I picked up the YinHex YH-X9 from the Wallbuys Chinese New Year reservation sale. For $46 (reg. mid-$80s) I got an APEX 5T6 host with five XM-L T6 emitters and probably the weakest drive you can imagine. I've driven one XM-L harder than all five of these were being driven. In fact, it is so bad that it's obvious the 'designers' were trying to pull a fast one here.
This mod is very involved. There are at least three different sections to it. The first part I'll cover the driver and emitter board upgrades. Next is the thermal path issue. Later I'll cover the optical and finally the tailcap upgrades.
But first, let's take a closer look at the light itself.
I have to say, this light catches my eye. I went for it based on the format alone. I had no idea about the innards.
In hand, it is thick but fairly comfortable. The grip is good. The tailcap lobes tend to get in the way. I'd say the body is similar in size to the Sky Ray King, with a slightly larger head.
Holding by the stainless steel handle initially doesn't feel secure, but it's not going anywhere. It's surprisingly sturdy for how thin it feels. This is the most comfortable way to hold it. I use the handle as a belt/pants clip as well and it stays in place just fine. You can one-finger hang this light at any angle from slightly upward to almost straight down.
Looking at the head, we can see a nice set of heatsink grooves that seem to do the job. This light gets hot fast, so the handle comes in handy, unless you need to warm up your hands.
OK, now what was I saying about the driver? Oh right, it sucks. Hard.
Part 1: Driver Mods
Behold! The YinHex solution for driving five 10W emitters.
Count em up. That's 10 AMC7135 current regulators, for a whopping 3.5 Amps. All that current is force-fed into five XM-L T6 emitters wired in parallel. The driver designers must have forgotten to do this step; 3.5A divided by 5 emitters equals 700mA per emitter. WOW!
OK, clearly I'm not going to leave it like this. So, I cook up something quick just to see what I can get out of this thing.
The picture below is mostly self-explanatory. I added 10 more 7135 chips on top of the original 10. Also, the PCB layout limits current flow through two rows of traces for Ground and L-. The grounds for each bank connect on one end, so i added a second path on the other end of each. For the L- The right bank was OK as it is close to the L- pad. The left bank has a long thin trace to pull current from L-, so I added an extra one on the far end to help things out a little. Since each bank regulates 3.5A each, I wanted to minimize loss as much as possible.
For the B+ to L+ connection, there are only 12 vias. Since my goal is 14+ Amps, This will be a bit of a bottle-neck. Unfortunately, there's no easy way to run a wire around to the other side. My solution, drill through to the center.
I had to drill as close to the center as possible so the wire would not interfere with the batteries. They occupy the space almost up to the inner edge of the red ring.
Looking on the other side, I drilled closer than I wanted to the chips. I drilled away most of the Vcc trace for some of the 7135 chips. Fortunately, enough is left so it still works.
The comes through and goes directly to the L+ pad. This adds a little relief for the vias.
OK, so that's 20 chips, where to put the other 20. Also, we need some control for modes. I wasn't ready to stack 7135 chips four high, so I added a couple NANJG drivers.
Thinks got a little hairy here. Each board has 10 7135 chips. The blue generic is a slave. I bought a few of these generic drivers from Fasttech and I do not like the modes. They are not clones of NANJG, have a very slow mode change timeout, and there's visible PWM. I've relegated them to slave use only. To make them slaves (or single-mode drivers), remove the 8-pin MCU.
The other driver is a real 2-group NANJG 105C with 2 extra chips. I ran out of 350mA chips so I put on 380mA ones instead.
I did mention that this is a pretty involved mod.
I drive this with 4x Sanyo UR18650ZT 2800mAh cells. These seems to cope with the load fairly well and they provide a little extra voltage which is good when dealing with these currents.
Five XM-L T6 emitters at 2.82A each. Emitter lumens would be close to 4200. Not bad. Unfortunately, a lot of this is lost due to heat sag. In the integrating box, I see this as a very rapid drop from 3280 lumens down to 2870 after just 10 seconds. By 30 seconds, it's down to 2685 lumens.
Update (Mar 22): User dorpmuller took a similar but more ideal approach, that I will probably steal adopt when I solve the thermal issues. Basically, five 10x7135 drivers slaved together, each driving one emitter. dorpmuller's comment is below, link here. My initial approach may be slightly cheaper/quicker, but I'd prefer to drive each emitter separately.
