D. I. Y. Watch Timing Machine.

[svorkoetter]

Just a teaser of the latest state of things:

 

[svorkoetter]

Another update. I've been busy working on the timing machine software, implementing features on my want list. Overnight last night, I ran a long term test on my Sturmanskie:

 

 

I started the test very shortly after 10pm. Notice that at 11:30, the watch starts to speed up, reaching a peak of about +12/d. This is during the date change, which puts more load on the mainspring, thus reducing the amplitude and increasing the rate.

 

Notice also that there is an hourly variation rate of about +/-1s/d, probably due to something being slightly out of round.

 

[bobm12]

I'm impressed! Great job.  I'm not familiar with your programming language but I bet you've been working hard and writing lots of code by now! Thank you for sharing.

[RosscoInOshawa]

Excellent build, I just stumbled across this after learning what a Timegrapher was.  Terrific job and another subject for me to read up on!

[matabog]

@ svorkoetter:

 

regarding http://www.stefanv.com/watches/restoring-a-russian-alarm-watch.html :

 

What formula did you use to obtain the amplitude from the tick length? ("With a 37ms tick length before the overhaul, the balance wheel had a swing amplitude of only about 85°")

 

I red your article but I didn't understand the connection between the tick length and the amplitude until I red http://www.witschi.com/download/Witschi%20Training%20Course.pdf 

 

Thank you!

Bogdan

[svorkoetter]

Hi Bogdan:

 

The formula is:  amplitude = liftAngle / sin(2 * pi * liftTime * beatsPerHour / 7200)

 

Where:

• amplitude is what we're trying to compute, in degrees
• liftAngle is the movement's lift angle, in degrees
• liftTime is the measured tick length (time from start of first sound to start of third sound) in seconds
• beatsPerHour is the nominal beat rate of the watch

I'm assuming the sin function is working in radians. If you're using a calculator or programming language where the trig functions work in degrees, replace 2*pi by 360.

 

So the example from my article you quote above is:

 

amplitude = 49 / sin(2 * 3.1416 * 0.037 * 18000 / 7200)) = 89

 

(I made a mistake when I wrote the article. I had the "2 *" outside of the "sin(...)" in my calculations. I've just corrected the article now.)

[matabog]

And there was light!

 

Thank you!

Bogdan

[svorkoetter]

Bogdan and I just had a PM discussion, and it turns out that the numbers I had in my article were actually correct, and the proper formula is:

 

    amplitude = liftAngle / (2 * sin(pi * liftTime * beatsPerHour / 7200))

 

In other words, the "2 *" does belong outside of the "sin(...)". I will now re-correct my article. :-)

[svorkoetter]

For those interested, here's how the formula is derived ...

 

If one were to plot the angular displacement, y, of a point on the balance wheel over time, it would be sine wave.

 

A sine wave of frequency f and amplitude A, the equation is y = A sin(2π f t), where t is the total time elapsed, measured in seconds. The frequency f is the number of full oscillations per second. The "beats" in "beats per hour" (bph) of a watch are actually half oscillations, so to convert from beats per hour to full oscillations per second, one has to divide by 7200. Thus the equation becomes:

 

y = A sin(2π (bph / 7200) t)

 

Now consider the following diagram, which shows one half oscillation (and small portions of the previous and next half oscillations).

 

 

 

The lift angle (LA) is the portion of a full oscillation during which the balance wheel is being propelled by the escapement, and is fixed by the design of the particular movement. The lift time (LT) is the amount of time this takes, and is measured by one's timing machine. During the beginning of a half oscillation, the movement is propelled through half the lift angle (LA / 2) for a period lasting half the lift time (LT / 2). Substituting these quantities for y and t into our equation above yields:

 

LA / 2 = A sin(2π (bph / 7200) (LT / 2))

 

Finally, rearranging to solve for amplitude (A) yields the equation from my previous post (with the variable names substituted by the ones I used in the diagram):

 

A = LA / (2 sin(π (bph / 7200) LT)

 

Q.E.D.

 

PS. Sorry Bogdan, I didn't use omega. I wanted to keep this brand neutral. :)

Edited August 12, 2014 by svorkoetter

[matabog]

Bravo!

Maybe people knew this but for me it was a revelation, this is why I insisted on Stefan to elaborate on the matter.

