Effect of Temperature on Watch Accuracy

[dwjquest]

Here are the results of a series of temperature tests I performed on some of my watches. The watch set contained both thermal-compensated and non thermal-compensated watches. The thermal-compensated watches included watches that used microprocessor based compensation (Citizen Chronomaster, Grand Seiko, Longines VHP) and watches that had increased quartz frequency (Omega Marine Chronometer and Seiko Perpetual) as the method of compensation.

The set of watches was held in a constant temperature chamber at either 82 or 92 deg. F for a period of about one month per temperature level. The drift of each movement with respect to the time signal from NIST was computed for each watch using a stopwatch and multiple time measurements. The clock used as the reference was a PC based clock which was synchronized at 2-minute intervals to the atomic clock at NIST. The drift of the PC clock over the 2-minute period between synchronizations was never observed to be greater than +- 0.03 sec. The estimated accuracy of the manual timings were estimated to be in the range of +- 0.05-0.07 sec. This was determined by observing the deviation among 5 timings taken one after another.

The temperature chamber used to maintain constant temperature conditions was a converted egg incubator (but that’s another story in itself). The temperature was easily maintained within +- 0.1 deg. F during the testing.

The first attachment shows the results sorted by the absolute value of the change in accuracy using 82 deg. F as the reference. The second attachment shows the same information but sorted by the absolute value of the accuracy at 82 or 92 deg. F.

The watch labeled “Walmart” was indeed a $7 Wal-Mart watch put into the set to provide an example of the performance of an inexpensive quartz movement.

As to what the data means, I will let the good members of the HEQ forum chew on that for a while. I have my own ideas, let’s see if you come up with the same conclusions.

[Bruce Reding]

Wow, David! Quite a job you did there. I will have to look it over more to digest it, but a first comment is that the way you lay it out nicely separates absolute accuracy from constancy. For marine chronometers in the days before GPS (or even LORAN) the latter was more prized.

[Eeeb]

Here are the results of a series of temperature tests I performed on some of my watches.
....

As to what the data means, I will let the good members of the HEQ forum chew on that for a while. I have my own ideas, let’s see if you come up with the same conclusions.

Good, Interesting work! DATA!! :thanks

OK, let me give it a shot.


1) Very interesting experiment on a variety of HEQ (and ChinaMart watches). Good to see two identical movements too.


2) As I remember it, the quartz crystal oscillators used in watches are usually calibrated at 24C... this being supposed to be the normal temperature of a watch that is being worn. (N.B. ETA requires the service centers to calibrate at between 20 and 30C.)

This assumption may be wrong, but data at this temperature would show an optimal operating environment and would be very interesting.

I would guess it would not show much difference between the ranking of the watches... but so much care was taken to provide a temperature controlled environment that getting the data at 24C would 'round out' what I believe is already one of the most interesting posts in HEQ.


3) From reading some of the other posts I have realized accuracy has two components. One is minimizing the absolute value of the measurement's variation from correct. The other is minimizing the range of error -- that is having a consistency that provides the ability to predict the error.

If a timepiece is consistently off x seconds per day and you know how many days since the last hack, then you can predict the correct time accurately by applying a correction to the displayed time.


With all this as a background, the data shows some interesting things:


A: The same movement can have different performances. (This would be the Longines VHP's contribution.) This shows the calibration of a movement can vary and not all variation in the data is due to differences in the design of the movements.


B: The thermocompensated movements have less variation in the range of the error at the two tempatures. The non-thermocompensated movements were less predictable in their error. This difference was quite pronounced.

(You have to throw out the Seiko Twin Quartz's data out to absolutely show this... but I strongly suspect there is something wrong with that watch... You might as well have the Wallmart as have that specimen of Twin Quartz... there has to be something wrong with that watch!!)


C: Thermocompensation is less a useful metric in trying to predict the absolute value of the error at any given temperature. They were better but their superiority was not as pronounced.

I suspect this would not be true if the movements were properly adjusted.

(At least one of the ETA movements has on-board adjustments that can be done with a wire and a 3V battery. All these movements were meant to be adjusted. And it is obvious they NEED to be adjusted!!)


D: Beyond these two observations, the data becomes interesting but not compelling. The Seiko that has the xtal that runs at 6x the normal frequency came out looking very good... but it's only one data point and it may have just been a good watch. The Seiko Twin Quartz has to be in dire need of hospitalization.


Very interesting... I'm looking forward to what others see.

[ausrandoman]

Well done.

I'd like to see more of this sort of testing. I have a Seiko twin quartz away being serviced at the moment. When I get it back next week, I'll let it run for a month or so, then put all my HEQ through a temperature test.

[Bruce Reding]

I have one more observation to add. With the exception of the IWC, the other non-TC watches have a similar order of magnitude change when moving from 82F to 92F. This is what I would have expected, given that the curve for a standard quartz crystal should be quite predictable. The IWC is a bit perplexing. I wonder if the crystal used is of the standard type?

I agree that the Twin Quartz is seriously deranged. Probably the absolute rate of one or the other crystal has varied. The resultant subtraction would still yield a linear Rate vs. Temp. response, but it would have a slope.

Any idea what the difference between the A660 and the A660H is? Is the latter in the newer model? I hadn't realized that they had modified the movement.

[ronalddheld]

I would prefer to look at more data, before making an conclusions. Could you ,at least, provide data points for the watches at 72 and 102 F?

[Eeeb]

I would prefer to look at more data, before making an conclusions. Could you ,at least, provide data points for the watches at 72 and 102 F?

It looks like these take several months to run... so, sometime this winter ok?? Yes, more data would be nice. But sometimes the art is getting as much out of the data you have AND NO MORE! (If this were to be published, we just don't have enough data for a 95% confidence level to say anything but we do have some things the current data at least indicates ...

If we are going to re-run, I'd like to see the units all calibrated... And I'd pick lower temperatures rather than hotter ones. Right now we are only seeing the error at the upper end of operational ranges...

[Eeeb]

Well done.

I'd like to see more of this sort of testing. I have a Seiko twin quartz away being serviced at the moment. When I get it back next week, I'll let it run for a month or so, then put all my HEQ through a temperature test.

You going to rig an egg warmer for your tests too??

I'd love to see some photos of the experimental apparatus.

... which brings me to a totally random thought that doesn't have much to do with this thread... well, I'll post it in it's own thread... I guess that's the proper thing to do... Have fun!

[ronalddheld]

These are good points, but I am in no hurry to see more data, so I can afford to wait.

[ToddG]

Well, I for one simply want to say thank you for taking the time to do all that data gathering with your collection of watches. It's definitely informative.