# How does this work?

Discussion in 'Electronic Design' started by Anthony Fremont, Jun 30, 2006.

1. ### Anthony FremontGuest

Being an amateur horologist/PIC-hacker that lives on a tight budget,
I've glommed together my own version of the professional clock/watch
timing equipment. I have optical, acoustical and inductive sensor
inputs. Using a PC and some Visual Basic coding, I can duplicate
virtually all the functionality of these highly expensive machines.
Everything is fine except for one minor issue. The professional watch
timing machines purport being able to determine the amplitude of a
balance wheel's swing. They manage to do this by purely acoustical
sensing methods. :-? :-? :-? :-?

For the life of me, I can't figure a way to listen to watch ticking
sounds and determine the arc of the balance wheel. These watch timing
machines have been doing this for decades, how? The interval between
the roller jewel smacking the lever aint it, that just determines the
rate (beats per hour) and the amount of beat error. When the balance
wheel stops and reverses, there is no sound created by that event. Any
ideas?

2. ### Ignoramus2330Guest

terms. Suppose that the swing is very heavy. Make watch float in
liquid (say on a little piece of styrofoam).

Then rotations of the swing would cause watch to "recoil" in the
opposite direction. Since you know the weight and momentum of inertia
of the watch, you could figure out a lot about the swing if you can
measure the watch's recoil.

i

3. ### LuhanGuest

Could you post a link to the company making these claims?

Luhan

4. ### John O'FlahertyGuest

Might acoustical sensing include doppler sensing of the maximum
velocity of the wheel? If it's simple harmonic motion, you would have
amplitude = vmax / angular frequency.

5. ### Spehro PefhanyGuest

It's done entirely acoustically, but the calculation involves the
"lift angle", which must be entered manually for a given movement.

Best regards,
Spehro Pefhany

7. ### NonnymusGuest

X-No-archive: yes

Elegant and practical way to do it. Actually, there would
be no need for calculations; all you would have to do would
be to watch and make adjustments until no rotation was
noted, I suspect.

Nonnymus
to reduce spamming and kooks

8. ### dAzGuest

done by calculating the time difference between the drop and unlocking
of the escapement, so its the sound of the roller hitting the pallet
fork and unlocking the escapement, the impulse as the escape wheel
slides across the pallet and the drop as the now freed escape wheel
tooth hits the locking face of the other pallet, those are the two
loudest sounds the unlocking and the drop.

have a look here for ideas
http://www.bmumford.com/microset.html
http://www.bmumford.com/mset/tech/watches/tickdsp.html

I do use this machine and find very good for clock work, I have yet to
buy the software but will do so soon, I have never bothered with a
amplitude meter for watches because I can tell just by looking at the
balance to whether it has good or bad amplitude, it would be nice to
have one but could not justify the cost of a gradoscope to go with the
vibrograph b200, the only real advantage I could see it could measure
the amplitude while the watch is still closed up.

9. ### Anthony FremontGuest

"Spehro Pefhany"
Ok thank you. I found some info mentioning what you just said, but
no specific equation. I must admit that my math skills aren't up to
figuring it out myself, so my quest continues to find the secret
formula.

So far I have only been experimenting with clocks and quartz watches.
Clocks tend to be much louder therefore they are easier to "hear" than
watches. I'm trying to come up with a good technique for hearing the
watch, but as little ambient noise as possible. Any ideas on a good way
to attach a microphone so as to hear thru mechanical conduction of the
sound and not the sounds traveling thru the air.

10. ### Anthony FremontGuest

"dAz"
Thanks for the reply dAz as always. After reading that, I'm thoroughly
confused as to how the drop factors in. It seems that mainspring
tension, train friction and other factors would determine the amount of
time that it takes for the escape wheel to advance. I don't see how
that would relate to balance wheel swing at all.

I could see how the duration of the impulse sound could give a clue to
the speed (and consequently the swing angle) of the balance wheel.
That is my intent exactly. Now for a really stupid question. Given a
300 degree swing, does that mean the balance swings 150 degrees each way
using the stopped/neutral position as the center reference? IOW, would
the maximum amount of swing (where the jewel continues around full
circle after flipping the pallet and then hits the outside of the fork)
be considered as ~720 degrees of swing or ~360 degrees?

11. ### Jon ElsonGuest

Get a crystal phono pickup, and attach the needle mount to the watch.
This is the "poor man's" vibration mike, known in the trade as a
vibration accelerometer. You might have more luck buying a complete
cheap phonograph at a flea market than just buying the pickup.
You can also try to get an Endevco accelerometer on eBay, but these
will be expensive unless you get lucky.

Quartz watches? With mechanical movements, or all electronic? It
may be possible to pick up the vibration of the crystal in a quartz
watch, but that sounds extremely difficult.

Jon

12. ### Anthony FremontGuest

Ok, thanks for the tip. I'll look into that.
I'm primarily interested in mechanical watches for this particular
puzzle.

Some quartz watch analyzers seem to be listening to the 32kHz
oscillator. The inductive pickup that I made from a Ford starter
solenoid coil doesn't really hear that frequency, but it's real good at
picking up the stepper impulses every second, you don't even have to
remove the watch. Just bring it within a couple of inches of the coil
and you're in business A PIC chip takes 24 bit readings (1uS
precision) of the intervals and feeds them serially to a PC. Perfect
precision for timing quartz watches, but kinda absurd (by several orders
of magnitude) for 100 year old spring driven clocks.

