Powering/Driving a Stadium type of light

[frippe75]

Hi!

A first post in an all-over-the-place type of fashion. Sorry for that. Maybe more of a mission statement than anything else.

Looking to design a stadium type light (portable/mast/battery driven).
Have been playing around using Cree CXB3590 COBS but discovered harnessing all of the lumens from a large COB is challenging (reflector).
I need a beam closer to 10-15 degrees but is getting closer to 20-25 with my 3D printed reflectors and spill light at 45-ish degrees.

So. thinking more along the lines of using 3030 or 5050 (metric) LEDS in an array formation.

First thought was to design a 5x5 light and then multiply that design until I reached my lumen goal (which is not really well known at this stage).

Here comes the first question.

I arranged the LEDs in strips of 5. Forward voltage is 3.2 giving me a total of 16V.
Batteries are "12V" LiFEpo4 types of lets say 20Ah per piece. Thinking I can then arrange the batteries for the best way I see fit.

Length of cable from battery to light will be in the range of 12 meter so using a higher voltage than those 16V is probably a good idea to keep down on cable diameter....

But also using roughly ~26Volts will probably lower total system effectiveness using a buck converter to do DC-DC from 26Volts->16Volts.
Then using constant current drivers or current limiting resistors needs to be factored in.

What I just realized is that choosing to do 5x5 is a bad idea just because the lens is 5x5. 25 LEDS are really hard to divide into something suitable.
Maybe better to do 20x5 (100 per lamp/light). That gives me four lenses but easier to divide up the LED "strips".

10x would be 32Volts. Maybe there is a pre-made driver for that?
As long as it divides up those 100 LEDS. And I don't want to go too high on the voltage resulting in a system considered high voltage (>50 Volts I guess)

Any input here?
Cheers!

 

[JustAnOldFashionedLEDGuy]

Commercial or hobby?

 

[frippe75]

Pure hobby project! But... A serious one
Looked at purchasing pre-made lights at a 200.000 lumen level with a runtime of 2 hours. Super expensive and virtually impossible to mount those lights on a home-made and portable mast. So ... I started a new hobby. Watched Youtube and all those worlds-brightest-flashlight channels and simply got interested in what I could improve upon. One thing lead to another and now many hours spent thinking and testing this is where I ended up.

Not sure in the end this will be cheaper than buying all ready existing products but... China made probably means crappy LED chips and much lower lumens than advertised or no real thermal management so they will brake quickly.

Buying better engineered products super expensive... So took the "fun route" of DIY ...

 

[LEDphile]

Start by figuring out what sort of performance you actually need - if you actually need a beam angle under 20 degrees, you're probably looking at high power LED packages, or big reflectors for CoB packages (looks like a 6" reflector will get you a beam angle in the 10-15 degree range woth an 18mm LES CoB). Both of those approaches will have implications for your fixture size, although high power LEDs should give you better power density with less thermal hotspotting (at likely higher LED/optic cost).

For what you're trying to do, you're going to find that there are many tradeoffs to be considered, and that there's a reason the commercial products cost what they do. Understanding what your needs are will help guide you through the tradeoffs to what is hopefully an acceptable solution for your application.

 

[frippe75]

Start by figuring out what sort of performance you actually need - if you actually need a beam angle under 20 degrees, you're probably looking at high power LED packages, or big reflectors for CoB packages (looks like a 6" reflector will get you a beam angle in the 10-15 degree range woth an 18mm LES CoB). Both of those approaches will have implications for your fixture size, although high power LEDs should give you better power density with less thermal hotspotting (at likely higher LED/optic cost).

For what you're trying to do, you're going to find that there are many tradeoffs to be considered, and that there's a reason the commercial products cost what they do.

Tradeoffs have appeared and I'm trying to sort them out. The matrix is simply overwhelming. I do understand that overall efficiency will have a price penelty.

But current iteration is using 3030 LEDs from OSRAM @ 184 lumens/watt. The highest lumen bin has forward voltage of 3.2 @ 700mA (~430 lumen). Spec on lens is 10-13 degree and size is 75x75mm.

