Is a quartz timepiece always more accurate than a mechanical one?

I have an old Omega watch (1964 Constellation chronometer) that I keep serviced and maintained properly. It keeps accurate time, to the point that it seems that my wife has to adjust her watch, a modern, but basic quartz watch more often to keep it right than I do.

Now, my old watch had the reputation of being one of the most accurate in the world in its day. I understand that modern mechanical / automatic watches are even more accurate now. However, I’d always heard that quartz is ALWAYS more accurate. I know that the science behind the quartz crystal timing is sound, but can it be let down by cheap components in a substandard modern watch?

So basically, is my wife’s watch a lemon? Or will a crappy cheap quartz watch always keep better time than a top-notch mechanical watch?

It’d be rare for a mechanical watch to keep better time than a quartz one that isn’t broken, but it can certainly happen. Shame on the manufacturer that made the quartz one, though.

A quartz crystal will oscillate at a specific frequency forever, as long as some electrical juice is passed through it. It is simply a fact of the structure of crystals.

No matter how finely crafted or maintained, a purely mechanical watch is an assembly of fine parts relying upon one or several springs to do the rhythmic oscillation work.

Over time, thin bits of metal ( or in the case of huge tower clocks, very large bits of metal ) do wear and lose their ability to…er…spring back into place.

As long as the components in a modern quartz watch are functional to the point where they can convert the oscillations of said quartz crystal into useful data points and display them, they won’t wear out.

Amazingly, this is the same principal that allows your heart to be beating as you read this post.


However, Quartz crystal oscillators exhibit both temperature sensitivity and aging. If these errors are not taken into account, they can keep poor time. I’m using an embedded controller whose realtime clock varies about +/- 2 minutes / month when used in an uncontrolled temperature environment.

>A quartz crystal will oscillate at a specific frequency forever, as long as some electrical juice is passed through it. It is simply a fact of the structure of crystals.

What beowulf said. Moreover, there are all kinds of further issues and opportunities for trouble.

The temperature sensitivity is a big issue. Generally, there are various methods to improve it. One is cutting the crystal at angles that minimize the temperature sensitivity, but those angles aren’t the ones with the highest Q (which would have the most stable frequency). Another is adjusting things in the electronic circuit that shift the oscillator frequency. Quartz crystal oscillators are often run at frequencies slightly away from their resonance, to compensate for the temperature (as in a TCXO or temperature compensated crystal oscillator), and to hit a specified frequency more closely.

And it isn’t true that the crystal oscillates “as long as some electrical juice is passed through it”. You have to build a circuit that oscillates, and include the crystal in the circuit, and let the resonance of the crystal dominate the oscillation. Though, to the extent that you are tuning the frequency, other components in the circuit have to have some control over the frequency too.

Crystals age for several reasons. They accumulate contaminants that make them heavier, so they oscillate more slowly. The electrodes (generally fine electroplating on a face of the cut crystal) gradually become detached due to stresses at the electrode-crystal interface. There are also sudden shifts in crystal frequency (not sure if anybody knows why this happens).

And there are other reasons crystals aren’t ideal. For example, quartz wristwatches often use a crystal cut in the shape of a tuning fork oscillating at 32,768 Hz so they can use a 15-stage binary divider to create a 1 Hz signal. But 32,768 is a pretty low frequency to try to make quartz vibrate at. It’s much more accurate to make the crystal vibrate at higher frequencies, maybe 1000 or more times higher, by cutting different shapes and inducing different vibration modes.

There are compromises all over this beast. It’s not as if quartz has some kind of natural frequency built into it (like cesium and rubidium atoms do in atomic clocks)!

According to “Fundamentals of Time and Frequencye” by Michael A. Lombardi, National Institute of Standards and Technology,, the Shortt pendulum clock, which had a master pendulum and a slave pendulum, had a frequency uncertainty of 0.1 parts per million on a 24 hour basis. He says the Harrison H4 spring and balance wheel chronometer of 1759, which allowed navigational measurement of longitude by sighting astronomical objects, had a 4 ppm uncertainty.

