Thermocompensation: Methods and Movements by Bruce Reding

This forum is for Horology and the art of watchmaking
Post Reply
User avatar
Posts: 40781
Joined: December 16th 2009, 11:00pm
Location: Oregon, Thanks for visiting! Now go back home!

Thermocompensation: Methods and Movements by Bruce Reding

Post by koimaster » March 10th 2013, 8:28pm

In Pursuit of Perfection : Thermocompensated Quartz Watches and Their Movements

The purpose of a watch is to correctly mark the passage of time. Thermocompensated quartz watches do this better than any other standalone watch ever made. Reasonable quality mechanical watches can be expected to keep time to within fifteen seconds per day. Standard quartz watches can be expected to keep time to within fifteen seconds per month. Thermocompensated quartz watches take performance to the next level. Current models are specified by their manufacturers to keep time to within fifteen seconds per year or better, and usually perform much better. To put this performance in perspective, fifteen seconds per year is five hundred parts per billion. This is the equivalent of measuring the distance from Geneva to Tokyo to within a car (or small boat) length. Radio controlled watches can outperform thermocompensated watches, of course, but only because they periodically synchronize to a reference clock elsewhere. While the radio control aspect of these watches is an interesting and impressive technology, the timekeeping portion of their movements is unexceptional. Thermocompensated watches, on the other hand, achieve their performance solely through the excellence of their movements. As the current champions in the centuries long quest for ever improved performance, they are impressive mechanisms indeed. While they may eventually be surpassed by new technologies such as Chip Scale Atomic Clocks (CSACs), their place in horological history is assured.

The moderators of this board -- Bruce Reding and George Palasti -- lay out what they know about thermocompensation in this article. It is divided into four parts. The first part reviews relevant background information about quartz watches. The second part describes the various thermocompensation methods. The third part lists thermocompensated movements, the watches that use them, their thermocompensation method (if known) and other relevant information. The last part contains useful references. While our article is factual in nature, we hope to convey an appreciation for the scientific depths and profoundly clever engineering embodied in these magnificent machines.

We have done our best to treat this topic completely, but there are gaps in our information. This is especially true of the thermocompensation methods used in specific movements. We have highlighted these gaps, and would welcome any information that would help us 'fill in the blanks'. Also, incorrect or questionable information about thermocompensated watches is floating on the web. We have taken care to ensure that the information in this article is correct. For the most part, our sources of information are the watch manufacturers or experts in the field, and so hopefully are definitive. If, despite our best efforts, some errors have slipped in, we will appreciate corrections. In a few instances, we engage in speculation, but we clearly state when we are doing so, and of course welcome confirmation or denial of same.

Before proceeding, acknowledgments and thanks are in order. First, sites put together by other enthusiasts were an excellent starting point in our quest for information. Four sites in particular, those of Gino Mancini, Gary Frazier (GMF), Pierre de Brial, and Bob Thayer (references 6, 8, 13 and 14 in the References section), were quite helpful, and are well worth visiting. Second, we would like to highlight the excellent tutorial on quartz resonators and oscillators that Dr. John Vig has made available on the web (reference 1 in the References section), and thank him for his kind permission to use material from his tutorial in this article and his willingness to answer questions on this topic. We would also like to thank ETA officials for their help in confirming details of past and present ETA movements, particularly Mr. Martin Bieri and Mr. Jean-Claude Robert. We also thank Robert Logie (Roba) for his editorial assistance and Japanese translations, Dr. Walt Arnstein for reviewing the article and providing information on ETA movements, and Gino Mancini for reviewing the article and providing information on the Omega Megaquartz Marine Chronometer. We were fortunate to be able to use the excellent photographs of Walt Arnstein, Reto Castellazzi, Gino Mancini, and Joacim Olsson (Jocke). Finally, we appreciate the support of our fellow high end quartz aficionados, whose enthusiasm spurred us to write this article.

Part 1: Background Information (or Everything You Didn't Want to Know About Quartz and Were Afraid to Ask)

Thermocompensation in a quartz watch is a way of ensuring a stable rate in the face of varying ambient temperature. It does not ensure an accurate rate. To understand the difference between the two, consider older mechanical marine chronometers used for navigation. The US Navy, for instance, required that the rate of its marine chronometers be correct (accurate) to within a relatively loose 1.55 seconds per day. Given that the chronometers were never reset during their years long deployments, this meant that their indicated time could be off by many minutes. The Navy demanded considerably tighter rate consistency (stability), however, with separate specifications on how temperature affected rate, how the degree to which the piece was wound affected rate (called isochronism), and intrinsic variations of rate over time. Having a chronometer with a highly stable rate, and knowing the rate error of that chronometer, a navigator could determine the correct time. Their process would go something like the following: 'The rate error is a positive 1.2 seconds per day. I last knew the correct time 20 days ago. Therefore, to get the correct time, I need to subtract 1.2 x 20 = 24 seconds from the time that my chronometer indicates has since passed.' Being able to count on the rate being the same every day -- not one second slow on some days and five seconds fast on others -- was the key to correctly determining the time.

Figure 1: 'Steady as she goes.' A Hamilton model 21 box chronometer. (Photos by Bruce Reding.)

Of course, it's not reasonable to expect watch owners to make correction calculations as navigators did. Therefore, makers of thermocompensated watches take care to set the rates accurately. Perhaps surprisingly, though, achieving an accurate rate is much easier than achieving a stable rate. A comparison with mechanical watches may help to explain this. For mechanical watches, changing the rate (referred to as 'regulating' the watch) is relatively straightforward. For most watches it's a matter of moving a lever. Achieving rate stability, however, is much more challenging, as it delves into areas such as isochronism, the effect of orientation on rate (the 'positions' one hears of), temperature effects, etc.. Minimizing these effects (referred to as 'adjusting' the watch) is the deepest test of a watchmaker's skills. The same is true of quartz watches: accuracy is a quick adjustment (usually done only by the manufacturer), whereas stability gets into deep issues of science and design. In short, for good timekeeping, stability is the key, stability is the challenge, and so stability is our topic.

