WatchUSeek Watch Forums banner

1 - 19 of 19 Posts

·
Registered
Joined
·
1,604 Posts
Discussion Starter · #1 ·
How do you calculate the rate?

Here's how I've done it so far:
  1. Wait for the watch to be exactly one second off. It is quite easy to see if the second hand jumps at the same time as the radio time signals.
  2. Make a video recording of the watch and a reference clock.
Both are not very satisfactory:

If your watch is quite accurate, it takes ages before it finally is a second off.

A recording only shows you when the second hand still was on one marker and when it is on another marker. If you do your recording with 25 frames per second, thats a 40 ms period of in-accuracy.

I'd love to know another ingenious way of working out the rate.
 

·
Registered
Joined
·
18,176 Posts
How do you calculate the rate?

Here's how I've done it so far:
  1. Wait for the watch to be exactly one second off. It is quite easy to see if the second hand jumps at the same time as the radio time signals.
  2. Make a video recording of the watch and a reference clock.
Both are not very satisfactory:

If your watch is quite accurate, it takes ages before it finally is a second off.

A recording only shows you when the second hand still was on one marker and when it is on another marker. If you do your recording with 25 frames per second, thats a 40 ms period of in-accuracy.

I'd love to know another ingenious way of working out the rate.
I have been doing some mini-experiments with various time references. I have 2 RC clocks, 1 RC watch, time.gov, and a shortwave radio that can get WWV at night.

I can notice display differences with clocks/watch/computer screen. Mostly this seems to be a function of how fast the display changes when a second advances... the watch and computer display the fastest... the clocks the slowest. And I can see these differences down to what I suspect is a tenth of a second.

But when I time the tick of a second hand, I find I can't relate that to when the display last changed beyond a third or maybe even a half a second of accuracy.

So any system that requires measuring differences of more than about a half second will have to involve some sort of automation.

I have a big timing project I want to do... but I confess the setting up and debugging the necessary video production has been slowing me up. I agree it's not fun... but I don't know any alternative.

I think the secret is to find the right video editing suite so you can do frame by frame comparison relatively easily. And to have a good lighting setup so you can easily read the watches, et.al. in the video.

But if I actually do it, I'll know more :-d
 

·
Registered
Joined
·
1,604 Posts
Discussion Starter · #3 · (Edited)
Great idea to compare different time sources!

On the Southern Hemisphere I have to do without rc clocks. My reference clock is a NTP synchronized computer. If it is left for a few hours to do its synchronizing, it is supposed to be very accurate. But I struggle to compare it with something else. Compared to the radio time signals on the hour, it is spot on. But I would not notice any differences under 0.1 second.

To use this computer clock, I had to write a small program that displays the time with a 0.01 second resolution: I could only find computer clocks that display the time with a one second resolution.
To the naked eye the clock looks like a blur with only the tenths of seconds displaying clearly. If I tape it, the situation isn't much different. It takes the monitor quite a while to settle on a new 0.01 second digit. Only a few video frames display the 0.01 second digit clearly.

Then I read that GPS receivers have the time to the millisecond accurate. Out came the old Garmin emap. It displays the time with a one second resolution but has a very slowly changing display.

Then I came across this article: http://time.qnan.org/

It talks about connecting a basic GPS receiver to a computer to synchronize its clock. But the best part is that it has an LED that lights up at the start of a second. Recording that shouldn't be a problem; it does not involve the slow monitor.

Now, do I really want to buy a GPS receiver for US $70 to get a clock that is accurate to the microsecond?
https://buy.garmin.com/shop/shop.do?pID=223&pvID=796
 

·
Registered
Joined
·
18,176 Posts
If the object is to time watches - getting their variance from a reference of consistent accuracy - then timing against the same reference is more important than the time displayed by the reference. If the reference is consistently 1/10th of a second off in it's display this does not affect the deltas from this reference.

