Sunday, January 6, 2019

The Hollywood Gas Station, or Robert Breaks His Unbreakable Rule Again

Final model

Robert’s Rule of Making Structures in a 3d Printer is simple: don’t make HO scale buildings with a 3d printer. Buildings aren’t particularly interesting for 3d printing; you don’t need multiple models, the models are too big for the printer, and the buildings usually have plain surfaces that are easier to make in other materials. The prohibition doesn’t hold for details - window or door castings can be used for other projects. Smaller parts that are hard to fabricate might be worth a quick 3d print. But don’t try printing a whole building.

Though even if it’s my rule… that doesn’t mean I won’t try.

The inspiring photo

The Hollywood gas station

Years ago, I’d run across this photo of an early gas station in Los Angeles. It appeared in Larry Harnisch’s “Los Angeles Daily Mirror” history blog; he’d found the photo in a back issue of the Daily Mirror from 1915, showing the filming of a new movie at D.W. Griffith’s Fine Arts Studio. Griffith later filmed his silent masterpiece, Intolerance, and the sets for that movie are visible in the background of the original photo.

But between the crowd scene and the sets for the future movie, there’s this tiny little gas station. It’s the dawn of the auto revolution in Los Angeles, and cars need gasoline. This corner (Hollywood and Sunset) is on the edge of suburbia - the sets for Intolerance over there are being built in a former fig orchard. Photos of the studio behind the photographer show scattered buildings and empty lots. In a few years, this will be a very urban corner. Little gas stations like this would have been scrapped for the huge service stations that would appear in the 1920’s.

One sign of its age is the lack of any gas pumps visible in the photo; it's almost as if the pump machinery was hidden in the posts.

There's little sign of the gas station remaining. The site of this photo was almost certainly 4500 Sunset Boulevard, Hollywood, right where Sunset and Hollywood Blvd come together. The former movie studio is now a Von's supermarket, and the nearby Monogram Pictures is now a Church of Scientology video studio. The corner does have a small gas station, but the style doesn't match (art deco), plus it's triangular instead of square. I'm guessing our little gas station got torn down when a road was widened... or when newfangled gas pumps couldn't fit in the arch.

For a model railroad, and a model railroad set in the 1930’s, it’s a neat bit of architecture. The building itself is remarkably simple - a fifteen foot square office and similarly sized roofed porch covering the fueling area. There’s a lot of ‘teens era gas stations that appeared to be simple boxes to protect the attendant when he wasn’t checking your oil or putting air in your tires. (What a crazy time!) As a modest gas station, it’s also appropriate for the area down by the railroad tracks - this isn’t a high rent building.

However, unlike those dusty country corner gas stations, the Hollywood gas station dresses itself up by stealing details from every flavor of Spanish Revival it can. The most obvious feature are the silhouetted bell gables on each side, traditionally from Romanesque architecture. The fueling area has looks like the arched porte cocheres seen in any of the stucco spanish revival bungalows getting built out in Westwood. The wooden beams sticking out are vigas, straight from New Mexico and Pueblo Revival architecture. Floor to ceiling windows light the office; I’m guessing there are similar french doors on the front side to enter the office.

Man, I’m really a sucker for Spanish Revival. Show me a Spanish Revival gas station, and I’ll try to build a model of it.

Construction Like all my 3d printed models, I designed this in SketchUp. The model is one piece - walls, bell towers, and port-cochere. I omitted the roof - it’s easy enough to do with sheet styrene or cardboard. The viga beams are styrene, set into sockets in the walls. The posts are hollow to limit the amount of resin needed.

Beyond the issues of architecture, this was an interesting model because it reminded me of the challenges of 3d printing and manufacturing. Getting this model printed involved a chain of challenges; as easy as 3d printing seems, there’s always snags trying to make more than one.

I’d initially sketched up this model after seeing Harnisch’s photo. I’d liked the model and thought it would be a good exercise to practice in SketchUp. That initial model messed up a few angles, making the model have some minor holes in it. When we try to print a 3d model, the slicing software needs to figure out which bits are the inside of the model (where the plastic or resin goes) and which parts are outside. A good 3d model is “watertight” - all the exterior faces touch, there’s no holes that will make the software confuse the inside and outside of the model, and there’s no extraneous faces to make the software question which counts as the exterior surface. Cleaning up the holes in an existing model is always a tedious process as you try to get rid of some incorrect angle or out-of-parallel plane without tearing apart the whole model.

