High Power UV LED Driver

[IrradiatedHaggis]

Hi all,

Hoping someone can help me... I'm working on a DLP resin based 3d printer. I have a modified LED Pico projcetor which I have removed the LEDs from and replaced them with a UV LED. Currently I have a 5W LED on it for a proof-of-concept. It works, but I need to upgrade to a more powerful LED for cure times to be at all reasonable.

I'm looking at using the CBT-120-UV-C11-G382-22 UV LED from Luminus Devices, which is explicitly designed for this purpose. It is an insane (and expensive) diode that is ideal for my purposes... With 11 Watts of actual optical power output it is one of the brightest LEDs on the market. The problem is, I need 18amps at 3.7 volts to drive it. So... My question is, how can I get this?

Ideally, I'd like to run off of 5V. Using a regular resistor to power it from 5V I need about 0.08 ohms capable of at least 25 watts (I think). That is going to get _Very_ hot and is simply an insane waste of power. Thats assuming my power calculation is correct.... Do I use the full 5V for power dissipation in the resistor, or the 1.3 volt difference between the diode voltage drop and the power supply? If it's the full 5V then I actually need a 90 watt resistor....

So obviously I need a switchin constant current regulator. A linear regulator just will not do here. But I cannot find an LED drivers that will work at 18 amps.... And those that approach it (16 amps is the largest I can find) will not work at output voltages below 15v since they're intended to drive multiple LEDs in series.  Does anyone know of any regulators that would work?

I did find one buck regulator that is intended to drive a white LED at 9A (on pcb-components.de). Hypothetically, could I use two of these in parallel? I'm thinking that with switching constant current regulators running two in parallel is probably not possible... I don't know how they would behave. Any thoughts on this?

Anyway, just thought I'd drop a line here and ask if anyone has any experience in this area or could provide some suggestions /advice for lighting those LEDS up.

Thanks,

Troy.


Sent from Samsung tablet

[Aaron Oliver-Taylor]

How about a constant current source - use an op-amp and high current mosfet (or several in parallel) to obtain a precise current.  This programmable constant current dummy load uses a constant current source like this, you could use this idea: http://hackaday.com/2012/10/01/test-your-projects-mettle-with-a-protected-dummy-load/

This also might be of use: http://hackaday.com/2012/08/30/building-a-driver-for-absurdly-high-power-leds/

Aaron
__________________________________________
Aaron Oliver-Taylor
Email aaron.oli...@gmail.com
Web   aayotee.wordpress.com

[Nigel Worsley]

> The problem is, I need 18amps at 3.7 volts to drive it. So... My question is, how can I get this?

One of these would be a good start:
http://uk.farnell.com/tracopower/txl-100-3-3s/ac-dc-3-3v-23a/dp/2080728

It is a 3.3V PSU but can be adjusted +/- 10%, which gets you up to 3.63V. According to the
data sheet for your LED this would drive it at about 15A, let it warm up by 35°C and this will
increase to the desired 18A. You will be very hard pressed to stop it warming up by a lot
more than this!

In practice you will want to modify the PSU to increase the maximum output slightly (probably
just a resistor change, you only need it to go up to +12% instead of +10%) and add a circuit
to make it accurately regulate the output to a constant current. This isn't likely to be as difficult
as it sounds, a sense resistor with an opamp driving into the existing feedback opto should
do the trick.

If the above sounds too scary then I can help you with this, electronics design is the day job
and this project sounds like fun :)

Nigle

[IrradiatedHaggis]

Hi Aaron,

I looked at the op-amp / FET circuit and considered it. The problem though is that it's a linear regulator. I'd be dissipating at least 25 watts through the fets. It's not hard to find fets that can handle this, but it's just so wasteful and basically the same result as using big power resistors.... I might just end up using power resistors to start with if I cannot find a better way.

Regards,

Troy.


