Adjusting beat when the stud is fixed…

[Zendoc]

Hello All,

I’m restoring an AS/ST 1686 movement.  After service, the beat error is 2.0mS and I’d like to adjust it.  The stud on this movement is fixed, so I’ll need to adjust the collet.  I’m doing this as practice in the technique as much as anything.  I’ve done it successfully before on a different movement, also with a fixed stud, but on that occasion I managed to fluke the direction I needed to turn the collet and reduced the beat error pretty much by accident.  I didn’t learn  how to determine in which direction the stud needs to be turned. 

On this current movement, the pallet fork looks pretty well centred between the banking pins to my eye, but clearly it’s not.

What tips can folks give me to work out in which direction I’ll need to turn the collet once I’ve taken the balance out of the movement?

I’ve read some suggestions on other forums but they left me confused rather than enlightened.

Looking forward to learning what others do.

With thanks,

John 

[Bonefixer]

Isn’t it the impulse jewel that needs to be centred, not the pallet fork?

If it looks ok I’d suggest trial and error. Move the collect a minute fraction one way. If the error increases go the other way. 

[HectorLooi]

Have a look at this video.

[Jon]

I worked out a simple formula to work out how much to turn a collet.

great video by Mark explaining well and showing how to set the impulse jewel between the banking pins and in the middle of the horn/notch of the pallet.

If the watch is running at 300 amplitude and it is a 18,000 beats per hour, then in every second the balance would turn 5 revolutions, or 5 tics, which would equate to 1500 degrees of travel.

So, 1500 degrees of travel of the balance happens in 1 second, or 1000 milliseconds, so every millisecond equals 1.5 degrees of travel.

If the movement is out by 3 milliseconds the collet needs to be turned 4.5 degrees (3 milliseconds x 1.5 degrees = 4.5 degrees)

If the watch runs at 270 degrees, the maths would be changed accordingly and a smaller travel in degrees over the space of one second would occur

[Bonefixer]

1 hour ago, Jon said:

I worked out a simple formula to work out how much to turn a collet.

great video by Mark explaining well and showing how to set the impulse jewel between the banking pins and in the middle of the horn/notch of the pallet.

If the watch is running at 300 amplitude and it is a 18,000 beats per hour, then in every second the balance would turn 5 revolutions, or 5 tics, which would equate to 1500 degrees of travel.

So, 1500 degrees of travel of the balance happens in 1 second, or 1000 milliseconds, so every millisecond equals 1.5 degrees of travel.

If the movement is out by 3 milliseconds the collet needs to be turned 4.5 degrees (3 milliseconds x 1.5 degrees = 4.5 degrees)

If the watch runs at 270 degrees, the maths would be changed accordingly and a smaller travel in degrees over the space of one second would occur

Is that right, or should it be half that amount? For a 3ms beat error, do you not need a 1.5ms adjustment which effectively makes the tick 1.5ms sooner, and the tock 1.5ms later, correcting the 3ms error. 
 

Just asking.

[praezis]

1 hour ago, Jon said:

So, 1500 degrees of travel of the balance happens in 1 second, or 1000 milliseconds, so every millisecond equals 1.5 degrees of travel.

Sorry, a naive fallacy, far from reality. What you are getting is an average speed, the balance wheel however is changing permanently between zero and maximum speed. Near the middle position - where we are looking here - it reaches maximum velocity, a multiple of average.

You can calculate the angle of course, but not so simple. Take in account the sinusoidal character of this velocity.

Clever people use PCTM - it shows exactly that error angle together with beat error as milliseconds

Frank

4 hours ago, Zendoc said:

After service, the beat error is 2.0mS and I’d like to adjust it.

I wonder what advantage you expect from this adjustment: a cosmetically nicer display number or more? 

[Jon]

The amplitude of a movement is from rest (impulse between banking pins and in the middle of the notch) to the full arc of the balance in one direction (either clockwise or ant-clockwise) The beat error is the difference between those swings from clockwise to ant-clockwise. You don't half the beat error in each direction.

3 minutes ago, praezis said:

Clever people use PCTM - it shows exactly that error angle together with beat error as milliseconds

So tell me what calculation you come up with?  I have measured this and it works. I'm not saying I'm going to win a Nobel prize with this calculation, but it gives an idea of how much to move the collet. I had a watch with a beat error of 4.5 milliseconds and moved and measured the collet spun 6.75 degrees to the nearest degree and brought it to 0.2 ms

[JohnR725]

12 minutes ago, praezis said:

Clever people use PCTM - it shows exactly that error angle together with beat error as milliseconds

Where would clever people find your software?

[Jon]

14 minutes ago, praezis said:

so every millisecond equals 1.5 degrees of travel.

