As long-time readers (or others who know me) know, I have a bit of a fascination with layout lighting and making it part of the overall operating day experience. Plus, it’s easier to work on a layout – either from a construction, detailing, or operations standpoint – when there’s excellent lighting. So in today’s post, we’re going to look at what I settled on for lighting the Copper River.Continue reading
In addition to track power, which is something most of us have provided by our DC throttles or DCC systems, most of us have a ton of accessories. That can be as basic as things like switch machines or a hodge-podge of building lights, animated features, signals, sound modules, etc. Nearly every single one of those is going to have its own requirements in terms of voltages, currents, etc. Most layouts I’ve ever been on solve this by a maze of power strips, wall warts, battery packs, and old DC power packs repurposed once the owner converted to DCC. It’s a mess.
As an electrical engineer, some things about how people build layouts bug me far more than they should. Messy, disorganized power systems are definitely at the top of the list. I thought I’d give you a look at how power is distributed around my layout to run everything that’s not the track.Continue reading
The real CR&NW was completely timetable & train order operations, with no block signals of any kind. There’s a lone, uncredited reference in Wikipedia about the CR&NW having at least one “wigwag” crossing signal, but I’ve never seen any evidence to support this. Given the limited number of trains operated and the generally poor condition of local roads at the time, I sincerely doubt that the CR&NW ever had a single circuit for anything. At least as of 1920, this is confirmed by the ICC’s “Annual Report on the Statistics of Railways in the United States”, where the CR&NW has nothing under the cost line items for “Signals & Interlockers” and “Signals & Interlockers – Depreciation”, and “Crossing Protection”. There’s the possibility of them coming later, but I still doubt it. (I would love to be proven wrong, however. Anyone?)
Update (Oct 16, 2017): Turns out, I’ve been indeed proven wrong! See Robert Hilton’s comment below. There’s a photo in an old Magnetic Signal Company catalog of an overhead, lower quadrant wig-wag in Cordova on page 9. Now I have a plausible reason to build a working wig-wag for the layout.
The model CR&NW however, having evolved into a modern heavy ore hauler, would almost certainly have block signals. In my “alternate history” leading to the present day, the railroad underwent extensive modernization and reinvestment in the late 1940s / early 1950s. Radio dispatch (which didn’t become widespread until the 1960s-1970s anyway) would have been nearly impossible, given the remote country and deep canyons traversed by the line.
For inspiration, let’s look to a pair of near-contemporary prototype ore haulers in the far north – the Quebec, North Shore & Labrador and the Cartier Railway (Quebec Cartier Mining or QCM). The QNS&L was built between 1951-1954 and was equipped with CTC from the start. The Cartier was built several years later, in 1959-1961, but it too was equipped with CTC from the start. Clearly equipping a remote ore line of a few hundred miles in length with CTC isn’t beyond the realm of feasibility. Plus, I have a serious fascination with signalling, so…
Given a modernization date in the 1940s/1950s, searchlight signals would have been the standard of the day. (Again, looking at the QNSL and QCM, it’s searchlights all around.) The US&S H, H2, H5 and GRS SA were both extremely popular and were the most common type of signal installed all over the US and Canada during the 1940s through about the 1980s. Recently they’ve been falling in record numbers, as their inherently mechanical color changing mechanism (a relay with three small color lenses) requires regular inspection, testing, and maintenance, as opposed to modern three-light heads. The preference for searchlight type signals works out just fine with me, since they’re probably my favorite signal type and they minimize the number of wires or fibers that need to go to each head. Showcase Miniatures / Century Foundary makes an absolutely beautiful N scale searchlight kit. They’re lit with fiber optics, which allows them to be very accurate in terms of scale. (Oversized N scale signals really, really bug me…) I’d purchased a couple of their kits some time ago, so I pulled one out tonight and built it. It really is a work of art and not nearly as hard to assemble as I’d feared. (I still have some fear of doing a double or triple head…) I didn’t feel like breaking out the airbrush, though, so it’s unpainted for now.
