Finished: Quad XML Mag 4C (Optics NiMH 3-mode) new everyman's recipe

[Techjunkie]

*update #3* The taicap resistors and improved regulator heatsinking now have the light dialed in perfectly. I can run on max current until the head is uncomfortable to hold without the regulators overheating. See updates in post #31.

*update #2* The KD V2 regulators overheat way too easily in this application (maybe it's the 380mA 7135 vs the 350mA ones) have to rework the drivers (having deja vu here. when will I ever learn?)

*update* Build complete. Beamshots in post #11.

I'm finally starting the build log for this light, but the truth is I've been collecting parts for it for months now. The finished specs will be:

• Mag 4C
• Glass lens
• Quad Optic (edges trimmed)
• Cu PES C heatsink machined flat (drilled & tapped)
• 4x 5000K CREE XM-L (my new favorite tint)
• One individual 3.04A regulator for each LED, mode slaved (KD V2 regs with 380mA per 7135)
• 3 modes: very low, medium very high (my new favorite mode group thanks to these new V2 boards)
• 4x AccuPower Evolution NiMH LSD C cells 4500mAH (4.8V pack voltage ensures full power available through nearly the entire discharge)
• 12A draw at the tail, mere milliamps through the mag switch

I've got a new twist on my Everyman's recipe from my other builds. It occurred to me that LED positive can be connected directly to the battery instead of coming across the switch. When the regulators are off, the circuit is broken. This reduces resistance by removing the Mag switch from the LED circuit and also saves the switch from carrying 12A of current. I've tested this already with one XML, one IMR22650, and one KD V2 regulator configured for H-M-L. The only oddity that I noticed in this configuration was that the mode board's memory seemed to take a split second to kick in, causing a quick flash of high mode when turning on in, or changing to, low mode. A light that I built using the same components and mode group, but with a traditional tail switch doesn't exhibit the mode switch flash. I'm not concerned at the moment.

A quick mock-up using the old diagram as a starting point (in actuality, the new V2 regulators with 380mA per 7135 will be used:



Some pics of the mechanical parts of the build so far...


5000K emitters from DigiKey SMD soldered to 16.5mm boards from KD (top)
PES-C Cu heatsink machined flat, drilled and tapped (left)
DX Quad optic with edge trimmed to recess inside Mag C head (right)



Underside of heatsink shows color coding corresponding to dots on MCPCBs. This little trick allows me to align the LEDs in the same orientation as when measurements for holes were taken.



MCPCBs fully mounted with AS5 and #3 screws. What you don't see here is that all 8 screws had to be trimmed in length and deburred and all 8 slots in the sides of the MCPCBs had to be made wider with a 3/32" bit. (What a pain.)



Closeup of the optic positioned over the LEDs. The holes that the four pegs stand in had to be drilled to 2mm depth to position the optic the same height over the MCPCBs to simulate the wider part of the pegs sitting on the MCPCBs. This puts the optic at the correct/optimal focal distance from the LED surface.



A preview of what it will look like fully assembled. (Nothing's wired up yet.)




I decided to weigh it with everything but the drivers installed. This sucker's heavy - almost 2 lbs.!


Build completed. See posts 9-11.

[simplec6]

Sweet build man.

[LilKevin715]

Lookin good so far. Is the primary heat path via the head, then to the body tube with AS5 on the threads? Or will the bottom part of the heatsink make contact with the body tube directly?

[Hill]

Hi TJ,

Nice looking project. I really like the idea of a multi-XML powered by NiMh C cells. The host is quite long, but given it is a C-body, it's not too fat in the hands. Nice machining work on the sink. In all my builds, this is the part that usually takes the longest, and I understand how much of a pain it is without fance tools (i.e. lathe, etc).

Looking forward to the final product!

[beekeeper5]

Will this be a throw or spot light?

[Techjunkie]

@LilKevin715: The way the HS is shaped, all of its underside, except the cavity in the center, makes contact with the neck and with the head. Besides that cavity, the only surface not making thermal contact with something is the spaces on the surface between the MCPCBs.

@Hill & Simplec6, Thanks

[fortean101]

Where did the quad optic come from? I have a quad XML mag as well but the head is not stock and I want to swap if for a standard one. Lovely build by the way!

[Techjunkie]

Will this be a throw or spot light?

4 XML in a 2" head will be unavoidably floody, but the optics will make it as much of a spot as possible with the given dimensions. (Not counting 4x20mm aspherics lenses, one suspended over each LED)

Where did the quad optic come from? I have a quad XML mag as well but the head is not stock and I want to swap if for a standard one. Lovely build by the way!

DealExtreme.com. Thanks.

