I had hoped for great accuracy from this clock. Torsion spring balances should have a very high degree of isochronism, or the ability of the balance to take the same amount of time per rotation of the balance, irregardless of the size of the arc it travels. This one also exhibits a high "Q", or low frictional losses, since the balance unpowered will turn almost as many times with the spindle rotating inside the spindle jewel as it does without the jewel installed. But accuracy was not forthcoming, after months of tests. It seemed that just when the clock seemed to be regulated, it would make a "jump" and gain several minutes over the period of just a few days.
I exhausted all possible sources of error, then decided that perhaps the balance itself was at fault. Could it be shifting somehow, due to temperature changes? So I decided to make another balance.
This one would be made of glass, which has a very low coefficient of expansion under temperature changes. It would also be mechanically ridged.
I used an ordinary clock bezel glass to make the new balance, sized at 3", or the same OD as the old balance. Since the glass is lighter than the old balance, it would turn faster, so to get the new balance to turn at the same rate as the old, I mounted five brass weights on the edge of the glass.
I used diamond drills available from Harbor Freight to drill the holes in the glass. Six holes are required: one for the center, and five for the weights on the perimeter of the glass.
I drilled all the holes by mounting the glass in a wooden "box" as demonstrated many times on this website. The wood for the box is mounted on a Taig faceplate, then turned with a shallow inset made a tiny bit undersize to hold the glass blank firmly.
I used the Taig lathe to drill the center hole, as shown in the photo. The bit is held in the tailstock. It's important to use a slight but constant pressure on the tailstock lever, or the glass will crack. It's equally important to use drops of water applied at the cutting location. I used an old clock oiler filled with water to do the job.
The best way to drill glass is to drill half-way through one side, then turn it over and drill all the way through. This minimized shell-shaped chip-outs on the exit side.
The five peripheral holes were drilled with the "box" on it's Taig faceplate screwed to the Sherline rotary table on the bed of the Taig mill. The principal is the same -- drill all five holes halfway through, then flip the glass over and drill the same from the other side. It is possible to align the glass with the bit for the second go-around on the glass just by eye, using a reference mark made on one side of the glass and on over to the "box" as shown in the photo.
It is prudent to have a couple of "backup" glasses available, since the process is error-prone. Again, go slowly and use water. (You can see the glassey water in the photo.)
Once the glass was drilled, I mounted it on an arbor through the center hole in the Taig lathe and used a fine diamond hone to polish the outside rim of the glass. I followed this with silicon carbide sandpaper, which gives a fine finish to the edge of the glass.
It's obviously important that the new balance have the same period of rotation as the old. I mathematically determined the volume of the brass in the rim of the old balance, and divided that by five to get the approximate volume of one of the peripheral weights. Then I calculated how long should be a piece of half-inch brass rod to get that volume.
So that each piece of half-inch brass was the same, I used a little inexpensive set of scales to weigh them. I found the smallest of the five pieces of brass, then cut down the other four to be the same weight as the smallest. Thus all five weights were the same.
Next, I made a new spindle for the new glass balance, using epoxy to cement it into the center hole. The spindle was made in the same way I made the spindle for the old balance.
Once the balance was finished, I suspended it from a test stand, using the actual regulator from the clock. Then I used a stop watch to get the approximate period of the balance set to 15 seconds per revolution, or four RPM's. It is possible to thin a Horolovar suspension spring with 400-grit silcon carbide sandpaper to get the balance period within parameters. Only a few strokes with the paper will slow the balance slightly.
Another interesting effect of the new balance is that the balance is much more efficient. Once the new balance was installed, when the clock was running the balance would over-turn to turn two complete revolutions, which would stop the clock. So I had to reduce the diameter of the barrel the weight cable is wound on to reduce the torque driving the clock. The clock used to run 7-plus days per wind. Now it runs 10-plus days per wind. The increased efficiency also contributes to accuacy, since it will reduce escapement error and increase "Q".
The clock is also more handsome with its new glass balance, I think.
Todays' links:
Glass with weights
Diamond Drills
Cutting Box
Drilling Center Hole
Drilling Outside Holes
Glass Blank showing reference mark
Smoothing Glass Edge
Glass Done
Scales
Spindle
Suspended
Finished 1
Finished 2