Gina this project looks exceptional.  I can't even begin to understand the details, but it makes me want a 3D printer.  Saying that I think I may stick to film photography & scanning.  Its always very interesting to see your posts. 

Latest photo of the parts I've designed and printed so far (not including the display drums).  It's not an easy project but I'm gradually getting there

There's been a delay in the delivery of the picture frame and until I get it I can't establish the exact dimensions of the box sides or back so since I think I currently have more chance of success in proceeding with this project than with the longcase clock and I want a working calendar to use, I shall make use of the old acrylic sheets and the framework to further this project at the expense of the longcase clock.  I already have my moon globe clock so the longcase can wait. 

I think I've gone as far as I can without fastening things to a solid base and using bolts to carry wheels etc.

This is as far as I could go without being able to mount things and check for clashes.

Some of the parts mounted on an acrylic sheet.  The combined hanging levers as springs may not work.  The main date drive bar one certainly doesn't work - there's far too much movement for that length.  The month drive one might work but I'm not at that stage yet.

More parts mounted and the opportunity to test the date advance.  Note - there is some distortion due to wide angle lens.

The end-of-month pawl (purple) wants a better shaped tip to fit the tooth.

To go off at a bit of a tangent, I've been thinking of much easier alternatives, interesting though this purely mechanical perpetual calendar might be.  The original thinking was as part of a mechanical clock with the only modern workings being auto-winding.  In that aspect this made perfect sense.  But if I'm driving the calendar off an RTC with all its facilities, maybe it's daft to just use the change of date to trigger the mechanism.  All the "clever stuff" of differing numbers of days in months could be handled in code quite easily.

I could make the mechanical perpetual calendar part of another clock project with just the simplest electronics.  For example, the 50Hz mains frequency could be used to time an Arduino which would then drive a stepper motor at a constant speed.  Since the mains supply is maintained to the accuracy of an atomic clock over a 24hr period the clock would keep perfect time - even better than using an RTC or quartz crystal.

It would be possible to make maximum use of an RTC and reduce the mechanical parts to a minimum even with a mechanical display.  The reason for wanting a mechanical display is to provide a large and easily seen display under all normal lighting conditions from full sunlight to electric light in the evening.  The best/easiest way of doing this seems to be to use drums with numbers and letters glued on.  In the past I have investigated and experimented with various types of display and the drum approach won out.

The minimal mechanical approach would be to use stepper motors to drive each drum, which means four of them.  Next would probably be to use the layshaft as in this project but driven directly (through gearing) by a single stepper motor, plus another for the day-of-week.  This relies on the special type gears used to turn the drums correctly.  I shall explain this shortly as I realise I have not done so yet.  The layshaft gearing has 31 teeth per revolution and is simply advanced one tooth at each midnight, except at the end of the month when it needs to be moved on extra notches for all months without 31 days.

I guess it's time I described how the 31 days is converted into the day-of-month in terms of two decimal digits and how the weekday and month are driven

The basic principle uses gears with missing teeth where no motion is to be transferred to a pinion on the drum.  eg. for the units the counting is one-to-one until we get to the end of the month at day 31 and the units show "1" but the next position corresponds to the first day of the next month so instead of going from "1" to "2" it needs to stay at "1" so the tooth that would turn the pinion is missed out and the units drum stays showing "1".

For the tens digit of the day-of-month, the drum wants to show zero or a space until the day reaches 10 when the tens digit wants advancing to show a "1".  So no teeth for 1-9 then a tooth at 10 turns the tens drum one digit.  From 11-19, again the drum must stay at "1" but a tooth at 20 moves it on to display "2".  Same again from 21-29, no teeth, but one at 30 to move the drum on to show "3".  Finally, when the date goes from 31 to 1 the tens drum needs moving onwards to show "0" or a space so here we have a tooth.