[Techjunkie]

Making progress...

Emitters wired up:


Regulators prepared:


It's hard to see in the pic above, but there's a drop of epoxy holding the mode signal wire in place on each board, and another drop covering the ground ring nearest to the LED- connection, to insulate that part of the ring and avoid an accidental short to ground while soldering the LED- wire in place.



Switch mods. Pass through for LED+, anti-bounce mod.

[Techjunkie]

Construction continues...



Looking at the pics of the wires hanging from beneath the heatsink, I'm reminded of a giant jelly fish. I think I might call this one the Man O' War.








And now for the moment of truth... tailcap mesurements

High:


Medium:


Low (like map-light low):

[Techjunkie]

And now, without further ado, the requisite beamshots. All shots taken at 1/10 sec. F3.5 ISO 200 WB=5200K (daylight).

The candidates for comparison. From left to right:
1) This Mag 4C with 4x 5000K XML in quad optic, driven @ 3A each
2) A Mag 3D with 5x MCE 4A in SMO pentaflector, driven to 2.8A each
3) A Mag 2D with 3x SST50 4500K in SMO triflector, driven to 4.2A each




Control, lights on:


Control, lights off:


This light (4x XML quad optic)


5x MCE pentaflector


3x SST50 SMO triflector

[Techjunkie]

Next project: Quad in a 2C, powered by 2x AW IMR26500, wired as two standalone strings of 2-in-series, using one KD V2 driver for each serial string of 2S. Board with modes powered from between LED 1 & 2 in string A. Board that regulates string B will be mode slaved from the first board. Not sure if I'll use another quad optic or 4 20mmx20mm SMO reflectors for that build.

[FRITZHID]

that SST is so bright in the pic it almost hurt my eyes!!! nice work!

[Techjunkie]

that SST is so bright in the pic it almost hurt my eyes!!! nice work!

Thanks for the compliment, Fritz. Look again though - this 4X XML is clearly way brighter than the 3x SST50 (XML left, SST right):

[LilKevin715]

As always nice work Techjunkie. What is the tint of the XM-L emitters? I am guessing somewhere around the 3X range like 3C or 3D. I like the heatsinks fior the drivers, should be able to dissapate the heat. Also, how long are you able to hold the light before its too hot?

[Techjunkie]

I discovered that within what seems like less than a minute of running this light on high, brightness would wane noticeably, to the point where high looked like mid and low won't even light. This feels like a symptom of the regulators overheating. A tailcap measurement with the light in this condition showed that on high, current draw had dropped to less than four amps. The head never even got very warm. I immediately took one cell out and measured draw on high with three cells and current draw nearly doubled, going up to around 8 amps. If taking one cell out improves output, then the dimming clearly isn't from the batteries sagging.

My next experiment was to let the regulators cool, then run the light for a prolonged time on medium (3-4A). I let it run that way for several minutes, until the head got hot and then tried high mode immediately while it was still hot. I measured 12A at the tailcap for a while, then within about a minute, brightness began to wane, the same as it had when run as long on high from a cold start. Once again, after the regulators are overheated this way, low mode produces no output.

I've concluded that the 380mA AMC7135 chips on these regulators don't handle their max output of 380mA each very well at all, at least not compared to their conventional 350mA siblings. I can't even blame a high input voltage for making them work too hard to burn off excess, considering that nominal pack voltage is only 4.8V.

With no space to provide the regulators with better heatsinking than I already have in here, I'm left with few options:

1. Do nothing and refer to this as one of those lights where high is actually a "nuke" mode meant to be used sparingly, and medium is the practical "high" mode. While this "solution" appeals to my laziness and protects from me damaging what I've already created, the perfectionist in me will lose sleep over it.
2. Experiment with adding some resistance in the tailcap, to offset the amount of resistance (and heat) required of the regulators to hit the target current. This might work because voltage sent to the regulator is not impacted by what mode the string is in the way it would be if it were in series with the LEDs like in the poor man's method. The challenge will be finding a resistor that's capable of 12A, the right Ohm value, and fits inside the Mag C tailcap.
3. Move the light engine over to a Mag 3C and recycle the fourth cell in another project. Running the light engine unmodified on 3 cells instead of four means that full blast would be 8-9A instead of 12A, but I'd be able to run on high constantly instead of in short bursts.
4. Replace all four AccuEvo cells with another brand of NiMH C cell that sags more under 12A load. (This is sort of a twist on #2, where the added resistance comes from the batteries.) The challenge here is identifying the right cell for the job.
5. Replace the four KD V2 regulators used here with four of the original (KD V1, download, shiningbeam) 8x7135 multimode regulators and hope I'm right about them being needing less heatsinking.
6. Pot the regulators I've used here in PC Fahrenheit thermal putty epoxy and hope that helps. I've been lucky with unsinked regulators not overheating before, and think that stuff may have been the secret to my success.

