DIY LEDs - The write-up

[kcress]

Kcress It that is true how many parallel strings of LEDs could you reliably run on one Meanwell? Even if it was just two it would save substantial cost on a large system. If it was 4 it would would also make controlling with an Aquacontroller much easier for me.

You could theoretically run 4 strings of 48/3.5 = 13 LEDs.

Or 52 total.

You would be limited to 1.3A / 4 = 325mA per string.

To do it right though you'd need to do some additional work.

It would consist of some detailed meter work.

You would set up a string on a Mean Well and set the string current to 325mA using an ammeter.

Turn it on and wait until the string is warmed up. As you wait, use a Sharpie to number every one of them. Once warm measure the voltage across each one and write it down in a numerical table.

Do this for all 52.

Now take this table and mix and match the values to end up with the same total voltage in each string. You could do this many different ways. Use, say, the highest 5 with the lowest 6 if that works. Or just match across one low one in each string then the next higher one in the next string, etc, etc.

Once you have them grouped build your 4 strings.

You need to build the strings normally BUT you need to add fuses in each string.

Something like a 375mA fuse. Digikey F1504-ND in a holder F1467-ND.

Now when a LED opens or one shorts the fuse will open protecting the rest of the string.

Note that if any fuse opens they will all open, so keep spares.

If you can't pull this off as described, don't run parallel strings.

 

[TheFishMan65]

There is another way

There is another way

Kcress,

I mentioned this once before. The theory is good, but I haven't tried it in real life LEDs. If anyone can say why this won't work in real life I would like to hear.

Assume your maximum current of and LED is 1 amp. You can run three strings at .333 amps. If you loose one string each gets .5 amps no damge, loose two string and you get 1 amp, no damage assuming you can handle the heat. Note since you can never exceed the current no fuse is needed. You also sort of have self correcting light levels, since as each string gets more current you get more light.

Now if you also run the LEDs in parallel across the string. I think you could skip the measuring. In the following drawing the letter number combinations are LEDs. The numbers are the series string and the letters are the parallel links.The - are wires and the + are connections

----+----A1----+----B1----+----C1----+----D1----+----
----+----A2----+----B2----+----C2----+----D2----+----
----+----A3----+----B3----+----C3----+----D3----+----

This way you have 4 (probably 6 or more) little voltage imbalances rather than 3 big ones. Since the A group is in series it will drop around 3.5 volts it may not be exact, but with the three in series it will have less variance than just one. Now each LEDs will have around .333 amps, they won't be exact becuase of the variance in the LEDs. But is should be ok, becuase they cannot reach the maximum. This is were real life may break the theory. Do three LEDs vary in voltage vs light output vary enough that you will see it? I don't think so.

No look at loosing and LED. If one (say B1) shorts (+ and - get connected) then the whole group of Bs will go out since all the current will go through B1. No additional current (if you are using a driver) through the other leds so no harm. Check each B series LED and replace the bad one. Say one (C1) opens (no conection between - and +). In this case each of C2 and C3 will have to handle the additonal current they will go up to .5 amps, which is still in range. Its easy to find the bad LED in this failure mode - only one is out.

The downside of this plan is a little more wiring. However wire is a lot cheaper than the drivers.

 

[Gomer]

Anyone thought of using high power UV emitters in the 365nm range?

Unfortunately there aren't economical solutions. this is indepentent of it being a good idea (I think it might be a bit short in the UV)

Does anyone know if XPE optics will fit correctly on XPG?

I doubt they will fit the same. XRE's have the dome ring. Phyically different geometry. Optically, they also wont work nearly as well as XPG optics for the XPG. Both the NA (cone angle) and the reference plane (emitter location relative to the optic's design image plane) are different between the XPG and XR-E.

Related question: I am going to try the whole drilling and tapping thing for fun when I do my sump light. So:

A) Which thermal paste are you guys using? Think it would be ok to do the two-dollar ebay stuff or should I go with a name brand?

B) Are nylon washers necessary when tightenting the screws down onto the LED or is metal fine? It looks like I should be fine with metal providing that I keep it clear of the terminals, but I can picture the potential benefit of having the nonrigid nylon..

a) thermal conductivity may matter. Thermal paste has horrible thermal conductivity compared to metal. It is likely the thermally limiting path if you are depending on the paste (ie non uniform surface and not much compression). Paste is cheap. Even the "expensive" and far superior arcticsilver is cheapwhen you compare to your investment with everything else.

b) My solution was (depending on the project) was to use either nylon screws or glass-filled polyurethane (which is non conductive and more ridged).

