Converting existing lighting to LED

[JDT]

I have quite a lot of small profiles and fresnels that use 500W lamps. Now that 50W and even 100W LEDs are getting quite cheap, I am wondering if it would be possible to replace the lamp with one of these.

 

I could build-in a DMX controlled LED driver so doing away with my old Strand dimmer panels.

 

Would the LED be optically compatible? Thinking that the LED only radiates in the forward direction, I could remove the rear reflector. This would give me space for a controller and heatsink. Anyone tried this?

[timsabre]

Would the LED be optically compatible? Thinking that the LED only radiates in the forward direction, I could remove the rear reflector. This would give me space for a controller and heatsink. Anyone tried this?

 

I have done some experiments.

 

I presume you're thinking of using COB leds (the ones with the big circle of emitters).

The emitting area of the high wattage ones is quite a lot bigger than the filament of the original lamp so you will run into some optical problems. You might get away with it in the fresnels but in the profiles (assuming they are simple ones) you will get an image of the LED dies. Profiles with condenser optics work much better.

 

Keeping the LED cool enough inside a fixture which is not designed for it will also be a challenge. You'll need a fan on the heatsink and probably one on the outside housing as well. Keep a close eye on the LED temperature while testing.

 

Post a report with some pictures if you try it, it would be interesting to know what happens.

[timmeh2]

Hi

 

Just to add what Tim has said and that I have found the great majority of 8-bit PWM drivers have rubbish curves and are generally hopeless for driving big chips at low intensity. If you're going to go through the hassle of all that R&D then I'd suggest you go with 16-bit from the outset.

 

All the best

Timmeh

[Jivemaster]

Remember that the filament is carefully placed relative to the lamp base(pins) to match with the luminaire optics design. If you are going to place chips then you will need to do this accurately (but not often!).

[gyro_gearloose]

Just to add what Tim has said and that I have found the great majority of 8-bit PWM drivers have rubbish curves and are generally hopeless for driving big chips at low intensity. If you're going to go through the hassle of all that R&D then I'd suggest you go with 16-bit from the outset.

Simply going from 8-bit to 16-bit dimming won't solve the problem with the dimming curve. You'd need to write the dimmer firmware such that the PWM output is non-linear i.e. 2% dimmer level should not necessarily equal 2% PWM pulse length. The pulse length should be shorter for low dimmer levels to give you more control over low intensities. I'd also add some form of damping so that the LED doesn't snap instantly (and it really is instantly, quite unlike any conventionally lamped fixture) on or off.

 

You should also be aware that some, possibly all, commercially available high power LED dimmer modules from places like Meanwell and EldoLED only dim down to about 2-5%. They won't dim to 0%, and I don't know enough about electronics to understand why. What you could do, depending on the LED you choose, is use several drivers for each LED. The LEDEngin LED that I used for my LED Miniscans was actually made up from two strings of several LEDs on one die. I drove each LED with two drivers, one per string, but didn't electronically dim them as I was using the Miniscans mechanical shutter for dimming/strobing, although the drivers could dim to 0% had I needed them to.

 

The emitting area of the high wattage ones is quite a lot bigger than the filament of the original lamp so you will run into some optical problems.

Really? I didn't know that there were single LEDs bigger than the filament on a T19 lamp :)

 

You might get away with it in the fresnels but in the profiles (assuming they are simple ones) you will get an image of the LED dies.

Sure about that? Using your logic you should be able to see the filaments on a conventional profile. I don't think you can, unless you've done something odd with the focus.

 

Profiles with condenser optics work much better.

Yes they do, regardless of wether you're using LED, tungsten, or discharge/arc lamps.

[timsabre]

The emitting area of the high wattage ones is quite a lot bigger than the filament of the original lamp so you will run into some optical problems.

Really? I didn't know that there were single LEDs bigger than the filament on a T19 lamp :)

 

There are.... I have got one on my desk. Like this 100W one from Sharp - emitting area is approx 40mm square (emitters are arranged in a square within the circular phosphor).

T19 lamp filament is approx 15mm square.

http://www.ledsmagazine.com/content/dam/leds/migrated/objects/news/10/2/10/Sharp02112013.jpg

 

They are not strictly speaking single LEDs, they are an array of smaller LED emitters.

 

You might get away with it in the fresnels but in the profiles (assuming they are simple ones) you will get an image of the LED dies.

Sure about that? Using your logic you should be able to see the filaments on a conventional profile. I don't think you can, unless you've done something odd with the focus.

 

I have done it and seen it - you get a sort of grid appearance where the LED dies are, I guess it is because the LED dies are very small with relatively large gaps between them. In a normal lamp the filament is arranged in a double zigzag to close up the gaps. You also get a square beam, even from a fresnel.

[gyro_gearloose]

I have done it and seen it - you get a sort of grid appearance where the LED dies are

Where was the LED in relation to the lens? I imagine you'd get a very good projection of the LED dies if the LED were in the profiles focal plane as thats how profiles work. What happens if you move the LED back?

[Brian]

...you get a sort of grid appearance where the LED dies are, I guess it is because the LED dies are very small with relatively large gaps between them.

 

Like this you mean?

 

 

...from an LED fresnel.

[timsabre]

...you get a sort of grid appearance where the LED dies are, I guess it is because the LED dies are very small with relatively large gaps between them.

 

Like this you mean?

 

 

...from an LED fresnel.

 

Yes, exactly like that. But on a profile spot you can get a niced focused circle round it.

[JDT]

Well, looks promising then. I will make it a new year's project. No time this side of Christmas.

 

I will design my own PWM led driver. No problem dimming to 0%. Interesting comment about damping to simulate incandescent lamp. Can do.

Of course the colour won't change as it dims. I call that a bonus!

 

Will report back with results.