Awesome! Thanks for the feedback - so you're saying it came off on the button end, not the PCB end? If so, I'll just squeeze them a little tighter with pliers so that doesn't happen.
If so, also just a tip for the other ones that are in route if you do run into this. On some of them the buttons aren't too hard to pull out using your fingernails, that could be an easier way to get to them.

Cheers!

Nice! I vote Road Race!

== PART VI - Post-Mortem ==

So, this afternoon I shipped the last 2 of the batch of 12.

IMG_4113 (resized).JPG

A few parts of my house were becoming sort of a mess, so first on the list is cleaning up.

IMG_4112 (resized).JPG
IMG_4111 (resized).JPG

I'll also need to finish shopping out this Wipe-Out so I can get it out of the way. As a sidenote, I picked up the worlds shittiest Wipe-Out this winter, thinking it was a good deal. When I showed up to the guys place with money in hand, I still picked it up but figured I'd just play it for a while then resell it, hopefully not at a loss.
But I couldn't even play it because the PF disgusted me

I worked on the playfield here and there, but the machine was in the way ever since in the workshop and I had to squeeze around it to work on the clocks. Pretty annoying.
Here's the before and after for the PF:

IMG_3389 (resized).JPG
IMG_4119 (resized).JPG

Once that's done I'll clean up the workshop too and that should be it for this run.

After that?

So building these 12 was really a lot of work, and the bad news for those on the waiting list is that I can't envision just starting another batch with the exact same build/process for what profit I made doing it.

The good news is that there are some specific aspects that were very time consuming and I think I can improve on the clock design itself to make the process easier. I'm still interested in the challenge of improving upon this iteration, so I'll probably be going back to the drawing board over the fall and winter, and potentially come up with another iteration early next year.
I'm glad I went with this number of units as it allowed me to find out issues I hadn't realized doing a single prototype yet I didn't dig too large a hole for myself that I couldn't get out of.

I'll keep updating this thread in the process to keep things interesting and keep people up to date, so keep watching!

Awesome! Very cool presentation.

I get asked by my friends "who the hell is gonna buy a 200$ pinball clock?!" and I tell them it is niche, but there are some interested people. The type of people who neatly collect 7 years worth of past versions of TurboTax in their library

Figured I'd give some news!
First, reading again my last post, Tortelvis , I hope you didn't take my comment the wrong way - I think your display is very cool and tidy looking. I found the TurboTax detail amusing in the corner of the picture there and it honours me that something that I created would be the centerpiece in what seems to be a meticulously arranged part of your home like this.

On my end, I've been able to free up the space in my less than tidy workshop

The Wipe-Out is now out of the way and is next to its two Gottlieb EM cousins.

IMG_4173 (resized).JPG

Looking a little bit less cluttered now:

IMG_4175 (resized).JPG

I've repaired the bed of my 3D printer which wasn't heating anymore, and I'm in the middle of printing parts for an improved mechanism prototype:

IMG_4176 (resized).JPG

IMG_4177 (resized).JPG

So far the results look promising but I'll wait until I'm a bit further along to let you guys in on the progress.

== PART VII - Post-First Batch Improvements ==

Hi everyone, it's me again. I'm leaving for a little vacation in Northern Florida where hopefully I get to visit great pinball places I went before such as Pinball Palace, Cafe One and also maybe checkout the Keg & Coin which I've got my eye on. Anything else, pinball-wise that I'm missing?
That's going to put most of the project on pause for a little bit, so I figured I'd let you in on the progress before I go.

As I was stating, there are a few pain points in producing the clocks that I want to do away with.
The first one is this:

IMG_4186 (resized).JPG

These are the switches that help the clock figure out when each reel has reached an initial position to which the digits will be relative. Although spec-wise I think it is well suited to the project, an issue I've been having is that unless they were adjusted just right, not all of them would have 100% repeatability. They have a pretty narrow range of operation, so that they sometimes require adjustment during assembly.
What this means is that I'll assemble a whole clock mechanism, go through the reset sequence a few times and realize that one of the switches either sometimes "over-registers" (is triggered by the inside of the reel before it hits the notch to that effect) or that it doesn't register at all because the notch doesn't move it in enough. So I'd take apart the whole clock, readjust the switch, and start over. Sometimes I'd go through the calibration process which in itself is pretty long, and realize an issue after that, so I'd have to start all over again.

