A pendulum clock's rate depends on the length of the pendulum and gravity, and both are measurable. This one runs to within two seconds a week against a radio signal in one hallway, and loses four minutes a day in my workshop.
I restore and service longcase clocks. Thirty one years, and I have had perhaps nine hundred through the workshop.
The physics is simple enough to be worth stating, because everything here rests on it.
A pendulum's period depends on its effective length and on local gravity. Nothing else of any consequence. A seconds pendulum, which is what almost every English longcase has, is about 994 millimetres and beats once a second.
You adjust rate with a nut under the bob, which raises or lowers it. A turn is worth a few seconds a day and every clockmaker in the world knows roughly how many.
Temperature matters, because a steel rod expands and lengthens the pendulum and the clock loses. It is a small effect, a second or two a day across a normal room's swing, and a good clock has a compensated pendulum that cancels it.
That is the whole of it. Length, gravity, temperature. Everything else is friction, and friction affects whether it runs, not how fast.
The clock is an eight day, five pillar movement, about 1780, in an oak case, in a farmhouse hallway.
The family have had it since it was made. That is not a claim I would normally repeat, but they have the bill of sale.
It keeps time.
Not adequately. It runs to within two seconds a week, which I have measured against a radio controlled clock over eleven weeks, and which is better than a domestic clock of that period has any business being.
I took it in for a service in 2016.
That is a routine job. It comes out of the case, the movement goes on the bench, everything is cleaned, pivots are burnished, bushes replaced where they are worn, and it goes back.
I set it up on the wall bracket in the workshop, in beat, with the same pendulum, to run for a fortnight before it goes home. Every clock does this and it is how you find a problem before the customer does.
It lost four minutes a day.
Four minutes a day is not a rate error. Four minutes a day is a clock with something wrong with it.
I checked everything. The pendulum was the pendulum that came with it, and I measure and record every pendulum before it leaves the case. The suspension spring was sound. The crutch was true. The escapement was in beat and I could hear it was.
Four minutes a day is what you would get from a pendulum about two and a half centimetres too long, and it was not.
So I did the thing you do, which is adjust the nut, and I brought it to time on the bench over about a week, and it then kept excellent time in my workshop.
I took it back and hung it in the hallway.
It gained four minutes a day.
What happened next has cost me more sleep than anything else in thirty one years of this trade, so it had better be set down exactly.
I put the nut back where it had been. Exactly: I count turns and I write them down.
It kept time. Two seconds a week, in that hallway, with the pendulum at the length it has presumably had since 1780.
I have done this four times, over nine years, because the family have it serviced properly and I am not going to stop taking their money for work that needs doing.
Bench: loses four minutes a day. Hallway: keeps time. And the nut in the same place both times.
Everything a clockmaker would ask.
Beat. A clock out of beat stops, it does not lose four minutes.
The case. A pendulum swinging in a case that is not upright rubs and stops. The case is plumb and I have checked it with a level and shimmed it once, in 2018, by about two millimetres.
Temperature. My workshop is eighteen and their hallway is about eleven in winter, and the difference works the wrong way and is worth two seconds a day, not two hundred and forty.
Gravity. This is the one an engineer always reaches for and I have done the arithmetic twice. To lose four minutes a day you need gravity to differ by about half a per cent. Between two buildings eleven miles apart in the same county it differs by something like a hundred thousandth of that.
The pendulum. Measured, at both places, with a steel rule and then with a vernier, at the same temperature. Same length to a tenth of a millimetre.
What I have tried.
I hung it in my own hall at home for a month in 2019, which is a third building. It lost three minutes fifty a day.
I hung a different clock, a good one of mine, in their hallway for a fortnight. It kept its usual rate, which is about eight seconds a week fast, and it did not become extraordinary.
So it is not the hallway making clocks accurate. It is that clock, in that hallway.
What I do about it.
I service it every three years. I do the whole job on the bench, I regulate it on the bench to keep time on the bench, and then before it goes home I put the nut back exactly where it was when it came in.
I have never once told the family. They think they own a very good clock, which they do.
And on the last visit, in the March, I did something I have not done before and will not do again.
I stood in that hallway with the case open and my hand on the pendulum, stopped it dead, held it for a count of about ten, and let it go.
It picked up on the beat it had been on. Not a stumble, not a half swing to find itself. It came back on the second like a man picking up a sentence he had been interrupted in, and the strike at the hour was where it should have been.
I have stopped and restarted several thousand pendulums. They do not do that.