Sent from Samsung tablet

[Mike Harrison]

>> Hoping someone can help me... I'm working on a DLP resin based 3d printer.
>> I have a modified LED Pico projcetor which I have removed the LEDs from and
>> replaced them with a UV LED. Currently I have a 5W LED on it for a
>> proof-of-concept. It works, but I need to upgrade to a more powerful LED
>> for cure times to be at all reasonable.
>>
>> I'm looking at using the CBT-120-UV-C11-G382-22 UV LED from Luminus
>> Devices, which is explicitly designed for this purpose. It is an insane
>> (and expensive) diode that is ideal for my purposes... With 11 Watts of
>> actual optical power output it is one of the brightest LEDs on the market.
>> The problem is, I need 18amps at 3.7 volts to drive it.

Probably a bigger problem is how are you going to cool it? 50 watts from a small emitter is an
awful lot of heat to remove from a very small space.

>So... My question is, how can I get this?

This is in the same power ballpark as high power laser diodes, so may be worth looking for any
appnotes in this area to get some ideas.

The PCB layout at these power levels is going to be extremely critical.

Bear in mind that a constant current supply is little different from a constant voltage one - the
only difference is how the feedback is sensed.
Most voltage regulators can be used as CC sources by putting a current-sense resistor in the
negative rail and using the voltage across this as the feedback instead of the divided output
voltage.
The main difference with SMPS chips aimed at LED applications is that they're designed for a low
feedback reference voltage (typically around 0.2V) to minimise losses in the sense resistor - most
voltage-oriented chips use a 1.2V feedback reference. You can use an amplifier to boost the feedback
voltage form a low-value reference resistor, but you need to keep an eye on loop stability issues -
at least your load isn't going to change suddenly so you can probably just damp the crap out of it
to keep it stable.

One approach that may work, avoiding the layout issues is to use a 5V mains PSU that has remote
output sensing, and drive the sense input from an amplified voltage across a low-value sense
resistor. If done properly, the input sense will use a differential amplifier - if you delve inside
you may be able to find it and increase the gain to avoid the need for an external amp.

If mains PSUs scare you, a similar approach would be to do the same with a brick type 5V DC-DC
converter module.

>> Ideally, I'd like to run off of 5V. Using a regular resistor to power it
>> from 5V I need about 0.08 ohms capable of at least 25 watts (I think). That
>> is going to get _Very_ hot and is simply an insane waste of power. Thats
>> assuming my power calculation is correct.... Do I use the full 5V for power
>> dissipation in the resistor, or the 1.3 volt difference between the diode
>> voltage drop and the power supply? If it's the full 5V then I actually need
>> a 90 watt resistor....

Bear in mind the Vf will probably vary with temperature, so you absolutely do need active current
regulation.

If you do want to roll your own, the LT3743 may be worth a look:
http://cds.linear.com/docs/Datasheet/3743fc.pdf

[Aaron Oliver-Taylor]

How about a pwm controller (i.e. a smps controller) to controll your mosfet.  This looks promising: http://www.onsemi.com/PowerSolutions/product.do?id=NCP3065  You can use it with an external transistor/mosfet for larger currents.

__________________________________________
Aaron Oliver-Taylor
Email aaron.oli...@gmail.com
Web   aayotee.wordpress.com

[IrradiatedHaggis]

Hi Nigel,

Your suggesting that I manually tune a power supply to dive it at constant voltage where it will give me the correct current? I had thought of this, didn't think I'd be able to tune ine precise enough to bump it just over the voltage drop of the LED and not get too much current. As far as I can tell this LED should be able to handle up to 30 amps intermittantly with appropriate cooling, so maybe it's not such a concern and 18amps is really my top target, anything over 13A is probably intense enough.

I'm no expert in this, I'm only an electrical engineer by hobby and have never done any formal training in it.... In particulair LEDs are a little bit of a mystery to me but the forward voltage drop of these resistors is listed at 3.7 volts. Doesn't that mean that I need to exceed that for the LEDs to light? I'm not sure what the internal resistance of the LED would be at this (I didn't see that / don't know where to look for that in the datasheet), but I believe that the voltage I need is 3.7 + V where V = 18 / r_LED.  As you state I guess the internal resistance would drift as the LED warms up. Obviously I have a lot of heat there that I need to get rid of, but I don't think that's too much of a problem. I have enough room to mount the LED to a CPU heatsink/fan. I was hoping a 40mmx40mm one would work, but I could probably fit up to 60mm in the space available. Also with a little bit of creativity (and a block of aluminium) I could mount a larger heatsink to the bottom of the unit.