The problem is then feeding light into them. Railroad signals have a unique color to them that’s often not captured by LEDs. The AREMA standards (Communications & Signals Manual, section 7.1.10 – “Chromaticity”) require green to be between 498-513nm, yellow to be between 589-597nm, and red to be 627-660nm. Very few 3-color LEDs hit this or even get close, particularly for green. One of the few that gets very close is the Bivar SMP4-SRGY. It’s a small PLCC4, with wavelengths of 525nm, 591nm, and 631nm. To my eye, it looks nearly dead on for the prototype colors. The PLCC4, while fairly small, would still look huge on the head of an N scale signal, and would need four wires running down the mast.
So, given that my signal models of choice are based around fiber optics, I created a board with two LEDs on board and holes for clip-in light pipe holders that fit perfectly over the LEDs. (The light pipes are Dialight part 515119200550F if anybody cares.) I can then drill a small hole in the light pipe and glue the fiber into it. The signal LEDs and their wiring (attached through an RJ45 jack for easy connecting) stay attached to the layout, and the signals can be installed and uninstalled with the ease of just connecting or disconnecting the fiber and light pipe.
Given their fragility, the actual signals will be one of the last things installed on the layout. I’ll build some temporaries for initial operations and testing. The LED boards, however, will be installed as part of the signal system. I did a temporary install (using the power of electrical tape to hold up the signal) at one of the block boundaries tonight just to see what it would look like. In the final install, the light pipes will be painted black to eliminate leakage, but as I said earlier – wasn’t in a painting mood tonight.
As many of you know from my previous post, I recently converted from Lenz to NCE, largely for the wireless throttles. Despite that, I’m going to put in fascia cab bus jacks just in case. While I don’t have any intentions of running the layout with wired throttles under normal conditions, I can foresee a day when it might be necessary. The last thing I want is to have to cancel an op session because one of my neighbours cranked up their 900MHz cordless phone and took down the wireless throttles.
The one thing I don’t particularly care for with NCE is the use of 6p6c (often incorrectly called RJ11, RJ12, or RJ25) connectors for the cab cables. You know, those little modular connectors commonly associated with phones. And it’s not just NCE – everybody seems to have gone to these now.
Why don’t I like the 6p6c modular connectors? I find the little plastic tabs hard to release from the jacks, that they often break after any significant use, and that wiring fatigue and failure often happens near the jack as a result of inadequate strain relief. Just a month ago I was at an operating session and we had two different operators lose control of their trains because the cable failed right at the 6p6c connector.
It’s not (just) a personal dislike based on anecdotal evidence. Often times manufacturers of 6p6c jacks don’t rate the number of insertion cycles, and those that do generally have minimum lifetimes in the 300-500 insertion range. Modular jacks just weren’t designed to be plugged and unplugged repeatedly. They were designed in 1975 by Bell to provide a cheap, uniform connector for telephone cords. Telephones don’t get plugged and unplugged all that often, unlike a guy following his train around the room.
The good news is that there’s a superior connector out there. The 5-pin DIN connector, standardized as DIN 41524, was designed originally for connections between audio equipment. The connector itself has robust pins and can easily be moulded to a cable with an integral strain relief. There’s no little plastic tab to break off. Plus, even the cheapest jacks are rated for 1000+ insertion cycles, double or more what the RJs can handle. I first encountered it years ago when I was using CTC-16e as my command control system, and I was pleased that Lenz had used it when I first moved to DCC.
NCE offered a dual DIN fascia panel at one time (the NCE UTP-DIN), but apparently they’re out of production and no longer available. (Update: I’m told via the NCE Yahoo group that they’re just out of stock, not necessarily discontinued as many of the retailers state. Apparently Tony’s Train Exchange has placed an order for 500 of them and should have them soon.) Lenz still offers the LA152, but they’re rather pricey ($25-35 typically) when they’re not out of stock. Apparently everybody else has decided they can live with the god-awful little RJ connectors. I can’t.
As Usual, Build My Own
That left me with – as usual – only one option that I was happy with: design and build my own. Fortunately, the cab panels are extremely simple – just a couple jacks for the cab bus, a couple DIN connectors, some mechanical board-to-faceplate bits, and a “cab bus power” LED.
