I've been thinkin' about an experiment...

[etmb61]

There's been a lot of discussion about different factors that impact the operation of torsion clocks. I'm working on an experiment to track and log some of those factors to see what shakes loose using available sensors and micro controllers. I actually started this some time ago and got side tracked.

So far I have a setup to record the date, time, temp, atmospheric pressure, and relative humidity. All the environmental stuff.


For the main spring torque I plan on using a load cell attached to a balanced beam on the winding arbor with the click removed. That way as the spring runs down I can record the change over time. (stock photo)


To sense the escapement drops I have a sensitive microphone and amp. So far, so good.

The difficult puzzle is how to sense the pendulum without interfering with it. I'm thinking I need its amplitude and direction but I haven't worked it out yet. I'll be using a disk pendulum clock so I can attach triggers and sensors around the circumference. Just not sure what to use yet.

There's no way to I can think of to record what time the clock says.

Any engineers out there?

Eric

 

[Raymond101]

2 hall effect transistor placed under pendulum. One on the start and finish of swing. Count swings per hour minutes. Good luck. Don't like the idea of removing the click . those springs have a high torque. Accident waiting to happen.
Have fun . Ps don't forget gravity.

 

[etmb61]

2 hall effect transistor placed under pendulum. One on the start and finish of swing. Count swings per hour minutes. Good luck. Don't like the idea of removing the click . those springs have a high torque. Accident waiting to happen.
Have fun . Ps don't forget gravity.

Yeah I should leave the click. I'll just position the sensor so that the click isn't taking the load. I'll plan on some passive restraint too.

I'm assuming gravity is constant.

 

[Raymond101]

Yeah I should leave the click. I'll just position the sensor so that the click isn't taking the load. I'll plan on some passive restraint too.

I'm assuming gravity is constant.

Gravity is not constant. Far from it.
These clocks are well known for pore time keeping. I have one it's keeps within a few minutes a month. A rotating pendulum is affected by everything. Moon as well. The winter my 1 keeps a better time, and it's in a sealed cabinet, and the temperature is almost constant . The main spring the movement is going to be small as it's a400 day and only takes about 8 turns to fully wind.
Your better to measure the swing cycle which very from make to make but about 3 or 4 rotation per minute. So if the swing is constant over a month you should be good. Any vibrations slamming doors or sudden air pressure will effect it.
I made my glass bell cover almost air tight and sealed in a cabinet on a solid wall and only open to wind it. That's about as good as it gets.
But good luck on your project it's interesting but I doubt you will achieve less than a few minutes a month. Moon phase as in tidal.

 

[etmb61]

Gravity is not constant. Far from it.
These clocks are well known for pore time keeping. I have one it's keeps within a few minutes a month. A rotating pendulum is affected by everything. Moon as well. The winter my 1 keeps a better time, and it's in a sealed cabinet, and the temperature is almost constant . The main spring the movement is going to be small as it's a400 day and only takes about 8 turns to fully wind.
Your better to measure the swing cycle which very from make to make but about 3 or 4 rotation per minute. So if the swing is constant over a month you should be good. Any vibrations slamming doors or sudden air pressure will effect it.
I made my glass bell cover almost air tight and sealed in a cabinet on a solid wall and only open to wind it. That's about as good as it gets.
But good luck on your project it's interesting but I doubt you will achieve less than a few minutes a month. Moon phase as in tidal.

Hi Raymond,

My object is not to make the clock run better but to see what affects different factors have on its operation over a period of time. This is just an exploration.

At this point I have no way to measure the affect of the moon's pull or of the changes in density of the earth's core or the positions of the planets and sun on the rotating pendulum. As far as I know there is no way to compensate for them either. Unless I can find something that can do these things for under $20 gravity is constant.

Eric

 

[KurtinSA]

Any way you can hook up a digital camera to take pictures and synch to the data in order to gather the time info.

