The Best Thermal Insensitive Movement

[dwjquest]

I am hoping that I do not get into too much trouble as the subject of the "best" movement could bring forth strong emotions. ETA vs Citizen vs Seiko vs Omega vs ???.

With the above in mind, I am going to keep to the facts as I have found them.

I have subjected quite a few watch movements to rate evaluations over the temperature range of 78 to 96 deg. F. Some thermally insensitive via crystal design (Seiko 8Fxx and Omega Marine Chronometer) and some via careful crystal selection and active compensation (Citizen A660H, Citizen E510, Seiko 9Fxx, and ETA).

The results are interesting and somewhat unexpected. That could be simply the result of a small data set or what is actually happening. I don't know for sure.

The first figure below shows the change in rate vs temperature using deg. 78 F as the zero point. Data for nine movements is shown:

• Citizen Exceed with E510
• The Citizen with A660H (2 movements)
• Longines VHP with ETA 252.611
• Longines VHP with ETA 255.563
• Longines VHP with ETA 252.561
• Omega Constellation Perpetual ETA unknown
• Omega Marine Chronometer with 1516
• Grand Seiko with 9F62

The Omega Constellation movement is unknown since I have not yet cracked the case to check the ETA number inside.

The second figure shows the set with the "bad actors" removed. Here a bad actor is any movement that deviated more than 10 sec/yr from the rate at 78 deg. F.

Loosing the Omega Constellation was not a surprise as I have found that the ETA movements perform better with respect to thermal compensation the better their base rate is adjusted. This particular watch has not been adjusted. The Longines VHP's all have been more closely adjusted.

The big surprise is loosing both Citizen A660H movements. Both movements drifted from the base rate consistently as the temperature was increased. By contrast, the Citizen E510 movement's rate deviated at first and then stabilized.

All three Longines VHP's did well as did the Omega Marine Chronometer. The OMC's insensitivity is quite amazing considering its age and the state of the art in reducing thermal drift in the 1970's.

The best performing movement was the Grand Seiko 9F62. It's ability to compensate for temperature changes was quite amazing. In fact, I had to use a bright color to get the 9F's line to show up from the zero departure line on the figure.

So let's have some discussion. What happened to the Citizen A660H movements? Bad examples or is the E510 movement superior. Maybe with a little extra love and care at the factory, the Citizen E510 could be the new Asian leader?????

I have also posted PDF files of the same images so the information can be downloaded and viewed.

[Eeeb]

I take it the difference between the two charts is you dropped the 660s from the second one. (If they represent 2 different samples you are getting a consistency I do not see in my data!)

It is interesting that some watches slow down and some speed up as temperature increases.

I would postulate the two 660s are 'overcompensating' for increased temperature.

I've been collecting similar data for any watch I have which has an inhibition period of one minute or less (the Exceed meets that criteria). I have not even looked at the data seriously but I can say non-compensated movements vary greatly on how they respond to temperature... but none are as accurate as your thermocomp movements.

Did you buy the oven controlled crystal add on for the timing setup?

For other folks who want to do this at home... I set my 'fridge' to 100F and managed to cook a nice Eta movement... strange as 100F is well within operating specs. Something now stops the second hand from doing anything but a feeble shake. Oh well.

Good data! Thanks for the effort!!!

[Hans Moleman]

What strikes me most is the 'natural' curve of the Marine Chronometer.
We're taking about a watch without a digital correction scheme here, right?
A higher frequency but apart from that it is free to do what it wants in any temperature?
The digital compensation showing as 'crude' corrections that vary by temperature.

I was hoping that the higher frequency would perform better. Good to see some numbers.

Maybe the temperature sensors are getting better?
Or some manufacturers spend more time in measuring their corrections?

Nice work!
Thank you.

[ronalddheld]

Should all of this data be archived?

[Bruce Reding]

Very interesting, David. Thanks for pulling this together!

