I have been working on a very old clock – an English tall case clock from somewhere around 1725. This clock was last serviced about ten years ago by a respected repairer. I went through the movement doing the usual pivot polishing/bushing work. After reassembly, the strike train had a couple issues, one of which is wear on the gathering pallet.

The gathering pallet is sort of like a one tooth gear that advances the strike rack so the clock stops striking after the right number of strikes. The gathering pallet has two components – the tooth that advances the rack, and the flag on the other end that contacts the stop pin to stop striking.

When the clock was made in 1725, the industrial revolution was in progress, but early on. The gears were probably machine-made, but the clock plates and other components were largely handmade, including our gathering pallet. It’s a pretty complicated part including a square hole, and a moderately tight tolerance on the other dimensions. I suspect it probably took the clock maker several hours. If I had to make one from scratch, it might take me an entire day, and that’s assuming I didn’t screw something up and had to start over. For the sake of maintaining originality and also to avoid exposing my marginal craftsmanship, I wanted to see if I could repair the wear rather than trying to replace the part.

The wear on the gathering pallet is on the tooth that advances the rack. It works marginally, but some of the time the tooth doesn’t quite engage enough with the rack to advance the rack, meaning the clock may strike too many times. The solution to this is to fill in the missing metal somehow.

There are several ways to potentially do this – silver soldering, or perhaps some kind of epoxy or modern material. The problem with these is wear resistance. Any material that could be applied cold would likely not last long rubbing on the steel rack. The best solution would be welding the part to build it up with the original material – more steel. 

The challenge to welding is the microscopic size of the part. Welding melts both the original part and the filler material. The benefit of this is a welded repair is likely as strong as the original part was. The challenge is that welding – particularly on a part this small, may warp or crack the part. Thankfully, I know a great welder – my father spent his career in metal fabrication and had a few tricks up his sleeve.

There are a lot of kinds of welding. One of the most refined is TIG welding – tungsten inert gas. TIG welding uses electricity passing from a tungsten electrode to the part to create an arc, and then filler material is melted into the arc to build up the part. With care and experience, it is possible to scale TIG welding down far enough to work on parts this small. The electrical power of the welder can be adjusted to vary the amount of heat. Normally TIG welding is done at around 150 amps or more; for this job my dad used a setting of 15 amps.

Here is the pallet after welding – a blob of steel has filled in the missing metal.

Grinding the weld back to the right form required some care. I carefully used a belt sander to remove most of the extra material, then I moved to a very fine file to match the original profile. I left a little extra material to compensate for any wear on the rack.

 

After a bit of polish to remove the discoloration from welding it went on the clock and is working great. Thanks to my dad for a great job that should last another couple centuries.

Here is the new pallet in action.

 

I have been working on an early American tall case clock. The click spring on the strike drum is pretty shot – the finger that makes contact with the click  is worn down and the spring is bent. Winding the clock causes the click to jump behind the spring, breaking the ratchet action.

Clocks this old were largely hand-made. Gears were cut with a machine, but much of the other work was done by hand. This click spring was filed out of a sheet of brass and riveted onto the gear. I think the spring was poorly designed – it is fairly tall which gives it its pushing force, but it is also very thin, meaning it bends easily in a direction I don’t want it to bend. For the replacement spring I wanted to make it out of a thicker sheet of brass.

My interests run the gamut of technology – from clocks built before the industrial revolution to modern manufacturing techniques. I wanted to see if I could make the replacement part using my CNC router.

I measured the gear and made a crude model in my CAD software. The outer circle is the outside of the teeth, the next two rings are the solid ring of the gear, and the inner circle is an approximation of where the contact point between the click and the click spring needs to be. I drew the spring profile with a couple Bezier curves.

My CNC router is primarily a wood-cutting machine, but it can also cut soft metals like aluminum and brass. This was my first attempt at brass.

There is quite a bit to the math of feeds and speeds and cutting geometry. This matters to varying degrees depending on what you’re trying to do. A feed rate that is too fast can break bits. Too slow and the heat of the cut isn’t evacuated in the chips and the bit overheats and dulls. If I was making a thousand of these, this stuff would matter (and I probably wouldn’t be making them on a woodworking CNC router), but for a one-off experiment, taking shallow passes and going slow is the path to success.