 

[jma1978]

Kcress,

I mentioned this once before. The theory is good, but I haven't tried it in real life LEDs. If anyone can say why this won't work in real life I would like to hear.

Assume your maximum current of and LED is 1 amp. You can run three strings at .333 amps. If you loose one string each gets .5 amps no damge, loose two string and you get 1 amp, no damage assuming you can handle the heat. Note since you can never exceed the current no fuse is needed. You also sort of have self correcting light levels, since as each string gets more current you get more light.

Now if you also run the LEDs in parallel across the string. I think you could skip the measuring. In the following drawing the letter number combinations are LEDs. The numbers are the series string and the letters are the parallel links.The - are wires and the + are connections

----+----A1----+----B1----+----C1----+----D1----+----
----+----A2----+----B2----+----C2----+----D2----+----
----+----A3----+----B3----+----C3----+----D3----+----

This way you have 4 (probably 6 or more) little voltage imbalances rather than 3 big ones. Since the A group is in series it will drop around 3.5 volts it may not be exact, but with the three in series it will have less variance than just one. Now each LEDs will have around .333 amps, they won't be exact becuase of the variance in the LEDs. But is should be ok, becuase they cannot reach the maximum. This is were real life may break the theory. Do three LEDs vary in voltage vs light output vary enough that you will see it? I don't think so.

No look at loosing and LED. If one (say B1) shorts (+ and - get connected) then the whole group of Bs will go out since all the current will go through B1. No additional current (if you are using a driver) through the other leds so no harm. Check each B series LED and replace the bad one. Say one (C1) opens (no conection between - and +). In this case each of C2 and C3 will have to handle the additonal current they will go up to .5 amps, which is still in range. Its easy to find the bad LED in this failure mode - only one is out.

The downside of this plan is a little more wiring. However wire is a lot cheaper than the drivers.

I'm not quite following you. Are you talking about hooking it up like they show in the product spec sheet?

 

[laverda]

You could theoretically run 4 strings of 48/3.5 = 13 LEDs.

Or 52 total.

You would be limited to 1.3A / 4 = 325mA per string.

To do it right though you'd need to do some additional work.

It would consist of some detailed meter work.

You would set up a string on a Mean Well and set the string current to 325mA using an ammeter.

Turn it on and wait until the string is warmed up. As you wait, use a Sharpie to number every one of them. Once warm measure the voltage across each one and write it down in a numerical table.

Do this for all 52.

Now take this table and mix and match the values to end up with the same total voltage in each string. You could do this many different ways. Use, say, the highest 5 with the lowest 6 if that works. Or just match across one low one in each string then the next higher one in the next string, etc, etc.

Once you have them grouped build your 4 strings.

You need to build the strings normally BUT you need to add fuses in each string.

Something like a 375mA fuse. Digikey F1504-ND in a holder F1467-ND.

Now when a LED opens or one shorts the fuse will open protecting the rest of the string.

Note that if any fuse opens they will all open, so keep spares.

If you can't pull this off as described, don't run parallel strings.

Cool! That would be worth the effort. Adding fuses would be worth it for peice of mind. That should save me a $200-300 on drivers for the systems I plan to build, depending on exactly how I dicde I want to do them. It would also save on Aquacontroller components and wiring required to control the meanwells. Less power supplies would also mean less power cords running everywhere.

 

[laverda]

I'm not quite following you. Are you talking about hooking it up like they show in the product spec sheet?

I as wondering the same thing. If wired this way what size fuses and where would they be located? My thinking is they would the not work with it wired this way, as if one fuse blue power would just go through the other fuses and LEDs. Of course that would cause the other fuses to all blow in short order.

 

[Skeptic_07]

I doubt they will fit the same. XRE's have the dome ring. Phyically different geometry. Optically, they also wont work nearly as well as XPG optics for the XPG. Both the NA (cone angle) and the reference plane (emitter location relative to the optic's design image plane) are different between the XPG and XR-E.

thanks. i just now received XPGs in the mail and they are much much smaller than the XPEs I've been messing with. The wierd thing is that I alos ordered a whole bunch of royal blue XPE and well.. they're the same size as the XPGs. this is very puzzleing because i already had some RBs from LED supply which are complete different. Let me get some pics up

 

[Skeptic_07]

the white XPG

the royal blue 'XPE' I got with some of the XPE I already had

\

testing each of the new ones

well on the upside I only have to order one kind of optics... hopefully they come in the 60 degree variety :hmm2:

 

[kcress]

TheFishMan65; That configuration, just so you know, is called "series parallel".