So I'm going to use one of my RC clocks and invest the $70 in another watch rather than a GPS receiver. But I realize folks on the underside of the planet many not have this option. ;-)

It seems to me you need to record three 'times' to do timings via video.
  1. The time reference - recording when it marks a new second
  2. A stopwatch - displaying 1/10ths or 1/100ths of a second continuously.
  3. The watch being timed - recording when it thinks a new second occurred.
When the time reference changes the second the stopwatch will record a value. When the watch being timed changes the second, the stopwatch will record another value. The delta of the watch from the standard is then the difference in the stopwatch values.

My cellphone will display a stopwatch in 1/100ths of a second. My camera will record .avi movies (but not in macro mode :-() If I can get the lighting right so the camera can record all these images, and can find software that will allow display of .avi's frame by frame (or at least in very slow motion) then I might find it isn't too much of a pain to record exact timings... but I fear it won't be THAT easy... it never is. Too many "ifs".

But if it was easy they would have already done it over in the Time Zone and then would would be the use of WUS HEQ?

(Didn't Bruce define an HEQ WIS as one interested in talking about timing experiments?? :-d)
 

·
Registered
Joined
·
1,604 Posts
Discussion Starter · #7 ·
Yes, a great idea this stopwatch!
As long as it has a display, digital or otherwise that is fast enough. I would not trust a digital display to be fast enough, but I could be wrong there.
Counting frames between one event and the other isn't very accurate. Sometimes frames are duplicated.

Give this man one more martini garçon!
 

·
Registered
Joined
·
638 Posts
Here is my technique.

I use a digital stopwatch with resolution of 0.01 sec.

I set the PC time using the time signal from time.gov (or other suitable secondary time standard) in the US using a free program that updates the time every 5 minutes. I have a digital clock on the screen that displays the time in 1 sec. intervals (just like my watches). I don't use the clock supplied by Microsoft as I have found it to be very slow to update the time from an internet time standard.

I start by getting into the rhythm of the seconds hand movement on the watch by lightly pressing the stopwatch time start button on each hand movement for at least a 5 sec. interval. I find that I can get into a very smooth rhythm touching the button almost exactly when the seconds hand moves. On the 5th movement, I press the stopwatch hand slightly harder to start the stopwatch. Then I turn my attention to the PC clock and pick up it's rhythm. I start my timing on the next 5 sec. mark, ensuring that at least 5 sec. have elapsed before starting. Then I mark the time in the same way as with the watch for 5 sec. to get a consistent rhythm going. Then, at the next 5 sec. mark, I press slightly harder to start the stopwatch and time for 5 sec. With this information, I calculate the offset between the watch and the PC clock. This gives me the time difference between the two.

Generally, my watches are within a few seconds of the correct time (I am an HeQ forum member after all) and the offset is easy to calculate.

I have found this method gives consistent readings within 0.07 sec. over short periods of time (when the PC clock does not drift). The largest error I have found is in the time signal from time.gov since internet delays can often throw the accuracy off by from 0.1 to 0.4 sec. Thus, even though my part of the calculation is reasonably consistent, the overall accuracy when depending of an internet time standard may produce on error of several tenths of a second.

The only way I have found to minimize the inherent error of this method is time. An inconsistent 0.2 sec. error in real time can give very inaccurate accuracy readings in the short term. But, if you give your watch a couple of months, then this inaccuracy will not place an undue error on your calculations.

The best time to do this procedure is when not much is happening on the computer. The more the computer is loaded, the faster the PC clock gets out of sync. In other words, don't be editing a video, downloading a large file, etc. if you want a good consistent time from your PC.

I am not sure how to get a better time signal without spending a lot of cash. I have seen rubidium time standards for sale, but they are a little pricey as are the GPS time standards.

I have recently purchased a Vibrograf M90 watch timer. It has a claimed accuracy of 0.01 sec/day. It will be interesting to see how this instrument compares with my hand calculation of watch accuracy. Of course 0.01 sec/day translates into almost 4 sec/yr or near the upper limit of error for a Citizen Chronomaster, but it will prove useful in bringing lessor watches into an accurately regulated state.
 

·
Registered
Joined
·
1,604 Posts
Discussion Starter · #9 ·
Thats another good way to get to the rate.

I guess my watch is getting into the 4 seconds per year range; its getting very difficult to get to the rate without waiting a few months.