Building my own allowed me to make a couple improvements. The biggest change from the NCE version is that these use 8p8c (RJ45) connectors for the cab bus that follow the NCE Cat5 pinout. I also added a self-resetting polyfuse and a terminal block for injecting power into the cab bus, should it be needed, along with a jumper to select whether to inject power or just pass it along. This replaces the 1/8″ audio plug on the back of an NCE panel where power can be injected, and adds a bit of protection (the polyfuse) against shorts.
Here’s a few pictures of the first assembled prototype, plugged into the CRNW’s cab bus a few hours ago:
As a ardent supporter of Open Source Hardware, this project will be released under a Creative Commons BY-SA 2.0 license, just like everything else I build personally. (Yes, the board says Iowa Scaled Engineering, but given that I am half of ISE, I can do that…) That license basically means that as long as you give me credit for it and you share any modifications you make likewise, you’re welcome to do whatever you want with the design.
For the prototypes, I’ve sourced the panels through a PCB prototyping service called OSH Park. I use them for all sorts of things, and they do high quality work. If you just want some v1.1 boards just as they are, you can order them directly via this link. You’re also welcome to use the design files below to create your own gerbers and have them made through whomever you’d like. The schematic and PCB are designed using the popular open source gEDA suite of design tools
Design Files and Bill of Materials
[table “” not found /]
Note 1: F1 and J4 may be omitted if you do not intend to inject power at this cab panel. Also, see note 2.
Note 2: J5 and the shunting jumper that goes in it (part 3M9580-ND) may be omitted if you don’t want to ever change if a panel can inject power or not. Just solder in a small piece of wire to permanently connect the proper two holes in the board.
Note 3: There’s nothing magical about most of these parts. The Digikey part numbers are provided as a reference, but pretty much everything except the DIN plugs are commodity parts that can be found from many sources and from many suppliers.
What About DIN Cables?
The other half of using DIN plugs in the cab fascia is that you have to actually have to have cables to connect them to the throttles. The standard offering from NCE has a 90-degree plug at the end. Again, while it should work with my panels, it’s not my favorite. I’m also not a huge coiled cord guy – I find they often get all tangled up.
My solution was to purchase a 25′ male-male DIN cord from Amazon and cut it in half. I then stripped the ends and crimped on a 6p4c connector to the appropriate wires. It took a little doing to get the cable forced in far enough that the casing would engage with the strain relief, but it is possible.
At least for the cable I used, the pinout worked out as follows (your mileage may vary):
1 – no pin
2 – White
3 – Blue
4 – Green
5 – Red
6 – no pin
Work has been keeping me insanely busy lately, but I have gotten the start of the electrical cabinet installed. So far, it’s mostly the three power supplies for the DCC boosters and the fourth power supply for the auxiliary power bus, the DCC system, and one set of DCC breakers, but it’s a start. And it’s enough to power up the helix and Nizina, along with bringing up the programming track.
The wiring is still a bit messy, but that’ll get cleaned up before it’s final. I just figure there’s no sense lacing things together before all the wires are in place.
I know, I haven’t posted many updates lately, but there will be about four coming in the next few days. There’s been lots of work, but little time to properly write things up.
I’ve decided to rev the LED lighting PWM boards again to accept a 24VDC input. 24V offers the advantage of more efficient large power supplies (the Mean Well RSP-2000-24 being the one I plan to use) and half as much current. That means smaller wires, less heat loss, and less inductive kick. While I could always put two 12V strips of the warm or cool white in series to make a 24V load, the RGB strip was always the problem, because there was no way to wire those in series. About a month ago I found a 24V RGB strip supplier, so I rev’d the control board to handle 24V.
I assembled the prototype on Monday and tested it last night. The results are spectacular. It works flawlessly. With confirmation it’s going to work, I’ll probably start putting up the final light bars before the end of the year. I’ll also post the control board design this weekend.
No, no, not the choice of DC or DCC. That’s not a question around here. There’s no way I could possibly hope to operate the CRNW on DC. The real question is “what kind of DCC?”
