Now to the fly-fan gear chain, taking what I have drawn above :-

1. Gear on main shaft - 105mm with 75mm forming the pair - ratio 105:75 = 7: 5
2. Next gear pair to fly-fan - 150:20 = 15:2
3. Total ratio = 15x7/(5x2) = 21:2 or a little over 10:1 making the fly-fan rotate at 10x the striking rate
4. With a striking rate of one per second the fly-fan wants to rotate at 10.5rps which from hand testing seems a good starting point.

The escapement.

Having fiddled about with my current anchor escapement including variations in the escape wheel as well as the anchor, with setting it all up then moving it to the clock case and back again after failure, I think it's time to make a test rig!  First diagram, new escape wheel and second, a failed idea due to being unable to print sharp teeth.  3D printed plastic (PLA) is very different from brass.  Apart from the standard escapements, I have a couple of ideas of my own I can try.

Here is my idea for an escapement.  It's nearly deadbeat but not quite.  The anchor part has "pins" in the form of miniature ball bearings just 6mm OD.  These lock the escape wheel on the leading sides of the teeth whilst a sloping top provides the drive action.  Having ball bearings rather than plain faces reduces friction.

Looking more at this design I've seen an improvement that makes it truly deadbeat (no recoil).  It is always the leading side of the escape wheel teeth that contacts either ball bearing on the anchor so I was able to make this surface an arc centred on the anchor pivot.  Thus in the "lock" phase, the escape wheel doesn't move.  Once the pendulum swings back the bearing rolls off the arc and onto the top of the tooth providing the "drive" phase.  Once the bearing and tooth part company the escape wheel rotates so that the edge of the tooth near the other ball bearing makes contact with the tooth.  The pendulum continues to swing a bit further and the bearing rolls along the tooth front side.  It swings outwards and then back (in the "lock" state) until the bearing reaches the slanted top of the tooth and enters the "drive" phase for that side.

Escape wheel for testing.  Thin cord will be wrapped round the boss to drive it.  The cord will go over a pulley and over the side of the table to a weight.

Printed escape wheel complete with bearings to fit an M5 bolt as axle.  This is just a test model and rather rough print which needs some tidying up.  I shall use a fine printing mode for the working version.

A new deadbeat escapement diagram.

Here is a screenshot of the model of the escape wheel in Slic3r.  Now printing this on my Mini printer with 0.2mm layers and white PLA.  Currently using a 0.4mm nozzle but this may need printing wit a smaller one.  I have a 0.2mm nozzle I can use.

I need to work out how long the pendulum rod needs to be as it varies with gravity which depends on latitude and height above sea level.  Using an online calculator gives it as 9.81098 here.

1. T = 2π√(L/g) = 2π√L / √g
2. 2π√L = T√g
3. √L = T√g/(2π)
4. L = {T√g/(2π)}^2
5. Locally g = 9.81098 so √g/(2π) = 0.498512815
6. For T=2 (period = 2x beat), L = (2*0.498512815)^2 = 0.994060106m  IOW 994mm.

I think this escapement might just fit in.

New deadbeat escapement.

The escapement needs far too much pendulum swing.  To make this design of deadbeat escapement work would mean a much bigger escape wheel so that 3 or 4 degrees of pendulum swing make a significant movement at the pallets.

Applying geometry, the amount of movement on the end of the current anchor arm of 90mm for ±2° swing will be 90 x tan(2°) = 90 x 0.035 = 3.14mm.  For a deadbeat escapement this means splitting this already small distance into two, for the two phases of lock and drive.  This might just be possible with very accurate design and construction.

The pallets are 6mm OD ball bearings which means the minimum drive distance is 3mm with pointed teeth on the escape wheel.  Any radius on the point (to make it 3D printable) will add to the drive phase distance and hence angle.  With a 0.4mm nozzle the "point" has a radius of at least 0.2mm depending on extrusion factor giving a minimum of 3.2mm drive distance, corresponding to a little over 2° of pendulum swing.  The maximum swing the pendulum can have without bumping into the case sides is ±90mm which corresponds to ±5°.  This is more than recommended for good timekeeping but that can be corrected.  I does need more drive power though.  Allowing ±4°would mean a ±6.3mm swing at the pallets ie. 3.3mm of drive and 3mm of hold.  This supposes an accuracy of a tenth of a millimetre - difficult! - but if we allow 1mm accuracy the hold phase could be designed for 2mm.

I have split the drive phase over two parts including an arc for the shape of the escape wheel tooth "point" making it easier to print accurately.  The hold part is basically 1mm or up to 5mm providing tolerance for the driving force.  The resulting escape wheel looks like this :-

Still to much overlap of pallets and teeth so have shortened teeth by 2mm to try again.  Now printing.

Still not working!  There are so many interdependent variables with the deadbeat escapement that I've decided to leave it for now and just make a recoil type escape wheel so that I can get the clock working.  I hope to return to the deadbeat escapement later.

Come hell or high water I shall get this clock ticking!!!  Just maybe not today!

I think I may be wrong in going for a very large escape wheel.  Metal longcase clock mechanisms have quite small escape wheels of 50mm or less so my original 100mm diameter wheel was already much bigger than standard pendulum clocks with a one second pendulum.  I have found out why my earlier recoil anchor escapement gave problems - the pendulum bob was far too heavy.  With the new lighter bob the earlier escapement may work.

CAD screenshots of clock "going train" - the set of gears going from the chain wheel to the escapement plus the drive to the hands.  Viewed from front, left side and top.

Next I shall be adding the parts of the striking mechanism.  The parts I had already printed were the wrong scale.  First component is the strike cam.  The spokes of the cam fit the spokes of the hour wheel.  The cam goes behind the hour wheel and will be glued to it.

The striking mechanism - more parts to add yet.  Shown about to strike 12 o'clock.  The snail cam rotates anti-clockwise and winds the rack up one notch whilst striking once per revolution.  The pawl holds the rack as the tooth on the snail cam leaves the rack.  Once the appropriate number of strikes has occurred the tooth on the small snail cam contacts the lug on the end of the rack and stops.  The mechanism remains in this state until just before the next hour when a cam on the minute shaft causes levers to hold the small snail cam and lift the pawl, releasing the rack, which drops until stopped by the peg contacting the large snail cam (brown).  On the hour, the small snail cam is released and the pawl dropped onto the rack and the next hour strikes.

A couple of changes needed - the peg on the rack lever wants to be smaller and the small snail cam also wants to be smaller.

A couple of views of the latest clock assembly.

More added.

With frame.

Update...