So I decided to nix the switches in favor of something else. After discussions with an electrical engineer buddy, I decided to try reed switches.
pasted_image (resized).png

How these work is that as you can see in the picture, they're tiny mechanical switches inside glass. When a magnet comes close, it will magnetize the switches which will in turn be drawn to stick to one another, thus closing the circuit.
Again, here, the name of the game is repeatability. I want the reaction to be the same every single time, so that the numbers line up perfectly every time.

So I redesigned the reels to have an enclosure to put a little magnet:
IMG_4185 (resized).JPG

And redesigned the brackets to accommodate a reed switch:
IMG_4184 (resized).JPG

I took the opportunity to add a third connection point between the brackets to make the whole assembly more solid with the little clip at the top.

I started with a test on a single reel:

So far so good.

Then I needed to check that the motors and magnets from nearby reels wouldn't interfere with one another, so here is a full assembly test (while watching some Seinfeld ):

Still good. So that's one thing taken care of - unless something comes up during the (potential!) next batch, that should save me a lot of time. Added bonus, I like having one less mechanical interaction going on.

Back from vacation, and back to work on the project.

To complement the little vacation side story, I ended up going up to Georgia, but this time around I went canoeing in the Okefenokee Swamp instead of going to the Pinball Palace. Since I was in the Jacksonville area, I did check out two new places that I highly recommend.

-The Keg n Coin, which has a good beer selection, good punk music and a small but well kept selection of recent Stern and JJP machines.
-The Leaderboard which has a large selection of mostly recent machines, including JJP, American, Stern and Spooky.

Now to come back to my "process" improvements. Another one of the building pain points is the sanding of the reels. If you go back to the previous pages, you'll see the come in really rough from where I order them and I have to spend a lot of time sanding them. I have little control on how they get printed, I only pick the material and hope for the best.

The home printer I have is very basic, with a small print bed. I figured I'd get a better and slightly larger one so that I can print the reels myself and have more control on quality, and potentially the rest of the mechanism as well.
After reading a bit on the topic I decided to get a Prusa Mini+.

pasted_image (resized).png

It should be a few weeks until I receive that.

In the meantime, since I've been doing all my prototype printing in black and sticking white paper to it when I want to try it out with actual numbers, I had the idea to try and get actual white numbers to go on the black just to see how that looks.

IMG_4197 (resized).JPG

IMG_4216 (resized).JPG

IMG_4217 (resized).JPG

Those are rough prints, not sanded, hence the shininess, but you get the general idea.

I think they look pretty cool, but I think I still prefer the regular white. Could be an option at some point, we shall see.

Another little experiment while I'm waiting for my 3D printer.
I had a clock sitting without a cabinet running on a table for a few weeks for testing, and I started thinking it looked pretty cool like that too, so I thought maybe a "naked" model would be cool.
Then I thought maybe with an acrylic box it would be even cooler, so I put this box together to try it out.

IMG_4218 (resized).JPG

For the buttons, I could drill the sides but I had a few EM switch stacks hanging around so I tried this instead.

IMG_4219 (resized).JPG

Here's what the result looks like - I put all that together real quick but it gives an idea. I didn't bother to calibrate the reels and I was really sloppy with putting together the box so the pieces aren't cut super clean and I had to sand and buff it to clean up some of the spilled glue which is why it looks sorta hazy.
I think it looks cool, but I have to figure out a proper way to build the boxes so as to not create another PITA for myself.

IMG_4221 (resized).JPG
IMG_4222 (resized).JPG

Some videos I took of ongoing stuff:

Adjusting the time with the EM switches:

An improvement I made to my calibration app to simplify my life again. Calibration used to be done by going over each digit, guessing the needed offset adjustment, then on to the next digit. Once I'd have done all 9, then I push the offsets and go over the whole sequence again, doing this until I get everything just right.
Now I can adjust them on the fly and see the results instead of guessing.

Thank you sir!

Ok a little bit more news - what I've been working on lately is making the calibration app a little more clean and user friendly in case anybody but myself would need to use it in the future.
Also, I've put a little bit more time into trying to see if the plexiglass case version would be interesting. Two things I did is make the LED strip less visible by getting a narrower strip and cutting out a groove to hide it. This isn't a problem in the regular version since it's hidden by the front panel. Looks better!

IMG_4236 (resized).JPG
IMG_4237 (resized).JPG

On that same front, I printed some jigs to help glue the plexiglas panels together cleanly. Still trying to get that nailed since the cement is very liquid and a little bit hard to work with.

IMG_4238 (resized).JPG

Also doing a few improvements to the cab plans to reduce post-processing.

Again, this is all while I'm waiting for my Prusa 3D printer to see if I can self-print the reels to require less post-processing.