Stability starts with the resonator, which is the heart of any watch. Counting its back and forth motions (called cycles, or beats) is how timekeeping is done. For the rate to be stable, each cycle of the resonator must take the same amount of time to complete. (In technical parlance, its frequency must be constant.) To achieve this, the resonators in all watches have a special property -- they are harmonic. This means that they have one special frequency (called the natural frequency) at which they prefer to operate. For a mechanical watch, the resonator is the balance and hairspring. Its natural frequency is determined by two things: the 'angular mass' of the balance (more properly called the moment of inertia), and the stiffness of the hairspring. The greater the moment of inertia of the balance is, the slower it will beat. The stiffer the spring is, the faster it will beat. (To understand this, think of the old schoolboy experiment of twanging a ruler on the edge of a table. As the ruler is moved in, the 'twang frequency' increases. This is because the part hanging over the edge is becoming both stiffer and less massive.) The resonator is fed energy by the escape lever, which also passes its beats on to the rest of the watch for counting and display. Because it is harmonic, the balance and hairspring resonator beats very steadily. (Its predecessor, the foliot, was not harmonic, and so the invention of the balance and hairspring yielded one of the biggest performance improvements ever made in watches.)

For a quartz watch, the resonator is the quartz crystal. Figure 1 shows a simplified picture of the crystal used in current quartz watches:

Figure 2: Typical watch movement quartz crystal. (Slide 87 in reference 1. Used with permission.)

As the figure shows, the crystal is shaped like a tuning fork. (The crystal orientation shown in the figure will be discussed later.) Many watch enthusiasts don't realize that, like the balance and hairspring in a mechanical watch, the quartz crystal actually moves. Specifically, its tines move back and forth, usually 32,768 times each second. Analog electronics serve the same purpose as the escape lever in a mechanical watch does: they feed energy to the crystal to maintain its motion, and they relay the beats of the crystal to other parts of the watch for counting. They do this via the piezoelectric effect, a property of quartz and some other crystals in which applied voltage results in a change in the crystal shape (and vice versa). While its motion is maintained and counted via electronics, the quartz crystal is actually a mechanical resonator: it physically moves, and its natural frequency is determined by the moment of inertia of each arm and the stiffness of each arm (the same mechanical properties that determine the natural frequency of the balance and hairspring). Given these similarities, quartz resonators and balance and hairspring resonators are brothers under the skin. We emphasize their similarity for two reasons. First, the potential causes of instability in the two are similar, so the parallel will help to illustrate these causes. Second, we hope to soften the view of some watch enthusiasts who feel that quartz is an alien and usurping technology. In our view, the quartz resonator is simply the latest extension of the centuries old but ever renewing horologic tradition.

Quartz watches are better timekeepers than mechanical watches. One will sometimes hear that this is because of their much higher oscillation frequency. While there's a grain of truth to this explanation, its incompleteness obscures the larger truth. Simply put, quartz watches perform better because the quartz crystal is far superior to the balance and hairspring as a resonator. How so? The reasons can be grouped into two major areas: intrinsic stability, and stability in the presence of external effects. On the intrinsic stability side, the vibration of the quartz crystal is almost a pure, single frequency. Using musical terms, it rings with a very pure tone -- far purer, in fact, than any musical instrument or bell ever made. Because of this very pure tone, there is essentially no spontaneous wandering of the frequency that can cause error. (In technical parlance, a resonator that rings with a very pure tone has a high 'Quality Factor', or Q. Another property of high Q resonator is that they ring for a long time after being 'plucked'. This means that quartz needs very little energy to keep it ringing, which is good for battery life.) Moving from the esoteric to the practical, unlike the balance and hairspring, the quartz crystal uses no oil that can get gummed up over time, nor does it employ mechanical fiddly bits like regulators that can move.

Stability in the presence of external effects breaks into a number of areas. First, because the quartz crystal oscillates at a much higher frequency, movements of the wearer's arm have essentially no effect on its vibration. In technical terms, the natural frequency of the quartz is much further removed from the extraneous noise frequencies, so it will be less affected by them. (This is the 'grain of truth' alluded to above.) Also, all the classic error sources that the balance and hairspring are adjusted for -- positions, isochronism and temperature -- have a smaller effect on the quartz crystal. Because it is very stiff, the quartz crystal is largely unaffected by orientation with respect to gravity (the positions). Also, its frequency is much less affected by variations in the amplitude of its oscillation (isochronism). Finally, quartz expands and contracts very little with changes in temperature. (In technical parlance, this is called the coefficient of thermal expansion, or CTE.) In fact, it has one of the lowest CTEs of any material. Because of its low CTE, the length and thickness of the fork tines on the crystal change very little with temperature. This in turn means that their stiffness and moment of inertia change very little, which (remember the ruler experiment) results in very little change to the frequency of oscillation. The temperature insensitivity of quartz is no small matter, as achieving a similar level of insensitivity in balance and hairspring resonators it has been the work of centuries -- and some Nobel prize winning efforts. Quartz is not perfectly insensitive to temperature, though, and temperature is in fact the largest remaining source of rate variability in quartz watches. Combating this variability is the main topic of the next section. Still, with regard to the classical error sources, quartz does pretty darned well. Had quartz watches been made at the turn of the century, they would have merited a proud inscription like the following:

Figure 3: The mark of the very best. (Detail of Illinois Sangamo railroad watch. Photo by Bruce Reding.)