This is an average as the balance is constantly changing speed, but as an average you can get close

[praezis]

1 minute ago, JohnR725 said:

Where would clever people find your software?

I have no doubt at all that you are one of those. And afaik you already have the free version. Beat error as degrees is available with pro version only, I am sorry.

Frank

[Jon]

16 minutes ago, praezis said:

Clever people use PCTM

I don't know if this is a language difference, but in saying 'Clever people use PCTM' is inferring that if you don't use this method then your not clever?

[praezis]

1 minute ago, Jon said:

don't know if this is a language difference, but in saying 'Clever people use PCTM' is inferring that if you don't use this method then your not clever?

Not at all. Maybe this direct translation has a slightly different meaning in your language. There are surely many many clever people who don‘t use PCTM. Please don‘t take a flippant remark too serious.

[JohnR725]

31 minutes ago, praezis said:

have no doubt at all that you are one of those. And afaik you already have the free version. Beat error as degrees is available with pro version only, I am sorry.

What I was getting at was if somebody didn't have the free version where would they find it? Then the other amusing problem is even though the browser I'm using is a Mozilla product it isn't always recognized so the last time I clicked on the link that I had to download the free version it comes up as a blank page. Although out of curiosity right now I did check and it looks like Mozilla Firefox works just fine and it comes up. It just isn't friendly towards all browsers.

so for those people looking for the free version it's available here

https://c.web.de/@337134913998293880/YuEh_TobSjaCfyBDix_1gg

 

 

[mikepilk]

My figures are not far from yours @Jon

There's an article on SHM and beat error/angle calculations here https://www.vintagewatchstraps.com/shm.php  (I paste some of it below)

I wrote a spreadsheet to work it out. For 18000 A/h, 270° amplitude. 2ms error is 4.24°

[Jon]

14 minutes ago, mikepilk said:

There's an article on SHM and beat error/angle calculations here https://www.vintagewatchstraps.com/shm.php  (I paste some of it below)

I wrote a spreadsheet to work it out. For 18000 A/h, 270° amplitude. 2ms error is 4.24°

Nice one. Thanks! this is really interesting

[LittleWatchShop]

Always fun when @JohnR725 or @praezis gives us a little snark.  It is truly a great day when we get them both on the same thread on the same day!

[Nucejoe]

On 9/24/2021 at 1:47 PM, nickelsilver said:

I have a tool that moves the roller table in very precise increments for fine tuning the beat error. I only used it on one caliber I used to do for a small manufacturer, it's not terribly practical for everyday use. It has a table with a collet in the center that clamps on the roller table, then a finger clamp that holds the balance rim. The table turns around the collet, with index marks so you know how far you went. You can very easily turn the roller table a fraction of a degree, and then go back exactly the same if you want.

 

For every day work, from the smallest ladies movements to marine chonometer, I set the balance with the cock on a bench block so the roller table is in a hole, balance on the block. Lift up the cock and move it over- not flipping it, just moving laterally, until I can see the slot in the hairspring collet, get in there and adjust (for tiny watches this is usually with an oiler, larger, a small screwdriver). Go back in the watch and check on the machine.

 

If you have no idea which way to move, like when you are very close and the fork looks perfectly centered, you can check like this: let off all power, hold the balance rim, then nudge the center wheel so an escape tooth contact a pallet stone. See where it contacts. Rotate balance so an escape tooth can contact the other pallet stone, observe. This is with zero power on the watch so the balance really truly finds its neutral position. If, for example, a tooth hits the front part of the entry stone, and the last part of the exit stone, you can rotate the balance to see which way it needs to go (here the balance need to turn counter-clockwise from its rest state, so the hairspring needs to go clockwise).

 Nickelsilver showed this tool, think he said is a propertiertory tool. Nice to have one !    ins't it.

 I wonder if a printed copy with just similar index marks helps finding where the stud is to be at, for beat error tuning.

 

 

Edited June 29, 2024 by Nucejoe

[mikepilk]

58 minutes ago, Jon said:

Nice one. Thanks! this is really interesting

I should have explained : I work out the max velocity using the equation shown.  Vmax = (2 x pi x A)/T

Then the angular error is (Beat error x Vmax)/2

[Zendoc]

The angular maths is interesting, but I’m still curious about the direction - are there observations you can make of the movement which will predict the direction in which the collet must be turned to correct the beat error? Or is the simplest way, as @Bonefixer suggests,  merely to determine this by trial and error?

@praezisasks if this 2.0ms correction is purely cosmetic.  If I understand correctly, a beat error under 1.0ms doesn’t affect timing?  So at 2.0ms, the answer is still probably yes, it’s cosmetic on this old movement where so many other variables are also  likely affecting the timing. But it’s also about me developing a better understanding and this old movement is an opportunity for that. I’m a hobbyist, so I take these learning opportunities when they arise  because many don’t present themselves often.  