The prescribed bit for cutting brass and aluminum is called an “o-flute” bit. This is a one flute bit with a curved gullet that is designed to remove material without the chips sticking to the cutter.

I started with a 1/4″ bit, thinking that would likely be more forgiving than starting with something smaller and more fragile. In larger work, I think I can probably take passes up to 0.200″ thick, but since these bits cost about 30 bucks, I decided I had more time than money and  took 0.006″ passes. When I ran the job it was almost comically slow – it took about 6 minutes to shave off 30 thousandths on a 1″ x 3″ area, but I didn’t break my bit! Part of the slowness was the stepover parameter – how far the bit moves over for each pass. The tool default was 10%, meaning that on a 0.250″ diameter tool, the tool only moved over 25 thousandths per pass.

The profile passes were similarly shallow, but unlike the clearance path, the bit has to cut full width to go around the part. This went fine for the most part, but it did leave a little nub of brass on the click spring because the bit broke the part away from the stock.

The finished click spring is a little rough looking, but it’s the shape and size it is supposed to be. I clamped it onto a piece of metal to get a feel for how strong the spring was. Since I wanted to change the design of the click spring, I needed to determine experimentally the right width of the spring. The first attempt was pretty stiff.

Back in the CAD software, a minute change to one of the Bezier curves made the spring slightly thinner in width. Small changes can make big differences in spring strength, so I wanted to go easy for the second attempt. Buoyed by my relative success the first time around, I decided to shift down to a 1/8″ bit so I wouldn’t waste as much material. With new toolpaths in hand, I headed back to the router.

The 1/8″ diameter bit is a better choice for the profile cut – a smaller cutter means less cutting force on the part, so in theory the chances of me firing it across the shop would be lower. However, a 1/8″ bit removes a lot less material than a 1/4″ bit, so the pocketing operation would take longer this second try.

The difference between the first and second parts is hard to see without them being side by side. I clamped up the new spring to test the spring force and this time it felt about right.

The click spring is pinned to the gear. The old spring was pretty easy to remove – the pins had worked themselves loose in the 170 years this clock has operated. After removing the original click spring, I mocked up new spring to see if this was going to work. Normally Murphy’s Law would assume that it would take three guesses to answer a yes or no question, but Murphy must have been busy – spring #2 was almost perfect.

I probably spend too much time agonizing about non-performance-affecting elements of clock work. I have no problem putting the work in to make a clock mechanically sound, however the tedious stuff like polishing wheels is not a source of joy for me. Out of the router, the surface finish of the new click spring was pretty rough. I sanded it with 120-grit sandpaper and then hit it with some Simichrome polish with a wheel in my rotary tool, and I think it looks good enough to be worthy of the clock.

With the spring and wheel still held in the vise, I scratched the outline of the click spring onto the wheel.  I re-clamped the spring and wheel in the vise to allow access to the holes. I wanted to reuse the original holes in the wheel, so I used them as a reference and drilled through the new click. I reamed the holes to size using a cutting broach. I made the holes in the click slightly larger than the holes in the wheel because I wanted to mount these with taper pins, with the wide side of the pin coming into the click spring. After a few minutes of fussing around, I had a good product.

I  cut the pins to size using a cutoff wheel in my rotary tool. The picture below shows the original click spring (top), my first router-cut part (middle), and the second spring riveted onto the wheel.

I have no doubt this was the long way to solve this problem. I could probably have traced the old click spring, cut it out with my scroll saw and been done in 1/2 the time, but I learned a lot doing it this way, and any project where you get to use a power tool is a good project. That and if I ever work on another identical tall case movement needing exactly this geometry of a click spring, I’ll be ready to go.

 

 

I last worked on Lockwood & Almquist back in November. That’s also when I last photographed the clock as well – please forgive the reprint complete with Christmas lights. I promise they’re not still up. It’s an interesting clock with an industrial strength movement. The last month or so I’ve noticed the clock running a bit slower than before and also needing to be wound tighter than before to stay running. It was time to take another look.

The Movement

I pulled the movement out to see how 5 months of runtime played out. I should say that it is unusual that a clock will develop issues after only months of running. A well-serviced clock should run at least 5-10 years. Something clearly wasn’t right when I first serviced the movement.

This movement is unusual in a couple ways. First, it is designed to run for 90 days on a wind. To achieve such a long run time, the clock is powered by two massive springs. The winding arbor drives both springs and is geared down to only require a reasonable amount of force. The consequence of this is that fully winding the clock requires over 250 turns of the winding crank.

