FAQ Time! But first, don't forget that after your first two months of free Skillshare trial goodness it's LESS than $10 per month if you go with an annual plan, which is like a good cup of coffee per week but instead you're building on your lifelong knowledge. Enjoy! skl.sh/diyperks2 Now, question and answer time! The first think I'll say is that there's been a surprising spread of discussion about various aspects of this build in the comments. It's great to see! I actually learn a lot from reading comments that you guys leave, and there are some serious golden nuggets of info here and there which I appreciate a lot for building on my own knowledge. I will first address a few common misunderstandings though. Firstly, some folk are assuming that the LED is 90%+ efficient, which would mean that it would only output 150w as heat and the rest as light. This isn't the case unfortunately, as even some of the best LEDs are only 40% efficient MAXIMUM. This figure changes depending on the colour accuracy, with more colour accurate LEDs being less efficient as they have to emit more of the visible spectrum. Let's assume the LED is roughly 30% efficient (which is probably fairly accurate, and may even be a bit generous) that means that 450w is being emitted as light, and the other 1050w is being wasted as heat. So, I built my cooling system around the assumption of needing to dump 1000w of heat into the air, which is why I had four large radiators and made an extreme cooling block for it. LEDs last longer when they're cooler, so it's super important. I also want the final unit to run quietly, not with maximum fan speed, so this extra headroom allows me to do that. Another thing that crops up is talk of corrosion. In water loops it's very important to use similar metals, otherwise you run in to galvanic corrosion where one metal will literally steal electrons from the other, causing the other one to corrode quickly as they're 'connected through the water'. Brass and copper are similar enough, but the lead in the solder may be susceptible to corrosion. However, adding an anti-corrosive additive to the liquid should solve this in the final build. The radiators are copper, not aluminium, so no worries there. As for the PSU's failing... there have been many theories but nothing solid. They are 12v PSUs, and are rated for 83A on that lane. When in series, they give 24v at 83A... 83*24=1992w. The reason for my wanting a higher voltage is that it literally halves the amps on the voltage booster, meaning they can run easily rather than at their absolute maximum ratings. Hit me up with any more questions! Anyway, I hope that clears things up. If you still see these questions popping up please relay the answers contained here for me please :)
Where do you write your ideas down? How do you come up with your ideas for your videos? I always strive to find new ways to stay creative and keep the ideas generating! Cheers!
Your power was off. Psu have fix power outpot. For eg. 12v at 83A that comes to 996W per unit. In series you get 24v but only half of the amps. (996/24V). If you had them in paralel mode you could combine there output power. I presume that booster need at least 6-36V input to work and that psu units are around 1000w.
If you're not aware, computer PSUs have an "80+" efficiency rating, where server PSU's are typically much more efficient than normal ATX desktop PSU's, however this is actually a curve, depending on loading, with maximum efficiency at around 60% load, which drops off significantly towards their maximum rating. It could be that at their limits, the cooling was not adequate to remove all the excess heat and the least efficient component failed.
@@diophantine1598 led diodes can not match suns output but pulsate certain spectra and those pulsations are almost like laser pulses. So an led light is unhealthy in comparison to the sun
"Houston? Yeah that strange light is back again! I can see it right there in England. Can you confirm there is no nuklear war going on down there? Yeah? What about an alien invasion? Now it's blinking! It's shining right against us. Ahhrgh I'm blind. Houston the aliens are attacking us." No seriously. I bet you can see this thing from space...
@@johntheux9238 You have to look up for the photo-electric-effect. UV is much more powerful than IR, red, green, or blue light. Never heard of that Vampires dislike heat or temperature in general. Pure light was also never an issue. UV-C literally burns you.
4.2M subs and multiple ads and this surprises you? This is like the home shopping channel on cable TV, resurrected and with the serial numbers filed off... there are boatloads of better options.
From someone who works in a data center: Server PSUs never run in series. They run in n+1 redundancy. So they aren’t very good at working together at the same time. They also get a good bit of cooling from the fans within a server chassis as well as their own so it was probably inadequate cooling. They also never push their full wattage capacity for extended periods of time or reach their full capacity ever really.
