[Hans Moleman]

How well does my Longines VHP perform?

Not an easy question actually.
In my quest to get its exact rate I measured how much it differs from a reference clock.
I took 1000 measurements and none of them is the same!

Plotted against time, it gives an interesting picture:

I took one measurement every 2 seconds. Each one is shown as a blue dot.
The x-axis is the time passed. Almost half an hour worth of measurements.
The y-axis is the difference with the reference clock. The grid is in 10 milliseconds.

You can see the difference steadily increase until after 8 minutes when the thermo compensation correction kicks in.

Without it, the watch would be a very poor performer indeed.

ETA states an inhibition period of 8 minutes for rate measurements. Makes perfect sense now.

Its done on my computer, which is not a real time computer. Some times its too busy with other things, and the measurements are way off. It shows as noise in the picture.

 

[Eeeb]

Re: My hunt for the milliseconds continues.

Very interesting. Have you collected similar data on other watches?

My suspicion is your fine detail (which is finer than any detail we have seen before) may be showing how Eta quartz watches work in general. Even non-thermocompensated watches have an 'inhibition period' I believe.

I'm going to have to dig through your past posts... I think somewhere you discussed how you collect data on this level. I can't find a decent timing machine for quartz and you may have the answer!?!

Again, Very Interesting :thanks

 

[Hans Moleman]

Re: My hunt for the milliseconds continues.

You're right!

I did the same for my Tag Heuer with the lesser ETA 955.114.

It has similar jumps. Its inhibition period is one minute.
Only 960 measurements here, so that works out to 32 jumps.
Bit more of a 'lean' this one.

I am just happy that it explains why I was getting so wildly varying measurements.
The only way I knew to get around that was doing lots of them and average them out.

Now we all know what 'inhibition period' means.
It matters when you look at your watch; it might be a few ms out!

 

[Hans Moleman]

Re: My hunt for the milliseconds continues.

Anyone can pretend to understand everything. You won't learn anything new that way though. Good for you to be honest about it!

Computers are easily distracted. They always find an excuse to leave what they're doing and start off on something else.
And when they finally come back to the job that really mattered they're too late.
Those are the non real time computers.

You can tell the computer not to run off to something else. And to start a certain job straight away. Not the 'I'll look into it when I've got time'.
Those computers are the real time ones. They'll finish the job. Everything else will have to wait.

Mostly you'll get away with the non real time ones, since they are not too busy anyway. And when they're late, its easy to spot.

My computer was not very busy doing that. But I am sure it hopped off now and then to wipe a bit of disk or whatever it thought was needed, but easily could have waited a bit.

My Mac is a real time computer and I did similar measurements there, and it did not display that noise.

I display the difference between a reference clock and my watch. That difference seems to vary over time. Seems to vary very regularly over time. So regular in fact that ETA advices people who measure the difference by machine to take only measurements every 8 minutes or multiple of 8 minutes.

I am intrigued to know why. The watch 'knows' its off by 10 ms but can't be bothered to display that. It only corrects the display every 8 minutes.
Sure, not of importance to a casual observer, but very important if I measure it every day and want to calculate the rate with that.

I am sure we'll discover more sooner or later.

 

[Hans Moleman]

Re: My hunt for the milliseconds continues.

Thanks Pedro!
I am surprised myself.

Try this for an afternoon science experiment:
Do you still have one of those radios that gets its signals from the ether?
Put it next to a quartz kitchen clock. The signals from the stepper motor are so strong that they can be picked up on the radio. FM reception works best for me. Try fiddling with the tuning a bit.

Picking up the stepper motor electromagnetic noise works much better than picking up the audible noise from the second hand movements.

That noise is fed into my computer where it is compared with a reference clock.

My reference clock is the computer clock that has been synchronized with NTP ( http://en.wikipedia.org/wiki/Network_Time_Protocol ). NTP can be compared with using time.gov. Only NTP adjusts the computer clock for you and uses more tricks to get the synchronization very accurate. That is my only reference clock. The COSC uses two atom clocks to do its stuff. One reference is never enough. One day I'll get a GPS receiver with time output. That will have an atom clock accuracy to use as a second reference.

Simple enough in hardware also: A coil to pick up the stepper motor noise. All the rest is done in software, including the reference clock.

Here is the thread that started it off:
https://www.watchuseek.com/showthread.php?t=95310

 

[highvista]

Re: My hunt for the milliseconds continues.