Kurt

 

[michael isaacs]

I think this sounds very interesting. You could overlay tidal influences after the fact.

 

[Wayne A]

For rotation how about a slotted disk and optical pickup. Could make the disk wide to get higher resolution.

 

[Raymond101]

For rotation how about a slotted disk and optical pickup. Could make the disk wide to get higher resolution.

Any thing added to the pendulum will mess it up. That why I said F.E.T. or proximity sence they are cheap . Or even cheaper use the lazer from a mouse then you could connect directly to you pc .

 

[Wayne A]

Any thing added to the pendulum will mess it up. That why I said F.E.T. or proximity sence they are cheap . Or even cheaper use the lazer from a mouse then you could connect directly to you pc .

Paper don't weigh much and theres always a rate wheel and for a test setup you can do custom weights. No limits.

Wayne

 

[Raymond101]

Paper don't weigh much and theres always a rate wheel and for a test setup you can do custom weights. No limits.

Wayne

Try it . I think that it will cause a small air turbulent . But try .
Upside-down mouse under the pendulum would work as well.
It's going to be a lot of trial by error

 

[etmb61]

Or even cheaper use the lazer from a mouse then you could connect directly to you pc .

I like that idea! That would probably give a good reading just aimed at the edge of the disk.

 

[Raymond101]

I like that idea! That would probably give a good reading just aimed at the edge of the disk.

They didn't call me the mad scientist for nothing

 

[TQ60]

Go optical and use marks.

Old school white out would make white marks with almost zero weight.

Or use a hole punch and punch holes in some label stock and stick them on it.

 

[Raymond101]

Go optical and use marks.







Old school white out would make white marks with almost zero weight.







Or use a hole punch and punch holes in some label stock and stick them on it.

Might work but any added weight to the rotating weights will effect it



I hade a hair on my weight and it's enough to put it out.



If you put for example a white dots on the start & finish weights you will have to put black dots on the other to counter balance.



I use chrome pens for marking as they leave very small mirror dots .



Size of a pin head and are easy to see .
You don't have to mark the weights mark the arms just below the regulator screw as this is above the point of gravity and will not effect the swing .
Anything above a pendulum mass will not effect it . In theory.

Good luck.

 

[etmb61]

I'm thinking optical is the way to go. I don't think any added mass to the pendulum will make a difference in the results as long as it doesn't interfere with the normal rotation. To be fair, back in the day "they" added or removed mass from pendulums when the proper suspension spring wasn't available.

I'm thinking of adding a light weight slotted disk on top of the gallery, and using a pair of optical sensors to capture the motion.

The load cell can be calibrated to compute actual torque, but if power is lost during the test so is the calibration. I'm going to try it out with an 8 day clock movement to see how it works before the long term test.

 

[Raymond101]

You seem to have mastered a solution. How about a few photos

 

[etmb61]

You seem to have mastered a solution. How about a few photos

I haven't anything to show yet but a pile of parts, and not even all the parts I need. I did sort out the calibration/power loss problem so now I should be able to plot computed torque over time. Here is the load cell and its amplifier. It's an aluminum bar that is cross drilled in the center to give it a little flex. The white blobs cover the strain gauges, one pair on top and the other pair on the bottom, that are wired together to make a resistance bridge.



In my working life I maintained and calibrated similar equipment used to evaluate loads on critical aircraft components.

I'm going to make a stand to hold the clock movement and all the sensors.

Eric

 

[Raymond101]

I also used a load of that type of load cells did some design work for mot stations to test bike brake force. Quite a few years back. Slight over kill for a clock. Good luck be interesting to see your results. Even though I have still got all my old lab test equipment. I find manual adjustment more rewarding. Yes trial by error watching and little adjustments every 24 hours. Till its within spec.

I like your line of thinking.

