Who made this tower clock, and seeking advice on restarting it.

[kmt]

Since I already have a Raspberry Pi installed at the top of the tower in order to run my webcam, I figured it might as well do double duty and monitor the performance of the clock as well. So I wired up a little IR sensor/emitter pair across the path of the pendulum, connected the sensor to one of the Pi's GPIO pins, and threw together a little monitoring program:



This is still very much a work in progress, but so far it's working quite well. Up at the top are the current actual time, the turret clock drift, and the turret clock time, as well as realtime beat statistics. The software isn't able to read the time off the turret clock, but once this has been set manually, it can track drift based on the beat detections, and it can do things like send an email if the beat stops.

So far it's just displaying the current information, but I will add database recording to permit visualisation and statistical analysis, which should prove quite interesting, and probably useful in predicting performance problems as well as potentially illuminating the cause of such problems.

This is a long way from being anything resembling a complete package, but once it is somewhat more featureful and robust, I will publish the source code for anyone who is interested.

Also, after I iron out some software issues, I will make the webcam livestream public.

-k

 

[FDelGreco]

Could you also record ambient temperature to see if BPH or clock stopping has something to do with temperature?

Frank

 

[kmt]

Could you also record ambient temperature to see if BPH or clock stopping has something to do with temperature?

I am planning to do that in fact. I have ordered a couple of sensors that I can stick on the Pi GPIO and track that data alongside the beat data. I don't have them yet, but perhaps sometime next week they will arrive. I am certainly feeling no rush on any of this. This is a relaxed pace project.

And in fact the clock did stop again today, shortly after my monitoring software showed a significant swing (as much as 0.2 s) toward the "slow" direction. I came back out to the clock and it required three starts before it kept going. Something definitely was a bit sluggish.

Monitoring showed everything in green and yellow zones (although my tolerances have been arbitrarily set, just to produce a variety of output as I work on the software), so I was getting ready to take the bus home, but then there was another trend toward a lot of red, so I've stayed out here for a while.

15:27:29.187 - beat detect (+2475 µs / 0.9988 Hz / 3595.6 BPH / -106.9 s/day)
15:27:31.189 - beat detect (+2350 µs / 0.9988 Hz / 3595.8 BPH / -101.5 s/day)
15:27:33.192 - beat detect (+2395 µs / 0.9988 Hz / 3595.7 BPH / -103.5 s/day)
15:27:35.195 - beat detect (+2520 µs / 0.9987 Hz / 3595.5 BPH / -108.9 s/day)
15:27:37.197 - beat detect (+2435 µs / 0.9988 Hz / 3595.6 BPH / -105.2 s/day)
15:27:39.199 - beat detect (+2395 µs / 0.9988 Hz / 3595.7 BPH / -103.5 s/day)

Getting hungry though, so we shall see how my patience holds.

-k

 

[kmt]

...and it stopped again just about half an hour after I got home.

I really need to get that motor mount straightened out, and now that I can get immediate data on the effect of whatever I'm doing, I can probably tell fairly quickly if that's related.

Also, someone I was talking to the other day pointed out to me that the motor can disconnect from the transmission/gear assembly, which makes both realignment and removal far easier than I thought, so it will not be that hard after all to test the system with the winder removed.

I noticed when I visited today that the whole thing just felt very "sticky" for lack of a better term. Like the oil just wasn't really flowing well. I'm using Priory Polish Turret Clock Oil, which claims "[e]xcellent low temperature performance, pour point -38c", but I am starting to wonder if maybe I should be using something a little bit lighter, given that we're as low as -25°C during the winter, although it's only about -5°C today, so it doesn't seem like this should be a problem, today at least.

I've noticed a little bit of damage on the weight line, no emergency or anything, but just the same, especially given the suggestion of inspecting all the pulleys, I do think I will replace the line, but I will wait until springtime to do that, because it's just too cold to be dealing with that right now.

-k

 

[Thomas Sanguigni]

-k, thank you for showing some links to the neighboring area. Was the 3d view of Nokia from the clock tower? It was a neat view albeit a bit industrial.

 

[kmt]

Was the 3d view of Nokia from the clock tower? It was a neat view albeit a bit industrial.