Here are some pictures of the stepper motor pulses from the HEQ watches I have hanging around.

ETA 255.561 (Omega SMP 200M):

ETA 955.652 (Breitling Colt Quartz II):

Citizen E510 (Citizen Exceed):

Seiko 8F35 (Seiko SBCM023):

Pulsar Y301A (Pulsar PSR-10):

 

[Eeeb]

Re: My hunt for the milliseconds continues.

As I understand it, what we are recording are changes in the magnet flux around the watch... is that correct? (Or is this sound vibrations?)

And the x-axis is milliseconds?

 

[highvista]

Re: My hunt for the milliseconds continues.

As I understand it, what we are recording are changes in the magnet flux around the watch... is that correct? (Or is this sound vibrations?)

And the x-axis is milliseconds?

The x-axis is in milliseconds. The visual scale is a bit different in each picture, since I tried to cut an entire pulse for each and the pulse width varied, especially for the 255.561, where the pulse was much wider.

Yup, this is a record of the electromagnetic flux around the watch. The stepper motor coil gets a shot of electrical current each second. This produces an electromagnetic flux from the watch's coil (more or less a radio wave) that induces a corresponding electrical current in the coil sitting next to the watch. This current is fed into the microphone input of the sound card, which doesn't care that it is a spike from a coil rather than a microphone and treats it like a sound pulse. This pulse is then visualized using a software package intended for doing sound analysis. Also, the induced flux is the greatest when the outside coil is placed parallel and close to the coil inside the watch.

The software I'm using also allows you to play back the "sound" that is recorded. Each watch's pulses are definitely distinct. A bit of time and you can easily distinguish between the movements by the sound.

The whole process is pretty much the same as a radio station's broadcast signal being picked up by the antenna in a radio and being used to produce a sound output from the radio's speaker. Very neat stuff!

 

[highvista]

Re: My hunt for the milliseconds continues.

I've just acquired a refurbished GPS-disciplined OCXO clock that outputs a 1 PPS signal that I'm feeding into my computer's microphone input along with my stepper motor pulses. The 1 PPS pulse is precise to within around 20ns according to the docs, so I've got a very stable reference to observe the drift of my watches via eye-balling the waveforms from the pulses--down to the millisecond. I don't yet have a way to automatically gather data points, but I've been able to use observation to see the inhibition periods for a couple watches so far.

My Omega SMP 200M with the ETA 255.561 is showing the same 8 minute inhibition period as noted by Hans for his Longines VHP. My Breitling Colt Quartz II with the ETA 955.652 is showing a 4 minute period--I would have thought it would be the same, having "evolved" from the older ETA line.

Tomorrow I'm going to set up the Citizen Exceed with its E510 movement and see what I get.

 

[Hans Moleman]

Re: My hunt for the milliseconds continues.

Sorry, that was a bit too cryptic.

Only every other pulse looks the same.
If you compare the graphs from for instance 11:02:01 and 11:02:03 and 11:02:05 etc, you see that they look the same.
Similarly you can match up 11:02:02, 11:02:04, 11:02:06 etc.

I was wondering if the timings differ as well for the pulses in the two sequences.
That would point to a different way they get triggered.

Not that it matters off course! :-d

 

[highvista]

Re: My hunt for the milliseconds continues.

Sorry, that was a bit too cryptic.

Only every other pulse looks the same.
If you compare the graphs from for instance 11:02:01 and 11:02:03 and 11:02:05 etc, you see that they look the same.
Similarly you can match up 11:02:02, 11:02:04, 11:02:06 etc.

I was wondering if the timings differ as well for the pulses in the two sequences.
That would point to a different way they get triggered.

Not that it matters off course! :-d

Hans, from what I can tell, sequential even and odd pulses have the same delta from the current time. It's a consistent slide from the top of the inhibition period to the bottom, at least for the ETA 255.561, which is my favorite test case right now.

 

[highvista]

Re: My hunt for the milliseconds continues.

I spent some time last night trying to figure out how the Citizen E510 1 second pulses behaved when compared against the accurate 1 PPS signal from my GPS clock. I expected to see some sort of easily observable inhibition period such as is obvious with the ETA movements. But, just by eyeballing the waveforms, I couldn't find one. When examining the pulses second-by-second, the watch pulse would alternate back and forth between two time values about 11ms apart. Over a 15 minute period, measuring every 30 seconds showed a similar behavior. This is where being able to automatically collect and graph the data points over a half-hour period would be ideal.