 

[etmb61]

I also used a load of that type of load cells did some design work for mot stations to test bike brake force. Quite a few years back. Slight over kill for a clock. Good luck be interesting to see your results. Even though I have still got all my old lab test equipment. I find manual adjustment more rewarding. Yes trial by error watching and little adjustments every 24 hours. Till its within spec.

I like your line of thinking.

I agree they would have been overkill a few years back, but now they're readily available and inexpensive so why not?

I'm thinking of doing a mainspring failure test too just for giggles.

 

[michael isaacs]

"I'm thinking of doing a mainspring failure test"

Now that would be something I'd like to see propagate in slow motion!! Please video

 

[MartinM]

How about just using the encoder wheel and optical sensor out of an ancient hand scanner or ball mouse?
Maybe a bit too small to make it feasible.

 

[Raymond101]

I found a pulsotronic kj4- proximity sensor that uses a Hall. Something I mentioned in my earlier post. I just connected it up to a 15dc power supply & the Output to my scope and it detects brass aluminum non magnetic metal as well . Out is Normally open it does have a led . Works ok. This would count the rotations this one was in my box . They are about $3 . Simply simple.



It wouldn't need any additional attachments to the clock .
Sorry about quality it's was a 5min setup.

 

[etmb61]

How about just using the encoder wheel and optical sensor out of an ancient hand scanner or ball mouse?
Maybe a bit too small to make it feasible.

I had the perfect encoder wheel from an old bathroom scale in my "keep this stuff forever because you might need it someday" box. Now, of course, I can't find it.

 

[Raymond101]

Funny you should say that I have so much stuff I don't thow away I'm a big believer in reusable .

I found this sencor bottom of a big box . Of course it's always at the bottom. It's has seen a timing fault on this clock . I could see the verge was slipping even though the beat looked ok .this sencor on the pendulum is seeing 5ns errors.
Ok a little over kill .

 

[etmb61]

Here is my plan for the test setup. Pardon my primitive CAD drawing (I did scan it with a computer). I think y'all will get the general idea.



Eric

 

[Raymond101]

Hi Eric ,
I like your CAD .
I know the clock i,m working on is not a 400 day, yes I do have one.
your idea has got me thinking . a lot of french clocks have a very slow tick ,beat .
this schind clock i had to make a new verge as the original was missing .
I have gone for proximity sensor as i had to hand .just connecting now to my HP frequency counter
it measures the swing @ 557ms or 3.8Hz or 228BPM /13680BPH ( for non tech people that's 0.557 seconds,) half a second apx.
I can not get my head round why on french clocks the escapement rotates at 1 rev per 28sec .
4 revs would be 112 seconds . is this a though back when they had 100 sec to a minute . even after the pinion 8 tooth this still is an odd number to get 60 min hour. The clock does keep good time about 1/2 minute out per 24hours . this is why I am trying to use some simple idea's like you to figure
out how to time and get slow beat clocks into timing My brocot has an even slower tick about 1 sec. and keeps now within 1 minute a week not bad for 160+ yrs .
...
I did think about using ardrino mini pro as I have a box of them .. But I hate programing .
So I am going old school being a pensioner I have lots of Time to Kill . with xlab equipment all 30 year old . I just love 10-9 accuracy .
Looking forward to seeing your completed project .

Ray.

 

[etmb61]

Hi Ray,

I found a new Hall sensor (A1335 made by Allegro Microsystems) that might do what I want. It can sense a rotating magnetic field and give a readout in degrees. I'm thinking I could attach a magnet to the center of the bottom of the pendulum and center the sensor below it. Now I just need to get my hands on one.

I like the programming. When I get going on something I often forget to eat or sleep.