2 is easy and inexpensive to attempt. 3 is slightly risky and not too expensive. 4 could get very expensive, as I tend to need to build lights for all my leftover batteries and who knows how many I'd try before I succeed or give up. 5 and 6 present the most risk of me damaging the other parts beyond repair.

[Techjunkie]

As always nice work Techjunkie. What is the tint of the XM-L emitters? I am guessing somewhere around the 3X range like 3C or 3D. I like the heatsinks fior the drivers, should be able to dissapate the heat. Also, how long are you able to hold the light before its too hot?

DigiKey lists them as 5000K. The mfg p/n ends in E3. Not sure beyond that.

I liked the way my driver heatsink looked too, but it appears to have been insufficient (see previous post). The light gets nice and warm, but never too hot to hold (I wish it would) - the thermal regulation issue I'm having with the regulators doesn't allow 12A mode to be run long enough for the LEDs to make the light get uncomfortably hot.

[Techjunkie]

I just read the specs of these new V2 regulators:

Voltage support: 2.5V-4.5V
Battery types: 1xLi-ion or 3xNiMH/alkaline

Doh! When did that come down from 6V, 4x NiMH?



I can still try #2 by adding resistance just to the input voltage on the driver (If I had any idea how few mA the driver itself draws or how much to drop it), but I have a feeling the only real solution here will be #5 - replace 'em all.

[LilKevin715]

The Vin range of the AMC7135 linear regulators is still the same irreguardless of 350 or 380ma versions of the chips (see spec sheet). From another older thread (scroll down to post #56) it appears that the chips are overheating big time. Those Accupower cells are being discharged at approx 2.6C, so they should be sagging pretty decently (would need to measure while under load). I'd imagine/assume Vin @ 12A would be in the neighborhood of around 4v. Assuming a LED Vf of ~3.3v (cree data sheet) the 7135 chips would have to burn off somewhere around 6-8 watts of heat.

My view on your options:
1) I wouldn't accept nothing less than a full 12A.
2) Kind of half... you know what I mean. It could be done though. I = V/R so a resistor in the neighborhood of .025 to .040 ohms, depending on how much the accupower cells sag under load (need to measure).
3) See #1
4) Ehhh, I'd stick with quality cells. A slight twist: go for a 3D mag with the 10A Accupower cells?
5) Won't make a difference IMO.
6) The C sized tube wont allow much room for heatsinking. I've potted my drivers directly to the main heatsink (although nowhere near as much heat as yours) in my builds (see SST-50 link in my sig).

[Techjunkie]

I've just exhausted most of my options...

I tried different cells of different resting voltages, then I played with different resistor values and happened upon one that made the regulators drive the light to 9.8A. After a very short while, it dimmed again. I unassembled quickly and promptly burned my hand on the driver heatsink. I've never felt a regulator that hot before, even the one that was overheating in my triple SST50 light that I fixed successfully with a similar heatsink.

I'm going for option #5... just not tonight.

[Techjunkie]

I did some rough estimate math and calculated that if the cells are dropping to 1.05 Vout under load (guessing at this figure based on what Eneloop AAs and Titanium PowerMax 1800 AAs drop to under 10A load ), that's 4.2V. Each emitter needs only ~3.3V to reach the target max current. That means to maintain max output, the regulators need to drop .9V @ 12A, and shed 10.8Watts of heat. I realize now that's asking a bit too much considering the heatsinking available to them in this space.

I've just ordered some 5W 0.15 ohm cement resistors. Two placed in parallel in the tailcap will equal one 10W 0.075 ohm resistor. Using the math above, that should drop the voltage on the pack 0.9V @ 12A and land the target voltage right where it needs to be, leaving the regulators with little to no work to do.

12A = .9V/0.075ohm
0.075ohm = .9V/12A

I'll have to heatsink the resistors in the tailcap as best I can, considering that they'll be run slightly over spec when the batteries are fully charged and the light is on high.

Before arriving at this (latest) conclusion, I had already desoldered the regulators, from when I was planning to replace them with the older versions. While I'm waiting for the resistors to arrive, I'll see if I can solder all four into a short section of copper tube, which should improve their heatsinking somewhat.

Wish me luck.

*EDIT* I think I can fit 3 x 0.22 ohm 5W resistors in parallel in the tailcap. At 12A current, they'll collectively drop voltage 0.88v, leaving the regulators very little work to do. Each resistor will run at 3.5W, which will be in spec for them. Each regulator will have to burn off 0.48W, or 1.44W collectively, which is waaaay cooler than the 10.8W they'd have to burn without the resistors.