I am not familiar with this PSU, but some are split into multiple rails of 12v and 5v, with each rail possibly having a lower amp rating. This looks like a Supermicro PSU and looking up the specs on a few different models, they only list a single rail. Here an example: www.serverschmiede.com/images/product_images/thumbnail/supermicro-1000w-psu-netzteil-pws-1k02a-1r-9994-m-2.jpg
yea, could the fans only be there for when the server is in standby mode and the server fans are off? then when the server is on and there is already a few mm of H2O of pressure supplied by the case pressure, then the fans it has are able to supply the full required airflow?
@@SuperBrainAK Doubt it. We have Supermicro devices (if they are, as Johnny pointed out), and the fans are on all the time. Servers are rarely in a standby state and when they are, fans still sit around 50-60% I'd say. It's hard to say the exact speed because we don't have access to case metrics since my job is mostly hardware replacements like sticks of ram, hard drives and PSUs. But the fans are definitely on all the time. I think this was just a case of them being pushed to full capacity. Even a consumer grade 750 watt PSU is not rated to hold 750 watts for an extended period of time so I can say that neither are server PSUs. Optimal range is approximately 80% capacity. These 1000W PSUs are probably meant to output 800W steady with a surge wattage of 1000 during startup. They certainly weren't designed to hold 1500W continuously.
mikimomo97 ah, you should be able to load the PSU to 100% because that is what it’s rating is, for PC building you always put a bit bigger PSU in do you get better efficiency but you can still run them at max power.
I think the PSUs failed because of inrush currents, since converters ususally have a ton of inrush current. Turning them all on at once is a bad idea. (If you have done that) If you want some references, check out Great Scotts Video on inrush current and soft starters. If you want some help designing your own psu, let me know, I am a hobbyist who does a lot of high power stuff (no degree yet, but I am working on that).
I don't think so. Half way decent PSUs should have current limiting and soft start build in. If you look at a server board, it has a lot of capacitance distributed all over (CPU+GPU power management). The problem was probably connecting them in series. It would be better to go with 3 boost converters per PSU and get a 24 V PSU or use 3 PSU with 2 boost converts.
power supplies died because the total wattage is a sum of multiple rails(voltages), you got all the power from a single rail. Power rail limits should be printed on the power supply back
@John -- I work for a County in Wisconsin and occassionally do work for the Sheriff Dept. I asked a Deputy once about the LED strobe lights on the cars. He said, "If you think driving by one stopped at an accident site is bad, try standing next to it for four hours."
I’m always amazed these COBs don’t rip themselves apart with the thermal expansion changes of substrate vs. semiconductor. A lot of material science must have gone into their R&D!
@@Silkroadgermany2 Simply as stated. When materials heat up they often expand differentially. If bound together this can cause stress often leading to fractures in either material. COBs are a layered structure of semiconductor and various mounting substrates that get very hot, so a prime candidate for physical failure if not implemented correctly. I hope that helps.
@@kipling1957They use ceramic and gold wire which can handle a lot of current without heating up. The gold wire is less then 0.002mm and can easily handle 5+A
I've got a few 10w ceramic cobs made by Cree which fell off their heatsink and kept running for a few minutes and did not fail from thermal expansion. The phosphor cracked and it got slightly less efficient than its peers, but it still works just fine, every single die.
Water cooling a power supply would be extremely dangerous as power supply’s have large capacitors that if touched and the connectors bridge it could kill
I don't think that would work though because even though they could even be attached to the same plug the neutral and earth's on the input side would be separate and could cause some issues with power stability. especially if he is using a multi rail power supply. still worth a try though? :)
@@Viddiecent the led also works when only one of the circuits is powered as you can see in the video. I presume that a little instability can be handled by the led as well
Make sure if you're doing water cooling to check if you're not mixing copper, brass or nickel plated copper parts with aluminium parts as this will cause much faster corrosion of the metals.
You can mix brass and copper without issues. Copper is in brass. Copper and aluminum mixing in a loop isn't a good idea unless you run anti-freeze (pre-mix or concentrate you mix with distilled water.)