Thats a serious machine, but it ignores the temperature control part of the watch if it only senses the quartz frequency. Or does it listen to the second hand moving?
 

·
Registered
Joined
·
638 Posts
I don't know. I purchased the machine and then went out of the country. I will have to learn how to use it when I get back in December.
 

·
Registered
Joined
·
18,176 Posts
Interesting. This closely parallels how I hack watches. Except I find using the aural clue of the tics of WWV works better for me than visual tics of the computer. Brains work differently.

But the scientist in me rebels at data collection where the experimenter is part of the apparatus... Not easily reproducible for the doubting Thomases. And I don't think my biosystems can react with consistency to 0.07 seconds of accuracy... Your method does have the advantage of less setup overhead |>
 

·
Registered
Joined
·
1,604 Posts
Discussion Starter · #12 ·
More experiments.

This time I put a coil next to my watch to see if it was emitting anything.
Turns out it produces quite a pulse every time the second hand moves. The pulse has a smaller peak a fraction later. I believe that must be the time when it hits the new marker. But that's a guess.

I feed this pulse into my computer's sound card and record it. I start a new recording every 5 seconds when the computer thinks a new second starts.
When I measure the distance from the start of the recording to the peak I can work out how fast my watch is.
If I do a few hundred of these measurements and take an average, the rate graph seems to have lost its noise, and all dots are getting in line a bit better than with my previous video frame analysis.

I am only doing this a few weeks now, but it looks promising so far.
The distance between two red lines is 0.1 second.
 

Attachments

·
Registered
Joined
·
2,929 Posts
Excellent! This should be far superior to video frame analysis. This could be the key to the best rate check method yet. :-!

Do you think one could use this in relative (rather than absolute) mode to detect the "tick" that periodically has the supressed count? Such a method could be used on the The Citizen to confirm our suspicions that that is their method of thermocompensation.
 

·
Registered
Joined
·
1,604 Posts
Discussion Starter · #14 ·
This method could be changed to detect if indeed a pulse exists in a every one second period.
If that was your question.

I am leaving my watch next to this coil for 10-30 minutes to get a good average. You would need to subject your watch to this indignity quite a bit longer to pick the period without a pulse.:)

What I am not sure about in this method is why there is a spread in measurements. I was hoping to get a constant distance between the start of the recording and the peak.
  1. One theory is that my computer hasn't updated its clock often enough so that what I think is the start of a new second is actually a few ms off.
  2. And another, less likely but far more intriguing theory, is that the watch keeps a time internally and displays another. It might itself check every few ms if the second hand needs to move or not. That way it would add a tiny bit of randomness to its second hand moving.
Looks like this is getting out of hand quickly!
 

·
Registered
Joined
·
1,604 Posts
Discussion Starter · #15 ·
The results so far.
The first graph uses measurements from video frame analysis.
The second graph uses the pulse measurement method.

Y axis in ms fast, and X axis in hours after regulation.

Yes, I am back in the over correction territory. Next correction should put the Conquest VHP in the sub second a year territory.

A good way to see how your watch is doing without waiting for a few months!
 

Attachments

·
Registered
Joined
·
2,929 Posts
Really outstanding technique, Hans! It looks like the one sigma deviations around your best fit line are on the order of a few milliseconds. Most excellent.
 

·
Registered
Joined
·
18,176 Posts
I too am impressed. Any chance of pics of the setup??
 

·
Banned
Joined
·
3,654 Posts
... I am back in the over correction territory. Next correction should put the Conquest VHP in the sub second a year territory...
Excellent work, Hans!:-!
 

·
Registered
Joined
·
1,604 Posts
Discussion Starter · #19 ·
Setup is easy enough. Just a coil next to the watch. On its left seems to be the loudest. The coil connects to the microphone input of a sound card in a computer.

My apologies for the scratches, the Longines was in continuous use the last seven years until I came across your forum with regulation advice.

The coil, just like a sensitive microphone I guess, can monitor the second hand for quite a while. That way I can take an average over a period and filter out any irregularities. That must be its main advantage.
 

Attachments

1 - 19 of 19 Posts
Top