13. ### The BaronGuest

What machines claim to do this?

14. ### Spehro PefhanyGuest

I don't think it's very challenging. You know from calculus & Hooke's
law that the balance wheel angle follows a sinusoid wrt time, right?
Do horologists refer to the peak angle or the peak-to-peak angle?
Sprung lever? Enclosure with foam rubber lining?

Best regards,
Spehro Pefhany

15. ### dAzGuest

well its worked on sound and duration I suppose, the unlock and drop are
the two loudest sounds the escapement makes, but to human ears we only
hear one sound as a tick

you could work out the amplitude by the time taken for the balance to
swing from the drop to the unlock, or the speed it takes from unlock to
drop which probably why the beat count and lift angles are needed for
that calculation.

well Ideally a watch balance swings at around 270degs, that is from the
neutral or centre point with the roller pin stopped dead in line with
the pallet fork to swing around 3/4 either way or 270degs, thats one beat.

again ideally this is with a movement in top condition, the mainspring
is delivering smooth even power for most of the running of the watch,
the hairspring is correctly formed to preserve the isochronism to
compensate for variations in amplitude.

under 270degs you start having problems with timekeeping, any shocks the
watch recieves it takes longer for it to recover the swing,

over 270degs you run into the danger of overbanking where the roller pin
hits the outside of the pallet fork and gets an acceleration to the
swing which again affects timekeeping.

16. ### dAzGuest

first generation quartz with trim caps you could have adjusted rate via
the frequency.

modern movements with a couple of exceptions don't adjust the quartz to
affect the rate, its done by the integrated circuit controlling the
motor steps, in the cheap movements there is a pad of tracks where one
or more are cut to set the rate, others use a once only programable chip
or the better ones use an eeprom that can be reset by the use of the
witschi timer.

commercial watch timers either go by the magnetic pulse from the
stepping motor or for LCD the pulse from the display.

with the eeprom fitted movements you cannot really rely on just the
motor pulses either, because the circuit constantly corrects the output
from minute to minute, you would need to take a reading for several
minutes then do an average in seconds per day to get the correct rate
before clearing and rewriting the eeprom, this is what the \$3000 witschi
machine does, it also measures coil resistance, power consumption etc.

a quartz timer is a waste of space, most quartz you cannot adjust
because they are fixed, if a quartz watch doesn't keep time you either
service the movement or fit a new one.

17. ### Paul Hovnanian P.E.Guest

That last page makes it pretty clear what is going on.

A watch escapement passes through an angle equal to the lift angle
between the time from the unlock sound to the drop sound (Td). Since the
escapement is moving at the beat frequency (Fb), it is easy to calculate
the total amplitude of the escapement given these figures.

Amplitude = 2 * Lift Angle / sin( 360 * Fb * Td )

Someone else suggested using a phonograph pickup to aquire the
escapement sound. From the description of the commercial device, you
might be able to use the sound card in a PC plus some cheap PC scope
software to sample the waveform. By making a time measurement, you
should be able to perform the same calculation that the commercial unit
does.

18. ### Anthony FremontGuest

Thanks Paul! That indeed seems to give me numbers very close to his
example. Not exactly the same for some odd reason, but within a few
degrees.

Perhaps I'm too dense to understand this, but I still don't see how the
amount of time it takes the escape wheel to advance directly relates to
balance amplitude. Once the pallet unlocks the escape wheel tooth from
the entrance pallet jewel, the escape wheel will advance freely and with
no influence from the balance wheel until it slams into the exit pallet
jewel. AFAICT, it advances purely based upon train friction, the mass
of the escape wheel and the amount of mainspring force trying to push it
forward.
I already have the interval measuring equipment (PIC based), I just need
to be able to amplify and distinguish these tiny sounds. The
calculations will be easy.

19. ### Anthony FremontGuest

Er umm no actually. I haven't much of a clue about calculus. :-( I
know what it can do, just not how to use it to get the job done. One of
the perils of not going to college I suppose. After a couple of decades
of relative non use, even my trig capabilities have suffered. Paul's
formula seems to be correct and certainly should be easy enough to
implement even inside of a PIC. I'll be the envy of all at the trade
shows with my \$20 pocket beat analyzer. ;-)
As you've probably already seen, dAz gave a nice description. Seems to
be peak angle. I've been messing with clocks for years, but watches are
a new hobby. Funny how similar and yet different they are.
By "sprung lever" do you mean something like an alligator clip glued to
the microphone?

20. ### Robert SwinneyGuest

Agree wtih Jon re. the crystal pickup. AFAIK, there was no "magic" in the
old chrono-analyzers. It seems to me they had to be only very sensitive
audio devices. Could they have used elaborate audio filtering to emphasize
certain elements of the sound? Possible, I suppose, but highly doubtful.
Perhaps the audio waveform could be viewed on an o'scope in such a way to
discern specific parts of the sound waveform. I haven't tried this. I do
know you can't read the 32 kHz signal on an ordinary service monitor.

Bob Swinney