 I am getting about 1.2C/W. Hopefully my thermocouple can help verify that.
a 70+Watt emitter with an thermal resistance of 1.2C/W is a bad joke
the thermal resistance must stay below 0.4C/W for the CXA3070.

Typical Thermal resistance for Bridgelux Vero 29 is 0.26 C/W
 

SupraSPL

Well-Known Member
Thanks for your input guod. That number would be the total thermal resistance of the path with invalidates my xml2 vs cxa comparison but even more in favor of the CXA. Unfortunately after all sorts of testing I am at the conclusion that I cannot be confident in either of my methods of Tj estimation. At 880mA (31W dissipation) my Tc measures 29C. I will have to work with that for now.

Kicking on the cooling fan knocked my Tc down 7.5 degrees C but it increased my vF by .18, which I would assume means a decrease in Tj of 18C. But that doesnt make sense with a Tc change of 7.5C so I am confounded :confused:

According to the data sheet a Tc of 55c = Tj 85C but that is the only info we have to work with.
 

Abiqua

Well-Known Member
Haha I have a similar radio shack meter, they forced me to buy it for an energy class (in which we learned very very little sadly). I am sure they overcharged us and the one I have does not measure amps. You are right the amprobe can be had very cheaply. I got my 520 for about $35 shipped.

Here are the results of the 9V 2A $6 switching wall wort:
1 fan - .252A - 9.2V - 68%
2 fans - .46A - 8.9V - 73%
4 fans - .846A - 8.74V - 75.4%
6 fans - 1.14A - 8.43V - 73.3%
8 fans - 1.425A - 8.31V - 74%
Really consistent voltage wise, we always want better efficiency, but that is not bad at all!

10 fans - it burst into flame - jk I don't have enough pigtails on hand for 10 fans
ha ha
You are right on both counts. The wall worts are not regulated (and the vast vast majority of them are only 50% efficient).Unfortunately I had a hard time getting good readings with my radio shack voltmeter and I think it was mostly due to the leads. Now I can't find it to compare against the fluke and amprobe.
Yes, they can be hit or miss, but I probably spoke to soon......just after I tested another wart [12v@1amp] I hit 11.96 volts with 1100mA of load and 12.01 volts @ 600mA of load. This is probably the one I will be using for a bit.

Please keep us updated on the power supply readings, if you do more, imho they are invaluable. Hopefully I can start to afford picking a few up here in the near future

Ciao..
 

PetFlora

Well-Known Member
I know, right? An outboard PS would reduce the heat, while improving the heat sinks ability to defuse the cob heat


That's similar to my build using Veros. 2x 3000k, 2x 3500k, and 1x 5000k. Active cooling.

ETA: Not sure of his choice to stick the drivers on top of the sinks.
 
Top