No, those photos were made from a drone. If you pan around, you can see the clock tower, on the yellow building near the access bridge.

Yes, it's an old industrial area, and is still home to Nokian Rengat (Nokia Tyres), which some of you who live in colder climes may know as the manufacturer of Hakkapeliitta winter tyres.

This area is just a small part of Nokia. The city is nice, and mostly much less industrial. Hopefully this area has some more cultural life in the future.

-k

 

[Thomas Sanguigni]

I looked again, and I spotted it. Neat looking building. We have this same problem here in the USA. We get areas that get 'old' and urban sprawl takes over. People move further away, and the area becomes less desirable. I hope it all works out.

 

[kmt]

Some progress and some interesting data! I've been working hard on cleaning up the cable spool assembly, which I discovered isn't actually painted black like I thought it was. That was at least 50, if not 100 years of grime buildup. As the spool has presented different parts at different times, I've gotten bits and pieces cleaned down to the metal, figuring that eventually I'll get to every part, if I just keep cleaning all the black bits as they come around to reachable places. Sure, a real pro would just take the clock apart to do this in a bath, but I'm not really to a point where I want to be doing that to this clock.

Most important, I got the spring assembly much cleaner than before, and now the cable spool finally moves smoothly and the spring compresses well. As a result, it appears the clock is now able to survive the more difficult times, and keeps running. Fingers crossed. Of course, this does not mean that whatever is causing the issues of inconsistent performance should be ignored.

To that end, I got the database portion of my monitoring software written, so now I can produce some useful visualisations of pendulum behaviour, like this one:



This graph is showing the pendulum timing skew, which I have defined as "how far away from 2 seconds is the period". It's plotting a five minute average as well as the maximum and minimum recorded skew during that five minute period.

A couple of things to note:

- Rapid temperature change definitely presents a challenge for the clock. The large mountain at the beginning is what happens when I turn off the heater and take the bus home. The pendulum got as much as 5 ms too slow. Watching the numbers on the bus, I was surprised that the clock didn't stop, because this was far worse performance than the night before, when the clock did stop. This is where I think the cleaned up spring saved the day.

- There is definite periodicity here. There appear to be (at least) a three hour repeated event and a six hour repeated event.

It will be interesting to observe what happens when the winder runs, since unless it happens to be rather unfortunate timing, this should indicate whether these slowdowns are coming from the winder end of things or the clock end of things. If it's the winder, the problem period will probably shift. If it's the clock, it won't.

Now that I have good data collection, and I know that the winder isn't so hard to remove after all, I'm inclined to start looking at that part of the system, although depending on when I am feeling inspired to do so, and on what behaviour is observed in the next few days, I may feel like just getting some more cleaning done first, especially if the clock will actually run for a while!

-k

 

[kmt]

It's been a week without an update, and it's definitely time to let everyone know how things are going, because the answer is "well".

I took the motor off the winding system, but quickly determined that it was not the source of the skew problems, as the axle coming off the transmission and into the motor was found not to be turning when the clock was running. So it was already going through a one way mechanism somewhere before that point. It also occurred to me that since the motor is connected to the clock through a worm gear, it isn't even possible that the motor was being turned by the clock, since worm gears will not run backwards.

Upon closer inspection of the remainder of the winder transmission, I determined that it cannot be disassembled unless the whole thing comes off the clock as a unit first. That is doable, but it's a two person job if it is to be done with adequate regard to the safety of the clock, not to mention the human(s).

However, I was able finally to get the winder attachment straightened:



And I am very pleased with this "before and after" plot of the pendulum skew:



It isn't perfect, and I would still like to know how things look with the winder totally removed, but this improvement is almost certainly enough to make the clock run without stopping, and for it to be adjustable to fairly good precision, a task I haven't begun at all yet. I am content for the time being with a slow clock that doesn't stop.

I have also been working on some cosmetics, and have been manually cleaning up the gears as they present themselves (some of them obviously make this process much quicker than others), so the mechanism is bit by bit getting nicer to look at. I also spent a lot of time working on one of the rusted support bars and turned this...

...into this...

I still plan to make the webcam stream public, but so far I'm holding off on that until I know the clock is running reliably, and also until a few more parts are polished up and presentable.

-k