Eric

 

[Raymond101]

Hi Eric , sounds good I looked at the data . I have one though if you put a small magnetic on bottom of the middle. It might turn your rotation into a compass. Most of the npn proximity sensor use an oscillator and works well on brass and non ferrous metals . I found that most of the old brass had enough fe to sense. The distance is reduced by about 40% . so 10mm for iron . I would look for the sensor that can work with NON Fe . also a magnet at the bottom will add a downward or upward force on the torsion spring depending on its material make up. some of those sensor are happy with just a thin disc ie a thin washer non magnetic .
The idea is good and if you set it up right will give a very clear output of rotational losses .
Sorry about the bad text layout i'm having trouble with the forums text editor . it adds space & and extra lines

 

[MartinM]

Beware Schrödinger's Cat.

 

[Raymond101]

Beware Schrödinger's Cat.

Don't see the connections

 

[etmb61]

Progress.

I've cobbled together a data logging device. So far it records from a load cell, a pressure/temp/humidity sensor, and a temperature stabilized real time clock, and writes it to a text file on an SD card. Right now it's grabbing data every second.

My question is do y'all think that's necessary, or should I just trigger the recording on the escapement?




I should be able to import the data to a spreadsheet and graph it out. The output screen shows the date, time, temp, pressure, humidity, and the reading from the load cell.

Eric

 

[Wayne A]

Nice work. Storing 1second data on things that move so slowly might be unnecessary and make for larger files.

Wayne

 

[etmb61]

Nice work. Storing 1second data on things that move so slowly might be unnecessary and make for larger files.

Wayne

That's what I'm thinking too. The normal period for a standard size 400 day movement is about 7.5 seconds. That's probably plenty of data. Plus using the escapement drop conveniently synchronizes everything.

Eric

 

[Raymond101]

Eric nice work. Just a thought is it possible to add a line comparison in the program so if 2 or more results are unchanged it removes the data but still keeping the sampling time. It would reduce the data size without Loosening quality. I'm not a programmer so I wouldn't know how. I do have a sampling program for my big meter and scope though old school GPIB that ignores data duplicate. I know data logging is a big storage problem. My program stores as a csv.

 

[KurtinSA]

What is the desired accuracy? I wonder if you have "count bounce" that will eat into your accuracy. That load cell value...wouldn't you expect that to be the same number over and over? Why would that change as often as it does? And the load cell value seems to be continually increasing...wonder why.

Kurt

 

[Raymond101]

Those big load cells are temperature sensitive even though they use a whetstone bridge .

 

[KurtinSA]

Full bridges are (should be) temperature compensating.

Kurt

 

[Wayne A]

Just a thought is it possible to add a line comparison in the program so if 2 or more results are unchanged it removes the data but still keeping the sampling time. It would reduce the data size without Loosening quality.

This is how old Honeywell plant historian worked to prevent overloading older computer systems. Stored database values were only updated on a points value change. You can overload even modern systems by asking for too much.

Wayne

 

[Raymond101]

I am old school 70 . All my equipment & scopes are 35 years old & full calibration. Yes I'm working with old school.
BTW tubes are starting to come back.

 

[etmb61]

Eric nice work. Just a thought is it possible to add a line comparison in the program so if 2 or more results are unchanged it removes the data but still keeping the sampling time. It would reduce the data size without Loosening quality. I'm not a programmer so I wouldn't know how. I do have a sampling program for my big meter and scope though old school GPIB that ignores data duplicate. I know data logging is a big storage problem. My program stores as a csv.

I'll write a stand alone program to process the data after the fact. I'd rather gather a bit too much than not enough to see trends. The micro controller is going to be very busy just sorting and converting into the form I want to see. I can change the file type to .csv no problem. It's still just a text file.

Eric

 

[etmb61]

What is the desired accuracy? I wonder if you have "count bounce" that will eat into your accuracy. That load cell value...wouldn't you expect that to be the same number over and over? Why would that change as often as it does? And the load cell value seems to be continually increasing...wonder why.

Kurt

Accuracy? Basement workshop accuracy!