This current catalog page from DigiKey, a very popular electronics vendor, shows quartz crystal oscillators for sale. They all have either 50 ppm or 100 ppm frequency tolerances. This is not even close to the pendulum clock or the balance wheel devices above.

So, certainly, some quartz crystal timepieces are less accurate than some purely mechanical timepieces, even given that they are all functioning and not broken.

Only empirical data but over the past thirtyfive+ years I have had quite a few quartz watches, both LCD and analogue mechanical quartz driven ones.

They have all been remarkably accurate up until they stopped altogether, mainly from severe abuse. The current one, a cheap Lorus, has been in use for seven years now, looks like shit but only ever needs adjusting to conform with daylight saving nonsense - every six months then.

They clearly do not operate in ideal conditions, lots of variation in temperature, orientation and mechanical shock. The current one is the third of it’s kind I have owned, chosen for simplicity, durability and accuracy.

I wonder if there isn’t some form of feedback loop built in to them to compensate for drifting due to temperature, or whatever.

Purely mechanical watches, albeit inexpensive ones, rarely were accurate to better than a minute a month and rarely lasted more than six months anyway.

Ed: Typo.

Most quartz movements are designed to be cheap, compact and have a long battery life. Very few are designed to be accurate.

>A quartz crystal will oscillate… It is simply a fact of the structure of crystals.

This isn’t correct either. The issue is whether the substance is piezoelectric or not. While many piezoelectric materials are crystals and many crystals are piezoelectric, there are nonpiezoelectric crystals (Google “nonpiezoelectric crystals” to see), and there are materials such as polyvinylidenedifluoride (a flexible plastic sold, often in sheet form, under the trade name “Kynar”) that are piezoelectric but not usually thought of as crystals. PVDF is crystalline in the recently redefined sense (by the International Crystallographic Union) of having a diffraction pattern in X-Rays, at least if you isolate a region of it that is oriented homogeniously, but none of us looking at rolls of the stuff are going to call them “crystals”. Bone is another example of a material we wouldn’t usually call a crystal that is piezoelectric.

The 50 to 100 ppm includes initial manufacturing tolerances which are calibrated out for the pendulum clock. If you subject the quartz clock you will find that the frequency drift is very low. If you calibrate the crystals you will find that they drift very little over temperature ranges that the mechanical clock operates.

>The 50 to 100 ppm includes initial manufacturing tolerances which are calibrated out for the pendulum clock. If you subject the quartz clock you will find that the frequency drift is very low.

Fair enough. This might be true for many XOs. Bur for the first quartz oscillator I found when I went to that manufacturer’s web site, the “frequency stability” is - 100 ppm to + 100 ppm. See first site below. Another model I looked at offered better stabilities of ± 50 and ± 25 ppm. Some models specifically said things like “Electrical specifications include Supply voltages of 5.0V and 3.0V with a frequency stability of +/- 30 ppm over temperature range of -10 to +60C.” This sounds to me like the drift with temperature is an important part of the total, and therefore still large compared to 0.1 or 4 ppm.

I found another manufacturer, Oscilent, whose web site offers the following explanation.

“Stability: The most standard Stability for a Crystal Oscillator is 50ppm. The Stability is most dependant on the Operating Temperature Range, and, hence, 50 ppm over 0-+70 degrees celsius in most common. 15-20 ppm is normally considered the tightest production spec available, but special accommodations are possible for tighter specs.”

So, it looks like dozens of ppm can be typical for the stability in particular, even if it does not include the initial manufacturing tolerance.

This is an excellent web site explanation of quartz crystal vibrational modes, cut angles, frequency dependences on temperature for different cut angles, frequency drift from aging (e.g. 1 to 5 ppm per the first year for metal-can packages and 0.1 to 1 ppm for glass packages), and all sorts of other excellent stuff. Very very nice. The frequency stabilities are often dozens of ppm over different temperature ranges, though some can be much better. Note that hitting a particular frequency curve is itself a manufacturing effort with tolerance on it, so that the strength of the dependence of frequency on temperature is itself a distributed variable.

