How to power up a DIY LED panel with a PC PSU?

[Sid55]

Thanks to anyone who likes to chime in and help out this newbie.

I bought some CC CV buck converters to power my DIY LED panel project using a perfectly working PC power supply unit that was collecting dust inside a fairly old and unused PC but unfortunately the buck converter does not output desired voltage to power up LED COB lights to their full potential.

My test set up:
* Load: 2 X 5 Watt LED COB lights (rated at 9 - 11V @ 400mA each) wired in parallel.
* DC power source: 12V (480 Watt max PC PSU)
* DC-DC 7-35V to 1.25-30V buck converter CC CV (15W max output, conversion efficiency @92%)

Test results under load when wired in parallel:
* Output voltage from buck converter drops to about 8.5V which is not adequate at all
* Output current from buck converter is just over 900mA which is more than adequate for both
* LED COB lights appear a lot dimmer compared to when powered with 11V battery pack without the buck converter

Questions:
* How to raise voltage output from the buck converter while limiting output current ?
* Should I use a much higher wattage CC CV boost converter instead ?
* Can any other arrangement be worked out ?

The ultimate goal is to power up the LED panel with 3 LED COB lights at 10-11V @400-450mA each

 

[Rive]

The usual cheap buck converter needs to have a minimal voltage drop. With 12V input the adjustable range of the output would be ~ 1.25 to 9V (this is ~ the usual range, I did not check the datasheet of the actual IC on the linked board).

12V -> 11V is not really good with a buck converter. Also, LEDs are not really meant to be driven with voltage sources. What I would do is to get a boost LED driver for 12V input and (adjustable) 400mA output, then connect the LED panels in series.

 

[berkeman]

I bought some CC CV buck converters to power my DIY LED panel project

What is a "CC CV" buck converter? It can be either Constant Current or Constant Voltage, but I don't know what is meant by mentioning both modes of operation. Is there a jumper that converts it between CC and CV operation?

* Load: 2 X 5 Watt LED COB lights (rated at 9 - 11V @ 400mA each) wired in parallel.
* DC power source: 12V (480 Watt max PC PSU)

A couple thoughts:

• You could just use two diodes in series to drop the 12V output of the PSU down into the input voltage range of the LED assembly. Use diodes that can pass the 800mA current (find ones rated at 2A, for example).
• Keep in mind that PC PSUs have a minimum output current requirement in order to stay in regulation (they can do that because they are always connected to a significant load in the PC application). They don't make good general purpose power supplies because of that minimum output current requirement. But if you can find what that minimum output current requirement is (either on a datasheet, or by experimenting with power resistor loads), then you can ballast up the output some if needed to ensure that your application draws enough current to keep the PSU in regulation.

Hope that helps.

 

[jim hardy]

PC supplies are rated for full power
but you can't get all that power from every output port.
The +5V is the strongest port.

My guess is your +12 output port is capable of only a fraction of an amp . What does the nameplate say ?
Did you measure it under load ? You haven't said.

A step up boost converter from +5 would probably be a better choice.

old jim

 

[Rive]

What is a "CC CV" buck converter?

Apparently it is about an adjustable current limit.
It seems to be a common 'ali' circuit, I could found it in several different online shops. example

 

[berkeman]

Apparently it is about an adjustable current limit.
It seems to be a common 'ali' circuit, I could found it in several different online shops. example

Hmm, I wonder how they are implementing an adjustable current limit. The LM2596 is a pretty vanilla buck converter:

http://www.ti.com/lit/ds/symlink/lm2596.pdf

 

[Sid55]

Hmm, I wonder how they are implementing an adjustable current limit. The LM2596 is a pretty vanilla buck converter:

http://www.ti.com/lit/ds/symlink/lm2596.pdf

Its current limiter and voltage adjustment functions seem to work fairly well within buck converter's specifications and as expected voltage dropped at higher current output.

 

[Sid55]

What does the nameplate say ?
Did you measure it under load ? You haven't said.

*correction*
It's 500 Watt max not 480 Watt max I incorrectly stated in my first post.

My guess is your +12 output port is capable of only a fraction of an amp .

You are most likely correct here.
Dialing back to 800mA gave me 9.7 volts max (pic below).
It never went over 8.5 volts at 900mA plus.