FYI, the projector I'm using is a TI DLP Lightcrafter, which is an evaluation unit that is designed for structured light applications so it's quite hackable. Perhaps I could bring it into the space sometime next week if you'll be around to discuss.

Cheers,

Troy.

Sent from Samsung tablet

Nigel Worsley <nig...@googlemail.com> wrote:


> The problem is, I need 18amps at 3.7 volts to drive it. So... My question is, how can I get this?

[Nigel Worsley]

> Your suggesting that I manually tune a power supply to dive it at constant voltage
> where it will give me the correct current?

As a starting point yes, but a few mods to the PSU will make it much better.

> The forward voltage drop of these resistors is listed at 3.7 volts. Doesn't that mean

> that I need to exceed that for the LEDs to light?

No, that is the typical forward voltage at the nominal 18A operating current. There
is a graph of voltage against current in the datasheet, it goes from 3.3V at 0A to
3.97V at 30A.

> I have enough room to mount the LED to a CPU heatsink/fan. I was hoping a
> 40mmx40mm one would work

I rather doubt it, the ones I have seen are rated for about 1.5°C/W, which would
lead to a junction temperature of over 140°C - perilously close to the absolute
maximum and reducing the LEDs life to less than a third of that for a junction
temperature of 80°C.

> FYI, the projector I'm using is a TI DLP Lightcrafter, which is an evaluation unit
> that is designed for structured light applications so it's quite hackable.

Looks good, not exactly cheap though.

> Perhaps I could bring it into the space sometime next week if you'll be around
> to discuss.

I don't get to the space much as it is a 2 hour round trip, but could probably get
over there one evening as I have a couple of other reasons to go there.

Nigle

[IrradiatedHaggis]

Thanks Mike,

Is that really all there is to a CC supply? I thought they would be more complicated. I thought that it should be possible to accomplish this using the sense lines on a bog-standard PSU, but I wouldn't know where to begin. Messing with a line PSU certainly doesn't scare me (my other hobby is tesla coils ;), but I just don't know enough about SMPS units to modify them. The closest I've come is a half-bridge MOSFET SSTC which is very, very simple compared to a real switching power supply....  

Cheers,

Troy.


Sent from Samsung tablet

Mike Harrison <mi...@whitewing.co.uk> wrote:



>> Hoping someone can help me... I'm working on a DLP resin based 3d printer.
>> I have a modified LED Pico projcetor which I have removed the LEDs from and
>> replaced them with a UV LED. Currently I have a 5W LED on it for a
>> proof-of-concept. It works, but I need to upgrade to a more powerful LED
>> for cure times to be at all reasonable.
>>
>> I'm looking at using the CBT-120-UV-C11-G382-22 UV LED from Luminus
>> Devices, which is explicitly designed for this purpose. It is an insane
>> (and expensive) diode that is ideal for my purposes... With 11 Watts of
>> actual optical power output it is one of the brightest LEDs on the market.

>> The problem is, I need 18amps at 3.7 volts to drive it.

Probably a bigger problem is how are you going to cool it?  50 watts from a small emitter is an
awful lot of heat to remove from a very small space.

>So... My question is, how can I get this?

This is in the same power ballpark as high power laser diodes, so may be worth looking for any
appnotes  in this area to get some ideas.

The PCB layout at these power levels is going to be extremely critical.