I’ve been a Lenz guy for the last decade, and overall I like the fact it’s reliable, well-built, and works well. Plus the throttles have a very solid, professionally-built feel to them. That said, Lenz doesn’t offer a wireless throttle, and their products haven’t changed since I bought my system 10 years ago. (CVP Products does offer wireless Lenz throttles that do work quite well, however.) Wireless throttles are a must – there’s nothing I hate more than tripping over other operators trying to get to the next throttle jack. On the up side, the system does everything I really need and – best of all – the throttle bus (XpressNet) is an open standard and fully documented. This is a huge deal to me, as I tend to build a lot of my own hardware and I’m a rather big open source software and hardware proponent. I don’t like being boxed in to any given vendor’s products, on the off chance they go out of business. Because Lenz has an open, well-specified bus, I can pretty easily build my own wireless throttles. if I wanted.
On the down side of sticking with Lenz is the facct they haven’t really done anything in the last ten years in the base station or throttle market. Plus, their US market share has seemed – in my experience – to be slipping since Debbie Ames retired as their North American rep. While, as I pointed out above, I could build my own throttles, I don’t really want to right now. I have a layout that needs lots of work to get to operational, and I don’t want to get bogged down playing with electronics.
Having played with an NCE system at a layout open house several weeks ago and heard nothing but good things about them for years, I finally broke down today and ordered an NCE Power Pro radio starter set. My plan is to run the NCE system for at least the first phase of the CRNW’s life. If it works well, it’ll probably stay for a good number of years, or possibly for the life of the layout. Expect a “first impressions” look at it in a few weeks.
(BTW, if anybody’s looking, Brooklyn Locomotive Works has an incredible deal on both the wired and wireless versions right now.) I also ordered a Cab06pr from MB Klein, since BLW doesn’t seem to have them.
I don’t plan to get rid of my Lenz system. I still like it, and some day may get around to building compatible bits for it. For that matter, part of me really wants to design/build my own DCC system based on MRBus, and I have no doubt I could do it. But now, in the formative years of a new layout, doesn’t seem to be the right time to pursue such interests.
Those of you who have been following along know that one of the core electronic pieces I’ve been working on for the CRNW is an LED lighting system tied into the fast clocks. Over the run of an operating session, I want to transition from “dark” (really dim blue light) through the brilliant warm light of daybreak, the bright white of midday, and the through the golden hour and sunset back to dark. The system will consist of the fast clocks – obviously – as well as a MRB-GIO to figure out color and intensity, and then a set of power booster boards that actually control the 2000W going to the LED strips. The power boosters will be local, each controlling <20ft of layout, to keep the amount of power being switched to a more manageable level.
The prototype power booster boards for the LED lights showed up a couple weeks back, but this weekend was really the first chance I’ve had to try integrating them with the full system. I connected one up to my test LED strip and mounted it back over Cordova, and then connected it to a MRB-GIO and did some basic programming to turn it into a lightning controller.
Results are promising – I need to do a bunch more tweaking on the exact light transitions, but my first try came off pretty well. I also did a few tests using a bunch more strips to increase load. The system was designed for up to 6A per channel, so I cranked it up to around that. Heating was actually less than I expected – the board only slightly warm to the touch even switching 6A on a couple channels.
I’ve posted a few photos of the new power control board, as well as some samples of midday, evening, and night light. Night isn’t that bright, I promise. It’s just the camera evened out the exposure.
On the benchwork front, I did get the top deck extended from Strelna over to the north end of the Chitina yard. The bottom deck is still being pondered – I’d really like to add a short Katalla Branch as a very low level. The problem is that I can’t figure a way to shove a helix under where the junction should be. I’m contemplating a train elevator along the wall (hidden behind the Miles Glacier Bridge area), since trains to/from the coal fields above Katalla would be short – 6-8 cars plus power. Regardless, I’m still pondering it.
Oh well, off to Memphis for the week tomorrow. I’ll figure out what I’m doing about a potential Katalla branch train elevator when I get back.
Honestly I got almost nothing done this weekend in terms of benchwork. Nearly all of my efforts went into other, non-model railroad tasks and then into working on the LED lighting system a bit more. I’m running another set of boards soon, and wanted to get enough groundwork done to get the LED driver boards into this batch.
