Led too white

[Donaldf973]

So I got the 21ledusa 35w par bulb a few months back and it's too white. It's supposed to be full spectrum with a hint of blue but color rendition is between an Iwasaki 6500k and an XM 10k washing everything out except algae. I'm looking to make it as close to 20k as possible so was wondering how would I go about replacing these specific leds.

First pic is what they're supposed to look like, 2nd pic is what my galaxy s7 picks up (iphone 5 shows EXACT same colors) and the led board.

 

[Reefing Madness]

 

[Donaldf973]

Anybody...?

 

[nautical_nathaniel]

You would first need to know how to solder and then you would either have to find replacement boards that fit onto the central power distribution board or get super creative and begin changing individual LEDs until you got the spectrum you wanted. It may be easier to find a different bulb that fits your needs better or go ahead and convert to something like the Kessil a80 which you would be able to tune to whatever color you need.

 

[oreo54]

Purchase at no risk. If this is not the perfect color and intensity for your tank feel free to return for a full refund or message me to discuss better options!

Easiest option..

 

[Donaldf973]

The ballast died in 2 of the bulbs so I figured I'd play around with them after I replace the ballasts. Soldering I can do without issue.

 

[oreo54]

so they list them a CREE chips..Looks like th 5x5mm ones...
Though that's a tough call looking at a picture..
anyways you need to first identify the diodes and see what "footprint" they use..
http://www.cree.com/led-components/products/xlamp-leds-discrete..

Then you need to determine if they are in series or not..
As to the "ballasts"..whats written on them if anything?
In simplistic terms probably an AC to DC circuit..A DC voltage regulator and a constant current driver chip..
don't say they are dimmable so probably fairly
straight forward..

All those components in the center of a "heater" is kind of dumb and def prone to failure..

how long b4 those failed btw........?

 

[Donaldf973]

Each lasted 3 months before dimming and then dying. I identified the +/- connections but the ballast is only wired directly to one set of pads. The rest I can't for the life of me figure out how they did it without any visible wire unless all 4 pads were specially designed in the cross pattern and connected in the center where I cant get to. Identifying the led's going to be a pain in the butt lol.

 

[oreo54]

Best guess is that each is a series string w/ each 4 in parallel..
you can make out the traces in the white "overlay"..

Just disconnect the pink and white wire off the main circuit board and discard all the rest..most of it is to convert the 110v to DC anyways.
Driver chip or voltage regulator is most likely the U1 chip (on heat sink picture 4.)

One suggestion is to get a LDD-1000HW and a 24V DC (or larger) switching power supply (could test w/ 12v but you might be short voltage)..
That should give you 250mA per 4 diodes @ about 12-ish volts per pad (
hook the output of the LDD to pink(+)/white(-) and of course input to the power supply.
All "yellow" diodes are whites of some sort.

Prior to re-powering them you should be able to probe each diode using a digital VOM set to diode or low Ohms..
This puts out 3V (possibly 8 if a Fluke meter) so most should light. Current is low.
put probes at the "dots" (or on those that have the small pads) observing the correct polarity.
Fairly easy to identify..
Working diodes should glow faintly.

The center piece that covers the "spider" should be removable by futzing w/ the plastic clip on the power side..
Looks like once you remove that round piece the other 3 plastic prongs can be moved in to release the domed round piece on the light side. Then the entire diode assembly can be removed if you want.
At the least, removing it may reveal more of the tracing..

Other possibility is all in series which would mean you need , roughly, a 16x 3.5V (56V) power supply and reduce the LDD to like 350mA.
That could be problematic..and the more I look at it the more likely it is..all one series string..

If you google "led lightbulb circuit" (no quotes) and click "images" you will see that in all likely hood someone has taken one of these apart. .somewhere

 

[Donaldf973]

Thanks for helping me out with this project I'm in the middle of finishing up an acid bath for 70lbs of rock, tearing down 3 tanks and setting 2 back up the way i wanted from the start so I'm all over the place right now. I'm going back to the KISS method its what worked best for the first 5-6 years of reefing. Once I switched to led I couldn't keep anything alive so started dosing random and **** everything went downhill for me the past 3 years. Figured a restart is needed plus rock is 9 years old. Sucks my small tanks need led cause of footprint issues.

On a positive I figured out why one of the bulbs died there was a lose/not soldered properly ground wire. But the other still dead so at least 1 good thing came out of this.

 

[Donaldf973]

Finally dettached the middle piece and the 4 led pads is 1 whole piece.

 

[oreo54]

Yea, so it does look they are all in series..
To prove it takes a little risk.. Like measuring the voltage at the pink/white wires w/ a working one energized..

https://www.cdiweb.com/ProductDetail/0--meanwell/598503/

Something like this would power if if so..
Still.. doesn't change the LEds though..nor allow the fan..
There are others..

 

[Donaldf973]

Guess I'll just leave the bulb alone too much work to modify it right now and I really don't feel like micro soldering lol. If I can get the other to work again and the numbers are solid I'll just retrofit 2 ati bulbs in a diy fixture to add more blue once I get some par/lux numbers.

 

[Donaldf973]

Just did lux to par conversion readings directly underneath with optics at 14"= 580 to 600 and 26"= 200. No optics 12"= 266 and 24"= 133. So if those numbers are right this things a dang beast.

 

[Donaldf973]

Correction: no optics 24"= 60-80 par

 

[Donaldf973]

So, I came to the conclusion I'm going to leave the led board alone but get rid of the ballast on both and get new dimmable ones. I'll diy a fixture, pick up a 2 bulb t5 retrofit and mount everything together but spread the led pucks enough so it can be used on anything uo to a 50 gallon breeder. Seems like the best way to use them.

 

[oreo54]

Sounds like a plan.....Those boards and flex wiring are kind of a PIA AFAICT..
Do recommend trying to figure the voltage out at the pink/white wire though..
Will guarantee success w/ drivers..
both min voltage and max voltage of the driver will play into this..

One curiosity is the exposed trace sections in the center.. Removal scrapes or were they there? Almost like test points..

 

[Donaldf973]

The exposed sections were already there. How would I go about checking the voltage?

 

[oreo54]

Well just hook a VOM to the output wires on DC Volts (range should be 100v-ish).

old trick is to jamb a pin through the wire for probe attachment points.
Light needs to be fully functioning (hooked to diodes) and also "live".. so be very careful. 110v will be quite exposed.

Hooking the light to a power strip w/ a switch will allow you to set up everything and be "remote" from it when energized.

If you want to get really critical, after testing voltage you can cut the pink and put a VOM in series in-line set to Amps (again w/ a wholly functioning light) . That will get you the drive current
Use 20A setting if you have one. this test isn't really necessary but it will give you a good idea of what current the heatsink/fan is designed for..
https://www.engineersgarage.com/tutorials/dc-voltage-current-testing

NOW THE assumptions:
Diodes are driven by a constant current driver chip..
circuitry converts the ac to DC. The DC level is variable and controlled by the driver chip which is itself controlled to only allow a set amount of current out..
DC wil ramp up/down to meet this set point..
The steady voltage out will be at this current level (DC voltage to the LED's will fluctuate a bit w/ warming/cooling of the diodes but not a lot).

my guess is somewhere around 60v 350mA...

 

[Donaldf973]

Well bulb started flickering again and finally died. So, did the multimeter thing and no power was going through to the leds so the ballast is definitely shot. I'm just going to gut them and build a led t5 hybrid and call it a day. Thanks for the help and info.