Description of Issue

Back in 2018, we had engines stop-and-go on turnout T161 for quite a while, especially steam engines, or short non-MU engines. The turnout was always in the normal position when this happened. Back then, it was determined the frog power routing on the Fulgurex created a resistance. This was solved in 2018-12-12 by replacing the Fulgurex. The issue re-appeared recently.

The “Train Vision” project uses my custom Vision software running on a laptop, displaying pre-recorded videos I made of trains running on the layout, mixed with live views of trains actually running on the layout.

All the videos come from my Youtube Channel playlist for the museum. I have a couple news train videos to add to that list: an SP Daylight, an Amtrak Superliner, and an SP Cotton Belt. Interestingly since I’ve recently switched to doing shorter videos in the 3-minute range, that works actually better for that kind of display.

One thing I’m particularly proud of is what is not visible here: this runs on a Lenovo Yoga 15 laptop folded over. It provides a good sizable display although it’s not huge so it doesn’t dwarf the layout. This runs the latest Debian Linux distro, and I have spent a lot of time ensuring it powers up automatically with the layout. All the museum staff have to do is turn on power to the exhibit room in the morning, and they don’t have to specifically deal with that computer. Same when they close in the evening. All automated, as it should be. Of course it’s an automation which means sooner or later it will fail to start and it will require manual intervention, however that will be newsworthy because it will be the exception rather than the norm.

This has been running for a couple weeks now and is performing fairly well. Based on my observations:

• Although some viewers get it instantly, some others are confused on which video shows a live view vs a recorded one. I already added a blinking “Live” sticker to the live views. I’ll follow up by adding a “Recorded” banner on the recorded videos. I’m thinking of a generic “play” triangle icon + record date + title.
• Originally my first prototype was streaming videos directly off my Youtube Channel playlist. I thought that would be a concern as the recorded videos would stop playing if the wifi was not cooperating. Thus I switched to downloading all the videos and playing them off the drive. Glad I did that because the first week the wifi was really sub-par at the museum although it seems to have somewhat improved since. I have a script I can run that will update my playing list using youtube-dl to download my own videos locally, so updating the playlist is still fairly easy.
• One thing I do look forward to is working with Tiffany from the Randall Museum Friends to make the display customizable for special events. Says they host a gala or some kind of social event, we can have the software download one or more mp4 (or a private youtube playlist) and play it specifically for that day & time slot.

The last item on the list is the quality of the live video feeds from the cameras on the layout. I was having some terrible frame rates over wireless, with some obvious mp4 dropped frame artifacts. I switched one of the cameras from wireless to wired ethernet, and I also switched the laptop to wired ethernet using a Linux-compatible Cable Matters USB to Ethernet Adapter (linux pro-tip: this one requires an apt install firmware-realtek from Debian 11 “bullseye”). That combination brought me the stable 15 fps I was looking for. I will wire the other two cameras once I have them in their final location.

Time for more maintenance. This time I adjusted the grade crossing light sensitivity with some room lights off, and it seemed to go better. In this case I had the fluorescent lights off, the single center spot on, and the side spots (by the walls) on medium.

More importantly we had an interesting case: a turnout malfunctioned on the Bridgeport lead track.

That seems innocuous except this somehow shorted the B450 block on the mainline (the one that leads to Bridgeport) on the Mountain Panel 2, and it also made blocks B330 and B360 seem occupied on the Mountain Panel 1 -- these two blocks are used by the automation and since the track appeared occupied the automated trains would not start. It did take me some time to figure all that out since I started in reverse (why the blocks seemed occupied) and I had to trace it back to the short on the other panel and then to the Bridgeport lead. I still don’t have a full explanation here. It just doesn’t quite make sense.

Anyhow, this is what I found on the Bridgeport lead:

What’s wrong with the point?

Yesterday was also time for some regular maintenance.

First order of business was to swap the engine for the automated UP passenger train. It has been running with UP 8736 for quite a while, and I swapped it for UP 8749. They are both identical SD70ACe engines, and I had previously programmed them to have identical running characteristics, so it was a quick one-line change in the automation computer’s event script.