Tasks left to accomplish were as follows:
- Work out mounting to layout and characterize thermal/electrical parameters in operation
- Solve the inductive ringing problem in the LED drive
- Plan how lighting control would be distributed and design power control board
Task 1: Mounting Considerations for LED Lighting
Some time back, roughly when I started benchwork construction, I found some great T-shaped galvanized steel drip edge (intended for roofing) at Lowes. Strong due to its folded shape, light, wouldn’t rust, metal (to provide some heatsinking and protection between a potential high current short and the wooden benchwork), and cheap ($~4 for a 10 foot strip when I bought it), it seemed the ideal material. Plus, the folded edge would reflect light back towards the layout and away from the aisles. I put a single strip in with one of my lumber purchases, and it’s been sitting in my garage for the last four months.
Today, I cut off a four foot section of the stuff, since that matched the LED strips on my test piece. Using a thorough coating of Loctite 300 heavy-duty spray adhesive (I’d previously tried my usual go-to adhesive – 3M 77 – and was underwhelmed), I laid out the four planned LED strips on the steel -warm white #1, RGB, cool white, and finally warm white #2. Each strip was harvested from the original test 2×3 board, so these have been used a number of times now…
With the strip prepared, I then mounted a thermocouple to the top of the steel to monitor the temperature. Using a bar clamp, I attached it to the layout (in the usual test position, over what will become the Cordova yard someday) and powered it up with an old bench supply. Ambient temperature at this point was about 23C (73F). Current was approximately 3A for 3x 4′ strips of white being lit simultaneously.
I left the lights alone for an hour as I did other things, so that I could see where the temperature would stabilize. Turns out, the answer is at approximately 45C (113F). The steel was warm to the touch, but the strips were notably cooler after an hour of running then when they were mounted to a piece of wood. It looks like the heatsinking works! LED longevity is closely linked with temperature, and once you pass an operating temperature of 50C, things start going downhill in a hurry. 45C for three strips running at 100% (more than I plan to run in actual operating conditions – my plan is currently at most 2 at 100%) is pretty darn good.
One interesting note – the operating current at 45C had increased to 3.3A. This makes sense – bandgap of LEDs decreases with increasing temperature. So, with less voltage drop across the LEDs, the ballasting resistors (56 ohms total, split into two 1206 resistors – see more about this later) allow more current. So, as a note, LED strip lights appear to have a 10% current rise over the 20C between room temperature and operating temperature. Almost all of that additional power will be dissipated in the resistors, so it’s not even going into providing useful light. (See analysis in the final section.)
Task 2: Fix the Inductive Ringing
When I initially investigated doing variable LED control for layout lighting, I noticed some reasonably nasty inductive spikes whenever the MOSFETs controlling the strips would shut off. The screen capture from my scope on the right is a pretty typical kick – the FET shuts off, and the drain side of the FET would ring up up obnoxiously high voltages.
Since I’m only planning on using FETs with a 30-40Vds rating, these spikes could easily turn into circuit-killers. Plus, they’re likely radiating lots of electromagnetic noise that will cause other issues down the line when I have 100+ amps of LED strip lights running and switching on and off.
My initial hope was that a freewheeling diode from the drain up to the+12V rail would solve my problems. I was operating under the assumption that my problem here was inductance in the LED lines, and that if I gave the current somewhere to go, my problem would go away. At low currents, it looked promising, but as I started pushing 2-3A through my test setup, I got the waveform on the left. Better (less ringing), but still with the 30+V spike before the diode got going. All I had on hand for reasonably quick diodes were some old FR155s. While great at reverse recovery time, I can’t find a spec on how fast they go into forward conduction. Regardless, a decent Schottky diode should beat them hands down. I just don’t have one to try.
An alternate approach I tried was to damp the system by applying a small capacitor between the FET’s drain and source. As little as 22nF was enough to significantly clean up the waveform, and 0.1uF damped it out completely. I don’t like the idea of a capacitor in there, though, because this will almost certainly need to be quasi-matched to the parasitic inductance in the lights. I’m going to try a faster diode first and see if that fixes my problem. If that doesn’t work, however, we’re back to some sort of RC snubber circuit.
Task 3: Plan a Lighting Control Scheme
Safety in designing this light system is paramount. Given that the layout will consume at most ~150A of current to run the lighting (at 12V, but 150A is still a lot of I2R heat in any short or questionable connection), fuses in all the right places are a key design feature so that any failure doesn’t end catastrophically.