I’ve been making quartz oscillators sound too bad. Lots of them aren’t meant to be so precise, but ones for watches in particular are.

“An obvious simplification would be to use the bare
quartz-crystal oscillator and to dispense altogether with the
stabilizer based on the atomic transition. The stability of
such a device is clearly the same as an atomic-based system
at sufficiently short times. The stability at longer times will
depend on how well the quartz crystal frequency reference
can be isolated from environmental perturbations, or how
well the effects of these perturbations can be estimated and
removed.2 Quartz-crystal oscillators can be surprisingly
good in this respect. Even cheap wrist watches have oscillators
that may have a frequency accuracy of about 1 ppm and
a stability 10 or 50 times better than this. The residual frequency
fluctuations are largely due to fluctuations in the ambient
temperature, and stabilizing the temperature can improve
the performance.”


Interesting stuff. My experience with quartz crystals is not limited to timepieces. Motion picture cameras hold synch by using quartz crystals. They are routinely taken from extreme heat to extreme cold in the space of days or weeks.

In 28 years as a film camera operator, I never never once heard of someone needing to replace the crystal due to heat aging. Or simple aging, though I am going to call Panavision New York in a few hours when they open up and get the straight dope on whether or not crystals are EVER replaced.

I do know that they can be damaged or go bad. Unfortunately I shot a documentary with a 16mm Arriflex BL with a bad crystal. It could not hold synch for more than 20 seconds.

It is possible that traditional quartzes are not used any more to hold synch. More about this later today.

I would guess that the tolerances in film are pretty loose, compared to the possible aging in a crystal.

You mechanical clock will not operate at 5 ppm over 0 to 70. And these numbers still include initial manufacturing tolerance. If they did not you would need at least two numbers. One for initial tolerance and another describing other variations.

>You mechanical clock will not operate at 5 ppm over 0 to 70. And these numbers still include initial manufacturing tolerance.

Gazpacho is right to point this out; the comparison is not perfectly fair this way. I think a nice balance wheel timepiece, though, might pretty reasonably be sold in “individually calibrated” condition, and be used over only a narrow temperature range. I’m looking for examples of high accuracy in balance wheel clocks and low accuracy in quartz clocks, as counterexamples to what is otherwise overwhelmingly typical: quartz timekeeping is generally way more accurate than balance wheel timekeeping. I’d put my money on quartz every time. But, if somebody thinks they are observing an example of a quartz timepiece keeping poorer time than a mechanical timepiece, including examples where one is pampered more than the other, they certainly could be correct, and it doesn’t necessarily mean the quartz timepiece is broken.

I designed and manufactured quartz crystals for nearly thirty years. Never lost my fascination with these devices.

If anyone has a real interest in learning a little something about quartz crystals, take a look at this.

LouisB, fantastic link.

So tell me about building the crystal oscillator product I’ve wanted to build for years.

What I want is the oscillator that will allow the best estimate of time possible, assuming you have access to a microprocessor and memory and an oven. Or maybe there’s an ovenized version and a nonovenized version, but still only two products.

I want the target frequency to be whatever makes it easiest to cut a crystal with predictable aging and frequency. I don’t care if it’s kHz or GHz. Whatever’s got the best combination of Q and predictable instability.

I don’t want to hit target frequencies exactly. let Q be as high as we can make it. My microprocessor will learn what its frequency is. Don’t trade away anything to tweak the frequency - microprocessors are better at multiplication than oscillators are, anyway.

I don’t necessarily want to minimize aging. Any aging that is predictable, the microprocessor will perfectly take care of. What I want to minimize is the uncertainty in the aging.

If a heat pump that lets you operate at cold temperatures lets you keep the crystal in a more stable regime, and one that ages more predictably (probably because it ages less), then the ovenized version doesn’t run hot, it runs cold. Peltier heat pumps retail at $10 or so in some places, and that’s big ones. Splurge!

So, what do you think of this product? You could make a more predictable timepiece if you moved some of those functions to the microprocessor, and made the oscillator best at oscillating, couldn’t you?

I think it’s a wonderful idea. I guess nobody will want to buy one.