LEDs in Showcase Miniatures Signals

[C855B]

One of the concerns expressed about the Showcase Miniatures searchlight signals was the reliance on fiber optics to light them up. An okay solution, but not ideal.

I've been playing around with chip LEDs for a few years, and couldn't help but think there was a way to get 0402-sized (0.04" x 0.02") LEDs in the head. In the meanwhile, 0404-sized RGB LEDs became available, which seemed like a more straightforward solution than trying to corral multiple grains of salt. So... after reaming out the fiber path from the front to 0.060" (but not breaking through the back), I was able to get this result:



It took a bit of intense time under the microscope to get leads soldered on, and a few RGB chips gallantly gave their lives before I had a working solution for attachment. There were three non-obvious "secrets" to the process. First was using 0.020" eutectic (63/37) solder. What this solved was reducing the amount of heat needed for a clean joint.

The second was to attach the wires flat against the chip and also that the cut end of the wire faced "out". This way the wires could be trimmed to not go outside the square of the chip, and the resulting leads could form sort of a bundle emerging from the center. As you can see by the poor iPhone snapshots, I don't have a means to show how the wires attached to the chip. I can draw something if somebody is puzzled by my awkward explanation. I only wired the R and G contacts since at the moment I have no use for the blue.

Anyway, the third trick was to paint the inside of the hole (I used white Scalecoat II) to insulate the white metal head from the wires and chip contacts.

I'm mostly happy with the outcome. It needs a lens, and I'm wavering between a slice of 0.060 fiber optic or a drop of clear paint. Once it gets the lens it will be mounted to the mast.

The remaining issue is yellow/amber. I lit red and green at the same time, which in theory is supposed to result in a yellow - not so much, you can see that the red and green are separate elements. Possibly a lens will provide some blending, but I doubt it will solve the entire problem. I have a PIC test rig in my signals project box, so one possibility is PWM strobing to fool the eye into seeing yellow. The end game may be contracting with the chip supplier to produce a R-Y-G chip.

Bottom line is that this technique appears to work, and with practice might be a reasonable solution to produce highly-detailed and accurate searchlights that light-up nicely.

[peteski]

Cool!  Do you have any pictures of the secret part of the process (showing the wires soldered to the back of the LED?

EDIT: Nevermind - I didn't read the writeup carefully enough.    No such photos exist.

I would be leery to depend on a coat of paint for insulation. I would either use a coat of CA glue, epoxy or maybe even coating the LED with few layers of clear fingernail polish (it is thicker than regular paint and quite durable since it is designed to be wear resistant).  For lens I would probably use some 5-minute epoxy (maybe mix some talcum powder into it to make it slightly milky to better diffuse the light).

[C855B]

Thanks! I wish I had pics of the lead attachments, and, yes, of course I'm regretting not buying the USB camera that plugs into one of the microscope eyepieces. It was possible to solder the leads to the corners similar to what you do with edges of the 0402 chips, but for what I was doing it added too much footprint for the hole. Think of bending the lead into a tiny 'L' with the crook of the L in the center of the chip, the leg attached to the backside of the solder pad on the chip rather than the end.

Oh... these LEDs are bright! The test rig is setup for 4mA, and is what you see in the pics. The chips are rated for 10mA, so factor accordingly!

To do a poor man's R-Y-G, I was playing with arranging 0402 chips in a triangle (some fancy logic to drive it), but that exceeded the 0.060" hole limit. Unless the R+G=Y solution works (which I doubt) custom 0404 is our only solution, it appears.

[C855B]

Now seeing your edit... how is a reasonable paint worse than the enamel coating on the wires? I was mindful to use a solvent-based paint and let it cure for a day; I wouldn't do it with an acrylic since as you suggest they remain soft. Since there really isn't much mechanical load here I didn't see much of an issue, and thickness is not our friend given the microscopic fitments.

[peteski]

If you had a way to measure a thickness of a layer of Scalecoat II and a layer of enamel on a magnet wire, the Scale coat paint is much thinner. It is also much less resistant to abrasion, especially since it is a metal-to metal contact (and one of the metal pieces could have sharp edges (like the wire you cut.  Just look at it under your microscope. 

I realize that I'm a bit anal, but that has never failed me. 

[C855B]

??? - under the microscope the paint was pretty thick - a lot thicker than the wire coating. It also may have helped that I used white, which is going to be heavier/thicker because of the higher pigment density. Anyway, I've had pretty good luck with paint before, so it still seems a viable solution. If one of the pads contacted the case it wouldn't be the end of the world, as it would take two failures before there was an issue. The way the chip rests in the hole, two points failing are slim enough to be a lesser issue than, say, breaking a lead in handling.