Per measurements taken at steady-state in the test fixture, any of the white strips are going to take 0.275A/ft at 45C. The RGB strip will need approximately 0.1A/ft for each of the three channels lit. So, figuring on an absolute maximum of three full strips (all three whites) on at any given time, I’m looking at 0.825A/ft maximum.
My initial plan had been to use larger modules, controlling as much as 30A through a single control board. However, that means larger FETs, larger freewheeling diodes and snubbers, and larger wires everywhere, since any potential short may have to soak up 30A @ 12V until the fuse goes. Given our 0.825A/ft metric, a 30A module would feed ~35 feet of layout lighting. The heavy wire that I would need to handle any possible shorts over such lengths quickly showed this as an impractical approach.
If I used the 10ft length that the steel flashing comes in as a module length instead, that’s 8.25A per module, maximum. That’s nicely under a 10A fuse by a decent safety margin, and most wiring that I would use could handle 10A for a short overload period until the fuse went out. (Mini-ATO blade fuses like I’m considering using will blow in <0.75s at 135% of their rating. So 13A would take it out in less than a second.)
My current plan is modules for every 10ft section, with one single input fuse of 10A. Each board will then have 6 N-channel FETs (likely 3x AO4882 duals), 6 FET drivers (likely MCP1416s), and six optocouplers, along with dual RJ45s for the control bus (one in, one out). A single Cat5 cable will run around the layout, delivering signal ground and 6 light PWM channels from a main lighting control board to the optoisolated inputs on each booster node.
I have the schematic done, but it’s already pushing 2300h and I’m tired. Have to go to that real job tomorrow, so I should probably get some sleep and put the board off for another day.
Addendum: LED Strip Lights and Voltage
As sort of an interesting side research project, I decided to investigate more about the relationships in LED strip lighting between input voltage, power dissipated in the two ballasting resistors, and power actually delivered to the LED lamps themselves. Some of my discussions earlier in the day with a friend had made me wonder what the relationship really was between these three things.
Step one was to just cut off a piece and decapsulate it so that I could get probes right on the pins. Not really that hard – with a little encouragement, the coating peeled right off. I then powered it up and realized that having three death rays shining in my eyes as I was trying to probe was going to be a real problem. So, I used a trick I learned long ago when dealing with warning lights on cars that shouldn’t be on but inexplicably were: grab the electrical tape and – poof – no more annoying light!
Internal construction of these strips (nominally rated 12V) consists of a 27 ohm, 1206-package resistor (rated for 0.25W of dissipation) between +V and the first LED, then three LEDs in series (each “LED” being a 5050 package with three LED die inside in parallel), and then another 27 ohm resistor to ground.
Procedurally, it was just a matter of punching each voltage into the power supply, measuring the voltage across one of the resistors, and plugging the result into a spreadsheet. Given the input voltage and voltage across one resistor, I could calculate the current through the circuit, and thus the drop in both resistors and consequently the power that must be dissipated in the LEDs themselves.
Here’s a pair of graphs of the results. The first is just power burned in each type of component – blue being resistors and orange being LEDs. In the second graph I stacked them, so that you can see how large of a percentage of the total is burned off as unproductive ballast resistor losses at high voltages.
It really makes me wish for a constant-current design, but I realize the additional complexity that would entail. Cheap constant-voltage LED strip is cheap – custom engineering my own is expensive. Darn.
Like any good model railroader, I spent New Years Day in the basement working on the layout. I was hoping my Fast Tracks jig would arrive on New Years Eve, but alas, it did not. So, I reverted to working on lighting, benchwork, and some electronics design.
The good news is that the entire layout room now has proper room lights – nine fluorescent 4′ dual T8 fixtures, to be exact. It makes it much less of a dingy hole in the ground and much more a presentable layout room. Now if only the construction disaster would clean itself up…
As far as benchwork, I accomplished a piddly 32″ – the upper deck between McCarthy and the Kennicott River crossing. The electrical took longer than expected, and I needed to accomplish some design work for Iowa Scaled for a new optical track detector we’re working on.
My goal for the rest of the week is the rest of the east wall (McCarthy to Kuskulana on the upper and Cordova to Eyak on the lower) and – if the stars align – the rest of the wall framing and clean up some of the random junk in the way of progress. We’ll see what actually happens.