 

[Sid55]

Apparently it is about an adjustable current limit.
It seems to be a common 'ali' circuit, I could found it in several different online shops. example

Yes.
constant current & constant voltage

 

[berkeman]

Dialing back to 800mA gave me 9.7 volts max (pic below).
It never went over 8.5 volts at 900mA plus.

That power supply looks to be plenty beefy (if I'm reading the label correctly). I'd suspect the problem with the minimum output current spec not being met. See if you can find the datasheet for it online, to check that specification. The unfortunate thing might be that it takes a 100W load to ballast the power supply enough to bring the rails into regulation...

 

[jim hardy]

Holy Cow ! Did you use the +12V1 (34 amps capacity) or the -12V (1 amp) ?

 

[rbelli1]

You may need to load one or more of the other rails to get the PSU to stay on properly. Some of them will shut down or otherwise go sideways if they don't see an expected load.

BoB

 

[Tom.G]

This from:
ATX / ATX12V Power Supply Design Guide Version 1.2
by Intel, 2000
pg.22

3.4.3. No-load Operation
No damage or hazardous condition should occur with all the DC output connectors
disconnected from the load. The power supply may latch into the shutdown state.

From your description,

rated at 9 - 11V @ 400mA

the LED assemblies probably have a current controller buit into them that operate correctly only between 9-11V.

For the easiest and most effective solution, I would go with the suggestion that @berkeman made, a couple series diodes:

You could just use two diodes in series to drop the 12V output of the PSU down into the input voltage range of the LED assembly. Use diodes that can pass the 800mA current (find ones rated at 2A, for example).

Cheers,
Tom

p.s. please keep us updated on success/failure with this.

 

[davenn]

Thanks to anyone who likes to chime in and help out this newbie.

No one has mentioned that you cannot just use a PC PSU on it's own outside a PC without a little mod
Without the mod, the PSU will not fire up correctly. This is a trick that many of us have been doing for many years
to use a PC PSU's as a bench-top PSU.

The ATX connectors, the 20 and 24 pin main connectors versions

Firstly ... the PSU requires a “Power-ON” 0V signal to start up

Secondly ... for the PSU to correctly regulate the +5V output voltage it needs to have a load attached, ~ 5-8W to
make the PSU think its attached to a motherboard ... An appropriate wattage rated resistor ... use Ohms law to work out the value

The second item isn't required IF you are not going to use the 5V rail

For the “Power-ON” 0V signal the usual trick is to ground that pin to a 0V rail line ( black).
Personally I have never directly grounded it, rather I have used a resistor in series ... anything from 100 E to 1k ... a few 100 Ohms is idealRegards
Dave

 

[Rive]

Dialing back to 800mA gave me 9.7 volts max (pic below).
It never went over 8.5 volts at 900mA plus.

The minimal voltage drop depends on the current. Especially when you have a current sense resistor in series. So these values alone are not spectacularly wrong as I see it.
What is your actual Vin for the circuit? Could you please give us a V_in, V_out, I_out list for some current values?

Again, you would do better with using a 12V_in boost LED driver with the LED panels connected in series. This task is just not for a cheap buck with 12V input.

Hmm, I wonder how they are implementing an adjustable current limit.

Well, this guy seems to be on a similar track and even the components seems to match (up to ~80%)
Ps.: this one seems to be the actual schematics, from https://www.minikits.com.au/LM2596-PSU-02As I see blaming the PSU in this case is just a dead end. Especially without measurements backing it up.

 

[essenmein]

Test results under load when wired in parallel:
* Output voltage from buck converter drops to about 8.5V which is not adequate at all
* Output current from buck converter is just over 900mA which is more than adequate for both
* LED COB lights appear a lot dimmer compared to when powered with 11V battery pack without the buck converter

Your answer is here. Seems to me when you connect a supply that controls for current you get the desired current (900mA), when you connect those same lights to a battery pack, nominal 11V, in reality likely higher, you get many more of the lumens, I'd put money on it being that the battery pack is pushing a lot more than 900mA when you direct connect your LED to it.