Bear in mind that a constant current supply is little different from a constant voltage one - the
only difference is how the feedback is sensed.
Most voltage regulators can be used as CC sources by putting a current-sense resistor in the
negative rail and using the voltage across this as the feedback instead of the divided output
voltage.
The main difference with SMPS chips aimed at LED applications is that they're designed for a low
feedback reference voltage (typically around 0.2V)  to minimise losses in the sense resistor - most
voltage-oriented chips use a 1.2V feedback reference. You can use an amplifier to boost the feedback
voltage form a low-value reference resistor, but you need to keep an eye on loop stability issues -
at least your load isn't going to change suddenly so you can probably just damp the crap out of it
to keep it stable.

One approach that may work, avoiding the layout issues is to use a 5V mains PSU that has remote
output sensing, and drive the sense input from an amplified voltage across a low-value sense
resistor. If done properly, the input sense will use a differential amplifier - if you delve inside
you may be able to find it and increase the gain to avoid the need for an external amp.

If mains PSUs scare you, a similar approach would be to do the same with a brick type 5V DC-DC
converter module.

>> Ideally, I'd like to run off of 5V. Using a regular resistor to power it
>> from 5V I need about 0.08 ohms capable of at least 25 watts (I think). That
>> is going to get _Very_ hot and is simply an insane waste of power. Thats
>> assuming my power calculation is correct.... Do I use the full 5V for power
>> dissipation in the resistor, or the 1.3 volt difference between the diode
>> voltage drop and the power supply? If it's the full 5V then I actually need
>> a 90 watt resistor....

[Adrian Godwin]

There are some 50A 5V supplies around in the hackspace. they could
probably be modified to control current or voltage as Mike and Nigel
have suggested.

[Nigel Worsley]

> There are some 50A 5V supplies around in the hackspace. they could
> probably be modified to control current or voltage as Mike and Nigel
> have suggested.

Worth looking at, but it might not be possible to run the output voltage
low enough. The controller chip is usually powered from a auxiliary
winding on the transformer that tracks changes in the main output,
turn it down too far and the under voltage lockout activates.

Nigle

[IrradiatedHaggis]

I was hoping that i could run the whole thing off of a single 5v PSU because my projector's logic runs off of 5V as does the beaglebone that I'm planning to use to control the whole mess. That would mean that my regulator would need to be a seperate component. It's sounding like that's not really feasible though. Oh well, the projector logic and beaglebone can run off of a small seperate 5v supply, their power requirements are quite minimal.

Where are these 5V supplies located? I don't normally go into the space that often, but I work in the city so I can stop by and take a look next week.

Thanks,

Troy.


Sent from Samsung tablet

[Mike Harrison]

On Thu, 04 Oct 2012 13:59:39 +0100, you wrote:

>Thanks Mike,
>
>Is that really all there is to a CC supply? I thought they would be more complicated.

Mostly. There are a few side-issues, like what happens if the output is disconnected, and stability
into different loads. However your load is pretty constant and non-capacitive, which simplifies
things substantially

> I thought that it should be possible to accomplish this using the sense lines on a bog-standard PSU, but I wouldn't know where to begin.

It is, but the problem is that the sense lines are expecting to see the nominal output voltage, not
the small voltage across a sense resistor - it's only a matter of scale. Thinking more about it,
most PSUs will be convertingthe sense at nominal voltage down to around 1.2v, whearas you will want
to be stepping it up from a few hundred millivolts, so it may not be practical to mod the internal
circuit, however adding an external amplifier ought to be pretty doable.

e.g. with a nominally 5V PSU and a 10 milliohm sense resistor to give 0.18V at 18 amps, you'd need a
gain of 5/0.18=27.7. At these levels you probably want an instrumentation amp, with a 4-wire
connection to the sense resistor to minimise offset errors.

You do need to be careful about how the amp behaves during powerup, to avoid spikes etc.