Then I had been told the grade crossing bell was ringing all the time with no trains, so I spent quite some time adjusting it. One of the sensors is way more sensitive than the others. I readjusted them all. Unfortunately I’m told today that the bell was again ringing for almost 20 minutes when the layout was turned on. I had noticed that before -- when we turn everything on in the room, the lights are not as bright and they take time to warm up. And that can be enough to trip the grade crossing light sensors. Ideally I’d have to adjust the sensor right when the lights are turned on in the morning.

So I need to adjust the grade crossing sensors again. I’ll see if I can reduce the dimmable spots or turn off some lights to reproduce that “warming up” state from the morning.

The “Train Vision” project: One part is my custom Vision software that runs on a laptop, displaying pre-recorded videos fullscreen. A second part are cameras located on the layout display live trains as they go by. When the live trains are visible, motion detection kicks in and the display changes from fullscreen to a 4-video quadrant with one view being the recorded video and the 3 other being the 3 live cameras.

I’ve spent the first half of the year reworking the Vision software from the prototype I did last year, and it’s now good enough to install it at the museum. It’s not perfect, yet we need to see it in real condition to adjust it.

The display is a 17-inch laptop. A lot of time was spent to make sure it would work as automatically as possible. It is now configured to start when the power to the layout is turned on, and it shuts off automatically with the layout power. It auto-boots Linux and my dedicated Vision program. The museum staff will not have to interact with the laptop -- no need to manually turn it on or off like they do with some other exhibits.

I quickly placed the 3 live cameras on the layout by the Fairfield passenger station. That’s temporary.

Description of Issue

The Bridgeport turnout needed fixing, and it is now fully functional again.

It uses a Fulgurex turnout, and recently this one had started shorting the entire turnout power supply intermittently.

This is a Fulgurex slow-motion motor. Connections have two wires for turnout power, and two sets of contacts (three wires each) that are commutated based on the position of the switch:

Description of Fix

And here it is after the fix, when testing it:

Lighting on the Stockton side:

The Fairfield main street with street lighting on, and the illuminated McDonalds:

I’ve been trying to update my documentation on the Building Lighting panel on the layout.

As a “recap of the previous episode”, under the Valley Panel 1 is located a row of toggles for “Scenic Lighting”:

Besides a few numbers (1 to 29), there’s not much indication on what these do.

Looking at my old emails, back in 2015-02-09 (before the GGMRC club donated the club to the museum in December 2015), I tried to understand this panel to light up the Fairfield town. Back then no lights were working and I “fixed” the power supply by replacing the fuse -- after finding they had the appropriate fuses behind the Stockton Yard panel.

From my notes, the fuse for the bulky 1997 power supply is a “5A 250V 313 Slo-Blow in a 31.75 mm (1 1/4") glass cartridge”. Here’s a mouser link for that kind of fuse.

So let’s try to decode all this. First, which toggles do produce any kind of lighting today? Allen and I could only find these below illuminating anything on the layout:

There’s been a multiplication of small-to-medium tasks recently, so it’s worth making an inventory of what needs to be done on the layout. These are more or less in the order I plan to address them.

As usual tasks fall into two categories depending on their “impact”, meaning how visible a task is. “High impact” tasks have the most impact on the layout as a museum exhibit, from a public/visitor point of view -- these can be either scenery improvement, automation improvement, or anything that enhances running for the Saturday Operators and public view.

I reinstalled the tablet displays on the layout, preparing for an eventual reopening of the museum:

Allen got a “Cracker Jack” illuminated HO-sized billboard, so we’re trying it out to see where it would look nice on the layout; here’s Allen first idea and it seems like a good option:

(Edit: And here’s an update on how I think these work and how I managed to ruin the electronics for that one by using the wrong voltage to power it, so now I’ve ordered a full replacement and I’m waiting for it; they also sell replacement controllers.)

I think we should power it from the building lighting circuits (see below). So I had to dig a bit into it. The building lights are controlled by a row of unlabelled toggles under the Valley panel: