The question to ask is why a French clock can run on a mainspring 0.009" thick, whereas many German clocks have mainsprings that are 0.014 to 0.016" thick, and British and American clocks have to be the most massively over-powered? Why could an American clock not be made to run on the same mainspring as a French clock?

The thinnest 3/4" mainspring available from timesavers.com that could be used on an American clock was 0.012" thick and only 72" long (part #16879), and would have 30% of the strength of the 0.018" thick spring normally recommended. The only way to find out if it worked was to try it. An Ansonia with the Brocot escapement became the first patient for this experiment, since it had the wrong time mainspring anyway, which needed to be 5/8" wide and not 3/4", a mistake commonly made by repairmen. The thinnest 5/8" mainspring available from timesavers.com that could be used on an American clock was 0.013" thick and only 69" long (part #16876), calculated to have a strength of 31% of the 3/4" mainspring it replaced. After overhauling the mechanism, I installed the new mainsprings (time: #16876 and strike: #16879). Despite the shorter mainsprings, this Ansonia clock ran for 11 days, striking correctly and keeping accurate time (after adjusting the pendulum). The error was about a minute in total during the first five days, after which it began to lose time (typical for spring-driven clocks).

The next patient was a Sessions clock, with a mechanism like one you would find in a parlor clock or a kitchen clock. With the new mainsprings, #16879 for both time and strike, this clock lost 5 minutes in the first 24 hours, so the pendulum needed to be adjusted up a little. This brings the issue of the two types of error. Constant error is easily corrected by adjusting the length of the pendulum: screw it up to speed it up and screw it down to slow it down. Variable error is not as easily adjusted. Timekeeping will vary with changes in temperature. Colder temperatures cause clocks to runs slightly faster, and vice versa. A pendulum with temperature compensation, such as the mercury pendulum, can reduce this error. Timekeeping will vary with changes in power. Spring-driven clocks will run faster at the beginning of the week and slower towards the end of the week, especially after the fifth day. Weight-driven clocks will run faster when the weight is increased, and vice-versa. Where the experiment becomes interesting is in observing how the variable error is decreased when the weight is reduced to a lower amount, as seen in the grandfather clocks above, and similarly when using a thinner mainspring. Once the pendulum was adjusted, the clock kept more consistent time, running and striking correctly for 9 days.

While the 0.012" thick mainspring I used in this experiment had 73% more power than a comparable 0.009" thick mainspring for a French clock, the 0.012" mainspring offered a considerable improvement as an alternative mainspring to consider to replace the massively powerful 0.018" thick mainsprings in American clocks.

The strike mechanism in this Sessions clock only strikes the hours, not the half-hour. The half-hour is simply the lift-and-drop method by the centershaft. This requires more power in the time train than lifting the strike lever. I believe this is why the Sessions ran for 9 days, whereas the Ansonia, which uses the strike mechanism to strike the hours and the half-hour, ran for 11 days. Both clocks kept more accurate time than before and lost less time after the fifth day. After adjusting the spring for the strike hammer, so that the hammer would strike the gong less than half as hard as before, the Sessions ran for 13 days.

There are three advantages to using a less powerful mainspring: more accurate timekeeping, longer durability, and less damage if the mainspring breaks. Furthermore, the clock will probably not run when the oil becomes gummy, which is good because the clock will wear faster if used after the oil no longer lubricates. A disadvantage is that a clock running on minimum power must be set up correctly and adjusted precisely for it to run.

The third patient was a Seth Thomas, similar to the #89, with a gong strike for the hour and a brass bell for the half hour. The escapement was a half-deadbeat. With the new 0.012" mainsprings, the clock ran very well, though the pendulum needed to be adjusted. In this clock, it was necessary to consider the other springs. It had a thick suspension spring, which I decided to replace with a thinner one. The spring for the strike hammer was very strong, so I decided to use another spring wire, in which I could adjust the tension easily by bending the wire. The springs for the strike count lever and the left lever were very strong, as I have found in many similar American clocks, so I decided to use another spring wire, which I could adjust for the minimum tension needed to obtain proper performance. This Seth Thomas clock was running and striking correctly for 13 days, losing time after 5 days. The point here is that these clocks are running for more than 8 days with the thinner mainsprings, as they should, not whether they run for 11 or even 14 days.

The fourth patient was a Gilbert with two mainsprings. All the clocks in this experiment were made around 1900 to 1910. Even though it had a strip-pallet recoil escapement, this clock was different in that the action of the pallets was visibly more symmetrical and correct than the others. Despite that, I was surprised to see this clock run and strike correctly for 15½ days with two 0.012" mainsprings.

The last patient was a Gilbert calendar clock with one mainspring. While I wanted to concentrate on American clocks with "H" plates, which look like the Seth Thomas 89, I thought this Gilbert was essentially of the same design as the other clocks on the time side. Before proceeding with the experiment, it was necessary to repair the escapement in this clock, even though it looked like the escapement in the other Gilbert. The problem was that the escape wheel did not rotate as the tooth slid across the exit pallet. Work done is equal to force times displacement. If there is no angular displacement, then no work has been done, which means that the pendulum received no kinetic energy via the exit pallet. Half the energy is lost in the escapement because of this problem. This is a surprisingly common problem in American clocks, suggesting that many repairmen do not know how to repair escapements, and make up for the loss by using brutally powerful mainsprings. After repairing the escapement so that the rotation of the escape wheel would be equal on both sides, so that the lock and drop were also equal on both sides, and so that the amplitude of oscillation of the pendulum were reduced to a lower level, this Gilbert clock ran nicely and kept good time for 15½ days with a 0.012" mainspring. It kept better time than the other clocks in this experiment, particularly after day 5, because it had a longer pendulum of 17 inches long instead of 9.

Since the other clocks in this experiment had ticking sounds which were considerably louder than in my French clocks, I decided to try a 0.009" thick mainspring, available only with a hole end, so I added a loop from another mainspring and installed this 0.009" mainspring in the Gilbert calendar clock. It ran nicely and kept good time for 13½ days. Note that the 0.009" thick mainspring has only 14% of the strength of the brutally powerful 0.018" thick mainspring, all other factors being equal.

The last step in this experiment was to go back to the Sessions and install two 0.009" thick mainsprings in it. The Sessions ran for 10½ days, striking correctly. Then I installed a piece to wire to keep the warning lever out of the way, and a piece of cardboard to keep the strike hammer out of the way, to observe how this Sessions would run as a time-only clock. It ran for another 1½ days, for a total of 12 days. I thought that the Sessions was the most important part of this experiment because it exemplified the typical American clock more closely, when considering all the American clocks I have repaired over the decades. It had the design imperfections I expected to find in this type of clock, the asymmetrical pallets, and the escape wheel which was ever so slightly out of true.

Six months later, however, the Sessions was stopping after 7 days, so some extra power is needed as power losses increase over time. The 0.012" mainspring is a better alternative.

To explain the reason why these clocks lose time after day five, consider how the strength of the mainspring decreases as the clock runs. According to Robert Hooke's Law (1678), which states that "the force F needed to extend or compress a spring by some distance X is proportional to that distance," we get a graph with straight lines. If we express the mainspring strength as a percentage and we assume that all the mainsprings have no strength after 17 days, we get a graph like the one below. You can see that the stronger spring has a much greater difference in strength after 7 days, compared to the weaker spring. By comparison, the line for the weaker spring appears to be almost horizontal by comparison, so the clock with the weaker spring will keep more consistent time, especially if it has a recoil escapement.

It is worth mentioning that the Timesavers catalog has a 3/4 x 0.015 x 170 inch mainspring for Japanese and Korean 31 day clocks from the 1970s and 80s, because some of their mechanisms look like copies of American clocks from the early 1900s. A 0.015" mainspring, when installed in a Seth Thomas 89, should have a calculated length of 169", and it would have 58% of the strength of a 0.018" mainspring. They were obviously onto something there.

This simple experiment is replicable and could therefore be subjected to "peer review." My sample of only five clocks is not enough to generate statistically significant data, though it does confirm that most American clocks of the type in this sample are massively overpowered and could run well with much less power. While not all American clocks will work well, such as some of those made before 1890, suffering from lower manufacturing standards, and those with escapement problems, which need to be corrected first, I believe that almost all American clocks of this type could be repaired to enable them to run with thinner mainsprings. The purpose of this experiment is to find how to do the best possible repair for the most durability and the best timekeeping. Could my Sessions clock ever keep time as accurately as a Vienna regulator? No, because it is not weight driven, nor does it have a Graham escapement. However, installing the thinnest mainspring possible would reduce the recoil action to a minimum and offer a terrific improvement in timekeeping.

Now for the disclaimer to keep me out of trouble: experiment at your own risk! Avoid irreversible modifications. Clocks with escapements which are badly worn or defective, with escape wheels which are out of round or have irregular teeth, would need more power to overcome such problems. A high-beat clock with a very short pendulum, like many Ingraham tambour clocks from the 1920s, will probably require more power. A clock with an unusually large and heavy pendulum may require more power, etc.

Part 2: Mainspring Strength and Timekeeping

The Origin and Evolution of the Anchor Clock Escapement

Clock Repair Main Page
Escapements in Motion
Links Page