Hopefully, when Hans has a chance to characterize a Citizen A660 with detailed graphed data, we'll see what's going on--since I'm guessing the E510 and A660 will behave similarly. In the meantime, as time allows, I'll try to do some manual data collection and graphing and see what can be found

 

[Hans Moleman]

Re: My hunt for the milliseconds continues.

This is what the 9F62A movement graph looks like.
Using Jim's movements. Thanks Jim!

One thousand measurements over a period of 41 minutes.

What I can see is a consistent addition of 20 ms every 'now and then'.
How's that for scientific?
Definitely no inhibition period inside the 41 minutes.

What is the significance of this 20 ms?
No idea, but it sure makes it harder for someone to calibrate it.

Don't be confused about all the 'noise' showing up under the graph instead of above it as before.
I've changed my calculation method.

 

[Bruce Reding]

Re: My hunt for the milliseconds continues.

very interesting! Clearly, it's a digital scheme. Apparently for the rate as well as the thermocomp. Looks like a pulse adding method, which is the same concept with different execution from inhibition schemes.

Can you show a blow up of the graph showing so that the seconds in which more pulses occured show up individually?

 

[Hans Moleman]

Re: My hunt for the milliseconds continues.

This is a section towards the end where the measurements seem stable.

I'll attach a .csv file as well. So you play around with it in a spreadsheet.
I had to change the file's extension to .txt though.

 

[Eeeb]

Re: My hunt for the milliseconds continues.

Glad the movement made it to NZ.

On the text tile, the first number is the time and the second number is the error from the time standard, I assume. (What is the scale of the second number? milliseconds?)

This sure looks to me like the same scheme Highvista was seeing... The Japanese are using an entirely different scheme than the Swiss.

It is almost as if they have a fast rate and a slow rate for the watch and by alternating between the two, and by throwing in an occasional 'wild' correction, they average to the correct rate. But I don't see a pattern to the alternation...

Very strange... gotta sleep on this...

BTW, keep the movement for a while. I don't need it yet and you may be able to extract more data...

I bought two inductance microphones on ebay... but I fear we are going to have to go to a digital scope to get some clean data... and I am lousy on scopes.

Oh well... as Pepys used to say "and so to bed" ...

 

[Hans Moleman]

Re: My hunt for the milliseconds continues.

The .txt file indeed has the format , <milliseconds difference from reference>
Positive difference: lagging behind the reference.
Negative difference: running ahead of reference.

Great! I've ordered a GPS clock. Let's see if that clears up the data a bit.
I am not too sure how accurate my reference is.

 

[Bruce Reding]

Re: My hunt for the milliseconds continues.

First, Martien, let me say that this is totally fantastic stuff. I think I speak for the forum when I say "Totally awesome, Dude!!!!!" (and thanks!b-))

Second, I was wondering about the apparently random pattern of 20 millisecond adders. Each data point represents the timing of the "tick" that occurs each second. There are 60X41 = 2460 ticks in 41 minutes. You have a thousand samples. So, it seems that we have subsampling. I'm wondering if this subsampling is obscuring a more regular pattern, either because the subsampling itself is not regular, or because the pattern is high enough in frequency such that the sampling is not above Nyquist. Can you tell us more about the subsampling?

BTW -- I am off to a racquetball tournament today. I may not get a chance to check back in until Sunday evening. Bummer, as this thread is "sick". (My teens assure me that this is a good thing. :-d) I'm looking forward to seeing a lot of interesting discussion. b-)

 

[Hans Moleman]

Re: My hunt for the milliseconds continues.

Here is another one.
A straight line if displayed in the same scale as before. That made me change the scale.

Again in ms behind the reference.
It has lost 40 ms in a day.

 

[Eeeb]

Re: My hunt for the milliseconds continues.

so it is loosing about 15 sec per year...

I have two curiosities...
1) can the movement be adjusted via what appears to be an adjustment point?
2) does the movement change rate when running at a different temperature?

I am actually more curious about the second. We have yet to prove this movement does not change rate due to temperature.

This is fun.

I like the new scaling. I assume the x-axis is minutes of observation. What are the units on the y-axis?