Before delving into the effects of temperature and how they are minimized, it's worth noting one other source of variability in quartz oscillators -- aging. Aging is a phenomenon in which the vibrational frequency of the oscillator slowly changes over time. It can have many causes. Contaminants in the container that holds the crystal could stick to the crystal surface (adsorb), increasing the mass of the vibrating structure, and therefore decreasing its frequency. Conversely, contaminants that were initially on the crystal could unstick (desorb), or gasses that had diffused into the quartz could diffuse back out from the body of the crystal (outgas) once its container is evacuated, either of which would lower the mass of the vibrating structure and hence increase its frequency. Also, the structures that hold the crystal or that bond the electrodes to the crystal could change over time, changing strains in the crystal that affect its frequency. Finally, the electronics in the analog circuit can be a source of aging as well. Ironically, the trimmer condensers (adjustable capacitors) that were used in older quartz movements to adjust the rate have been reported by some hobbyists as being particularly guilty in this regard. Reference 2 discusses the causes of aging in detail.

If great care is not taken to avoid these causes of aging, the rate can change significantly over time. As is noted in reference 3 below, the quartz oscillators used in digital electronics (which are used for synchronization rather than for timekeeping) can drift up to 5 parts per million (ppm). While this may sound stable, consider that it is the equivalent of the rate changing by 13 seconds per month each month. If the rate were dead on in January, it would be 13 seconds per month fast in February, 26 seconds per month fast in March, 39 seconds per month fast in April, and so on. In a watch this would be unimpressive performance indeed. As is evidenced by many dozens of patents, however, the manufacturers of quartz crystals for watches have worked diligently to minimize these effects. The result is that they have essentially eliminated aging in their best crystal packages. We have seen numerous reports of thermocompensated watches keeping their rate constant to within a few seconds per year over a number of years. Note that one sometimes hears statements among high end quartz enthusiasts that manufacturers age their crystals, presumably to let transient effects settle out. This is plausible. This practice would only be a part of the overall solution for aging, however. Thoughtful design and extreme diligence in the manufacture of the crystal and its package are the main ingredients for success. Indeed, the conquest of aging is probably one of the epic but unsung battles in the saga of quartz development. (Were it only the case that a similar victory had been won against aging in watch enthusiasts.)

Part 2: Thermocompensation Methods (or Some Like it Hot and Cold)

It is a common misconception among watch hobbyists that manufacturers simply pick the lucky few best crystals from the millions they make and set them aside for their high end movements. This is not the case. How the frequency of the crystal varies with temperature is completely determined by its design. All crystals with the same size, shape, and crystal orientation will have the same response to temperature. This is a bad news/good news story. The bad news is that manufacturers cannot cherry pick crystals that are temperature insensitive by luck. Every crystal of a given type will have the same temperature sensitivity. (Note that luck can play a part in the actual performance of a watch. One occasionally hears of the miraculous Timex that keeps time to within a few seconds a year. This can happen due to a fortuitous cancellation of errors. If it is not thermocompensated, though, the performance of this watch will change significantly when placed in a different thermal environment.) The good news is that the effect can be anticipated and planned for in design. This is where the engineering magic comes in.

The most obvious way of reducing the effect of temperature on rate is to keep the crystal at a constant temperature. This strategy is used for laboratory grade oscillators, in which the crystal is kept in a (very small) oven that is held at a constant temperature. In industry parlance, this is an Oven Controlled Crystal Oscillator, or OCXO, where the X stands for crystal. This approach, because of its necessary size and power consumption, is not practical for a wristwatch. It's worth noting, though, that some manufacturers suggest a poor man's version of this strategy by recommending that their watches be worn regularly to ensure best performance. Seiko, for instance, states that watches with their 8F movements should be worn twelve hours daily to ensure performance to their twenty seconds per year specification. Citizen recommends the same wear time for their The Citizen models to ensure performance to their five seconds per year specification. So how does wearing your watch help? First, the temperature of the watch will vary less than the ambient temperature does because it's attached to a constant temperature object (you). In fact, for the same reasons that your beer cools down faster in a bucket of ice than it does in the fridge, your watch will be closer to your body temperature than it will be to the ambient temperature. So wearing it reduces the magnitude of its temperature swings by a good bit. Second, by wearing it, you maintain your watch at a temperature that it 'likes' to be at. (This will be explained shortly.) Note that, just because a manufacturer recommends a minimum wear time for their watch to ensure performance, one should not conclude that the watch is not thermocompensated. Regular wear can only improve rate stability for any watch.

A second, more sophisticated method of reducing the effect of temperature on rate is to use an orientation of the quartz crystal for which temperature effects are minimized. As is true of most crystals, quartz is anisotropic. This is just a fancy way of saying that its properties, including stiffness and coefficient of thermal expansion, differ as a function of direction within the crystal. These properties determine the natural frequency of the crystal, so the trick is to find a crystal orientation for which, as temperature changes, the changes in these properties cancel each other out in terms of their effect on frequency. Quartz is blessed with a number of 'cuts' (orientations) for which temperature effects are minimized due to such cancellations. One of these cuts is shown in Figure 1, and is used in all standard quartz crystals. Figure 4 shows how the rate of a tuning fork type crystal with this cut varies with temperature:

Figure 4: Daily rate error vs. temperature of the typical 32 kHz watch crystal. (Slide 132 in reference 1. Used with permission.)

As can be seen, the rate is at a maximum at around 28C, which is referred to as wrist temperature in the plot. This maximum is where the changes in the properties that affect the frequency perfectly cancel. Since the curve is parabolic, the rate change grows as the square of the difference from this temperature. Let's 'make this real' by putting some numbers to it. If the crystal is initially at 28C and then gets 1C cooler, then, using the Temperature Coefficient of Frequency of -0.035 ppm/C^2 (noted in the figure), the rate will change by (-1)^2 x -0.035 = -0.035 ppm, which is -1.1 seconds per year. If, instead, the crystal gets 10C cooler, then the rate change will be (-10)^2 x -0.035 ppm = 100 x -0.035 ppm = -3.5 ppm, which is -110 seconds per year. In summary, as long as the crystal temperature remains close to 28C, or 'wrist temperature', the variation of its rate will be small. Large deviations from this temperature result in much larger rate changes. (This explains the earlier comment that the crystal 'likes' to be at this temperature.) So, by dint of clever engineering, even standard quartz watches have a reduced sensitivity to temperature changes.