My main issue now is how to power them efficiently and how to decide on how many LEDs go into each lamp.
Using narrow beam angle would benefit from building a larger amount of lights/fixtures to distribute the light where it's needed.
Since its battery operated I guess I cant afford to spill light.

So my first objective is trying to balance efficiency of driving the light/fixture against be able to model the light distribution.
Have created a simple 3D model in Blender but its not accurate enough to give me a final answer... I got a tip to look at Dialux but haven't had the time. Probably should invest a few hours in that direction.

Since I'm 3D printing a few parts I'm trying to think in terms of modules so I can reuse them in a new iteration.

Haven't built the LEDs version yet since I'm waiting for PCB/parts and lens. But dug into the world of Fluid Dynamics and modeled my 100 LEDs version to see if I could handle the heat generated, hoping to keep them under 50 degrees.. Total lamp weight now is less than a kilo but electronics for driving it is not included since I'm stuck on what to do here. Weight is top-three priority wise (not a fixed list yet)

This is the CXB3590 (LES = 30mm) driven at 7watt against the roof. 4.5" reflector. Larger reflectors is not an option since the total surface area will make it too sensitive to wind when number of lights are scaled up. Reflector = 102 cm2 and Lens 81 cm2 inc housing. Reflector 25% larger. 6" reflector will have 124% larger surface area than the lens one.

Spot @ 26 degree and Spill @ ~50 degree

So I have paused the COB one for now. Figuring out how to best drive my lens version is current obstacle
And efficiency is key. Not sure if I can reach above 90% ?

 

[frippe75]

Just found this advise in respect to driving LEDs cheap.

Also look up the "Poor Man Regulator" option- basically you use LEDs to drop the voltage and then put a linear regulator on the last one. The linear regulator keeps the current fixed and dumps the excess voltage as heat...

But I would have to be really close in voltage on that last one not to wated too much as heat?'

Maybe not much better or cheaper than using a current limiting resistor with proper watt rating?

 

[LEDphile]

You do not want to treat LEDs as a constant-voltage load, which is what that linear regulator based approach is doing (assuming a typical 3-terminal linear regulator or LDO). The voltage varies with the LED bin, temperature, drive current, and possibly also LED age. A current-limiting resistor is a better choice if you don't want to use a proper LED driver (but you really should use a proper constant-current driver, as they are simple, inexpensive, and readily available), as it allows the current to adjust to match the voltage, instead of trying to force a voltage and hoping that the current doesn't run away.

And note that if your LED string voltages are relatively low, the minimum dropout voltage across a basic linear constant-current driver may result in lower efficiency than you like.

 

[des-con]

I agree with LEDphile, a constant current design is the way to go. Linear is easiest to design, but a switching regulator will run cooler. I use an old Bose 1801 heat sink piece to cool the substrate on a 15000 lumen constantly on light and have tried both switching and linear. Ebay, Aliexpress or Banggood have smaller, lower power versions for cheap to play with. Some are more efficient than others...

 

[frippe75]

Ok so a CC-driver is probably not avoidable. Then simply think about how to best arrange those 100 LEDs into strings and simply tune the parameters of the CC-driver to match. @des-con, that heatsink looks like a beast in terms of weight and I'm really looking for lightweight ;-) solution.

I'm trying out water-cooling.
Here a model with a 2mm sheet of Alu and water cooling block center-mounted. 29-44 degrees after a run time of 300 sec.
Flowrate of 1.5lit/min @ 15 degrees. The model does not account for the 2 degree per watt increase from the solderjoint in the datasheet as well as thermal resistance from the Arctic silver paste on two layers but that's probably negligible.

 

[des-con]

Nice model. You are approaching a "high bay lighting" application. Note that you could also use 5 groups of 20 LEDs each. Depending on which LEDyou select that could require a source voltage of over 65-75VDC at 60mA to 150mA (5630). While it might be tempting to choose 20 groups of 5 LEDs each, and to modularize the electronics, complexity does impose size and cost concerns, and each individual regulator will suffer some efficiency loss. You may specify the circuit board with an aluminum core to reduce heat and make assembly easier. If you do go to the lower voltage you might start with a source voltage of 20VDC. Current setpoint can be adjusted during testing with sense resistor and gain settings. If you have a micro in the system linearity compensation becomes easy.

