I understand that there is a main spring and a balance wheel that release force at regular intervals, but what allows it to be so exact? What physical phenomenon allows the balance wheel to always swing at the same rate?
The back-and-forth rotation of the balance wheel, attached to the hairspring, is similar to a pendulum in that it can have a very precisely tuned period - each cycle of the balance wheel releases a step on the escapement wheel - so the tuned oscillation period of the balance wheel is converted into finite movements of the rest of the mechanism.
The mass and size of the balance wheel and the physical properties (material, stiffness, dimensions etc…) of the hairspring.
The fact that we can make these very accurate does not surprise me. The fact that they could do a pretty good job of it a hundred (or hundreds) of years ago does surprise me. I have a Hamilton 992b railroad pocket watch (85 years old) that is as accurate as the best mechanical watch made today (about a second per day).
I have a fusee verge pocket watch from about 1760 that, while not very accurate by modern standards (about 15 minutes per day) is still pretty damn good for when it was made. It could probably be adjusted to be better but I don’t have confidence in my amateur skills to take it apart and service it.
Specifically, it was the search for some inherently periodic mechanism other than a pendulum that led to the invention of the first marine chronometers. The heart of which is the Balance spring - Wikipedia, which with compensation for temperature variations can be made inherently periodic.
P.S. according to James Burke’s “Connections”, it was the search for a way to produce steel homogenous enough to use in balance springs that led to the development of the reverberatory furnace, with all the effects that had on the industrial revolution.
The resonance is determined by the angular moment of inertia of the balance wheel and the spring constant of the balance spring. This means that, in principle, it is immune to changes in orientation. Unlike a pendulum.The balance spring is very long, with many turns, and the beat is adjusted to maintain a constant range of angle, so the effective spring constant remains very constant.
The design of the escapement is such that there is very little friction or other mechanical losses. The bearings and escapement pallets are usually ruby.
The torque applied to the going train (the series of gears from the mainspring down to the escapement) is maintained at reasonably constant levels. So the balance wheel doesn’t change in speed due to small changes in torque from the escapement.
The things that will affect accuracy are second order effects. Changes in orientation my change the friction of the system. Magnetic influences can cause drag on a metal moving spring, changes in temperature can change the diameter of the balance wheel and thus moment of inertia. The biggest is slow degradation of lubricants.
There are a whole lot of optimisations that address these problems. Silicon springs, careful choice of metals, advanced escapement designs, systems to maintain very constant torque on the train. That and very careful assembly, getting the mechanical system just right.
But as famously known, Harrison was able to build a successful chronometer once he perfected the basics with quite humble tools.
A quartz oscillator in a basic quartz watch isn’t a great deal different. Essentially it is a mass/spring resonator that just happens to combine the two parts into one lump of material. And due to its piezoelectric properties the escapement mechanism can be achieved with an applied electric field and some electronics. It vibrates (usually) at 32768Hz (so it divides down to 1Hz if you divide by 2 15 times). But it achieves resistance to outside influences in much the same manner, just with even better effect.
A mechanical watch may see beat periods of between 2.5 and 5 Hz. (Although there is a remarkable movement that uses a MMS silicon balance that oscillates at 40Hz.)
Thank you for those explanations, Francis.
I am very fortunate to have a Grand Seiko “snowflake”, a retirement gift to myself. It is rated at ± 15 seconds per month. Granted, it adds some modern technology to achieve that; I am sure there are other mechanicals that are more accurate, but not likely at this price point.
Well a Snowflake is a bit of an outlier. The Spring Drive is something else again.
Funnily enough I have promised myself an Omiwatari as a retirement present for myself. So we have similar taste.
Oh yeah, that Omiwatari is nice and clean!
Some folks bemoan the Snowflake’s power reserve on the face, and I see their point, but damn it sure is handy.
On-topic: after syncing with time.gov mine runs about 12 seconds fast per month, measured over a couple months.
Yeah, the spring drive is a quartz/mechanical hybrid. No battery needed like a normal quartz watch; all power is provided by the mainspring by an analog watch. The mainspring drives a relatively standard gear train, but also creates a tiny amount of electrical energy. That energy powers a quartz oscillator and a very small circuit that measures the beat rate of the mechanical component of the movement and compares it to the reference signal from the crystal.
The circuit controls an electromagnetic brake that it uses to grab and release another part called the glide wheel in the movement to act like the escapement in a traditional mechanical watch.
The gears are known as the Train of Wheels.
The gears take the accurate motion and translate it to hours, minutes, seconds.
The seconds was an early problem. It was a in a separate area (6 o’clock) on the dial. I think it was the 1920’s before watchmakes found a way to include all three hands on one stem.
I found a link to a website that I had forgotten about. If anyone reading this thread, even has a fractional level of interest and understanding how a mechanical watch works, this is the gold standard. It uses animated 3-D models to build a watch from first principles. Truly outstanding.
Kind of amazing that mechanical devices of this complexity could be made with stone knives and bearskins.
In short, clever mechanical design allows them to keep time. @Mangetout, @TriPolar, and @mozchron pretty much detail how the watches work.
Thing is though, they’re not that accurate by modern standards. Most of your mechanical watches are accurate to about +/- 5 seconds a day, while your el-cheapo, bog-standard quartz watch is accurate to something like +/- 0.5 seconds per day.
Generally speaking, mechanical watches are more works of art and/or paeans to clever engineering, artistic vision, and exquisite craftsmanship, not really practical timepieces. I mean they keep time well enough, but nobody’s buying them because they really need an accurate watch. You buy a cheaper quartz watch for that.
Remember all the movies set in WW2 where the team of commandos synchronize their watches before splitting up? That was so all the various explosions would go off within a second or two of the expected time.
A question about tuning fork escapements: AIUI they were mainly used with electromechanical watches; but could a purely mechanical one have worked?
I was about to say “of course not, it needs an electromagnet to keep the tuning fork vibrating” but then I thought about it and I can’t see any conceptual reason why the energy from a mainspring couldn’t be used to keep the fork vibrating.
Main reason is probably because by the time they were developed the transistor had just been invented making it a cost-effective way to regulate. But it’s an interesting thought!
In case anyone doesn’t know what these are, n a tuning fork watch it’s the physical movement of the fork that drives the movement. As it vibrates, it pushes a minuscule pawl and ratchet to turn a wheel which drives the train. If the fork is vibrating at the correct speed, the train turns at the right speed and thus the hands = good timekeeping.
Bulova developed the system in the late 1950’s/early 1960’s. A few other companies (Omega) also made them licensing the patent from Bulova. Until the invention of the quartz oscillator, these were the most accurate timekeeping devices in existence. Used in spacecraft etc…
I have a 1963 Accutron that I love. Had it professionally restored by a specialist in Thailand (one of the few guys left who still works on these). It runs about 15-20 seconds fast per month, which isn’t phenomenal by today’s standards but was about the very best possible in the early 60’s.
Hey, before the invention of digital computers, or even the application of electricity to calculation, all-mechanical devices that could do harmonic analysis were devised for predicting tides:
For a portable watch anyway; obviously naval chronometers did better.
Yeah i’m talking wristwatches 
This was a really great explanation. I see that the balance wheel is set up in such a way that the force is applied at the same point in its cycle every time. Is this what allows it to keep the same frequency even though different amounts of force are applied?
Loosely speaking, it’s the length of the spring and its stiffness, coupled with the mass of the balance spring that defines the oscillation rate of the system. That brief impulse provided by the escape wheel to the pallet fork shouldn’t change the rate, only the angle that the balance rotates through. And pendulums have the same rate, no matter whether other swinging through a wide arc or a narrow one.
That said there’s a lot of interest in keeping the energy transmitted more consistent through the wind, which is the main reason the main spring has kind of an oddball shape and extra pieces on it. It’s all about trying to maintain a constant torque through the entire wind period. .