It's a lot like real home improvement, except with more straight lines and flat planes than reality.

The next big challenge was how to print the model. By default, the Form One wants to print models on a support structure. You take your model, choose the face-up direction, and the Form One automatically chooses how to place supports (sprues) to support the first few layers as the surface is built. Support structures are important because it lets us build items that aren’t flat; it also lets us build hollow objects without pressure from the liquid resin pushing walls out. However, support structure require a lot of material - sometimes as much resin as the model, and the bottom of models isn’t always flat.

First attempt

I’d printed an initial version of the gas station on a support structure, but it doubled the amount of resin needed, and I ended up with a not-quite flat base. If I instead printed straight on the build platform - ok for surfaces with a flat bottom large enough to hold the model to the build platform - I could cut resin use and get a flat bottom surface.

So I tried it - I printed one directly on the build platform, but fluid pressure (as the build platform peeled the part away from the tank then put it back against the tank) caused one side to blow out, and the window muntions to break. If I did things the way the manufacturer intended, and wasn't trying to cut corners to save resin and time, I'd have better results, but if I'm concerned about economics, I might try pushing the machine a bit harder than it really can take.

On a second attempt, I gave up and printed again on support structure. The windows again didn’t print perfectly, but I made new windows by drawing white lines on clear plastic using a technical drawing pen.

Failed print

That final model was good enough for me, but if I wanted to sell the models, I’d need to do a lot more work on the process so the models were perfect coming out of the printer. Making these efficiently would also mean cutting the resin needed, and that means limiting the support structure. Finishing On the first model, I used my usual trick of white glue and gesso, stippled onto the model. It dries quickly and adds a lot of texture, but I found the surface much too rough to my eyes. I ended up coating the second model with an acrylic gel with pumice (from Golden Acrylics) which was much more subdued. In both cases, I had to be careful to only coat the stucco surfaces with the fake stucco.

Just like resin building kits, cast structures are great for assembly, but frustrating to paint. My first attempt at the lanterns left black paint everywhere. For the second try, I painted the lanterns orange, then used a very fine tip to color the metal parts black. Similarly, painting the bells, inset into the walls, definitely required a bit of care.

So now, I've got two very cute 1915-era gas stations that mix up way too many architectural styles. They don't quite have a place on the layout, and I don't really need two, but they'll be great reminders not to break "Robert’s Rule of Making Structures in a 3d Printer".

Building an Interlocking Machine For West San Jose Tower

As I’ve mentioned before, I like model railroading as a hobby because of the mix of projects I can do. I’ve got a friend who’s big into wargaming figures. His photos of some of his painted miniatures shows great work, but I always wonder “what does he do when he doesn’t want to paint?” Luckily, I’ve got no such problem; when I'm tired of one kind of project, I move on to another. Lots of projects go unfinished because I’m not quite in the mindset to spend time on them. Some times the project just gets delayed, and sometimes it gets rethought. I’ll switch and do something else for a while, and eventually I’ll come back. Maybe I’ll do the project as I originally conceived it; other times, I’ll throw my old ideas away and go in a completely different direction.

Take the Western Pacific crossing on my layout, for example. The WP crossing was a key part of my track plan, not because it was an active part of switching the canneries, but because it both helped set the location, and because it tied my railroad to the larger world. I had plans years ago for how I wanted to build it, but those plans never worked out. A while back, I rethought what I was going… and ended up with a new plan that sounds like much more fun.

The Crossing as Model

The crossing of the Western Pacific and Southern Pacific tracks in West San Jose isn’t much to see - just a set of tracks crossing between the Del Monte cannery and the Standard Oil spur. There’s an interchange track that gets a couple cars switched every operating session. I built a model of the WP’s 1920’s era switch tower years back, and I keep having ideas of putting the Virden Cannery next to the tracks, just like in real life. However, for the operating crews, there’s just not much there. Littering freight cars across the crossing when switching Del Monte is quite a common occurrence, and would have infuriated the real tower man for the Western Pacific Railroad on the real railroad.