To advance the month, one single tooth where the date goes from 31 to 1 engages with a pinion on the month drum to advance it to the next month.  The months drum pinion has 12 teeth for the 12 months in the year.  The units of date has 10 teeth (no surprise there) but the tens doesn't have 4, it has 8 because a 4 tooth pinion would not work - the tens drum has two sets of 0123 and 8 positions.

The day-of week drum has a ratchet with 7 teeth.

Here's a collection of photos and diagrams that show the principle and test rigs.  These also show that where no teeth are required the "gear" can be reduced to just what's needed to carry the wanted teeth.

To use this mechanical perpetual calendar as the project stands would not need a Real Time Clock, just a source of good timekeeping such as the 50Hz mains frequency.  This would simply need counting for 50x60x60x24 to produce action every 24 hours.  The action would consist of creating sufficient stepper motor steps to rotate the cams one revolution.  The 50Hz input would use interrupts to avoid the stepper code from causing missed counts from the 50Hz.  Of course, the Arduino sketch would need synchronising with real time so that the calendar advance occurs at midnight.  A simple push button pressed just once at midnight one day could do this.

The picture frame arrived this afternoon so maybe tomorrow I'll sort out the box.  Having gone this far with the mechanical parts I think I may as well stick with it - I seem to have a bit of a habit of changing my mind and scrapping many hours of work!

Printing drum axle brackets.

I feel like I might get back to this and have a change from the couple or three projects I have been working on.  Making up the box from timber will be a nice change from messing about with 3D printers

Well I didn't but I have done a bit of thinking...

With this fascinating mechanism it seems a shame that it will be stationary most of the time - only operating at midnight and only doing "clever" things at the end of 30 day months or Februaries.  And it will only be me that sees it.  It seems a lot of work just for that!  A practical perpetual calendar could be much simpler - all the "perpetual calendar" data is already in the RTC module and it only needs stepper motors to drive the drums.  With the Arduino RTC library all the data is easily read out and could simply be arranged to control the stepper motors that drive the drums.

It seems to me that my time might be much better spent on other things.   While the perpetual calendar was going to be part of the longcase clock it made sense as everything would be mechanical but on its own I don't think it does.  OTOH I guess this mechanical perpetual calendar is almost finished and may take less time and effort to finish than starting again.

I quite fancy getting back to this project but unfortunately it has a lower priority...

Priorities... nuts!!!  Need a break from all the hard stuff

Decided the month advance bar that turns the days-in-the-month wheel looks wrong across the 31 days wheel so I'm thinking of taking it across the bottom, clear of the 31d wheel.  Here is a rough diagram.

Another redesign to reduce clutter.

Another idea - simpler (I think).

Printing a trial month advance lever to see how it fits the rest.

That design had a few problems so yet another.

This is a very difficult thing to design empirically so it's a matter of trial and error.  Fortunately filament is cheap - even the best quality and reduced fill density keeps both cost and time down

This is silly!!  I'm at it again, going against what I said earlier about not going for a mechanical perpetual calendar but using the facility already built in to the RTC chip.  Drums driven be a stepper motor each is simplest mechanically and pretty certain to work.  I have several spare control units from 3D printers that I could use one of.  These are Arduino Mega 2560 plus RAMPS shield containing up to 5 stepper driver motors plus other stuff.  That would easily fulfill the stepper motor driver needs and would just need an RTC module added.  The I just need to find the cheapest source of stepper motors.

Ordered a set of five NEMA 17s from Amazon with Prime delivery tomorrow - £39.  OK so they're far more powerful than needed but the size of the market for these makes them relatively cheap.  I really want a calendar ASAP and this seems the quickest route

Decided NEMA 17s are too big for this job - I have 3 NEMA 11s and a NEMA 14 I can use instead.  With this arrangement I should be able to put the calendar below my Moon Dial Clock if I move the clock a few inches higher up.  That will leave room on that wall for my weather station analogue wall display, which can make use of the NEMA 17 motors.