Today Allen and I finally got the “Cracker Jack” illuminated HO-sized billboard powered by the layout’s building lighting panel:

Even with the full room lights on, it’s still quite visible and should be a nice animation for visitors:

Back in June, Allen got the billboard and we temporarily installed it, yet it remained unpowered. The sign can be powered via a battery and instead I wanted it connected to the layout’s building lighting circuits using an Adjustable Voltage Regulator.

That project took a bit longer than expected -- first I plugged the voltage regulator in reverse (thus increasing the voltage instead of reducing it), and that killed the billboard’s control board. So I got a replacement for that, and this time wired it up correctly. Allen and I made the building self-contained with the voltage regulator and I tried adding an orange LED inside to add some scene lighting, which honestly didn’t work and is quite sub-par.

Concluding 4 days of work, this morning I installed the new version of the vision software on the Vision computer at the museum for the Train Vision Project.

The project has three cameras looking at the layout and displays live videos when it detects trains. In the initial version, that video was really choppy. It was fine when I was experimenting with this on the wifi at home, but the same cameras + laptop resulted in a really bad frame rate when using the sub-par wifi in the train room.

So I spent the Thanksgiving break entirely rewriting that part of the software. In the previous version I was using the FFmpegFrameGrabber to get the cameras’ RTSP h264 feed, analyzing the frames for motion, and then feeding them to the display by drawing them directly with a Canvas in a JPanel. That was really not optimal. In the new version, I replaced the custom viewers by instances of VLCj in Direct Rendering mode, so now I have VLC driving the RTSP feed, and I use the reverse where the VLCj decoded frames are fed to the motion analyzer. They all work in their respective threads with blocking queues for synchronization.

The result is that the videos are really better looking now. Not choppy anymore. I also reconfigured the cameras to boost the contrast and saturation as the image was rather bland -- I overdid that part a tad much, I need to tune it down now.

The only downside is that CPU usage really went up, which makes sense since I now have 3 VLC viewers in 1280x720 @ 15 fps doing their work where before I had 3 meager 640x480 feeds at some very irregular frame rate. But that’s fine, because now I can underclock the analyzer threads and run them at 5 fps -- they don’t need to detect motion on every single frame. Overall the software now runs at 50-75% on the 4 CPU i7 cores which is on par with what I had before, but with better quality.

The last thing I’ve done is add a statistics collector for debugging purposes that can export the motion events to the chrome://tracing json format:

That should allow me to fine tune the motion detection thresholds.

I’ve done some work on the Vision computer to get more reliable networking. The USB-to-ethernet adapter hasn’t been behaving as stable as I’d like to, and I fixed the wifi so at least I got that.

But maybe it’s not just that USB-to-ethernet adapter that needs some more work though -- last time I went by the museum, all the port lights were off on the ethernet switch till I power-cycled it so maybe I should look at that first.

I have also updated the highlighter’s logic in vision, and actually I like the updated logic even less than the original. Look at these stats:

That’s terrible. Too many highlights. The thing is too sensitive, I need to change it to have some kind of adaptive threshold. Some of the highlights happen when no trains are moving, probably because the RTSP h264 stream is of subpar quality and very noisy, not to mention noticeable inconsistent frame rate. That’s why I need that wired ethernet to work.

And what’s up with the lack of stats after noon? I know the laptop was running at least till 4-5 pm. If it were a 12/24 hour mismatch, the “12 pm” events should look like midnight (12 am) and the 4-5pm events should look like 4-5am, which is not the case here. More mysteries.

A few things have happened this week-end.

First the visible yet inconsequential thing: the Bridgeport Ballon Track #2 now has the proper Auto-Reverser.

The story about this is that I had ordered two Tam Valley Dual Frog Juicers to be installed and configured as auto-reversers, one per balloon track. Unfortunately in the order I was sent a DCC Hex Frog Juicer by mistake. The Dual Frog Juicers can use up to 4 amps per track, whereas the Hex Frog Juicer can only only 1 amp per track and that’s not enough when running multiple engines in consist with sound. So now I just got around to installing the proper board. It was a rather trivial swap since I had done all the prep work when installing the Hex Frog Juicer, so really only terminals to unscrew/rescrew. Probably took more time to take the pictures and write this post than actually installing the board. But hey, one more task is complete on my long list.

The more exciting news is that I got around to update my Layout Presentation Documentation. I know, exciting, isn’t it? “See? Nobody cares.” Well I do. I added a whole chapter on the computer setup used in the “train room exhibit”, and updated the pending work list, and I think that’s pretty neat. I’ve also updated the Known Issues list.

Speaking of disasters waiting to happen, this is the mainline track at T330 Sonora:

What we can observe here is that both rails from the turnout lead are not soldered to the mainline track anymore. There’s some remnant of solder, thus I’ll venture a guess they used to be joined. Moreover the track lifts by about half a millimeter and can flex up and down freely. It’s amazing we don’t derail here. And that’s the mainline where every single train runs..

Description of Issue

As explained earlier, we’ve been having issues with the train stopping on a dead spot on the mainline block B321, just after the Branchline turnout T322. I posited the issue was a broken solder joint, and it was indeed the case. A repair has been made -- actually two of them --, we’ll see if that holds.

On the track schematics, repairs were performed at the approximate locations shown below, on the red-side rail.

Trains used to stop dead in that area recently, intermittently. Typically that’s due to a solder joint that is broken and now makes a bad contact. As the track shrinks or expands with humidity, temperature, and the weight of the rolling stock, contact can appear or vanish during the day. The test was simple: with the train dead at that spot, just putting some slight pressure on the track was enough to power the train again. Problem is, that track has been naturally weathered by decades of usage, and it’s nearly impossible to know where the solder joints are, or for what. There can be many, from track feeders, to rail jointers. Not to mention the voluntary block isolation gaps that I need to preserve. Due to the lack of documentation, it means that when I find a gap (or a broken rail that looks like a gap), I don’t really know if it’s on purpose or not.

Description of Fix

First repair at spot #1 was done somewhere near the middle of the block. There is a track feeder under the layout going to both the black and red rail. It’s worth pointing two important details at this point:

The motion sensor finally has a little cover to keep it in place:

Description of the Motion Sensor and how it is hooked into the NCE AIU01:

https://www.alfray.com/trains/randall/about.html#h.mnaglpwxrgon 

For reference, here’s how the sensor is calibrated:

Recently a dead spot has developed on the mainline track #1 just in front of Stockton Station.

And last week we had another one: the automation trains are stopping on the mainline between the Branchline Interchange and the Sonora turnout.

As with (the one) previous short we had on the mainline, this one is intermittent, which makes it so fun to find and debug. It works for a while and then suddenly stops. That’s the typical sign of a cold / broken solder joint that moves during the day as the rail expands or shrinks with humidity or temperature. So now I have to look for a solder joint. Could be a track feeder, or it could be a rail joiner, who knows.

The “nice” thing is that I can temporarily fix the dead short just by pressing on the outer rail. So that gives me a clue...

It’s been raining. And we’ve been reminded the joy of being in a basement:

There’s a full inch of water on the floor. Not exactly a novelty, although when it happened a couple years ago it was supposed to have been taken care of.

The new addition is that water is also dripping on the Napa yard; we guess it comes from the ceiling:

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.