Another way the clock is unusual is the back plate – it is thick cast iron rather than brass. Most clocks use brass for the movement plates and steel for the pivots. In addition to looking pretty, this is done for a reason – brass and steel are chemically compatible for wear. Steel on steel can be problematic. This is mitigated to a large degree with oil, and this clock obviously worked fine for the first hundred years of operation, so clearly it can work just fine, but it’s an unusual setup.

When I originally serviced this clock I approached it the way I approach any clock. The basic steps are to ultrasonically clean, polish the pivots, repair any wear with bushings, reassemble, and oil. Bushings get polished as well with a smoothing broach. If a pivot hole isn’t sufficiently worn to require a bushing, I normally just scrape out any leftover gunk the ultrasonic cycle missed with some peg wood.

At first glance the oil on the winding gear pivots was a little dirty. This surprised me. Also, the pivots in the back side of the movement looked dry. There was a bit of oil there, but not as much as I expected.

I disassembled the movement to check the condition of the pivots. They looked fine. I checked the verge pallets, and there were a couple minor marks. I don’t believe I polished the verge the first time around, so I did it this time.

Before reassembling the movement, I tried to address the pivot holes in the cast iron back plate.   I’m not entirely sure what the best way to approach this would be, but I took some time with some peg wood and got some more junk out. I did this the first time too, but apparently not adequately. I reassembled the movement and added a bit more oil to the rear plate pivots.

The Pendulum

The other thing I wanted to readdress this time around was to add even more weight (!) to the top of the pendulum rod. A quick refresher (more detail in my first post) – with the bob in the center of the adjustment range, the clock ran 5 minutes per hour too slow. Somewhere along the way the movement got altered or replaced, and the pendulum as I found the clock was too long for accurate timekeeping. For aesthetic reasons I didn’t want to just cut the pendulum rod shorter – it already seemed too short for the case, so I opted to add a bunch of weight to the top of the pendulum rod, which increases the speed of the pendulum, allowing me to move the bob down to a reasonable location on the rod.

The clock kept time with the added weight I settled on last time, but just barely – the bob was all the way at the top of the adjustment range. I wanted to drop the bob to the middle of the adjustment range if possible.

Space behind the movement is limited, so the weight needed to go mostly to the side of the rod. Last time I used multiple layers of 1/8″ thick 1 1/2″ wide steel bar. This time I used a piece of 3/8″ thick by 2″ wide bar.

The pendulum rod is just under 3/4″ wide. A 3/4″ end mill was almost a perfect fit. I cut the slot in two depth passes, leaving the slot just shallower than the thickness of the pendulum rod.

Here you can see the original weight arrangement the clock ran with for the last 5 months and the larger machined block.

I didn’t want to do anything to irreversibly alter the pendulum, so I decided to clamp the weight to the pendulum rod with a piece of the same 1/8″ thick 1 1/2″ wide bar I used previously.

I drilled and tapped holes (including one broken tap) for screws to clamp the pieces together. Other than the broken tap, it turned out great. Even the casual eye may notice something strange about my drilling setup in the photo above. A real machinist would have measured offsets from the edges of the parts and calculated the hole positions for drilling. As I am not a real machinist I am free from those constraints, so I took the “close enough” expedient alignment method of eyeballing it and using a C-clamp to hold the parts in position while I drilled the pilot holes. The parts were just long enough for me to be able to flip the part around to drill the second side without having to reclamp the C-clamp position.

 
 

This is a really strange situation – the top end of the pendulum weighs more than the bob.

Pendulums need to be plumb to run correctly. This means the suspension point needs to be directly over the center of mass, otherwise the pendulum will do funny things. When I reassembled the clock, the pendulum was arcing in the horizontal direction – the leading edge arced toward the back of the case. I bent the hanger slightly backward to better align the center of mass under the suspension spring and it started running normally.

I timed the clock and put the dial and hands back on. Amplitude seems to be better than it was, so hopefully I’ve addressed whatever issue I didn’t handle adequately on my first go at the movement. It’s a little unsatisfying to not find a smoking gun, but much of success or failure in clock repair comes down to little problems adding up. Hopefully my second look will give this guy another decade of good service before I need to look at it again.