What I can say at this point is that barring some major unforeseen event, I'm almost definitely doing another run. What I still need to figure out is what options there would be (assembled, kit, plexi-case), what price, how many and when. I don't want to commit to anything at this point to avoid putting unneeded pressure on myself, but once I'm good and ready I'll get in touch with people who told me they were interested.

Stay tuned!

Yes, this is in my backlog - however there's code I need to rework to implement it cleanly so I can't make it a quick fix. Will let you guys know when it happens!

Yes, I have you on the list. I'm not sure how long the next run will be but there's a decent chance you'll be able to squeeze in

Received a new batch of PCBs this week. Went with black for this one.

IMG_4241 (resized).JPG
IMG_4243 (resized).JPG

Yay!

IMG_4244 (resized).JPG
IMG_4245 (resized).JPG

The Prusa is up and running.

IMG_4247 (resized).JPG

I'm really happy with my purchase, it feels like a high quality product.
To the left is the reels printed with the Prusa - to the right is the reels I had outsourced for batch 1.

IMG_4246 (resized).JPG

Here's what I get after some sanding. Same end result, but a lot less sanding required.

IMG_4249 (resized).JPG

### Batch 2 coming ###

Now with this out of the way, I've pretty much derisked all parts of the build and I can safely say I'm going ahead with another batch. I've begun ordering parts, but I'm still waiting for some quotes before I can make up my mind about cost and schedule and then get in touch with the people who have stated interest, in the order they did so.

### Auto-calibration / Automated Testing sub-project ###

Now, waiting for parts isn't the most exciting stuff to read about, so I figured I'd work on another sub-project and post about the progress here to keep things interesting.

I come from a software background and usually do TDD and ATDD (Test Driven Development and Automated Test Driven Development). As the name implies, it means that you write test code that will test the functionality code you're writing first, and then you write said functionality code.

Since this isn't a pure software project, I had the idea of using a camera to watch the clock, then have code to validate that what it's seeing is what's expected.
These building blocks could also potentially be used to automatically calibrate the clock digits, as opposed to me doing it manually.

pasted_image (resized).png

I will post progress on this while waiting for things to fall in place for batch 2.

So, first thing was to be able to capture the webcam from my custom C# software.

The blue lines will be so that the software has a reference point as to where the numbers should line up.

AutocalStep0 (resized).jpg

Next, I figured I'd try some Optical Character Recognition so that the software can then find the numbers on the image. I'm not sure that's going to be necessary, as I could also just see how much black goes beyond the lines in either direction (for calibration), but for automated testing it would be helpful to see whether the correct numbers are showing up.

Tesseract seems to be the standard OCR framework so I figured I'd give that a shot.

First, with the sample image they provide:

pasted_image (resized).png

The results I get:
AutocalStep2 (resized).jpg

Ok so the sample stuff is working. Let's try the reels now.

I figure I need to clean up the picture a bit to "help it out". Here's what I start with:

AutocalStep1 (resized).jpg

and here's the cleanup:

AutocalStep4 (resized).jpg

No worky:
AutocalStep5 (resized).jpg

I gave this a try just to be sure that my problem isn't the font itself:

pasted_image (resized).png

That works:

AutocalStep3 (resized).jpg

So I guess I'll try to go with pure black and white:

AutocalStep6 (resized).jpg

No worky:

AutocalStep5 (resized).jpg

Ok this had me stumped but I thought what else is different... the spacing maybe? Let's try that!

AutocalStep7 (resized).jpg

And the results:

AutocalStep8 (resized).jpg

YAY!!!
Doesn't look all that confident (0.26), but I'll take it.

Now I did all that processing by hand, I'll need to implement it in the software.

Since not everyone posted directly to the thread and it's not all public info, I'd rather people message me directly if they want to know what their spot is. What I can say is that if you told me you were interested, either by message or posting here, I did write your name down.
In your specific case, you're at #11 and I'm ordering 15 units worth of parts, so you're definitely in there.

Updated the software so that it does the filtering.
I drag the blue lines to outline the numbers - then when I hit the Test button it takes a snapshot and stitches it together. Next step, feed that to the OCR code.
pasted_image (resized).png

Ok, OCR code plugged in... all good, moving on! Right...?

pasted_image (resized).png

uhh... QIWy..? (Circled in red is the character recognition results)

Yeah ok, QI I can see, Y maybe,... W???