Part 2: Thermal Issues
Heat does transfer to the head quickly, just not enough to keep the emitter cool. There's 46W at the emitters, and around 38W or more is heat. It takes a lot of contact to move that much heat. Unfortunately, we only have a ring of contact:
And on the emitter plate side:
That's not much contact for about 40W to pass through. I cleaned it up and used silver thermal compound when I screwed it back in.
The emitter plate was screwed in with a total of two screws... I had to drill the head and add three more. I also added more wire for each emitter as the PCB traces ran quite far and didn't appear to balance out the current well. The LED on the left would get more than the LEFT next to it in the clockwise direction. I scraped the traces and soldered 20ga wire so there was a little better current spread.
OK, back to the thermal problem. Fortunately, the emitters are close to the edge. This does help, but the emitter plate is about 25C hotter than the head when running. Not good. Add in the dielectric separation for the emitters and they are probably 45C hotter than the head. This design screams for copper. To do this properly requires some machining to make a copper base for the LEDs. I think copper SinkPads would be a plus too. ;)
Thermal test results:
Time | Temperature (C) |
0s | 20.2 |
15s | 22.8 |
30s | 25.2 |
45s | 27.0 |
60s | 29.4 |
90s | 33.0 |
2m | 36.0 |
3m | 42.4 |
4m | 47.8 |
5m | 52.2 |
Part 3: Optical Issues
Updated (Mar 22)
Basically, there's a problem. Everything in front of the reflectors is unnecessarily blocking light. Sorry, bad phone picture. This is how it came; nothing was aligned which made things worse.
First there's the reflector gasket (the black thing with five holes) made of smelly rubber. It is about 1.5mm thick with holes cut without a bevel, and slightly smaller than the reflector inner diameter.
My solution? add a bevel by cutting away some of the material. Steady hands (locate a surgeon) and a really sharp X-acto knife (ask him to bring his scalpel) required.
Repeat. Five times. This mod added about 130 lumens OTF. Not bad for 15-20 minutes of surgery.
Next, the lens is standard uncoated glass. Better than plastic, but AR coated would be nice. Wait a minute! I ordered two of the Leica UCL lenses for the Defiant 3C Super Thrower. I still have one left! Will it fit? A quick measurement... YES!!! It should fit. Dig around a bit, find it, and drop in the lens. SWEET! It's thinner than the stock glass, but it applies enough pressure to hold things in place. Perfect! This gave me another 100 lumens OTF.
So far, 230 lumens gained for $6 I already spent; not bad.
Another lumen killer is the bezel itself. It has a rather wide O-ring lip that covers a small part of the reflector. it also scatters some of the spill. Tests with and without the bezel show a potential gain of 100+ lumens. I have not attempted anything here yet. Machining things is not my strong suit. I could use a sanding drum and slowly remove material, but for now I will leave this one alone.
Part4: Tailcap
Updated (Mar 22)
The tailcap mods are pretty standard stuff. I decided the only thing I had to do was add copper braid to the springs. We start by opening the tailcap. Not easy. It's glued and the threads are mashed. They didn't want the tailcap to turn. I think this was for maintaining proper alignment of the handle. In any case, the handle helped me use leverage and brute force to crank it open. Heavy rubber gloves with a lot of grip helped.
Inside the tailcap is a plastic retainer for the switch boot. it's nice to know that a damaged boot can be repaired.
On the body, we have a switch mounted on a PCB with four screws holding it to the body. Three of the screws were loose. These are what provide electrical contact to the body, so I made sure they were tight when I put things back.
I touched up the switch solder joints too.
Removing the four screws and we can access the other side to braid the springs.
Once it was all back together, I checked the voltage drop across the switch while running on high. I measured about 80mV at 14A, which works out to 0.005 Ohms. Not bad at all for a 1.5A rated switch!
Conclusions
The light as shipped is terrible. I would not recommend it for anyone planning to leave it as is, especially at regular price. For the non-modder, I hear the Blackshadow Terminator is good. Or save a little money, and get a good Sky Ray King; no handle and smaller.
As a mod host, I recommend it for someone looking for a challenging mod on a light in this format.
More to come; stay tuned.
Thanks for reading! searchID8935