Yes it would work. Would it work well.. It could. It could also work poorly. It really depends on the consistency of the Vf of all the HBLEDs. A light manufacturer would typically run every single LED thru a voltage binning process. With a million LEDs in bins they would just pull ALL the LEDs for a light out of the same bin. Us on the other hand get only the number we need and they are going to be all over the place on Vf.

What then happens is you will get groups of three that have one member hogging more current than the others. You as a human won't be able to see this optically as your eyes are overloaded and can't detect a %50 difference anyway.

The failure method could also be a problem as you likely won't even notice a single failed (open in this case) LED and the others will increase markedly in brightness. Could this burn some expensive coral? Possibly, in fact likely. Might not matter in some cases though.

Running series parallel with unmatched Vfs will result in a patchy illumination field.

As for limiting the current to below the LEDs max the scheme should work OK. The only caveat would be if your cooling systems is good for 500mA and you end up with 1A running thru a LED it will end up just as dead as if you over amp'd it.

The setup I described above would essentially shut it all down on the first failure of any LED. I'd prefer that except if the whole thing took a dump as you went out the door to your week vacation. Also the method I described above would assure you of having each LED having the same current run thru them - if that matter to you.

The series parallel method is used all street light fixture, that I am aware of, so someone likes it. It's to provide that 'run-on damaged' functionality.

If you could somehow measure all your LEDs and bin them yourself so you make up each triad of equal Vfs then I would go with the series parallel setup. You would likely need to buy 10% extra LEDs to pull it off.

Since you would save $$ on fewer drivers you would come out ahead probably.

 

[skippyreef]

Unfortunately there aren't economical solutions. this is indepentent of it being a good idea (I think it might be a bit short in the UV)

DO you mean the wavelength is too short????

 

[Gomer]

Yes.

 

[skippyreef]

Thanks for the clarification on that.

 

[TheFishMan65]

jma,

Yep - but they draw much nicer pictures.

Kcress,

You really think there is that much variation in LEDs. I really have no idea, so out of curiosity I will ask if you have seen any stats. You will note that I did mention the heat sink at the 1 amp point. I did not think of burning coral so that was a good point.

I guess it is just hat people are comfortable with.

 

[kcress]

Kcress,

You really think there is that much variation in LEDs.

If you ever saw the pass/fail marks on a semiconductor wafer after a test you would understand. Yes, there will be a difference.

 

[laverda]

Kcress and TheFishMan65
Thanks for the information. I have 26 LEDs and 2 meanwells coming from the Nano_reef group but to expariment with. If I had know I did not need a meanwell for every string of 13 LED's, I would have ordered more. Although I really could not afford more at the moment. I will measure them and see how much they vary.
Steve

 

[der_wille_zur_macht]

the royal blue 'XPE' I got with some of the XPE I already had

The LEDs to the left are XR-E or XR-E knockoffs. XP-E and XP-G are smaller like the ones on the right.

As far as "binning" your LEDs when running parallel - YES please do it! I've measured significant differences even in the same batch. Also note that some of the LEDs we use are actually binned according to forward voltage by the manufacturer (in addition to color and brightness), though us hobbyists rarely get that info.

 

[Santoki]

DWZM,
how do you know they are knockoffs?
All builds I have seen prior to this (acquired from several different vendors) were built using supposed XR-Es, which look exactly like the ones on the left. The smaller format on the right, I have only seen as supposed XP-Gs.
-R

 

[chriscobb]

Has anyone run the leds in with MH's? How would the heat from the MH's affect the leds?

 

[der_wille_zur_macht]

DWZM,
how do you know they are knockoffs?
All builds I have seen prior to this (acquired from several different vendors) were built using supposed XR-Es, which look exactly like the ones on the left. The smaller format on the right, I have only seen as supposed XP-Gs.
-R

XR-E or XR-E knockoffs - there's really no way to tell, except either trusting the vendor you got them from, or measuring output and spectrum against a known-good LED. I suggested knockoffs because there seemed to be some general confusion from the OP above - there's really no way to tell just by looking at them.

 

[TheFishMan65]

Well thank you for letting me know aobut real life. I would not hve expected that much variation.