Figure 5: Thermocompensated? Well... that's a stretch. But certainly 'thermo-desensitized'. (Photo by Bruce Reding.)

One final comment about non-thermocompensated quartz watches is that manufacturers typically set their rates so that they run slightly fast when at wrist temperature. This is because, since variations around this temperature can only reduce rate, setting the maximum rate fast makes it more likely that the average rate will be closer to being correct.

As cleverly as it is designed, the frequency of a standard tuning fork type crystal will change significantly if the temperature is substantially different from the nominal 28C. Given that temperature may affect its rate by 100 seconds per year or more, if watchmakers wish to guarantee performance to a few seconds per year, then they must supplement the crystal orientation scheme, or use a different method altogether. This is where thermocompensation comes in. Thermocompensation methods can be separated into the following broad families:

1. Single crystal methods:
  • high frequency crystal
  • forced constant frequency (often referred to as TCVCXO)
  • digital count adjustment (sometimes referred to as count suppression)

2. Dual crystal methods

3. Other methods?

(Note that the terms 'forced constant frequency' and 'digital count adjustment' are not standard terminology in the industry as far as we know. We use them in this article because they are descriptive and unambiguous.)

The first of these methods -- the high frequency single crystal -- might be more fairly described as thermo-desensitizing than thermocompensation, since it is a further extension of the passive crystal orientation method used in standard quartz watches. The Seiko 8F movements are the only high frequency movements currently available. They have 196 kilohertz crystals (six times the normal 32 kilohertz frequency). We speculate that the higher frequency enables a different cut than that used for standard 32 kHz tuning fork crystals, and that this cut is less temperature sensitive. Seiko specifies the performance of these watches to twenty seconds per year, and only if worn regularly. This is not cutting edge performance for a modern thermocompensated watch, but it is better than a standard quartz movement. Figure 6 shows a nice example of a watch with this movement:

Figure 6: Seiko with 8F 196 kHz movement. (Photos by George Palasti and Reto Castellazzi respectively. Used with permission.)

Movements with very high (Megahertz range) frequency crystals, on the other hand, can have cutting edge performance indeed. Figure 7 shows the rate vs. temperature curve of a different crystal orientation called the AT-cut:

Figure 7: Frequency vs. temperature of the AT cut. (Slide 135 in reference 1. Used with permission.)

In this figure, the x axis is temperature, and the y axis is the frequency change in ppm. The colored curves show the frequency change as a function of temperature for a number of closely related crystal cuts (their cut angles differ slightly, as is marked in the graph). Because the frequency vs. temperature curve for this cut is S-shaped, if the S is appropriately oriented, it yields a broad range of temperatures over which frequency is essentially constant (note the green curves). Using such a crystal in a watch would yield a rate which was similarly constant over a broad temperature range. For the AT cut, the oscillations occur in the body of the crystal itself, rather than between tines. Because a solid quartz crystal is much stiffer than a tine shaped one, these oscillations are very high frequency. Two early quartz watches -- the 4 MHz Citizen Crystron Mega and the 2.4 MHz Omega Megaquartz 2400 -- used AT-cut crystals. They were specified to vary less than 3 seconds per year and 12 seconds per year respectively without thermocompensation. This rivals the performance of the best thermocompensated watches today, which is impressive considering that these watches were designed thirty years ago. Figure 8 shows one of these early champions:

Figure 8: The Omega Megaquartz 2400 (Photos by Gino Mancini. Used with permission.)

We are not aware of any current watches that use AT-cut crystals. Given their outstanding performance, it's natural to ask why this is so. One reason is that higher frequency crystals consume more power. Batteries lasted for only one year for both the Citizen and Omega models. We speculate that cost plays a role as well. The Marine Chronometer version of the Omega cost nearly $2000 in mid seventies dollars, and the Citizen cost an astonishing $15,000 in mid seventies dollars. It's reasonable to believe that their exotic crystals drove much of this cost. (An AT type crystal is very unforgiving. As an example, an error of a few thousandths of a degree in its cut angle will substantially degrade its temperature insensitivity. Manufacturing to such tolerances is probably very expensive.) Tuning fork type crystals, on the other hand, are more forgiving. Also, given that billions of them are now made annually, manufacturing technology has been put in place that makes them quite inexpensive. We suspect that manufacturers have determined that standard tuning fork crystals augmented by active thermocompensation yield similar performance for lower manufacturing cost. While they were expensive, these watches are appealing from an engineering point of view. Their approach is very clean, and it highlights the virtuoso potential of quartz.

The next thermocompensation method in the single crystal category is the forced constant frequency method. This method uses a Voltage Controlled Crystal Oscillator (VCXO). This is a special crystal package that is designed so that its frequency can be adjusted over a small range by varying an input voltage. (Note that the crystal itself is still a tuning fork.) The designers then create an analog compensation circuit for which the output voltage varies with temperature in an 'equal but opposite' way. When the output of the compensation circuit is coupled to the input of the VCXO, the frequency of the quartz oscillator is forced to be constant over a wide temperature range. This overall circuit is called a Temperature Compensated Voltage Controlled Crystal Oscillator (TCVCXO). Because this method achieves thermocompensation via purely analog techniques, it was realizable in the pre-digital era. The Rolex 5035 and its variants used in the Oysterquartz are the pre-eminent (and possibly only) examples of movements that use this method. While elegant, this method has some disadvantages. First, the temperature variation of the compensation circuit and the crystal must be precisely matched. Mismatch of only a few percent will lead to errors on the order of seconds per year. Such precise matching is difficult for an analog circuit. Second, adjusting the trimmer condenser changes the temperature curve of the oscillator, again compromising the match between it and the compensation circuit. Finally, because their frequency is adjustable, VCXOs are more prone to low level noise. Still, these watches are very appealing for their unique approach.