The Chronomasters are a tad disappointing. OTOH, it's good to see how well the OMC did. An elegant approach that holds its own even 30+ years later.

[dwjquest]

Right, I just dropped the 2 Citizens and the Omega Constellation. I thought that this would make the data a little easier to view.

I did not buy the oven controlled crystal as I am using the GPS receiver as my standard in much the same way as Hans Moleman does.

[dwjquest]

Very interesting, David. Thanks for pulling this together!

The Chronomasters are a tad disappointing. OTOH, it's good to see how well the OMC did. An elegant approach that holds its own even 30+ years later.

Yes, disappointing. I am going to do a long term test on one of the A660H movements to see if there are any adaptive changes going on that could improve the overall compensation. Forum member RPF gave me the idea. May be a few weeks or months before I have enough information for a definitive answer.

I have taken some long term data to try to determine the stability of several movements when held at a constant temperature. I will be posting that info soon.

[South Pender]

This is an interesting little study, dwjquest. Very thought-provoking. I have a question about how you arrived at the dependent variable in this analysis, "change in rate." You obviously didn't conduct the research over a period of a year, so could you tell us how you got "yearly" values? Did you notice a small change over a short period and prorate this to an annualized figure? Or do you have an instrument that does this automatically? Another question might be whether you gave the movements long enough to stabilize or reverse their over- or under-compensation for the temperature change. Still another question I might have from the present results is just where the undesirable responses to temperature change begin to appear--at 78 deg. F, or perhaps sooner than that (say at 70 or 72). So there's lots of room for further analysis!

I think, however, that an important limitation in this study that needs identifying--as valuable as the analysis undoubtedly is--is the fact that you have, except for the two A660H movements, an n of 1 in each of the categories of the independent variable (watch movement). Let's, for a moment, imagine an experiment in which you took 9 different examples of just one of those movements, say the E510. It would not surprise me in the least to see the same pattern of response to the changing temperature among the 9 E510 movements as you obtained with the 9 (actually 8) different movements in this analysis. Sampling error would, I expect, produce that. As a result, it would be completely untenable to infer movement differences with respect to response to heat on the basis of the present results. In other words, it would be a mistake to infer from these results that--in the population of A660H, Seiko 9F62, ..., movements--we can expect this kind of performance from the A660, this from the 9F62, .... This is a nice start towards such an inference, but although very interesting, the results fall far short of being definitive.

I don't mean, by the previous paragraph, to diss this analysis in any way; this is the kind of thread that provides the HEQ forum with the really interesting, empirical information that watch-accuracy fans crave. And I fully appreciate the phenomenal difficulty in conducting a controlled study with several "subjects" in each category (movement type). My only point really is that we not generalize from data that don't (yet) permit this.

[Eeeb]

I assume the methodology is the same as David outlined in his first post.

I agree a larger sample would be nice... over time we may see individuals providing their own timings and thus we may get the larger sample.

Until then, we peer through our knotholes in the fence... (at least we aren't astronomers... they never get to do experiments...)

[South Pender]

I agree a larger sample would be nice... over time we may see individuals providing their own timings and thus we may get the larger sample.

Yes, that's what we need. I'd like to find some way to get owners of HEQ watches on this forum to send in their accuracy results. There must be many Chronomaster owners, for example, out there. I think we could really get some definitive results if we could perform a larger-scale analysis of movement accuracy. This excellent analysis of dwjquest is one example where multiple data points for each movement would add generalizability to the results, but the study I'd most like to perform (at least as a first analysis) would be a between-movement analysis of average accuracy prorated to annualized form. There's probably enough within-movement variation that statistically significant results would require a reasonable sample of each movement. Still, I for one would be very interested in knowing--in a more definitive way--just how each of the candidate movements stacks up: Citizen A660H, E510, ETA thermocompensated movements, Seiko 8J, 9F, 8F. I'm betting that, with a little effort, we could gather sufficient data on each movement to do the required analysis. Again, I offer my services if enough forumers cooperate.