New Searchlight Signal Idea

[dcutting]

Hey guys,

I saw a big gap in the scale a while back. What we need are low-cost and high detail signals. What I can't take are the giant RGY LEDs on the back of the signal head, or having to pay $10 for a single LED that will fit inside the head. I think I have the solution.

I recently put together a test with a tiny RGB LED on a PCB that is the same cross-section as the signal head. The LED can be had for under $1 and anyone can solder it, no microscope required. Now, you might be asking "why RGB?" It's simple. With RGB you can blend the colors to make any color you want, just like a computer monitor. Instead of blending red and green to get yellow, you can change the amount of each and add some blue to get the exact color you want. But in order to adjust the colors, you need a PWM output.

So crafted a small, inexpensive circuit with a very small and cheap microcontroller that will output PWM signals to the lights. With this scheme you could display Red, Green, Yellow, Lunar, etc. elements all in one signal head.

Andrew Kyle has run several tests with this setup, and we have concluded that it will be the most versatile and to-scale approach to signals, in addition to being affordable.




I designed the first version of our new signal system around this idea, and here is a photo of the CAD design. More photos to follow as time goes on.



I'd like to hear your comments and critiques, as well as some feedback on other signals you'd like to see done this way.

David

[auburnrails]

I think it looks amazing.  I wish I wasn't modeling a 1.25 mile long industrial spur now!  Seriously, though, seeing how some of these guys are buying and installing signals by the dozen on their sprawling empires, this could be a big money saver and a good product as well.

-Dave

[C855B]

Green and red are easy. My experience with RGB chips used like this is they have have poor convergence, and are unable to show a true signal yellow without perceiving the component color artifacts. These chips are intended to be used in arrays to build larger displays, where viewing distance provides the convergence.

Please prove me wrong - I'm not happy with the $10 solution, either, but I was less happy with the poor yellows from the RGB chips.

[dcutting]

The RGB chips have had pretty good convergence as they are so small and we bought the shrouded lens version. I am going to have Andrew run some more tests tomorrow and get some pictures. The lens we are using also isn't perfectly transparent, more just translucent. I think we should be fine overall.

[Philip H]

I'm all for anything that makes n scale signals a more prototypical size. That said KCS in my area and era was a three light pre-'type D signal territory. So this doesn't help much for mainline signals. It COULD help for hooded dwarves I see chemical plants and on yard leads and Spurs.

[C855B]

OK, when you have demo photos of yellow and lunar I'd like to see them.

As far as Type SA signals in general, with the exception of the LED cost I think the Showcase versions are excellently rendered and reasonably priced. They also happen to hit the prototypes I'm interested in. That said...

I'm with Phillip. If you want to talk about a vacuum in N fine-scale signals, that is going to be decent three- (and two-) light heads. US&S one-piece cast-backs, especially. As a signals fan (and long-ago signal maintainer), I am not getting the preoccupation with searchlight types, which seem to be done over and over. Near-scale three-light signals in US&S, General Signal and Safetrans aluminum and plastic backs would be killer. And with 24", 30" and 36" targets.

Just sayin'.

[dcutting]

Thanks for the tips. I hope to bring out some of those other types soon. I started this signal as a project for a friend.

David

[peteski]

At the LEDs size David is using (tiny), the convergence should not be a problem. Even is it is, place a drop of epoxy with some talcum powder mixed in on top of the LED trio to diffuse the light.

What I don't care for is that you need a freakin' micro-controller for each 3-color signal.  To me that is crazy (even if it only costs $5).

[C855B]

The microcontroller, done right, can actually serve a purpose such as simulating vane movement - flash of red between green and yellow, and settling into red, plus approximating the filament warm-up/cool-down when in flashing modes.

[peteski]

The microcontroller, done right, can actually serve a purpose such as simulating vane movement - flash of red between green and yellow, and settling into red, plus approximating the filament warm-up/cool-down when in flashing modes.

I realized that, but in my mind that is just an overkill. But yes, if you want the ultra-realism, then that is the way to go.  I seem to recall that another TRW member has such a system on their layout.

[railnerd]

Interesting…  seems to match up with my own experiments around MSS + NeoPixels.  The normal big issue— which you seem to have solved by using a really small LED— is that the larger RGB LEDs don't mix color well without a diffuser.

What cost are you projecting?

-Dave

[sp org div]

I dont think its the $10 price of an LED that slows me down, as much as it is the assembly needed of putting multiples of signal kits together... I'd gladly pay more for slaves to build a quality assembled H2 signal, and I think that would fill the biggest void... I got a box of Showcase signals that have been patiently awaiting assembly, but it is the construction of them that has currently slowed any further progress.

Jeff
http://espeeoregondivision.blogspot.com/

[randgust]

Hmmm.

To a certain extent I think you've invented the wheel here, although the original wheel may no longer be available?    I used BeNscale etched parts combined with a Tomar tricolor LED (you can buy the heads separately as a special order, prewired, for the ham-handed among us).  End result, however, is pretty darn spectacular on the layout.   http://www.randgust.com/winsig5.jpg

Yes, signal 285.5 (Winslow) has the head on so that the ladder is in the front.  And they are abnormally high.

The Tomar LED's have a rather nice yellow at about a foot away, closer than that and you can see convergence issues. 

Stainless is definitely the right material as anything else just won't take any on-layout abuse at all, let alone the casualties likely during assembly.   I'll admit this is on the ragged edge of what I can assemble from parts and make functional.    I really wouldn't want to have to solder up a control board just to get a yellow aspect, don't need anything more.   I'll pay $5 more just to get the LED properly soldered onto leads though, and also provide a table of resistance values based on input voltage.

I've put the LED's only into heads from BLMA, those worked wonderfully well too.  Did two full ATSF BLMA signal bridges and I really ought to redo the first one I did that had single-color LED's in two different heads.

[dcutting]

Hi Everyone,

Just wanted to give you another update on this project. At this point most of the design is all final and I have moved into the optimization stage. Basically trying to reduce my costs so you can buy them cheaper. That meant a redesign of the PCB to control the LED. Here's a picture:



I added some new important features this time, the first being that on the bottom of the board underneath each section there is a solder jumper that allows you to select operation with a common anode or a common cathode system. So your system will work our PCB, even if it is the atlas system or whatever oddball system you may have.

The second major change is the addition of 3 head drivers to the board. Our first version was the same size as this board, and only had one head controller. This version can control 3 heads. If you buy a one or two head version, one or two sections of the board will not have components soldered.

The section all the way to the left is the power supply. Basically you can feed any kind of power into this thing, be it DC, DCC, or AC (6-16V) and the power supply will filter it and regulate it to the correct voltage. That means that modular groups can use this signal system without having to purchase separate regulators, which in the end is another cost-saving measure.

With this version we sourced a new microcontroller which costs about 1/3 of the first one we used, but still has enough peripherals and memory.

The rest of the design is also coming along. We have several more versions to create in terms of heads, but work is progressing in getting the injection molded components together at the lowest price possible. I'll have more updates soon.

David

[dcutting]

The microcontroller, done right, can actually serve a purpose such as simulating vane movement - flash of red between green and yellow, and settling into red, plus approximating the filament warm-up/cool-down when in flashing modes.

/>

Thanks for the idea! Once we get our hardware all put together I'll start testing this in the code.

David