Figure 9: The Rolex Oysterquartz -- a unique breed. (Photos by Jocke. Used with permission.)

The next major step in thermocompensation -- the digital count adjustment method -- eliminates many of the issues of the TCVCXO method. In this method, the crystal frequency is allowed to drift with temperature, just as the crystal in a non-thermocompensated watch does. However, an independent sensor (probably a thermistor) is used to measure the temperature of the crystal. Given a precise knowledge of the frequency vs. temperature characteristics of the crystal, the digital count that is derived from the crystal oscillations is then adjusted to correct for the temperature effect. For a movement that counts to 32,768 every second (i.e., a standard 32 kHz movement), subtracting one count every 16 minutes (960 seconds) will increase the rate by two seconds per year. Specifically, if the second hand were advanced each time the counter reached 32,768 for 959 times in a row, and in the 960th second, it was advanced after counting to only 32,767 (one less than normal), then this would increase the rate by 1.0 seconds per year. With this in mind, a conversational version of the digital logic might go something like this: 'The thermistor tells me that, for the last 959 seconds, the crystal temperature averaged 20 C. I know that, at this temperature, the crystal vibrates 1.7 ppm, or 54 seconds per year, more slowly. Therefore, since one count in 960 seconds equals 1.0 seconds per year, I will need to adjust the count by 54. So, for this 960th second, I will tell the stepper motor to move the second hand after the crystal has vibrated 32,768 - 54 = 32,714 times.' Note that the exact digital scheme (adjustment intervals, lump vs. distributed corrections, etc.) used by various manufacturers may vary. Still, the method works substantially as described.

The digital count adjustment method has several advantages over the forced constant frequency (TCVCXO) method. First, because the resonator frequency cannot be tuned, it is less susceptible to noise. (It is also simpler and therefore less expensive.) Second, because the compensation is digital, it can be made to precisely match the temperature dependency of the crystal, as opposed to the approximate match that would be yielded by an analog compensation circuit. (The digital compensation could be as sophisticated as a lookup table with the exact adjustment needed for that exact crystal for each degree C. We don't know how far each manufacturer goes in this regard, however.) Such precise matching is important, since seconds-per-year performance over a reasonable temperature range requires compensation to be correct to within a few percent. Note, that this method requires much in the way of digitally stored data and logic. Therefore, while it is capable of yielding better performance, it was not introduced until at least a decade later than the TCVCXO method. The ETA Thermoline movements (inclusive of the Breitling Superquartz line) use this method of thermocompensation. Figure 10 shows an excellent example of one such model:

Figure 10: Longines Flagship VHP Perpetual Calendar (Photos by George Palasti.)

In the dual crystal method, one of the crystals is the standard 32 kHz frequency, and one of the crystals is a higher frequency. In the enthusiast fora/literature, one will sometimes hear that the higher frequency crystal 'disciplines' the lower frequency crystal. This should not be taken to mean that the higher frequency crystal is correct in an absolute sense. (In other words, that it does not vary with temperature. Indeed, if this were the case, one would wonder what the role of the lower frequency oscillator was. For the dual crystal movements that we are aware of, the high frequency oscillator is well below one MHz, which means that it cannot be of the AT or similar cuts, and will therefore vary with temperature too.) So, how do they work? In a nutshell, while both crystals vary with temperature, they vary differently. For any given temperature, there is a unique difference in the frequencies of the two crystals. As such, the frequency difference is effectively a measure of the temperature, and can be used as input to the compensation circuitry. Let's make this more concrete by examining an older dual crystal method used by the caliber 9923A movement in the Seiko Twin Quartz. This method is described in reference 9, from which the following figure was borrowed:

Figure 11: Seiko Twin Quartz dual frequency thermocompensation method. (From Seiko Technical Guide for Caliber 9923A.)

In Figure 11, the 'master quartz crystal oscillator' curve shows how the rate of the 32 kHz crystal varies with temperature. The 'auxiliary quartz crystal oscillator' curve shows how the rate of the high frequency crystal varies with temperature. The high frequency crystal was chosen so that its curve is shaped the same as that of the 32 kHz crystal (both are parabolas with the same quadratic coefficient), but its peak is at a different temperature. Subtracting the two curves yields a straight line, which is denoted as line a in the right plot of the figure. Line a is effectively a measure of temperature. As is indicated in the figure, line a is then digitally transformed into curve b, which is the compensation curve. Because compensation curve b mirrors the master quartz crystal oscillator curve, summing the two yields line c (shown in the right plot), which is a rate that does not vary with temperature. Note that this is essentially a digital count adjustment method: temperature is measured (via the frequency difference of the two crystals), and the count from the 32 kHz crystal is then adjusted based on this temperature. The method of inferring temperature in this technique is quite clever in that it is inherently digital, and so immune to the kind of drift that might occur with an analog temperature sensor. Also, the digital processing is advanced for a movement designed in the mid seventies.

The other notable dual crystal movement is the ETA 255.561, which was first used in the Longines Conquest VHP (the older, non perpetual VHP). We have less information about this movement. Fundamentally, though, it undoubtedly works by the same method -- i.e., it uses the frequency difference of the two crystals to determine temperature (implicitly or explicitly), and uses this temperature signal as input to their correction electronics. Because this movement was designed a decade later than the Twin Quartz, substantially more sophisticated digital electronics were available to the designers. This allowed them to move towards a more fully digital system. For instance, they chose digital rate trimming as opposed to the analog trimmer condenser used in the Twin Quartz. It's possible that they incorporated other digitally enabled improvements as well, such as more sophisticated methods of determining how many compensation counts are needed, which would enable more exact compensation of the temperature curve of the 32 kHz crystal. Also, they may not have felt constrained to use a higher frequency crystal whose temperature curve was the same as the lower frequency crystal. (Choosing a cut that was more temperature dependent would yield an enhanced, albeit nonlinear, temperature signal.) These last two comments are speculation, however. Beyond broad strokes, we actually know rather little about the details of the thermocompensation method used in this movement, and we would greatly appreciate any information on same from other enthusiasts. Figure 12 shows a nice example of a watch with this movement:

Figure 12: Krieger Marine Chronometer (Photos by Walt Arnstein. Used with permission.)