I’d had ideas to make crews better respect the crossing. Somewhere around here, there’s an Arduino with a sound card to control some animation. It would play a soundtrack occasionally - a factory whistle, some cars going by, a far away whistle, and finally the sound of a WP train approaching. Some nearby signals would change to red (to announce the arrival of the train), and LEDs in the roadbed would flash as the phantom train passed. If that wouldn’t keep crews from blocking the tracks, nothing would.

The project that didn't work out.

The plans never quite worked out; I didn’t have a place for the speakers and was never happy with the soundtrack. The idea of making the crossing obvious and important during operating sessions did linger.

History of the WP Crossing

When the Western Pacific was built in the 1910’s, the Southern Pacific already had tracks in all the obvious locations around the Santa Clara Valley. The potential business from San Jose’s fruit industry encouraged the railroad to find a way to get past the SP’s tracks. The WP’s line from Fremont and Niles to San Jose had to parallel the SP for much of its length, then swing far south of San Jose only to approach the city from the south. The route required crossings at Niles Junction, at the crossing of the SP’s Coast Line at Valbrick, and a final crossing of the San Jose - Los Gatos branch at West San Jose.

The tower - West San Jose to SP, and Tower 17 to the WP, was built in 1922. The railroad signalling trade rag commented on the construction: a Saxby and Farmer interlocking machine controlled the semaphores, with 20 levers controlling signals and switches. The crossing was quite substantial, with WP’s branch line crossing the SP’s main line and two drill tracks right in the middle of the cannery area. Like all railroad towers, the second railroad to arrive at a place paid for it all - the track crossing, the tower, and staffing the tower. The WP never got the traffic it expected from the San Jose branch, and had little interest in staffing the tower on an unused branch line. By 1938, the tower was out of use. In later years, WP trains had to stop and check the SP wasn’t coming before dashing across.

James Barriger got a decent photo of the area around the tower in the 1930’s, capturing an SP switcher right behind the Virden Cannery. Although he didn’t capture the tower, he did show the trenches for pipe rods controlling derails on the drill tracks.

The Interlocking

On a railroad, a switch tower is a manned location that controls where several tracks come together, and where the track, switches, and signals are controlled to ensure safety and minimal delays. There’s usually a person present; he sets switches and signals to allow trains to safely move through the section of track based on train schedules, dispatcher orders, and the arrival of trains. Controlling those tracks is often done by an “interlocking machine” - a mechanical computer that ensures only non-conflicting routes can be set up through the stretch of track it controls. Interlocking machines usually have levers that control switches and signals, one lever per device. For the WP crossing, that means that the signals, switches, and derails can be arranged to let an SP train to cross the WP tracks, or allow a WP train to cross the SP tracks, but not both. (It also enforces safe order - the tower man can’t set a signal to green unless the switches and derails on the through route are set correctly, and derails on the crossing track are locked down.) Interlocking machines use a set of sliding bars connected to multiple levers to ensure that if lever A is thrown, lever B cannot be thrown.

As mentioned, the WP tower’s interlocking machine had twenty levers - were there really that many things to control? From various sources, we can guess what the 20 levers in the Saxby and Farmer machine controlled. The Barriger photograph shows piping for derails, suggesting all the tracks had devices to stop a runaway car on a track that wasn’t expecting a train. Railroad valuation map shows that the SP had distant (one mile before) and nearby signals closer to the crossing. Track diagrams show one mainline track and two drill tracks on the SP to handle switching the canneries in the area. One likely guess at the purpose of the levers would be one derail, one local signal, and one distant signal on the SP and WP main tracks in each direction (3 * 4 = 12 levers) + a signal and derail on each SP drill track in each direction (2 * 4 = 8) for a total of 20 levers - just what the trade rag says.

Locking bars, tappets, and tappet blades on the interlocking machine at Santa Clara tower. Chuck's photo.

So it would be neat if I could actually model the interlocking machinery, and give my operators an appreciation for everything involved with the tower - the rules about how train crews got permission to cross the diamond, the need to communicate to the tower man where they wanted to go, and the actions the tower man needed to do to line up the crossing. That means I need to build an interlocking machine - not a standard kit at my hobby shop. I’ve seen articles on how to make an interlocking. Model Railroader had a set of articles by Paul Larson and Gorden Odegard in the January-June 1961 issues of Model Railroader, but it wasn’t quite a step-by-step project, and the suggestion that the authors needed to build a wooden mock-up beforehand to test out the logic suggested it wasn’t a project for the faint-hearted.