If you watch their latest video with glitter, you can see they have a similar LED for that reason. I wonder how it compares. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-BtYKDamqo2I.html
The server psu has been designed with a metric ton of static airflow through the server chassis, therefore i recommend you switch that up. It might not holdup all to well otherwise.
@@38911bytefree When a power supply is listed as isolated, it is normally referring to input to output isolation. This isolation is typically achieved by the use of a transformer. A non-isolated power supply is generally using a type of chip conversion. When a multi-output power supply has isolation between the outputs, it means there is an electronic barrier between outputs and not tied to a common ground.
To get double the Current, you need to parallel the power supplies. Wiring them in series gives you double the voltage. You might have stressed the current limitations of you power supplies.
Yes, this is what I came here to say. Really, he need four PSUs: two in series for the voltage, and then those in parallel for the wattage. Then the boosters inefficiency has to be taken into account, which is usually 80% or so.
He wants double the voltage, not the current. These supplies when run on 240V (he's in the UK) can output 82A on the 12V rail. Put two in series and you get 82A at 24V. Even if the converters were only 80% efficient this would be (just) adequate. Instead I think the issue is that he's only taking the 12V output from one side of the 12V pads on the PSU interface circuit board. You have to connect to both sides to get the rated amperage.
@@paulgrimshaw6301 If the amps stayed the same, then you would be doubling your wattage. If these are 1000W power supplies, they stay 1000W when put is series. If you had 82A at 24V, you would have 2,000W (again, these are 1,000W). Instead, you would have 41.67A before taking the 80% efficiency into account from the boosters. Remember to use Ohm's law when calculating these things.
He doubled the voltage on the output of the second power supply in the line while still drawing the full amp load, this is way out of spec. If you want to run them in serries you can only pull half the rated wattage. He needed 4 power supplies, two sets of 2 power supplies in series linked in parallel in order to stay in spec at the component level for the second in line PSU. Thinking you could run double the rated wattage thought the output of the second PSU was naïve in the extreme. Those must have been good quality server PSUs to have managed to perform so far over spec for the first run.
"power supply fans are really starting to ramp up" *can still hear himself think* yeah, those server power supply fans haven't ramped up yet buddy: you can still hear yourself think
Doesn't seem like he's quite familiar with server noise levels. Those PSUs are made for outputting a constant 24/7 full rated load, so of course the fans ramp early - but they really didn't sound like they were anywhere near maximum speed either.
@@Kraaketaer The fact that it was just the 24V rails he was using when the whole tandem build was rated at 1500W and the thing didn't explode immediately was impressive.
The whole time I'm watching this, I trying to think of what kind of task it's designed for. The best I could come up with is "stadium light". So I was like "Oooh!", when you said it was for stage lighting. Some of my sillier ideas, was for holding a beach party at night, a tanning booth, and a full sized easy bake oven. (For when you want to allow your kids to be able to cook a full sized thanksgiving turkey with a lightbulb powered toy oven.)
I think you missed something here. The total wattage of the power supplies you used is 2000W however that number is the combined load of 12v, 5v, and 3, Rails. As you are only using the 12v rail to power the LED. You need to use power supplies that have at least 1000w each on the 12v rails alone. I think this is the reason why your one failed. I hope this helps, best of luck!
Was just about to comment something similar to this. He would likely be better using using a transformer from a microwave or similar. Something that can push the wattage needed at a consistent level and use heavy duty resistors to lower the voltage. I think the cooling solution is over kill too.
@@ThatTaytoGuy yeah the cooling solution was pointless. He could have done it with a single rad easily. I think he forgot that the water block he made is basically a mini radiator also lol.
On top of that he was using the older used power supplies, and as most people that mess with those things know, their power output degrades as they age.