These are the thermocompensation methods used in watches that we know of. It is very possible, however, that our list is incomplete. A number of watches remain veiled in mystery. Citizen's A660 movement used in their 'The Citizen' models -- the current front runner in specified performance among watches -- is a prime example, as is the Seiko 9F movement. Also, one of us (Bruce) has seen a number of patents for thermocompensation methods that are entirely different than those described above. Among these was a composite crystal with multiple crystal orientations. Another was for a three tined crystal that combined torsional with flexure modes. Most of these alternative patents are from the early eighties. We suspect that they are a product of the early ferment in quartz watch technology, and have not survived to this day. Still, it's possible that these or other methods are still in use.

As clever as these methods are, we are unaware of them approaching fundamental performance limits of quartz. Laboratory quartz oscillators are stable to 0.01 seconds per year or better. While the design constraints for watches (power, size, etc.) are more severe, it's not obvious why, with sufficient ingenuity, thermocompensated watches couldn't be improved from their current best of under five seconds per year to less than 1 second (or even 0.1 seconds) per year. One obvious path would be to pair AT cut crystals with active thermocompensation of the digital count adjustment variety. There may be other, more subtle paths as well. Sadly for us techno-obsessive watch hobbyists, there is no market demand for such next generation movements. (Frankly, there is little demand for even the current generation of thermocompensated quartz watches, which is why they are such rare beasts.) There is hope, though. Other applications such as wireless communications will drive development of higher performance standalone timekeeping methods that are low cost and portable. Quartz might be a viable contender, along with Chip Scale Atomic Clocks and possibly other technologies. We can only hope that, whatever the winning technology, there is some trickle down to our beloved watches.

Part 3: Details of Thermocompensated Movements (The Low Down on High Performance)

This section lists the thermocompensated movements that we are aware of, with information on the thermocompensation method used and timekeeping performance as specified by the manufacturer. (Note that we also include high frequency watches that are not, strictly speaking, thermocompensated. They merit a place, however, as they are 'thermo-desensitized', and their performance is better than that of standard quartz movement.) We also note whether the movement has adjustable rate trimming, independently settable hour hand, perpetual calendar and long life power source, as these have a bearing on the ability of high accuracy watch obsessives (us and presumably others that have read this far) to adjust and track the performance of their watches. Note that, when we state that a watch is 'user adjustable', we assume that the owner has the tools and ability to remove and replace the case back. We do not list information such as jewel count, other functions, etc. We do note incorrect or questionable information commonly seen in enthusiast literature/fora. Manufacturers are listed alphabetically. Pre ETA Omega is listed separately.

Thermocompensated movements made by Citizen:

Caliber: Citizen Crystron Mega (movement number unknown)

Technology: Single high frequency (4MHz) AT-cut crystal

Annual accuracy: ±3 seconds (We do not know if Citizen recommended a minimum wear time to assure performance to this specification.)

Rate adjustable?: Most probably yes, by the Service Centre in Japan. We do not know the specific adjustment mechanism.

Watches that use this movement:

Citizen Crystron Mega

Note 1: This is the tightest specification ever claimed for a watch!

Note 2: The battery lasted one year.

Note 3: Only 3000 were made. This is not surprising considering the 4,500,000 yen price.

Note 4: This watch is sometimes erroneously referred to as the Citizen Mega Quartz. The picture below shows the correct phonetic translation:

(Translation courtesy of Robert Logie. Photo from Citizen website.)

Source of information: The history section of Citizen Japan's website. (Reference 5 below.)

Caliber: A660

Technology: Single 32 kHz crystal (thermocompensation method not known)

Annual Accuracy: ±5 seconds

Rate adjustable?: yes, by the Citizen service centre in Japan

Watches that use this movement:

'The Citizen' (since 1995)

Note 1:This is the tightest specification among currently available watches. (Citizen does recommend daily wear for 12 hours to ensure best performance.)

Note 2: This movement has a perpetual calendar and independently settable hour hand.

Note 3: 'Chronomaster' was added to the dial in 2005, although the name of the line remains The Citizen.

Sources of information: Citizen sales literature, Citizen 'The Citizen' owner's manual (Reding)

Caliber A690

Technology: Single 32 kHz crystal (thermocompensation method not known)

Annual accuracy: ±10 seconds

Rate adjustable?: yes, by the Citizen service centre in Japan

Watches that use this movement:

Citizen Exceed (selected models)

Note 1: From pictures in Citizen sales literature of this and the A660 movements, the A690 and A660 are apparently identical. The crystal package in the A660 may be enhanced in some fashion, or it may simply be regulated and adjusted with more care.

Note 2: For the models that have a calendar, the calendar is perpetual.

Note 3: A number of models in this line use the Eco-Drive (solar) power system.

Source of information: Citizen sales literature

Caliber: E410

Technology: not known

Annual accuracy: ±10 seconds

Rate adjustable?: not known

Watches that use this movement:

Citizen Exceed (selected ladies models)

Note 1: This movement is clearly different from the A690, as it is smaller. Also, the date change is gradual vs. near instant.

Note 2: The model that we know about has a perpetual calendar and independently settable hour hand.

Note 3: The model that we know about uses the Eco-Drive (solar) power system.