Building a Modratec Interlocking Machine

Luckily, there’s folks who can help. Modratec, in Australia, sells kits for making a working interlocking. The price isn’t quite an impulse buy - about US$225 for 12 levers and electrical contacts, but it’s a pretty great little kit. To get an interlocking kit, you download their SigScribe4 software for setting up the constraints, define out how you want the levers to work, then mail off the interlocking details. You’ll get a kit back - all machined and ready to be bolted together, with a bit of metalwork to set the locking machinery to match your intent. I’d been considering this kit for a few years; I’d tried a couple times to get started, but never quite got it. A couple years back, I finally took the time to understand the software well enough to describe West San Jose Tower. The interlocking turned out really nice, and gave me a much better understanding of how real interlocking machines work.

Designing the Interlocking

The biggest challenge was just getting the interlocking designed. It took me several tries over a couple years to figure out the SigScribe software to get a working interlocking. Running through the tutorials multiple times helped. Once I understood the software, designing a new interlocking for my Market Street layout took only an hour. Don’t be surprised if it feels cryptic, or if you find yourself starting from scratch multiple times.

The general steps are:

  • Research your prototype to understand how the interlocking may have been laid out. Decide on signals, switches, and derails.
  • Draw the track plan in SigScribe, mark the location of signals and switches, and describe the configuration of each signal.
  • Associate levers with signal blades or switches.
  • Define a route for each signal lever indicating what switches must be set (or locked) to allow a train to proceed through safely.

Planning the interlocking involves a bunch of choices - how far out does the interlocking go? Where did the prototype have signals? What additional safety is required, such as derails or pointing an incorrectly-proceeding train away from active routes? Doing a bit of research helps you lock down what you’re building.

Track diagram for the West San Jose interlocking I built.

For the West San Jose tower, I started by looking at photos and other documents. The Barriger photos showed that the interlocking had derails to keep an incorrectly moving train away from the crossing; representing these adds a bunch of extra levers to the interlocking, and reminds operators about all the extra machinery needed to protect the crossing. Valuation maps pointed out the need for distant signals a mile away - something I chose not to represent because of the lack of space. The California Railroad Commission documentation on the tower mentioned the 20 levers, which confirmed I’d accounted for all the devices around the real tower.

Track diagram for Fourth Street Tower in San Jose.

I also tried building an interlocking machine for the Market Street layout (though I haven’t ordered a kit for it yet.) The San Jose Market Street station had a switch tower at the east end of the station where the lines up to Oakland and down to Los Angeles diverge. For the Market Street layout, I again used valuation maps and photos to figure out the signals and switches that existed. There were no derails in the interlocking trackage. However, Modratec’s documentation did mention that sometimes particular switches would be forced to be set in a particular way to keep runaway trains out of the way of a chosen route, so there were some places where I could explicitly insist a switch had to stay pointed away from routes in use. (Specifically, I designed the interlocking so switch 6 would need to be pointed towards Oakland whenever a train was coming or going from the train shed.) For Market Street, I also had to decide which switches would be controlled by the interlocking. Southern Pacific timetables mentioned whistle signals to get access to nearby industries, suggesting these switches were under the tower’s control. A crossover just east of the station train shed, however, was outside of the track protected by the various signals according to the valuation maps, suggesting those crossovers were manually controlled. I left them out of the interlocking.

The Fourth Street tower was also complex because of the need for separate signal arms for each possible route through. I ended up making the easternmost signal (near "To LA") a three blade semaphore to control which diverging route would be chosen. I'd been curious why they needed a separate signal just west of switch 6; it provided a way to indicate whether the switch was lined for the mainline or the route into the yard without adding extra blades to the signals further east.

Once I had a handle on the track and signal arrangement, I started describing the interlocking in the SigScribe4 software. I drew the track diagram and placed symbols, connected up the levers, then set up the routes - about an hour of work now that I understand things.

Here’s some quick tips for using SigScribe4.