Using air based heatingsinks is slightly problematic. Very often your cooling water starts to accumulats "stuff" from the inside of the radiators and pipings. This can start to block the small fins and cause some channels not to get enough fluid throughput and cause localized overheating. Removing some of the fins should solve this. Check your lamp with a terminal camera for hotspots. (This effect takes month/years to manifest)
6:53 Dude, if you want to know if something is watertight, just check if it's airtight first. If you just block one end, and an adaptor to a standard bike pump on the other end, you can pumpit up with the bike pump and watch to see if the pressure gauge on the bike pump goes down. It it doesn't, it's water tight. If it does, crank the pressure and listen for leaks. Use some soapy water in a spray bottle to detect leaks by looking for bubbles when it does. It's a MUCH cleaner way of checking for water-tightness.
@@LEO-xo9cz It really wouldn't. You'd only need two extra fittings. 1. An end cap. 2. A 8MM to whatever side he was using adaptor. That's all. Easy peasy.
Was guy on dutch pc forum 10yrs go who place truck radiator out side of his house to cool his OC game pc. Car and truck rad are super cheap a pc rad can easy be 100 euro
Aluminum and copper in the same loop isn't good if you want to keep using it for long. I suppose you could use automotive coolant and hope the anti corrosion additives can prevent it.
@@Jaker788 I highly doubt he bought 4x480mm copper radiators for this system, he was already talking about cost of copper in making the watercooling block
A few things to check: - A PC power supply may be rated for 1000w, but typically cannot provide all of that on the 12v output. Verify the watts the power supply can output on the 12v rail. eg: 12v max 50amps is only 600w on the 12v rail. - A PC only starts running once the PWR_OK signal from the supply goes high. This lets the power supply stabilize before the PC starts drawing full current. You should start your dc converters via a relay connected to this signal, so they don't draw current too soon. - Consider starting the dc converters one after another. They have several large capacitors on each of them, which typically means they will draw heaps more current for a split second when starting up. Starting them sequentially will decrease the maximum load on the supply. If you have an oscilliscope you can measure the startup time. It might be on the order of milliseconds. - Not all power supplies can be stacked in series the way you have done due to the way they are designed. An unscientific way to test this is to hook an independent load to each supply and measure the current. Then tie the supplies together while having the same independent load on each. If the current draw from the wall changes, the supplies cannot handle it. It is not guaranteed safe, but is a good first test. Given that your setup was only drawing 1600 watts it seems like these supplies were working okay in this configuration.
These are server power supplies, they only have 2 power rails, 12v and 3.3v for only standby use. So its entire rated power is available on the 12v rail. And yes how you wire things in series needs to be looked at carefully. I have a feeling they were old power supplies (used from a server that was running 24/7 for at least 3 years) and all the thermal paste had dried up, so the heatsink could not cool the mosfets well enough anymore.
You need to specify when you say the pwr_ok signal goes high, to justify whether or not a relay on its own is suitable as it may need complimentary circuitry to drive it, like a transistor.
For a server grade power supply full load shouldn’t be an issue even at 12v. I’m guessing the main issue was going from 0 to 1650w instantly and the fans not ramping up fast enough to cope with the heat thus overheating. Or he didn’t give them enough time to cool down after the first run and before the second.
As a designer of custom COB LED Light Panels I hope to sell soon, I appreciate the same exact challenges (although not quite at the scale you're on) in learning to power and cool your LED creation. Only working with a 3-watt 10V COB, the very first time I finished soldering wires to this postage-size beast, when I powered it up - UGH - I thought I'd go blind. I wasn't quite expecting so much light intensity from such a tiny source. For the next 20 minutes, I kept seeing dark spots before my eyes akin to staring at the sun too long. I cannot fathom a 1000-watt light so BE CAREFUL!
Thoughts regarding the PSU: As far as I know PSUs can have multiple 12V rails. They can't deliver all of the power on a single rail. Could this be the problem?
This could really well be the problem. Not qualified in any means, but I think usually computer psu's use the different rails to keep the different voltages as accurate as possible (all voltages can be "run" through one rail because of some electrical knickknacks(this is usually found in cheap psu's))
@@lukasstadler6594 good quality psu's can be run with higher wattage (good components) than promised and it probably would work but its different matter if it's good to do so and how long you do so. Computer psu's have to be able to sustain the advertised wattage, so I don't think the peak power is an issue.