Sources of information: Owner's manual (Reding), Citizen sales literature

Thermocompensated movements made by ETA:

Caliber: ETA 255.561 (from the ETA Flatline series)

Technology: Dual 32kHz/262 kHz crystals (thermocompensation method not confirmed)

Annual accuracy: ±10 seconds

Rate adjustable?: yes, via digital touch point (user adjustable)

Watches that use this movement:

Longines Conquest VHP (Longines caliber L.276) between 1984 and 1995

Krieger Marine Chronometer (approximately 1994 to 1997)

Note 1: Although not directly confirmed by ETA, we have strong reason to believe that this movement uses the digital count adjustment method.

Note 2: We believe that Longines participatd in the development of this movement, and that they were the first to use it.

Note 3: This movement has an independently settable hour hand.

Note 4: The 255.511 is a companion movement. Functionally, the only difference is that it does not have an independently settable hour hand. We suspect, but do not know, that the timekeeping portion of the movement is identical. We do not know of any watches that used the 511 movement.

Note 5: It is sometimes reported that variants of this movement had a 2.1 megahertz high frequency crystal. This is not confirmed, and the reports are questionable.

Note 6: The Krieger watches were offically certified as Marine Chronometers by Besancon.

Sources of information: Longines caliber number confirmed by Longines service center. The frequencies confirmed by Mr. Jean-Claude Robert of ETA. Functionality of the 511 vs. 561 from ETA technical information sheet.

Caliber: ETA 252.611 (ETA Thermoline)

Technology: single 32 kHz crystal (digital count adjustment method with thermistor)

Annual accuracy: ±10 seconds

Rate adjustable?: yes, by the watchmaker during service

Watches that use this movement:

Longines Conquest VHP Perpetual Calendar (Longines caliber L.546) between 1996 and 2002

Longines Flagship VHP Perpetual Calendar since 2002 (Longines caliber L.546)

Piquot Meridien Octanis Marine Chronometer

Note 1: This movement has an independently settable hour hand.

Note 2: This movement uses a 3 volt 10 year Lithium battery.

Note 3: The Longines Flagship VHP Perpetual Calendar is apparently no longer available, although Longines still lists it on their web site.

Note 4: The Piquot Meridien watches are officially certified as Marine Chronometers by Besancon. They are no longer available.

Sources of information: ETA sales literature, Longines sales literature, Piquot Meridien sales literature, Longines authorized service centre, Longines user's manual

Caliber: ETA 252.511 (ETA Thermoline)

Technology: single 32 kHz crystal (digital count adjustment method with thermistor)

Annual accuracy: ±10 seconds

Rate adjustable?: yes, by the watchmaker during service

Watches that use this movement:

Omega Constellation Perpetual Calendar (Omega caliber 1680)

Omega Constellation Double Eagle Perpetual Calendar (Omega caliber 1680)

Note 1: This movement has an independently settable hour hand.

Note 2: This movement is identical to the 252.611, except that it uses a 1.5 volt, 5 year battery.

Sources of information: ETA sales literature, Omega sales literature, Omega authorized service centre

Caliber: ETA 251.232 (ETA Thermoline)

Technology: single 32 kHz crystal (digital count adjustment method with thermistor)

Annual accuracy: ±10 seconds

Rate adjustable?: yes, by the watchmaker during service

Watches that use this movement:

Breitling Aeromarine Colt Chronograph SQ (Breitling caliber B73)

Breitling Aeromarine Avenger M1 Chronograph SQ (Breitling caliber B73)

Breitling Professional Emergency Mission SQ (Breitling caliber B73)

Note: Breitling has designated all its thermocompensated watches as 'SuperQuartz' (SQ) since2002. For reasons that are unclear, Breitling specifies its SuperQuartz movements to 15 seconds per year, even though we have no evidence that they differ from the ETA Thermoline models which are specified to 10 seconds per year.

Sources of information: ETA sales literature, Breitling sales literature

Caliber: ETA 956.152 (ETA Thermoline)

Technology: single 32 kHz crystal (digital count adjustment method with thermistor)

Annual accuracy: ±10 seconds

Rate adjustable?: yes, by the watchmaker during service

Watches that use this movement:

Breitling Windrider Callistino (Breitling caliber B72)

See note for ETA caliber 251.232.

Sources of information: ETA sales literature, Breitling sales literature

Caliber: ETA 955.652 (ETA Thermoline)

Technology: single 32 kHz crystal (digital count adjustment method with thermistor)

Annual accuracy: ±10 seconds

Rate adjustable?: yes, by the watchmaker during service

Watches that use this movement:

Breitling Aeromarine Colt Quartz SQ (Breitling caliber B74)

See note for ETA caliber 251.232.

Sources of information: ETA sales literature, Breitling sales literature

Caliber: ETA 956.652 (ETA Thermoline)

Technology: single 32 kHz crystal (digital count adjustment method with thermistor)

Annual accuracy: ±10 seconds

Rate adjustable?: yes, by the watchmaker during service

Watches that use this movement:

Breitling Aeromarine Colt Oceane SQ (Breitling claiber B77)

See note for ETA caliber 251.232.

Sources of information: ETA sales literature, Breitling sales literature

Caliber: ETA 955.452 (ETA Thermoline)

Technology: single 32 kHz crystal (digital count adjustment method with thermistor)

Annual accuracy: ±10 seconds

Rate adjustable?: yes, by the watchmaker during service

Watches that use this movement:

none known

Source of information: ETA sales literature

Caliber ETA 988.352 (ETA Thermoline)

Technology: single 32 kHz crystal (digital count adjustment method with thermistor)

Annual accuracy: ±10 seconds

Rate adjustable?: yes, by the watchmaker during service

Watches that use this movement:

Breitling Professional Emergency SQ (Breitling caliber B76)

Breitling Professional Aerospace SQ (Breitling caliber B75)

See note for ETA caliber 251.232.