 

For the balance mathematical and phisical basic concepts, one could read http://watchguy.co.uk/balance-mechanics-or-a-brief-excursion-into-the-science-of-watchmaking/ I found it quite interresting.

 

Again, Thank you Stefan!

 

Bogdan

p.s. ...still wainting for the beta test version of the "do it yourself timing machine"  :D

[matabog]

Hello!

 

Stefan, could you provide more info on the microphone and the amp you used on your diy timing machine?

I found an application for windows, biburo (http://tokeiyade.michikusa.jp/download.html) and I want to try it. For this, I think I need a mic and an amp.

 

Thank you,

Bogdan

[colinh]

Wonderful stuff! :)

[svorkoetter]

I'm finishing up the first version of the software now, and will be sending it and a prototype mic/amp to a watchmaker friend.

 

I still haven't decided on a distribution scheme. Most likely free and pro versions of the software, and the pro version can also be purchased bundled with a mic and amp.

 

For those who urgently want to build their own now, the mic element is a 27mm piezo disk, readily available on eBay. Power is provided by a 9V battery, which is good for about 50+ hours of continuous use.

 

Here's the current version of the circuit I'm using, along with a stripboard layout. Please don't redistribute it on other forums and such for now.

 

 

This amplifier will work with Biburo too, but I've found Biburo to be pretty bug-ridden and unreliable, which is one reason I decided to write my own software.

Edited September 25, 2014 by svorkoetter

[matabog]

Thank you Stefan!

I found a kit-for-dummies I could buy for less then the equiv of10$. Do you have 2 min to take a look at it and tell me if it would work?

 

http://www.velleman.eu/support/downloads/?code=K1803

 

 

Thank you very much!

Bogdan

Edited September 25, 2014 by matabog

[svorkoetter]

Bogdan, that kit only provides about 40dB of gain (i.e. 100x amplification) compared to the 73dB (4500x) that my circuit provides. You could possibly wire two of these in series, but you may run into noise problems, because this kit also doesn't include any filtering. My circuit applies about 30dB of attenuation of mains hum, and also about 12dB/octave above 11kHz. My circuit also isn't really that much harder to build than the kit.

[matabog]

I will, of course, try building your circuit and let let you know how it goes.

[rb0087]

Great job! I've been wanting to do this myself for quite a while now, and I have put together code for the timing analysis in MATLAB, and written the shell of a GUI in C++/Qt, but I haven't put it together like you have. I like Qt, particularly because I run mostly Linux but I know that almost all others who might want to use it would run Windows or OSX, but the audio input api for Qt isn't the easiest to use for this sort of thing in my opinion. 

 

What all did you use for the edge detection algorithm? Part of the reason that I haven't gotten as far as I'd like to is that I spent a while toying around with the idea of using wavelet analysis for finding the pulses edges, and I got nowhere with that (I'm very experienced with Fourier analysis, but haven't ever done anything useful with wavelets). When I gave up on doing anything fancy, though, and I just went to high pass filtering and finding threshold crossings, peak values, etc., things came together quickly. Just wondering what approach you took. 

 

Anyway, great job, and best of luck!

[svorkoetter]

I briefly considered using Lazarus for the GUI and PortAudio for the API, and I may switch to that in the future, but for now I stuck with what I know, which is Delphi and the Windows WaveAudio API. I prototyped the algorithms in Maple since it's a lot easier to debug there, and then translated them to Pascal and optimized them for real-time speeds when they were working.

 

Pulse detection is pretty much as you described it, with some cleverness up front to automatically set the levels. It works remarkably well. Pretty much any watch in any running condition will give you a trace.

[Vich]

Hello Svorkoetter,

 

Read through the postings with interest. I have just finished the Accutron variable power supply from the drawings by Rich Helzer and Rob Berkavicius (http://members.iinet.net.au/~fotoplot/Modified%20Bulova%209920_6604%20Variable%20Power%20Supply-l.jpg) had to convert it over to stripboard and after a few schoolboy errors it worked, the LM324 was a swine to use on a stripboard - there were more cuts and solder bridges than I care to mention but it works really well and I feel the urge for a new project to keep the brain working.