The load cell is for a kitchen scale. Turn it on, zero the tare, weigh your food, turn it off. The specs say the values will creep 5% over 3 minutes. My biggest concern with it is plastic deformation of the aluminum frame under the load over time, but I'll figure out a way to deal with that. I'll probably ignore it. It is not temp compensated but I could do that. I figured that by recording the temps with the readings I could factor that in. What I'm hoping to see is a reasonable representation of the mainspring load over time that I can relate to a known quantity like inch-ounces.

When I had the load cell by itself under a load it looks stable. The fractional changes could be just noise. I'll probably just ignore that too.

Eric

 

[Raymond101]

We also had a problem with noise load cells. It's a few years back .I think I added a screen wire and some very small ceramic caps .the problem we had was with arm transmission those big wip antenna are not perfectly SWR. They don't care . So I had to use rf suppression as well . Mains hum would be your noise. 50 or 60 hz
Just keep wiring short and twist the cell wire in pairs. CAT 6 cable is good or 5 .

 

[etmb61]

We also had a problem with noise load cells. It's a few years back .I think I added a screen wire and some very small ceramic caps .the problem we had was with arm transmission those big wip antenna are not perfectly SWR. They don't care . So I had to use rf suppression as well . Mains hum would be your noise. 50 or 60 hz
Just keep wiring short and twist the cell wire in pairs. CAT 6 cable is good or 5 .

My workbench has a 1950s florescent desk lamp. It puts out enough interference I think my ears are buzzing!

Eric

 

[Raymond101]

My workbench has a 1950s florescent desk lamp. It puts out enough interference I think my ears are buzzing!

Eric

That's good the led light are better for light . But they are the worst for line noise. Out of interest does have a original starter bulb .they were really cool. One thing about old school stuff it was made to last for years.

 

[etmb61]

That's good the led light are better for light . But they are the worst for line noise. Out of interest does have a original starter bulb .they were really cool. One thing about old school stuff it was made to last for years.

No starter. You have to press and hold the on switch until the bulbs warm up.

 

[etmb61]

Hi all,

Here is the sensor device I plan to use to capture the pendulum data. It senses the position of a close magnetic field and outputs that position in degrees. Who could ask for anything more? (#2 pencil for scale if y'all still use them) If you have electric power steering in your car, a chip like this makes it work.


It works with a rotating magnet closely centered over the chip. All I'll need to do is fix a magnet to the bottom of the pendulum and put the sensor below it. (Bubba CAD!)


Here is what the output looks like from my testing. The top and bottom lines on the right (green and red) show the movement extremes. The second line (yellow) it the difference between the two extremes. The third line (blue) is the position of the magnet. I was turning the magnet by hand so the curves are a mess, but you get the idea.


I think that's all the information I want about the pendulum movement. The only one that needs recorded is the second line unless y'all can think of something else.

Eric

 

[etmb61]

Here are a couple of better pictures using an actual pendulum with a magnet stuck to the bottom. The movement it's hanging from doesn't work so this was a short test.

The setup:


and the output:


I'm beginning to think this will work.

Eric

 

[Raymond101]

It looks really good. I think you are on to a winner.
to protect from wind , drafts etc .
Maybe but it into a perspex box .
Fish tank as protection..
Before any one says something. Empty & without water & fish.
You have put a great deal of effort into this . Really looking forward to seeing the final results.

 

[etmb61]

Well no joy sensing the escapement. I have a good running Schatz that I serviced some time ago set up and connected to a beat amplifier. The amplifier picks up the impulse and the drop, "tick-tick, tock-tock." Actually they all sound the same, more like "clunk-clunk, clunk-clunk." I'm not sure what I can do with that.

I looked up what I could find on clock and watch timers. Microset uses an optical sensor and a flag on the pendulum for torsion clocks, avoiding the escapement altogether. Not really what I was looking for but it was my original plan. Timetrax appears to use a sound sensor and an algorithm to figure out when the beats occur. Hmmm. Either way it looks like I'll need a third processor in the mix.

Eric