@@rasmusjauri as far as my knowledge goes, PSUs have different rails that are all rated for different amounts of power they can supply. on normal consumer PSUs there is even a big sticker on the side that tells you which rails can deliver what amount of Amps and at what Voltage they do that. so on a 1000W PSU your 5Vrail may be able to push 100W, 12VRail can support most of the power, probably in the neighborhood of 600-700W, and so on and so forth. a good PSU also splits the demanding rails(mostly 12V) in two, so one rail can only supply 300-350W. so even it these where good PSUs, they were working far out of spec and had to die. a 1000W PSU just can't push all of that through one rail
@@TheDraggerPlay that's the case in your pc psu. Server psu is different. The 12v rail has almost 100% of the rated power. And just to clarify a little more, in domestic pc, the psu has various 12v (for ex.) cables, and you might think that's from different rails, but no, all come from the same rail.
True and It's a matter of time when well see him working on the so called "Tritium" to create a fusion reaction this will in theory according to Spider-man 2 give him *the power of the sun* 😅😆
Please turn this into a fake-sunlight "window", with a parabolic mirror and a slightly-blue transparent window board (which you can make yourself with resin). Maybe one of those can be used for a whole house, instead of for just a room.
That would be a real concern, if he was using metal tubing to circulate the water. With rubber hose, he has some dielectric isolation between the aluminum radiators, and the brass/copper heat sink. There is, however, some concern for galvanization between the copper, and the solder, though, my whole house uses soldered copper plumbing.(lead free solder) and have not had any joints fail in 50 years.
@@dash8brj Even if that's the case, using non-metallic tubing between the radiators and heatsink, offers dielectric isolation. He probably won't have too much trouble with that. If he's concerned, automotive antifreeze offers protection against corrosion and galvanization.
@@dash8brj "Brass is nonreactive." Same with copper. Problem is the tin and other alloying agents in the brass, and considering the fins are copper... Ye no. He needs a sacrificial anode.
@@Dumdumshum Fittings for water cooling stuff are all brass, and are used with copper heatsinks and radiators, and are not a problem. No need for any sacrificial anythings. The only issue here is potentially the solder in the DIY block.
Each power supply has rails for different pc components, they all have their own rating of power; manufactureres quote the peak combined power of the PSU so that could be why they failed.
while thats true, the 12v rail is by a long shot the main rail, power output is probably not why they failed, my guess would be that those psu's aren't floating which means they cant be used in series
@@cj09beira Yep the 12th rail is the bigg'un by far. Also a good PSU can still put out a decent bit more than its rating without blowing up....mainly just devouring it's life expectancy.
Correct. These "rails" offer 3.3v, 5v, 12v, and -12v. All of these voltages will have a seperate current capacity, but the power supply rating will combine the max wattage of all rails. Look into the specs of the power supplies you plan to use and add up the wattage of the specific voltage you need.
Yeah, wiring DC switchmode power supplies in parallel or series is not a good idea. One of them will end up supplying the majority of the load, and when it overloads, invariably the other ends up being overloaded and they both fail.
Series is fine; by definition they will have the same amount of current going through them. Parallel is a different story, and switchmode supplies tend to not like switched loads, so I think that may be the issue. But from about 300 hours of run time, my series server power supply setup has worked fine (but I'm not maxing it out, either).
I don't know if any LCD would withstand this brightness as any LCD screen will heat up so hot and will break instantly. Only possible image source for this led is to use DLP which requires complex engineering!
@@rocifier that'll cause either just frosting issue which will produce blurred and distorted video, or will crack due to huge temperature difference. Your best bet if you really want to use LCD is to water cool it with absolutely zero bubbles and a custom transparent cooling block. Even with those things, the picture created will have warped or distorted image. Remember: you are cooling the picture producing part of the LCD. altering with it even with slightest bit may lower the image quality projected. That's why lcd projectors aren't as powerful as DLP projectors.