Sources of information: ETA sales literature, Breitling sales literature

Caliber ETA E20.341 (ETA Thermoline)

Technology: single 32 kHz crystal (digital count adjustment method with thermistor)

Annual accuracy: ±10 seconds

Rate adjustable?: yes, by the watchmaker during service

Watches that use this movement:

Breitling Professional B1 SQ (Breitling caliber B78)

See note for ETA caliber 251.232.

Sources of information: ETA sales literature, Breitling sales literature

Thermocompensated movements made by pre-ETA Omega:

Caliber: Omega 1510, 1511, 1515, 1516

Technology: single 2.4 MHz crystal with AT-cut (lenticular disc shape)

Annual accuracy: ±12 seconds

Rate adjustabe?: yes, via trimmer condenser (user adjustable)

Watches that use this movement:

Omega Megaquartz 2400 (1974 to 1978)

Note 1: Omega stipulated that the watch must be worn daily to perform to the stated specification.

Note 2: Caliber 1510 was the first non-marine chronometer version of this movement. It was succeeded by the 1515.

Note 3: Caliber 1511 was the first Marine Chronometer version. It was succeeded by the 1516. Both were officially certified at Besancon.

Note 4: The hour hand is independently settable. Omega claims this to be the first watch to have this feature.

Note 5: Owners report battery life to be a year or less.

Note 6: One will sometimes see claims that this was the most accurate watch of all time. (Omega makes this claim on their web site.) However, a number of watches then and now surpass it in terms of specified performance.

Sources of information: The history section of Omega's web site and references 6 and 7 below.

Thermocompensated watches made by Rolex:

Caliber: Rolex 5035 (and 5055 for the Day-Date model)

Technology: single 32 kHz crystal using the forced constant frequency (TCVCXO) method

Annual accuracy: around ±60 seconds (Rolex has never stated an official accuracy specification.)

Rate adjustable?: yes, via trimmer condenser (user adjustable)

Watches that use this movement:

Rolex Oysterquartz (1977 to 2001)

Note 1: We are unaware of any other watch using TCVCXO technology.

Note 2: While discontinued, we have heard that Rolex plans to introduce a new quartz movement with update technology and perpetual calendar. More can be read about this in reference 8 below.

Source of information: reference 8 below.

Thermocompensated movements made by Seiko:

Caliber: 9923A

Technology: Dual crystal (The lower frequency crystal is 32 kHz. We do not know the frequency of the higher frequency crystal.)

Annual accuracy: ±20 seconds

Rate adjustable?: yes, via trimmer condenser (user adjustable)

Watches that use this movement:

Seiko Twin Quartz (first introduced in 1978, long discontinued)

Seiko King Quartz

Note 1: Seiko recommended that the watch be worn daily to achieve the stated performance.

Note 2: One sometimes sees the statement in enthusiast literature/fora that Seiko claimed a ±5 second per year performance for the 9923A. This is erroneous, as can be seen in Seiko's literature in reference 9.

Source of information: See reference 9 below.

Caliber: 9F

Technology: single 32 kHz crystal (thermocompensation method not known)

Annual accuracy: ±10 seconds

Rate adjustable: not known, but probably yes, by the service centre in Japan

Watches that use this movement:

selected Grand Seiko quartz models

Note 1: Seiko re-issued a limited edition of the 'Astron' in 2000 that used a special version of the 9F movement that was rated to ±5 seconds per year.

Note 2: The 9F movement is reportedly designed to run fifty years before it needs servicing.

Note 3: Other Grand Seiko quartz models use the 8J movement. Even less is know about this movement.

Source of information: Japanese retail sites

Caliber: 8F (4F for ladies versions)

Technology: single high frequency 196kHz crystal

Annual accuracy: ±20 seconds per year

Rate adjustable?: yes, by 'pattern cutting' (a non-reversible method)

Watches that use this movement:

Seiko sells many models with this movement.

Note 1: These watches are not thermocompensated. Seiko states that they should be worn 12 hours daily to meet the stated performance specification.

Note 2: These movements have a perpetual calendar.

Note 3: The GMT versions of this movement have an indpendently settable hour hand.

Note 4: Some models have a ten year battery.

Source of information: Japanese retail sites, Seiko authorized service centre, Seiko perpetual calendar (8F) service manuals. (See references 10 and 11 below.)

Caliber: Y301 (Y302 for date version)

Technology: single 196 kHz crystal

Annual accuracy: ±10 seconds

Rate adjustable?: yes, by pattern cutting (non-reversible)

Watches that use this movement:

Pulsar PSR-10 (by Seiko) no date, discontinued

Pulsar PSR-20 (by Seiko) date, discontinued

Note 1: These watches are not thermocompensated. Seiko states that they should be worn 12 hours daily to meet the stated performance specification.

Note 2: The caliber Y302 uses similar technology, but was rated to ±20 seconds per year.

Note 3: According to numerous reports, the PSR-10 often does not achieve its specified performance.

Part 4: References

The following references are general sources of information on quartz crystals, oscillators, and general timekeeping/frequency control:

1. The following is one of the best sources of information on quartz resonators/oscillators available on the web: target="_blank">

2. The following is an in depth discussion of aging in quartz oscillators:

3. The following is an excellent, concise description of quartz oscillators:

4. The following is a good general reference to topics relating to timekeeping and frequency control:

The following references are good sources of information on specific movements:

5. Citizen Crystron Mega: Babelfish can be used for translation.

6. Omega Megaquartz 2400: Exellent in depth information.

7. Omega Megaquartz 2400: . Good supplemental information.

8. Rolex 5035 Oysterquartz:

9. Seiko 9923A Twin Quartz:

10. Seiko 8F/4F:,33A,35A.pdf

11. Seiko 8F/4F:

12. Pulsar (Seiko) PSR 10/PSR20 caliber Y301/Y302:

The following links are excellent general sites related to this topic maintained by other high accuracy watch enthusiasts:





“Your heart was warm and happy

With the lilt of Irish laughter

Every day and in every way

Now forever and ever after."