Constant Voltage Thermal Runaway Test #2 – High Current

In my previous test, I ran a couple of Vero 18 COBs at their typical rated current to see if I could get either of them to go into thermal runaway. Over the 4 hours I tested them, each of the COBs ended up pulling an additional ~50mA each than what they started at, but stabilized at this level. Curious to see if I could get one of my COBs to go thermal, I decided to take another stab at this and hit a couple different chips with considerably higher currents than they’d normally be run at.

Here are my results:

Test 1 – Vero 18 Starting at 1,800mA

The data sheet for these gen. 7 Vero 18s says their typical current is 1,170mA, and their max is 2,340mA, so I wanted to pick something above typical but not too close to max. 1800mA is close to halfway between these 2 points so I figured that’d be a good place to start. The data sheet showed that I could expect this current somewhere around 36.5V, so I set voltage accordingly.

Setting driver output voltage to 36.5V.

The COB pulled 1,828mA off the bat.

After 40 minutes, the COB was pulling an additional 25mA. It stabilized at this point and maintained this current over 4 hours.

I was again surprised at how stable this setup was. Even at a relatively high current for this model of Vero, it didn’t run away on me. The active heat sinks are doing their job excellently.

Test 2 – Cree CXB3590 Starting at 2,400mA

Having messed with the Vero 18s a couple times, I figured I’d switch to a CXB3590 and throw some real power at it. This COB is on a passive heat sink, so it lacks the extra cooling that the fans I’m running on the Vero heat sinks provide.

I adjusted voltage to start this test at 2,400mA, which is listed as the test current for this COB at its typical voltage. Again, hardly anyone runs at currents this high with these chips – most often, 1750mA is the highest growers go for CXB3590s.

I saw a rise of 270mA in half an hour… finally, some results!

The COB was puling 315mA more than what it started at after an hour of run time.

After 2 hours, the COB had stabilized at about 2,700mA. It remained around this current for the next couple hours.

In this case, while the COB did pull quite a bit more current over time (300+mA), it DID stabilize and refused to pull any additional current after that point. Time to blast it with even more.

Test 3 – Cree CXB3590 Starting at 3,000mA

The absolute maximum current rating for this COB is 3,600mA. After seeing how the CXB acted at 2,400mA, I expected a few hundred mA of rise in current for this test and set the voltage on the driver to produce 3,000mA of current to start. This would give me enough room to let it rise and keep it under its max.

Starting at 3A.

In 10 minutes, this thing had shot up almost 300mA.

I gave it another 5 minutes, and shut it down. Current was climbing very quickly and showed no signs of stopping.

Well, this did the trick. Apparently, 3 amps is enough to get one of these guys to go thermal and quite quickly. I’m glad I was able to get it to happen, but it took a lot of current to do, and I doubt anyone in their right mind would be feeding their shiny new COBs this much power.

Test 4 – Cree CXB3590 Starting at 3,000mA with Active Cooling

When I was finishing up the 3rd test, I turned on a fan I have in the room and aimed it at the heat sink of the COB just to see what effect, if any, it’d have on the rapidly rising current. I noticed right away that current actually started to drop, and figured I should reattempt a test with 3,000mA+, but with active cooling on the heat sink.

I rigged up the fan to blow directly over the heat sink. This is probably a lot more air than most people would have for an actively cooled heat sink, but I wanted to see if it would be enough to stop the COBs from running away again.

Setting up the fan to blow over the heat sink.

Another attempt at 3,000mA+ begins.

Only a few mA up after 30 minutes. Nice!

Holding steady after an hour.

And the exact same draw after 2 hours.



So, what I’ve gathered from this is that yes, thermal runaway can definitely happen when using a constant voltage driver. However, chances are, if you have proper cooling and run at a relatively low drive current, you shouldn’t have a problem with it. It was not until I boosted current way up near the COB’s max that I ran into problems, and even when I did have a chip run away on me at a really high current, I was able to stop the process just by improving my cooling.

The more I test these constant voltage drivers, the more I like them for this application.


  1. Fantastic just as I figured. 100% needed this to bad rollitup down 🙁

    • Yeah, what’s up with that? I read their hardware had probably died… shitty.

    • LEDGardener

      February 20, 2017 at 7:51 pm

      Yeah, what’s up with that? I read their hardware died… shitty.

      • Wonder what the new spot is. Btw im the one that you were talking to via email about the rsp-1500 bout a month again. Thats why these tests are prime and prove what we were talking about. Now the iceled ultras i ordered are perfect to maintain temp and stop the thermal runaway.

        • LEDGardener

          February 20, 2017 at 10:14 pm

          Yeah, I recognized the name right away! Those ultras should work nicely and I don’t expect you’ll have any problems. Looking forward to seeing your build come together.

          • Its a toss up right now i may actually Do a din rail system with there 480s and 960’s. Im looking at Phoenix contact power supplies they seem really good the voltage range on them are really nice.