 

[DIWdiver]

Just off the top of my head...

Low efficiency means more waste heat, means more heatsink, means more weight.
Low efficiency means more batteries, means more weight.

So a high priority on weight translates to a somewhat high priority on efficiency.

As a rule of thumb, the best efficiency comes from a high source voltage, with buck regulation at the load. However if you use inefficient buck regulators because they are cheap, it might be possible to match or even best them with linear regulators. However, this would require the battery and LED voltage to be pretty well matched.

For example, let's say you stacked 4 of your '12V' batteries to get 48V. Then let's say you took your 5x5 array and dropped one LED from the middle, then wired it as 12S2P. That would give you a 38.4V LED array (nominally). Your actual battery voltage would be 3.2x16 = 51.2V, and your efficiency would be 38.4/51.2 = 75%. You should be able to do significantly better with buck regulators.

But let's say you managed to get strings of 15 LEDs. That would give you a 48V load, and 93.8% efficiency. That would be hard to beat in a dirt simple regulator. You'd want a low-overhead regulator because at end of discharge the battery is going to drop to 48V or less. Interestingly, the theoretical efficiency rises to 100% at this point. While this is not achievable in practice, it is possible to get pretty close.

 

[DIWdiver]

Another reason to go higher voltage is to reduce power loss in your wiring, and/or wire size. Doubling the voltage from 12V to 24V (assuming the same power level) cuts the current in half, thus reducing the power loss in wiring by 75%.

We don't know exactly the output you are looking for, but you did mention 200k lumens. At 183 lm/W that's over 1000W. At 50V that's over 20A. You mentioned 12m cable length, so 12x2=24m of wire. I'm going to call it 75 ft since that's what I'm used to working with. 14 AWG (2mm^2) wire would handle the current for this application, and it comes in at 400 ft/ohm. 75/400 = 0.1875 ohms. I^2*R = 75W. So that would be a 7.5% loss before you even get to the regulator. Running at 24V would quadruple that. You'd really need quite heavy wire to make the system feasible at 24V. I would think you'd want 8mm^2 (8 AWG) at least.

 

[frippe75]

linearity compensation

That was an interesting topic! Wasn't planning on have a mcu for that reason but need to understand exactly whats non-linear. Not sure if you meant that available input voltage will vary (which is will due to battery discharge) or that current draw could vary due to temperature. vf is around 3.2V * 20 = 64 V ... Don't think I will get above 48V getting me into the realms of an high voltage application as well.

Appreciate the input!!

 

[frippe75]

Just off the top of my head...

Low efficiency means more waste heat, means more heatsink, means more weight.
Low efficiency means more batteries, means more weight.

So a high priority on weight translates to a somewhat high priority on efficiency.

As a rule of thumb, the best efficiency comes from a high source voltage, with buck regulation at the load. However if you use inefficient buck regulators because they are cheap, it might be possible to match or even best them with linear regulators. However, this would require the battery and LED voltage to be pretty well matched.

For example, let's say you stacked 4 of your '12V' batteries to get 48V. Then let's say you took your 5x5 array and dropped one LED from the middle, then wired it as 12S2P. That would give you a 38.4V LED array (nominally). Your actual battery voltage would be 3.2x16 = 51.2V, and your efficiency would be 38.4/51.2 = 75%. You should be able to do significantly better with buck regulators.

But let's say you managed to get strings of 15 LEDs. That would give you a 48V load, and 93.8% efficiency. That would be hard to beat in a dirt simple regulator. You'd want a low-overhead regulator because at end of discharge the battery is going to drop to 48V or less. Interestingly, the theoretical efficiency rises to 100% at this point. While this is not achievable in practice, it is possible to get pretty close.

Interesting! Heatsinks are minimized since it's water-cooled. But efficiency is truly important to limit batteries (for cost reasons mostly and long runtime). They will not be up in the mast.

Stacking 4 batteries will cause a larger voltage span during battery discharge between 46.4 - 50.4 volt maybe??
The trick of ditching the center led could be useful! The only thing would be that my 5x1 LED PCB would have to change.
And yes. Placing the batteries on the ground creates the need for long and quite thick cables. A lot of parameters to balance around.