  • On a Mac, regular mouse clicks only do selection. You’ll need to do mouse clicks while holding down additional keys to do some of the actions. Select a square in the track diagram, and drag with the alt/option key down to draw a line. When you’ve selected a square in the track diagram (and see the multi-colored square), then shift-click on any of the eight cells to indicate the direction a track line should exit the square. Shift click in the center of the square to finish editing that cell.
  • Select a cell and press V repeatedly to show signal options for that cell, or H for derails, level crossings, and other non-signal options. For each signal, open the detail view (right click or command click and choose Detail View) and hit H and V to indicate the kind of semaphore blade in the signal or to show multiple blades when there are multiple routes available.
  • When connecting levers or routes, first select a lever, and then right click (or command click) to get the context menu and select “Connect” or “Define Route”. Connect all the switches associated with that lever or route (right click and choose "Connect" on each), then press the big button at the bottom of the screen to commit the change.
  • Make sure to define all levers and set their correct color (black for switches and derails, red for signals). If you need to change them, you're likely to lose all previous work.
  • Modratec mostly caters to English-style modelers, so it’s worth reading up a bit on either the Modratec website or british signaling website to understand their terminology.

Once I had a design, I tested it to make sure it worked correctly. I tried each route and double-checked the correct levers were locked and unlocked. I then saved out the model, checked the number of locking bars and levers needed, and got an estimate on price. Once I was ready to get the interlocking, I sent off an order and the file describing the model; Harold, the owner, sent e-mails about status, and let me know when the kit was on its way. Total time from order to kit arriving was about 6 weeks.

Assembling the Interlocking

I spent three days assembling the kit. The first day was doing the majority of the assembly. Most of the interlocking machine just needed to be assembled with screws; it all went together smoothly. The next two days were for making the locking mechanism: the tappet blades and locking bars. Finally, I completed assembly and made the track diagram to show which levers to throw.

The two non-trivial bits of work was the locking mechanism. An interlocking machine is set up so that conflicting movements can’t be made; it does this by mechanically blocking tappet blades (bars moved by the levers) with tappets in locking bars. The locking bars that set restrictions between levers are square bar-stock, and come pre-drilled where there would pegs to block the levers from movie. Brass rod needs to be pressed in and cut off to form the tappets, and filed flush with the top of the tappet blades. The tappet blades, controlled by each lever, are brass bar stock. Each tappet blade needs to be filed at the correct location to ensure the mechanism works properly. It’s straightforward but careful work; I was constantly assembling and testing to make sure I was filing in the correct location.

Next Steps

Although I’ve finished the interlocking machine, I still need to install it on the layout, attach it to signals, and tell crews how to use it.

The first step will be adding the model components; I’ll need signals to indicate when it’s clear to proceed, switch machines to control operation of switches under tower control, and derails to mark tracks that should not be crossed. The prototype signals were semaphore signals, located around 500 feet east and west of the crossing. Tall and spindly semaphore signals wouldn’t survive well in this area where people are constantly reaching in when switching; instead I’ll use dwarf lighted signals to indicate when the mainline is safe to cross. The derails are another important part of the interlocking; although I could try to build working derails, it might be easier to just add red LEDs near the track to indicate when the derail is set incorrectly.

The interlocking itself will be inset into the layout so it’s easy to reach, but won’t interfere with movement around the layout. It’s only six inches deep, so it should be easy to hide near one of the Del Monte buildings. I’ve had good luck with Team Digital’s programmable logic boards, though they all appear to have been discontinued. An Arduino board would be easy to program; each lever would throw one switch which would go to the Arduino; the Arduino could then control the signals and Tortoises.

Underside of interlocking machine with electrical switches added.

Once the interlocking is installed, switch crews will start having to work around the interlocking. Scheduled passenger trains will have it easy; the tower man would know the timetable, and could make sure that the signals were clear as the train approached. Freight trains would have a harder time without a schedule; on the real railroad, trains on the mainline would have to stop, whistle “one short and two long” to get the attention of the tower man, and get the switches set correctly. Crews on the drill track would probably need to stop, chat with the tower man, and get the switches thrown appropriately. Everyone would need to set the levers back before leaving. That leaves the WP trains; although I could automate it, I’ll probably just occasionally throw the levers to let a WP train through, preferably when a crew is about to switch in the area.


Thanks to Chuck who inspired me to build an interlocking machine, and shared stories and photos of the work he’d done restoring the interlocking machine at Santa Clara tower. I’ll miss him.

Information on the WP Tower from Jeff Asay's "Track and Time: The Operational History of the Western Pacific Railroad". I know I've seen more details in the California Railroad Commission decision allowing the WP crossings, but can't find references right now.

I have no connection with Modratec other than building this one kit.