          • LEDGardener

            February 21, 2017 at 11:39 am

            That’s a sweet PSU and yeah, that’s quite the voltage adjustment range. I work with din rail and rack gear for my day job and I think you could do a really clean install either way. Holy shit though, is that 20A unit ever expensive here in Canada though.

  2. nice detailed post as always =] this is something i never took into account up to now, thanks!

  3. Good stuff man, so basically the only way you will have a failure is if your cooling actively and you have a fan failure, and if that happens you will probably loose a chip no matter how your powering it. With the avaibility of big used heat sinks out there I see no reason to not go passive, just takes the worry out of it and your not drawing down you efficiency with fans and wall worts, unless you have space limitations that is. I’m definately going to start looking into building a second light soon just not sure what chips I’m going with yet might step up to the 50v 1812’s but for a few more % efficancy I don’t think they are worth it untill you get into the clu58 that they are claiming are up to 60% but then your talking 44 bucks a cob same as a 3590, and from the charts the bigger chips lumens per watt are lower for some reason. Still working on talking the wife into letting me set up a tent though before I can have a place to put a second light.

    • Yeah I agree that passive is the way to go. I guess this just showed that in some high-current installs, active cooling might be required to keep current levels in check if passive is not dissipating enough heat.

      I think those big chips shine in high power applications. They’re likely much more efficient at 100W and up than smaller ones that are way better at low current. I’ll have to go over the 058 datasheet to confirm this – I haven’t looked too closely at it yet.

      A tent might be a tough sell… there’s nothing that screams “I’m growing weed in here” like a big black tent. Are you planning on trying different veggies with your potential new space, or just more of the same that you already grow?

  4. eBay man so much cheaper for the Phoenix ones

  5. Someone is selling the 10a in Canada right now 😎

  6. Very nice tests Gardener. I am about to build a parallel COB setup and the thing which first bluffed me was the result of connecting a single 3590 (active cooled) to a MW 120h-36B (only current adjustment possible). Led was pulling 78W, which means less than ~2A and it didnt go higher, which I first expected (would be 3.4A max for this 100% open PSU). I then understood that it hit 36V (I hadnt a multimeter atm) and wont go higher. So going paralell on 36V PSU should not harm the 3590 leds (as far as you have proper cooling for 2A) even if PSU has plenty of current – it was news for me ;). I will check differencies/ thermal runaway between 2 paralleled 3590 leds next weekend, as well as a Citi clu48 setup (4 leds on 150h-36b).

    PS could you pls write an article what are the cool options for what to grow? I think about chili, but I would be happy to hear about other wide accepted and somewhat cool plants.

    • LEDGardener

      February 21, 2017 at 9:41 pm

      Good stuff! Did you say only current was adjustable on that first driver? That’s odd. Please share your results when you test your CXB and CLU chips next weekend!

      I’ll have to look into finding out what the cool kids are growing. I think it’d be neat to try to grow a pineapple or something similar. I’ll see what I can dig up 🙂

      • Sorry, I could’t post my results since I went quick and dirty with available gear and mounted 2x clu048 and 1x cxb3590 on one radiator and am driving them with one PSU. So I don’t expect any thermal runaway on this setup. Still have to find out how they share the current (Cree seems to be significantly brighter) but I don’t promise any deadlines this time 😉
        The only thing I measured is that the cree (if driven alone) hits 78w on a 36V Mean Well (6W is the minimum) and with added citizens the PSU maxes out at 156W, so the citis are doing their job too.

        • LEDGardener

          March 15, 2017 at 9:24 pm

          Nice! I look forward to hearing how they’re splitting it up. I wouldn’t be surprised if the CXB is drawing considerably more current. than the CLUs at that voltage.

  7. Yes, the Mean Well HLG-xxxH-xxB version can be user adjusted with the current only. But… the voltage regulates itself (below available max voltage) to the value needed for this current. The xxB value is the max voltage of the PSU, pretty much the same as for xxA version, but for B it is the real max, which you cant slightly up; check the datasheets for the options (these are voltage values between 20 and 54v and aproppiate max current for given power xxxH of the PSU).
    The real advantage of B series are the 3 dimming options – external potentiometer or 1-10v input voltage or PWM. So you are able to dimm as you want, e.g. 30c external pot placed where you want or an arduino.

    As for plants – would be nice to read here what are the popular tasty/healthy/cool ( chose one or more) options, preferrably the ones which are not available in every grocery for cheap. I’m personally much better at reading datasheets than looking for plants and I see you have already this experience and could maybe share on your site.

    • LEDGardener

      February 22, 2017 at 6:59 am

      Oh of course – I’m just dumb. I missed the “B” part of it and I was thinking you were talking about an “A” with only a current pot for some reason.

Leave a Reply. For help with your own build, please make a thread in the Forum

© 2017 LED Gardener

Theme by Anders NorenUp ↑