Created the tables to see number of leds and available forward voltage groups and battery voltage levels.
But I only created rows with #leds that results in a even number when divided against a total of 100 leds.

 

[frippe75]

But since I'm still waiting for parts I took some time to test out water cooling on the Cree COB.
Ran it a 130 Watt and it maxed out at 52.5 degrees on the LED. Highest temp was recorded on the 3D printed collimator.
Probably an indication the spray paint is not reflective enough. And being printed in PETG it does not get any benefit from the water cooling.

One thing I note from this is that it differs from my FD model which is about 10 degree lower.
Not sure if I'm too theoretical here. I will look at the spec of my thermal compound and at it in to the model. Don't have arctic silver at home yet so I'm using a lower quality brand.

 

[frippe75]

Another reason to go higher voltage is to reduce power loss in your wiring, and/or wire size. Doubling the voltage from 12V to 24V (assuming the same power level) cuts the current in half, thus reducing the power loss in wiring by 75%.

We don't know exactly the output you are looking for, but you did mention 200k lumens. At 183 lm/W that's over 1000W. At 50V that's over 20A. You mentioned 12m cable length, so 12x2=24m of wire. I'm going to call it 75 ft since that's what I'm used to working with. 14 AWG (2mm^2) wire would handle the current for this application, and it comes in at 400 ft/ohm. 75/400 = 0.1875 ohms. I^2*R = 75W. So that would be a 7.5% loss before you even get to the regulator. Running at 24V would quadruple that. You'd really need quite heavy wire to make the system feasible at 24V. I would think you'd want 8mm^2 (8 AWG) at least.

Ahh.. That was an important catch @DIWdiver! I had calculated the wire and embarrassingly enough I only calculated one-way :-(
It would have been nice to avoid having the batteries on the ground for weight reasons. But I guess that wire will amount to some weight as well. The mast I'm looking at is a telescope-type of mast so it's hard to get the batteries mid-mast.

But if I got into the battery-industry as well I could potentially place the batteries in a different format and get them into the mast. But that's where I draw the line

 

[LEDphile]

But since I'm still waiting for parts I took some time to test out water cooling on the Cree COB.
Ran it a 130 Watt and it maxed out at 52.5 degrees on the LED. Highest temp was recorded on the 3D printed collimator.
Probably an indication the spray paint is not reflective enough. And being printed in PETG it does not get any benefit from the water cooling.

One thing I note from this is that it differs from my FD model which is about 10 degree lower.
Not sure if I'm too theoretical here. I will look at the spec of my thermal compound and at it in to the model. Don't have arctic silver at home yet so I'm using a lower quality brand.

View attachment 17508

Thermal imaging of illuminated LEDs is very difficult to get accurate, due to the visible light emission from the LEDs, and the different emissivities of the various parts in the image. So, although the thermal imaging can be useful for finding hot spots and places to put thermocouples, you'll need to verify with contact measurements to be certain about the temperatures.

 

[frippe75]

Thermal imaging of illuminated LEDs is very difficult to get accurate, due to the visible light emission from the LEDs, and the different emissivities of the various parts in the image. So, although the thermal imaging can be useful for finding hot spots and places to put thermocouples, you'll need to verify with contact measurements to be certain about the temperatures.

So maybe let it heat up to a contant temp being illuminated and then turn it off and snap a new thermal image?

 

[frippe75]

So I rigged it up and let it go back to steady state.
Stopped water cooling. And then turned off power and snapped this about 1 sec later.

That is just soo much closer to my existing FD model

 

[JustAnOldFashionedLEDGuy]

So what does the commercial unit cost?

What is this going to be used for?

200,000 lumens, 2 hours run time is never going to be cheap. That's 3000 w-h of batteries, not to mention 250,000 lumens of LEDs, but the LEDs will be cheaper (much) than optics and everything else, like 1500-2000 watts of power management, etc.

I expect anything commercial would be quite expensive -- for a reason.

How many are you planning to build?

I can't see any situation where rolling your own would give you a cost advantage over upping the batteries 50% and just running metal halide (or a small generator).