Gotta put some more work into the image recognition

Ok, so you can easily give the Tesseract engine a whitelist of characters. I gave it the list of numbers, and here's what I have:

pasted_image (resized).png

pasted_image (resized).png

pasted_image (resized).png

I still get some misfires, as you can see in the first pic.
Seems that when I play with the contrast and lighting I can get it to be pretty consistent. I know that I could potentially retrain the AI to recognize the characters on the clock themselves, but good enough for now, let's add another fun part!

I'll try running the clock at accelerated time and have the software validate that time advances in the proper sequence.

I now have a basic testing application!

It runs the clock at x10 speed, and it looks to see that the clock shows what it should. The green background behind the "Read" text means that it matches what's expected. If it didn't it would turn red and the bad match would be added to the list.

Currently the testing app doesn't "talk" to the clock, I just start both at the same time. I'll add some communication to test the various commands as well, then I'll look into also testing for misalignment.

Another piece of code to work towards auto-calibration / alignment validation.
In a first step the code detects the top and bottom alignment of the displayed numbers (black lines). Using the slider to the left you determine what the acceptable threshold is.

pasted_image (resized).png

It then detects when something goes outside this threshold and turns the line red.

pasted_image (resized).png

I might need a little more resolution, but that's the basic idea - I'll find out when I try to plug everything together.

A saturday night update - plugged the code that looks for misalignment into the hour validation sequence.
Last 2 pieces of that puzzle will be to have the app talk to the clock, and writing the adjustment sequence.

Exception at the end, still bugs to iron out

Ending the weekend with a little gin tonic to celebrate the working auto-calibration app.
I made a video that explains and shows the thing, enjoy!

Next updates should be when parts start coming in, not sure when that's gonna be... stay tuned and happy holidays!

*Edit: Hey look at that, 100 favorites! I'll drink to that too!*

A happy, positive, constructive New Year to everyone!

I've had the time to work on some stuff over the break which I'll share with you.

First, I've printed a bunch of parts.

IMG_4298 (resized).JPG

Then, parts on order have begun coming in, so I should be able to start putting together reel assemblies in the coming weeks.

IMG_4300 (resized).JPG

In the meantime, I've been working on another peripheral project. Since this whole thing started out as me wanting to reproduce score reel clocks without having to source actual score reels, I've always had in mind that it would be cool to go back the other way and try to use the score reels in actual pinball machines.

I've refactored the code (which I'll tell you about in a coming post, for those who are interested in this kind of thing) so that the score reels can be used in that kind of context.
As a clock, it's really a slow and low workload - change happen once a minute, the reels move at a relatively slow pace, and button input doesn't register while the reels are moving.
For a game situation, the reels need to move faster and I can't lose inputs.
I believe the code can potentially still be optimized to be more reactive - currently what I'm doing is that I'm accumulating inputs in a buffer while the reels are turning, and as soon as they're done I process the buffer.

For the moment I just have the two inputs and it seems to be working properly but the real test will be to hook it up to an actual machine:

IMG_4303 (resized).JPG

I plan on using these signals to trigger the Arduino:

IMG_4304 (resized).JPG

More on this in the coming weeks, plus updates on the clock builds. Not ready to start pre-orders yet, will let you guys know!

Here is what the code currently looks like on my score reel development branch.

ClockUML (resized).png

To the left is the code that runs on the PC - both the manual calibration app and the vision app use a single CommLib library which handles serial communications with the Arduino.

To the right is the Arduino code.
-An EventHandler class handles the communications with the PC (only necessary during the calibration process or other configuration adjustments).

-There is a StateMachine class which handles in what state the score reel is in (reset state, alignment at 0000, running state)

-There is a Button class which handles button input (there's currently two instances of this, one for each button)

-I've refactored the Clock class to split it into a base ReelAssy class, which now has 2 specializations: Clock and ScoreKeeper. What this means is that basically both the Clock and the ScoreKeeper (aka Score reel used for keeping score) exhibit the same behaviours and use the same logic, except for a few differences, which are coded in each of the specific class.
The main class doesn't need to know which one it's dealing with, it just knows it's dealing with a ReelAssy, and that ReelAssy (actually Clock or ScoreKeeper) knows what to do. This is called polymorphism and is the basis of object oriented programming (OOP).
This would also allow me to have another class for another usage, although when you get into trees of inheritance things get cumbersome and it often sounds better in theory than in practice.
There are other potentially ways to do this which I won't get into. The whole OOP approach in itself is debatable for this kind of project in any case.

-The Clock class has an RTC class which handles the Real Time Clock.

-The Config class handles saving and reading the configuration from the Arduino EEPROM (electrically erasable programmable read-only memory)

-Finally, each ReelAssy has 4 instances of the Reel class which handles the mechanics of turning each reel.