I’m learning  a lot here.  Thank you.  

 

[mikepilk]

For me, it depends on the movement. If it's a fairly expensive watch, where hairsprings are rare as rocking horse poo, anything within 2ms would be fine. Any tweak you make risks damaging the precious hairspring. If it's a cheap AS movement, I'm not so worried about trashing the hairspring, so might tweak a bit more. But I wouldn't be over worried with 2ms.

[Nucejoe]

8 hours ago, Zendoc said:

 I’m still curious about the direction - are there observations you can make of the movement which will predict the direction in which the collet must be turned to correct the beat error? 

 Slightly tweak to push the coil to one side, push right after the bend of terminal curve, you can actually see the balance rotate a bit .   Do this with the movement on TG while running, observe !!     see  if BE get smaller or wider.

If BE gets bigger ,  your beat was advance, and vice versa.

---Slight push is so you don't cause a permanent bend , you want the coil to bounce back unbent.

https://redirect.viglink.com/?format=go&jsonp=vglnk_171969890606813&key=7384788ce4aeab830f00b905bbcc0552&libId=ly0qhzdj010050gq000ULau7hbs8i&loc=https%3A%2F%2Frwg.cc%2Ftopic%2F188093-yuki-3135-hairspring%2F&v=1&opt=true&out=http%3A%2F%2Fimg.timezone.com%2Fimg%2Farticles%2Fhorologium631675494030170118%2Fmark12-2.reg.schematic.numbered.jpg&ref=https%3A%2F%2Fwww.google.com%2F&title=Yuki 3135 hairspring - General Discussion - RWG&txt=

 

Tweak section one.

Play with beat error in the name of practice is Okay.

In practical watch repair, I wouldn't'  touch a 2ms BE.

 

Good luck

Edited June 30, 2024 by Nucejoe
correction

[VWatchie]

6 hours ago, JohnR725 said:

Where would clever people find your software?

6 hours ago, praezis said:

I have no doubt at all that you are one of those. And afaik you already have the free version.

5 hours ago, JohnR725 said:

What I was getting at was if somebody didn't have the free version where would they find it?

I guess to figure out what you were getting at John you'd have to be clever!? 

6 hours ago, praezis said:

I wonder what advantage you expect from this adjustment

I would expect a higher amplitude and we can't get enough of that, can we?  Also, it will be easier to get the balance going once it stops and we wind the mainspring up from 0. I haven't personally investigated if that's true but it sounded reasonable when Kalle Slaap described it. Maybe the difference is negligible when the beat error is 2.0ms or less?

 

Edited June 29, 2024 by VWatchie

[Zendoc]

@mikepilk - yes, this is a cheap movement and I have a donor movement with balance intact if all goes wrong with the tweak.  I’d leave it alone if it were anything irreplaceable. That’s why I thought it would be good for practice

This balance self-starts when winding from zero, so clearly a 2ms BE isn’t high enough to prevent that. I’d be tweaking it for sure if didn’t. In fact, that’s exactly the reason  I needed to adjust the BE of the last movement with a fixed stud I serviced - it had trouble starting at a low wind and without shaking it. 

@Nucejoe - thanks for the tip.  I’ll give the hairspring a prod while running on the TG - gently - this afternoon.  

[Neverenoughwatches]

10 hours ago, Zendoc said:

The angular maths is interesting, but I’m still curious about the direction - are there observations you can make of the movement which will predict the direction in which the collet must be turned to correct the beat error? Or is the simplest way, as @Bonefixer suggests,  merely to determine this by trial and error?

I was going to comment on this as i didn't actually read an explanation to this main part of your question, it started to become a " get em out and lets see whos got the biggest " .  But it is explained in the video so i left it. So anyway looking from the escape wheel end if the impulse pin is to the right of the center of banking pins then the collet is turned in the opposite direction to which the pin would be turned to correct it, so from above the balance that would be turning the collet anticlockwise.

Edited June 30, 2024 by Neverenoughwatches
Additional info to make it more clear

[Zendoc]

Yep, the video is very clear, but it supposes you have the movement disassembled enough to enable a very clear view of the position of the impulse pin in relation to the banking pins.  And even then, I imagine that eyeballing the out-of-centre position  would become very difficult as the beat error becomes smaller?  

I’m starting to see that trial and error might be the easiest/quickest way to work out which way to turn the collet - might just mean you need to take the balance out of the movement one more time if you get the direction wrong on the first attempt. It’s an iterative process anyway, so one more time isn’t too onerous.