I did a lot of debating with myself about the insulation issue before arriving at the paint conclusion, including your CA idea. At this size there's not a lot of tolerance for "floating" the chip in a sea of CA, and immersing the fragile chip/wire assembly in something with more body just felt like it was asking for a wire to break or pull off a pad. EDIT: Also, inserting the leads and chip assembly "wet" was asking for too many stars to align, and much opportunity for mess or error. With the dry paint inside I could concentrate on feeding the #38 wires and nesting the chip gently. Seems to work, so I'm not going to ask too many questions at this point.

I sent more detailed project notes to Debbie and Walter at Showcase earlier this evening. They've been trying to make something similar work, so I expect to hear back from Walter in the morning when he has a chance to peruse my process.

[peteski]

When I deal with assemblies like this I usually "paint" the exposed electrically active metal areas with CA glue (applied with a sharp toothpick).  Then with another toothpick I apply some BSI brand accelerator. I use regular thickness CA (not thin or thick versions). I might repeat this process couple more times to build up some more thickness (if needed).  That gives me enough insulation I feel comfortable with.  Once the metal is insulated I could comfortable immerse the assembly in a drop of epoxy inside the signal (to make it a permanent installation.

[C855B]

Here are the results, a finished signal:



This is still with the 0404 red-green-blue chip. I created a lens with Tamiya "clear" transparent paint, and it did nothing for color mixing to fake a yellow. So next on this project is a R-Y-G made of 0402 chips.

This kit is advanced stuff. While the head is the same one pictured earlier, I had to abandon the original assembly of the full signal since I couldn't get all the ladder supports to align for gluing if they were threaded on the mast first. Well, maybe I could, but it would take five hands to do it. I also learned it was a bad idea to pre-paint subassemblies... the fitment tolerances are exact, and even the thickness of a single coat of paint meant certain parts would no longer fit together.

I like the results, which I expect to improve with practice.

[ryan_wilkerson]

That looks great! How much material did you have to take out of the inside?

[peteski]

This is still with the 0404 red-green-blue chip. I created a lens with Tamiya "clear" transparent paint, and it did nothing for color mixing to fake a yellow. So next on this project is a R-Y-G made of 0402 chips.

Instead of just clear paint you could have tried (like I mentioned earlier) making a lens from 5-minute epoxy  with some talcum powder mixed in.  That would result in a milky-translucent lens. Kind of like the "diffused" resin the use for the larger 3mm or 5mm dia. red/green LEDs.

[Sokramiketes]

I've been messing with the RGB 0404's as well.  I've found a white diffuser, like .020" styrene, works better than clear.  The LED is plenty bright to shine through, and the white diffuser helps hide some of the color mixing going on.  We're playing with an add on board to drive these new LED's off our existing signal driver, but first need to nail down the proper resistor values for each shade.

[C855B]

That looks great! How much material did you have to take out of the inside?

A fair amount, but it wasn't a complicated process. Basically I reamed out the fiber path from the front with increasing-sized drills until getting to a #53. The 0404 drops into the hole with clearance. Trick was not drilling through, just leaving a cone in the back of the casting. There was enough of a hole there to route the wires out. A side benefit to the reamer/drill approach from the front was gradual thinning of the visor, which as cast is sort of chunky.

New version with three 0402s I'm working on now is done by drilling all the way through from the front, then from the backside carving out the casting. I'm using a #83 drill as a mill bit. Only broken one so far. Plan is to use a bit of .005" styrene for a new back, gluing the LED "pack" to it to ease installation.

Pete, I wasn't ignoring your idea, just didn't have any talc handy. I ham-fisted this thing enough to consider it "proof of concept". I'm more inclined at the moment to pursue the 3-LED approach with the mast common anode, so once that is thumbs-up/thumbs-down/thumbs-punctured-with-a-#83-drill-bit, I'll decide then if the RG with translucent lens approach is better.

Mike, rather than the RGB 0404s, I am thinking that 0402 red-green pairs back-to-back like the old-school discrete red-greens has an advantage in only needing two wires down the mast, doing yellow with AC. Certainly easier to solder versus the tiny lands on the 0404 RGBs.

[Sokramiketes]

I gotcha. I'm striving for a better amber than the old bi-color LED approach. That's what led me to using the PLCC4 package (BeNscale) style LED with signal green and good yellow. I was willing to sacrifice the head detail for the right colors. Now I hope the RGB 0404 will allow both scale and color.

[peteski]

I gotcha. I'm striving for a better amber than the old bi-color LED approach. That's what led me to using the PLCC4 package (BeNscale) style LED with signal green and good yellow. I was willing to sacrifice the head detail for the right colors. Now I hope the RGB 0404 will allow both scale and color.

Are you actually planning to use the blue LED in the mix?  Or just green and red?  The only possible use I can see for blue is to get a better green hue.

[C855B]

Hmm. If it matters, the 'G' in the 0404 RGB chip I'm using (Rohm) is already 527nm. That's signal green, no blue necessary. That's why I thought the back-to-back R/G 0402 chips instead would be a reasonable solution. :shouldershrug: