A Bluffers Guide to the Watch Escapement Analyser and Timer

A small (size 4) movement on the
Vibrograf mount / microphone. The bed
can be moved in 3 axis to check timing
in any position. It will easily hold an
oversized S18 case.

Timing machines were originally hybrid electro/ mechanical devices, the "tick" of the escapement was amplified and used to drive an inked pen that drew a trace on paper moving over a drum turning at a fixed rate appropriate to the expected speed of the balance.

Modern computerised versions do much the same thing but give the results on a computer screen, sometimes in a similar format to the paper tape trace but with additional information. Most are dedicated units costing anything from about £650 to £7,000 or more and are orientated more to wrist watches than pocket watches and some do not, for instance, work with a slow train (16,200 VpH)  pocket watch movement and may not accommodate large cases.

I use a PC based product developed by Graham Baxter that uses the PC's audio sampling clock (usually at 96KHz) which, when suitably calibrated, gives an accuracy of better than 1 / sec per day with a high degree of stability and confidence. This software currently costs £299 with a clip on microphone which is more than adequate for most users. I have upgraded to a refurbished Vibrograf dedicated mount with a new microphone which is rather more robust for constant use, more resistant to background noise and is easier to use in some situations, the software also runs on a dedicated (cheap) tablet / laptop, mainly for convenience.



Click for an enlarged view.

The sample trace here is from an exceptionally good restored 7J Waltham Traveler undergoing a test in 5 positions. The vertical lines in the centre moving up the simulated paper (the coloured band)show the odd and even ticks, here they are very close together showing that the watch is "In Beat", as the detail on the bottom left shows almost perfectly so with an average error in 5 positions (Face Down, Face Up, Pendant Up etc) of only 0.13 mSecs, about 0.07% on this quick train (18,000 VpH) movement.

A trace moving off to the left (up the screen) indicates the movement is losing time and moving off to the right it is gaining. Again in this computerised system that is also shown numerically, the old systems did not.

The trace that looks rather like an e.c.g. printout is the sound of the escapement unlocking, the impulse jewel going through the lever and the escapement locking again with two ticks showing - one to the left and one to the right so that a fault occurring when the balance is moving clockwise or anticlockwise can be seen, unfortunately it can't tell you which is which.

The distance between the three blips gives an indication of the amplitude or power of the movement, if they are close together (in time) the balance is moving quickly (strongly and with more turn) if far apart then slowly. Some clever mathematics, knowing the geometry of the escape wheel and other factors, allows the amplitude to be calculated and displayed in terms of the degree of rotation of the balance wheel.

With experience a lot of faults can be pinpointed by irregularities in the trace such as knocking, the hairspring rubbing the balance cock or balance wheel, dirty or miss placed pallet jewels, banking pins set incorrectly, etc., (If I remember I'll add an example the next time I find a good one) More detailed views and explanations of the display can be found on Graham's web site

The initial faulty trace from an 11J size 4 half hunter.

An advantage of a computerised system is that the readings can be logged and displayed in different formats and scales. This trace taken over an hour or so (not all shown) shows a movement with a problem, the top trace is showing time keeping errors in seconds per day measured every 2 seconds. As you can see it is varying quite considerably averaging around a 20 seconds / day gain but varying from 0 to 60 s/Day.

The lower graphs show the amplitude and the beat error (the later looks bad but the scale, that I left on automatic, is such that the deviation is tiny and quite immaterial).

By looking at the periodic nature of the trace we can see that there is 3 minutes between the peeks of slowness and also between the peeks of fastness of the movement. This has to indicate something wrong with the train. The centre wheel

And after the first attempt at resolution

turns once per hour so it is unlikely to be that, similarly the seconds wheel turns once per minute so is not likely to be the problem. That leaves the wheel in between.

Taking the movement apart the wheel was checked for cleanliness, deformed teeth and flatness and some adjustments made. After reassembly the test was rerun. Things have improved with the deviation less and the peaks now 6 minutes apart. Again the movement was taken too pieces and some more fettling done, mainly getting the wheel exactly at right angles to the arbor (axel). And the variation in timekeeping was reduced to a few seconds per cycle with average time keeping (at full wind) within 5 secs per day, which is pretty good for a watch made in 1891.

Unfortunately that was not the end of the story, checking the watch in different positions it was fine face up and face down but gained 2.5 minutes per day pendant up and lost a minute pendant down. Another hours work was required to fix that and before the advent of timing machines that would have been several hours over many days. But adjusting for positional errors is well out of the scope of this piece and takes up many pages in most good books on watch repair.

The final result

Left to run over night to settle down it was checked at about half wind over a fifteen minute period and gave very consistent results (the blip at the beginning was the result of moving the mount for the photo at the top of the page). Job done!

Or not done, having put the motion work, dial & hands on and put it into its case I found it was loosing 20 minutes per hour due to a slipping cannon pinion so all that had to be undone, the cannon pinion removed, adjusted and refitted (these 3 operations all requiring the use of the staking set on this design of watch) and everything put back together.

Thankfully on this occasion the first adjustment did the trick.