 

I tried Kello on my iPhone but it was hopeless trying to get the amplification needed, I note your concerns for the software as well.  The Korg contact mike with your set up seems to be a good choice even though it is a tad more expensive than DIY. Still loads cheaper than buying a timing machine.

 

Hope you give us the chance to fund your efforts, I don't mind paying for good work and your software looks soundly thought out and what is better it comes down to ease of usage.   Not all watch tinkerers are maths experts or programmers but most can get their heads round the instructions for use of a timing machine even if they don't actually understand what is in the box.

 

I look forward to following your progress - really interesting thread !

 

Cheers, Vic

Edited September 28, 2014 by Vich

[svorkoetter]

The Korg mic works well, but the one drawback to it is that it's not so practical for regulating a watch, unless you can figure out a way to keep it clipped to the watch with the caseback off.

 

Speaking of instructions for use of a timing machine, writing the manual is one of the outstanding tasks remaining.

 

While we're at it, I should mention the great 1950 book, "Better Watch Repairing -- Faster", by Lewis and Lee. This 300+ page book is just full of two-page spreads of timing machine printouts, with the left page showing the printout, and the right page describing how to interpret it to determine what is wrong with the watch. Here's a link to an excerpt: http://www.stefanv.com/watches/wus_photos/faster.pdf

 

The copyright on this book has expired, so I'm seriously considering scanning the whole thing and bundling it with the software (with due credit of course).

[svorkoetter]

Hello Svorkoetter,

 

Read through the postings with interest. I have just finished the Accutron variable power supply from the drawings by Rich Helzer and Rob Berkavicius (http://members.iinet.net.au/~fotoplot/Modified%20Bulova%209920_6604%20Variable%20Power%20Supply-l.jpg) had to convert it over to stripboard and after a few schoolboy errors it worked, the LM324 was a swine to use on a stripboard - there were more cuts and solder bridges than I care to mention but it works really well and I feel the urge for a new project to keep the brain working.

 

I love working with stripboard and doing stripboard layouts. Somehow it's more fun that making PCB layouts. The one I'm most proud of is this one,

 

 

which was part of this project: http://www.stefanv.com/electronics/paia_stringz_n_thingz.html

 

I did two large stripboards for that one and managed to not need a single diagonal wire.

 

Looking at the Accutron supply schematic, I think I would have dispensed with the LM317 and used a Zener diode and voltage divider to feed VR1, and then used an LM358 instead of LM324 for the voltage follower (same op-amps, but only 2 per chip, and only 8 pins, so a lot less unused stuff to work around).

[Vich]

Thanks for the insight, it would have been good not to have had to fiddle around with seven pins in a row in parallel but it's sorted now and working.

A lot of your design and theory is beyond me but luckily one of the good things on this forum is that there are others here that you could bounce ideas off.

I am ok building stuff though and will definitely try out your amplifier and the way that you have designed it makes it easy for an interested hobbyist and you can rightly be proud of that achievement.

I feel sure there is going to be a way to attach the contact mike, I suppose it will split in two so I would start with something incredibly simple like rubber bands over the lugs moving up to sillier stuff like suction cups or temporary adhesives that will not affect the watch finish.

Anyway I am going to make my parts list up[emoji3]

[bobm12]

Or Vich, you can try a watch case vise to hold both the watch and the mics?

[Vich]

Yes Bob a cheap holder could be easily modified to copy the wooden holder in the earlier posting, the plastic/polythene ones are good candidates.. 

 

I made an incorrect assumption at the start. I thought the mike would be in contact with the back of the case to do a check on the watch.  However Svorkoetter posted that the case back will be off and I went back and checked the photo and it seems that the side is actually in contact so the back will be available probably for adjustment purposes.  Perhaps Svorkoetter can elucidate for me and forgive me if this is a rather base question.  I personally have never used a timing machine and am tired of checking and adjusting ad infinitum until I eventually get it right.  Thats why I am so interested in this posting which looks like much more fun than just buying a box off the shelf

[svorkoetter]

And in fact, my mic stand was modelled after a modified movement holder that I saw somewhere on the Internet.

 

For best audio transmission, the mic should be in contact with the crown, since the crown is part of the movement. In a watch with a soft plastic movement holder and an o-ring around the crown stem, the amount of sound you can pick up from the case is much less than you can pick up from the crown.

Edited September 29, 2014 by svorkoetter