This led is emiting probably 40%(*) of power in light and 60% in heat so you dont need 1500W loop to cool it *Edit. Efficacy of this led is really low (80lm/W) so more like 20% power is emited in light
There are single cobs out with over 200lm/w around 100w peak, but I think it is too expensive and difficult to build big led arrays like that, good binnings are probably not frequent in the selection process. It is way cheaper to produce lower power output leds with like 0,5w with 250lm/w. But I don’t really see a purpose for such a high power led (1,5kW), the small surface area makes it more difficult to cool it, the lifespan is probably also reduced and the price is insane I think?
@@dalivanwyngarden3204 He says the purpose in the video. Let's say you're on a movie set and you need to make it look like it's midday except it's night. This will cast shadows the same way the sun will while offering the same sort of output.
I think I know why the PSU went out. Assuming 92% efficiency, there were at least 100 Watts of waste heat. The fans were going at full tilt and you just killed the power to them, so it just sat there cooking with 0 active cooling. Instead, you should have disconnected the LED from the PSU and let it slowly reach a safe temperature and then power it off. That's just my 2 cents.
He probably should've pulled the led but you definitely got your numbers wrong, remember these are rated for 1500W which means at 92% efficiency they would waste 120W, but they were running at half power so that's 60W of waste heat for both units and remember that these are server grade so it's safe to assume they're a little better made and efficient than even high-end consumer desktop PSUs, they should've barely been warm, they must've had some power back fed or those boost converters doing something to upset the PSU or it's also possible it was from having multiple mains ground points and they became capacitively coupled.
vgamesx1 Not wrong numbers. I said at least 100 Watts. No, they are likely only 80+ gold-rated. So it might even be only 90% at *peak* effeciency. And server PSUs are meant to have 6000 RPM 50mm fans blowing on them like a cat 5 hurricane. So that was way out of usecase.
Hey Perks Nice video! But I have a question about the Server Power Supplies. The thing is that in my knowledge that you can't connect SMPS Power Supplies in series. And here is the reason, the whole outer Heat dicipassion Surface is grounded and one more thing is that I saw some PC power supplies uses this. Now in this way if you series the two power supplies then you will get 24v no issues but there will be higher possibility of the damage of the 2nd series connected Power supply. (If you run the power supply long time.) I have a great experience about this because I have damaged my several power supplies. 😢. So from my perspective I will highly suggest you to check wheather any of the mains wire is heating up or not. I have a highly doubt about that. If you have time then let me know what you think 😊 As always nice video 👌 Love from India 🥳 Regards, Creative Creator
".....whole outer Heat dicipassion Surface is grounded " Do you mean Earthed? The problem is with cheap SMPS which SHOULD BE isolated but have a crappy EMF suppression capacitor which causes leakage which causes problems.
As a heads up, there are 2 possible issues that I can see. 1 - the boost converters, used to convert 24v to 30v. These tend to draw lots and lots of current and while the overall power may be within the specification, if these boost converters are trying to pull too much current, it may cause problems. 2 - if, as you stated, you were using 2 power supplies in series, this may cause issues depending on the type of power supply. Switch-mode power supplies (which are the majority of all modern power supplies) convert power from 120 / 240 volts mains electricity to another specified voltage by switching on and off rapidly. If one power supply is "on" and while the other turns "off", the second power supply may in fact suck up power. Some power supplies (but not all) have a "synch" input which allows them synchronize their on/off phases. However, it is not super common and you will have to see if your specific power supply will support it.
Yeah that particular converter has a low input voltage cutoff which can be set. if I try to power mine from a 24v supply, it wants to turn on when the supply gets to 10v and trips the overcurrent protection every single time. I have to set the low voltage cutoff to like 23v, then it waits to turn on until the power supply is fully turned on and doesn't violate the current limits. (the low input voltage cutoff control is right next to the boost converter input terminals, I don't see him adjust it in this video. The default setting out of the box is the lowest setting intended for 12v battery input. The failure also could have avoided if he turned turned their output voltage down before turning the power supply on and then slowly brought it up one at a time like he did the first time.
I just bought a few led torches and was impressed with the little 100 lumen keyring torches and then I stumble on this. The internet is burning out my brain, in an enjoyable way, mostly, sort of.