[FRITZHID]

IDK if its my screen or what but the SST may be smaller, but it looks brighter... maybe it's just the color temp?

[Techjunkie]

IDK if its my screen or what but the SST may be smaller, but it looks brighter... maybe it's just the color temp?

I'll grant that the SST has brighter spill, but that's because the XML has almost all that spill concentrated into the giant, brighter hotspot. The ceiling bounce test says its no contest. I was shocked at how much brighter it made any room in my house compared to the fixed lighting.

On to the updates...

As I mentioned, I'm re-doing the same drivers while I wait for the resistors for the tailcap to arrive. Side note - It occurs to me that if the XML weren't so low Vf, and/or if I'd not bypassed the switch, and/or I'd used cells with more internal resistance, I probably wouldn't even need the resistors to bring the pack halfway down between 4 cells and 3 cells.

Rather than just reassemble them as they were before, Here's what I've come up with in an attempt to improve heatsinking...

Run 20awg stranded copper through the three sets of large via holes at the outer ring, and solder to both sides of every board, connecting all the boards thermally. Also solder the foundation of every AMC7135 chip to the outer ring and those wires stringing the boards together, using liberal amounts of solder:


Next, cut a piece of copper pipe, customizing diameter to fit inside heatsink base. Press new driver stack and wires in place, applying liberal amounts of HS paste between the inside of the tube and the new thermal bridges created with all that wire and solder.


Fit into place, making sure the switch will line up with the switch hole when everything is fully tightened.



I'm very happy to report that everything survived this operation. I've also noted that on high, my amp meter shows 11.6A in the forward direction, and -12.9A when wired in reverse. I see now that it's accuracy above 10A is shaky, unless I take the average of those absolute values, which is just slightly above 12A.

I'm less happy to report that installing four fully charged cells has confirmed that all this new driver heatsinking has alone not solved the problem. Here's hoping that the resistors do, or I think I'll just end up swapping the 4C tube for a 3C tube and call it a day.

[LilKevin715]

The AMC7135 linear regulators are relatively cheap and flexible (somewhat), but they do have their limitations as you have found out. Have you tried the light outside yet? I am curious how much throw those optics provide.

[Techjunkie]

The AMC7135 linear regulators are relatively cheap and flexible (somewhat), but they do have their limitations as you have found out. Have you tried the light outside yet? I am curious how much throw those optics provide.

I have, but I haven't taken beamshots outdoors yet. Throw is on par with my triple SST-50, but with an entirely different beam profile. The hotspot of this light is a huge rounded square that reaches as far, perhaps slightly farther, than the smaller hotspot of the SST light. These optics gather a lot of the light that would be spill in a reflector, and add it to the hotspot.

[CKOD]

IDK if its my screen or what but the SST may be smaller, but it looks brighter... maybe it's just the color temp?

The image is saturated/washed out in the center. its impossible to compare the brigtness unless he stopped down the exposure or added filters

[netprince]

I like this build a lot. That optic is really interesting. A shame about the overheating drivers. Have you considered a different driver all together? My first instinct would have been one of the taskled drivers, directly attached to the bottom of the copper heatsink.

BTW, sorry if I missed it, but what is the sku on that optic? I might like to try it sometime...

Thanks!

[Techjunkie]

The image is saturated/washed out in the center. its impossible to compare the brigtness unless he stopped down the exposure or added filters

I agree. I've charged the cells in all the lights and will take some better beamshots for comparison soon, maybe tonight.

I like this build a lot. That optic is really interesting. A shame about the overheating drivers. Have you considered a different driver all together? My first instinct would have been one of the taskled drivers, directly attached to the bottom of the copper heatsink.

BTW, sorry if I missed it, but what is the sku on that optic? I might like to try it sometime...

Thanks!

Thanks, netprince. I don't think I could fit 4 Task LED drivers into one light. (Keep in mind that pack voltage is only 4.8V, and I want each LED regulated individually at 3A, not 2x2 in parallel at 6A or 4 in parallel at 12A. That's why four are needed.) Also, I'm not sure how to do the mode-slaving on them in the same way as I'm doing on these, and cost is a concern too. DX sku for the optic is 01915.

[Techjunkie]

New stopped down beamshots with batteries fully charged on all lights. Current draw shown in photo captions confirmed with talcap measurements immediately following the shoot.

[LilKevin715]

Nice beamshots. Those optics really do provide a good amount of throw. Not to mention they are more efficient (~90%) compared to regular reflectors (~75-80%).