[IrradiatedHaggis]

Yes, the projector is expensive for what it is. It's a pretty low resolution, but the important part is that it doesn't have any safety interlocks to disable in order to make it work. Any other projector would not work with the main light-source removed and replaced. Also, this thing works right off of USB with a simple command set. I've gotten that part working so I can upload the layer exposure images to it in bitmap format directly. This lets me run it off of a device that doesn't have video out - specifically a beaglebone single board computer is what I have in mind.  I'm going to write a web interface for it that will allow me to upload sliced models and control all aspects of the printer from a webpage. I can even put a commandline slicer such as Slic3r (which already has DLP support in it) right on the beaglebone and add the ability to upload STLs directly to the printer to slice itself. Another feature I would like to add in the future is a webcam streaming to it's control page so I can log into the printer from any computer in the house and check on how the print is actually going (and stop it if there is a problem). I'm not sure about exposing that webpage to the outside world, but in theory I could let it out so that I can check on it from work... I currently don't let my reprap print when I'm out of the house.

Cheers,

Troy.




Sent from Samsung tablet

Nigel Worsley <nig...@googlemail.com> wrote:

[Nigel Worsley]

> I was hoping that i could run the whole thing off of a single 5v PSU

It should be possible to hack a VRM to do this, but you would need to
make sure it was designed for a 5V input and can be modded to deliver
the right output. If you know which chip it uses then this info can be
worked out, but they all seem to be designed for a specific motherboard
so you would have to buy one and then find out if it can be used :(

Nigle

[Troy]

Yes, I came across some VRMs that looked reasonable, but again, I
wouldn't know where to start for hacking one. I think they're usually
buck converters with a programmable voltage, right? But I'd need a sense
mechanism and some automatic feedback to dial the voltage in for the
required current. It would be the same as modding a full PSU for sense
feedback as previously suggested I suppose. Perhaps I will investigate
that line further.

Cheers,
Troy.

[Ken Boak]

Troy,

You need 66 watts to illuminate this LED. Trying to get that from a 5V supply is counter productive - unless you have some old PC power supplies lying around.  You will be trying to control 15 or more amps at that voltage.

Have a look at the Bridgelux LED modules from Farnell - they are insanely bright.  I put four of the 30W output modules on a heatsink as a sun simulator. You need welding goggles to look at it.

http://uk.farnell.com/bridgelux/bxra-56c9000-j-00/rs-mega-9000-lm-cool-wht-array/dp/2070609

Consider upping your voltage,  and downing current. It all becomes a lot more sane

K

[Mike Harrison]

The problem is that for projection you typically need a small source, so using multiple series
devices may not be an option.

[IrradiatedHaggis]

Hi Ken,

My requirements are very specific. I need a UV LED in the 360-400nm wavelength range and it has to be a single point source. That limits options. There are a couple of other LEDs that would work, but the others all have lenses that increase divergence which means more light will simply be lost to the walls of the optical path. This LED was specifically designed to replace a projection bulb for lithography purposes, which is exactly what I'm using it for. 

I think I'm going to get the LED and start by running it from my CC benchtop supply at 5A for testing. Then, if that goes well I will continue construction of my printer and maybe get one of the 9A drivers to start with while I look at alternative methods. Modifying a VRM or building a sense/control circuit for one seems reasonable. Also, the 9A CC Buck regulator that I'm looking at can apparently be modified (according to the web it's been tried up to 11A). I will see if I can figure out how it operates and if the same chip used in that could be used to build an even higher current device.

Thanks everyone for your help. I think this project will be cool. It certainly will not be the cheapest homebrew DLP Stereolithography printer around, but with everything I have in mind it will be one of the most feature-rich. Maybe there's a kick-starter in this ;) It should easily match or exceed the Form1 which is on kickstarter and selling for around $2500USD I think.


Sent from Samsung tablet

Ken Boak <ken....@gmail.com> wrote:


Troy,

You need 66 watts to illuminate this LED. Trying to get that from a 5V supply is counter productive - unless you have some old PC power supplies lying around.  You will be trying to control 15 or more amps at that voltage.

Have a look at the Bridgelux LED modules from Farnell - they are insanely bright.  I put four of the 30W output modules on a heatsink as a sun simulator. You need welding goggles to look at it.

http://uk.farnell.com/bridgelux/bxra-56c9000-j-00/rs-mega-9000-lm-cool-wht-array/dp/2070609

Consider upping your voltage,  and downing current. It all becomes a lot more sane

K