 

[Sid55]

The minimal voltage drop depends on the current. Especially when you have a current sense resistor in series. So these values alone are not spectacularly wrong as I see it.
What is your actual Vin for the circuit? Could you please give us a Vin, Vout, Iout list for some current values?

Some new results after this modified set up

Set 1
* 2 X 5 Watt LED COB lights (rated at 9 - 11V @ 400mA each) wired in parallel
* DC-DC 7-35V to 1.25-30V buck converter CC CV
* DC power source: 12V from ATX 24 pin connector of PSU

Set 2
* 2 X 5 Watt LED COB lights (rated at 9 - 11V @ 400mA each) wired in parallel (exact copies of set 1)
* DC-DC 7-35V to 1.25-30V buck converter CC CV (an exact copy of set 1)
* DC power source: 12V from one of peripheral molex connectors of the same PSUTest results under load: Both Set 1 and Set 2 were lit up at the same time
Set 1 input voltage: 11.53V
Set 1 output voltage: 8.7V
Set 1 output current (set on buck converter) : 790mA

Set 2 input voltage: 11.53V
Set 2 output voltage: 8.5V
Set 2 output current (set on buck converter) : 790mA

* Under this modified set up each individual LED COB ran noticeably and satisfyingly brighter compared to the previous set up mentioned in my previous posts
* On the down side, all LED COBs ran considerably hotter compared to the previous set up
* Similarly both buck converters ran noticeably hotter compared to the previous set up
* No idea why buck converters ran noticeably hotter with this set up

 

[Sid55]

Your answer is here. Seems to me when you connect a supply that controls for current you get the desired current (900mA), when you connect those same lights to a battery pack, nominal 11V, in reality likely higher, you get many more of the lumens, I'd put money on it being that the battery pack is pushing a lot more than 900mA when you direct connect your LED to it.

Would have loved to do the battery pack set up as LED COBs were amazingly bright but without any current limiter my batteries were not ideal. They were pushing several times more current than rated current of LED COBs.

 

[trurle]

* Output voltage from buck converter drops to about 8.5V which is not adequate at all
* Output current from buck converter is just over 900mA which is more than adequate for both
* LED COB lights appear a lot dimmer compared to when powered with 11V battery pack without the buck converter
The ultimate goal is to power up the LED panel with 3 LED COB lights at 10-11V @400-450mA each

Your LEDs seems to be overheated by at least 40 degrees over design specifications (as evidenced by over-nominal current and under-nominal voltage). Pay close attention to to thermal contact to radiator, using thermal grease if necessary.

Also, it is not clear if you have ballast resistors in your LED packages (likely you do not have). With the high currents you have, connecting wires parasitic resistance (if larger than 0.5 Ohm) may be enough for currents equalization though. 1 Ohm, 0.25W resistor in series with each LED would bring you on safe side.

Finally, for 3 LEDs, you need at least 0.04m2 of radiator area (if both sides of radiator are free-convection cooled) to avoid burns if both sides of radiator are free-convection cooled.

 

[Sid55]

Firstly ... the PSU requires a “Power-ON” 0V signal to start up

It pretty much did the job.

I have no idea how much load to put on this 12V 34A rail.
Hoping that load of LEDs would bring it into regulation if I draw power to light up 10 X 5 Watt LEDs all at once as I have noticed people putting 50Watt dummy load to bring PSU into regulation.

 

[jim hardy]

...

I have no idea how much load to put on this 12V 34A rail.

a cheap and easy dummy load for 12 volts is car lamps. 1157 taillight bulb has two filaments. One draws around 0.6 amps, the other a little over 2 amps.
Lamp and socket will cost just a few bucks at an auto parts store, probably free at a junkyard. .

http://automotivemileposts.com/taillightbulbcomparison.html

 

[Sid55]

Also, it is not clear if you have ballast resistors in your LED packages (likely you do not have).

Those LED COBs have none.

With the high currents you have, connecting wires parasitic resistance (if larger than 0.5 Ohm) may be enough for currents equalization though. 1 Ohm, 0.25W resistor in series with each LED would bring you on safe side.

This could be the reason why my second setup generated a lot more heat while the first set up barely did.

My previous setup had very long wires connecting to those LEDs but I cut 2/3 of their initial lengths for the second set up. I did not measure connecting wires parasitic resistance for either set up.

Thanks for pointing out this issue.

 

[Sid55]

a cheap and easy dummy load for 12 volts is car lamps. 1157 taillight bulb has two filaments. One draws around 0.6 amps, the other a little over 2 amps.

Wow!
24 Watts of dummy load right there.

It would be super nice if I could get a light bulb of similar load that also emits same color temperature or something close to 6000K. I see some opportunity of using all these lights for photography and video work indoors in future. 5000K - 6000K light sources are great for creating neutral daylight color temperature for photography and video work.

 

[Tom.G]

...a light bulb of similar load that also emits same color temperature or something close to 6000K.

You might have a bit of trouble with that. The Tungsten in an incandescent lamp turns liquid at 3695K, might work if you put it in a pressure vessel. The highest melting point element is Carbon when it is in the form of Diamond, 4713K... however that is at a pressure of 12.4Gp or 124 000 Atmospheres. That would need a rather strong 'bulb'.

It looks like you will need one of:

• Filter in front of an incandescent lamp
• Gas discharge
• Electric arc
• Fluorescent lamp that comes 'sorta close'
• Sunlight

I did run across a compound supposedly good to 4400K, but as of 2015 it was hadn't actually been created yet.
https://www.asminternational.org/web/cmdnetwork/home/-/journal_content/56/10180/25655039/NEWS

Oh well, at least it was an interesting search.

Cheers,
Tom

 

[jim hardy]

I think 6000K is asking a lot of incandescent lamps. EDIT as @Tom.G said (we crossed in the E-wind again)
From https://www.lumens.com/how-tos-and-advice/kelvin-color-temperature.html

Take a look at Phillips LED automotive fog lamps.

http://download.p4c.philips.com/files/7/78719719/78719719_pss_aenus.pdf?locales=,en_us

old jim

 

[Rive]

Set 1 input voltage: 11.53V
Set 1 output voltage: 8.7V
Set 1 output current (set on buck converter) : 790mA

If your setup is on CC mode then it is just OK. With voltage drop at 2.8V it is around the minimal value for this kind of cheap DCDC module, but since it could deliver the ~800mA it is acceptable. I think your problem will be the LEDs themselves:

5 Watt LED COB lights (rated at 9 - 11V @ 400mA

For me this says it needs CC 400mA and at the rated current it will have 9-11V => 3.6 to 4.4W. There is no 5W in them as I see. Guess it was cheap, right?

You might try add come cooling (mounting them on a flat metal surface and providing some air current might be enough) and overdrive them a bit to get 5W, but I still recommend to use a 12V input boost LED driver and connect them in series.

 

[Sid55]

For me this says it needs CC 400mA and at the rated current it will have 9-11V => 3.6 to 4.4W. There is no 5W in them as I see.

I agree. They seem to output 4 - 4.2 Watts for an extended operation with acceptable levels of heat generated on LED mounting boards. They can be pushed past 6 Watts and those mounting boards heat up to high levels within minutes, but I saw no pointing in doing that.

but I still recommend to use a 12V input boost LED driver and connect them in series.

Would you care to explain the difference between wiring LED COBs in parallel vs wiring them in series ?
Like boost converters supplying a much higher voltage to LED COBs wired in series ?

You might try add come cooling (mounting them on a flat metal surface and providing some air current might be enough)

This 8-inch x 24-inch x .025-inch aluminum sheet metal could be one of ideas for using it as a heat sink and also as a light reflector after it has been cut to a slightly smaller size.

 

[Rive]

Would you care to explain the difference between wiring LED COBs in parallel vs wiring them in series ?

Right now you have a voltage drop (2.8V) which is barely fits the step-down converter working limits. If you have to modify the circuit later, then there is a good chance that you get some issues. For example you have to replace one of the modules, and you get one with the higher voltage limit (11V). That will be out of the limit of your DCDC, it will never work at the specified power and it will push more current on the other panel in parallel causing it to run at higher power.

With a step-up driver you start at 17V: the voltage gap is >5V. Drivers are available up to 50+V. You can easily add more panels as you want and with connecting the panels in series you don't have to bother with the current balance since they will get the same current anyway.

Regarding the heatsink: don't forget the thermal grease.