What is a good beat error for a watch?

A good beat error is one inside the movement maker's limit, and the one maker that prints the figure sets it at 0.6 or 0.8 milliseconds, depending on grade. Beat error is the difference in duration between the balance's swing one way and its swing back. A timing machine shows it as the gap between two traces, in milliseconds. Zero is the ideal, and no movement holds it in every position and state of wind.

The question exists because the numbers in circulation disagree. Owners read that anything over 0.5 ms is a fault, see 0.7 on a new watch and wonder whether to send it back. The figures below come from documents that state a limit, set against one that does not, followed by the conditions the reading must be taken under.

Beat error limits by movement

Sellita prints a beat error limit in the escapement table of each technical document, next to the rate and amplitude limits. The term is easy to miss because the English column mistranslates it. The French says repère max. and the German Abfall max., both standard trade words for beat error, but the English reads "adjusting mark" in the SW200-1 document and "adjusting max" in the SW210-1. Search a Sellita PDF for "beat error" and you will find nothing.

Three things stand out. The first is that the figure holds across a decade of Sellita documents. The 2014 manual-wind SW210-1 and the 2024 automatic SW200-1 both allow 0.8 ms in their working grade, and only the dearer grades of the SW200-1, adjusted in five positions instead of three, tighten it to 0.6.

The second is that 0.8 ms is not a Sellita quirk. Archer, a watchmaker who answers service questions on OmegaForums, wrote in June 2024: "The beat error tolerance is typically anything less than 0.8 ms for most brands, so that's not alarming." That is one professional's summary, not a document, but it lands on Sellita's figure.

The third is Seiko. Its guide for the 6R35 labels the stud support "Beat error adjustment" on its regulation diagram, so the part is plainly there to be used, and then tests the movement on rate alone: plus or minus 10 seconds a day dial up, 15 seconds a day in the two hanging positions, all at full wind. A Seiko at 1 ms that meets those rates is inside everything Seiko publishes. Judge it on rate and positional spread instead.

How to take a reading you can compare

A limit means nothing unless the reading is taken the way the maker took it. Sellita's notes to the escapement table spell out its conditions, and Seiko's 6R35A guide gives its own.

Turned into steps for a home timing machine reading:

1. Demagnetise first if the rate looks wild. A magnetised hairspring sends you chasing the wrong fault, as the case below shows.
2. Wind fully, then wait. Let the watch run for an hour or more before you read it, which puts you inside Sellita's 1 to 3 hour window.
3. Lay it dial up. That is the only position Sellita specifies for beat error.
4. Give it 20 seconds, then average 40. Read the averaged figure, not the number flickering in the first few seconds.
5. Leave the complications off. Stop the chronograph and keep the date from changing while you measure.

The lift angle setting does not matter for this reading. Beat error comes from the timing of the ticks themselves; lift angle only enters the amplitude calculation, so a wrong setting (see what lift angle to set) spoils amplitude and leaves beat error alone.

Know your instrument's step size, too. Witschi's Watch Expert III displays beat error to a resolution of 0.1 ms, and its Chronoscope M10 production tester states an accuracy of plus or minus 0.1 ms for the same reading. So 0.5 and 0.6 ms are one count apart on a professional machine, and 0.0 means "under the last digit," not perfection.

Why the same watch gives different numbers

Beat error is a time, but the fault behind it is an angle. The balance's rest position sits slightly to one side of the line through the escapement, so the balance crosses the point where it unlocks the escapement early on one swing and late on the other. That offset stays fixed until someone touches the hairspring. How many milliseconds it costs depends on how fast the balance passes the middle of its swing, which depends on two things.

Amplitude. A balance swinging wider passes the middle faster, so the same offset costs less time. Near the centre of the swing, the reading is close to inversely proportional to amplitude. A watch that reads 0.6 ms at 300 degrees will read about 0.9 ms when the mainspring has run down to 200 degrees, with nothing in the movement having changed. This is why Sellita measures at full wind, where amplitude is highest and the reading lowest. A reading taken the morning after a day off the wrist can sit well above the maker's figure on a watch that is in specification.

Beat rate. A slower balance also passes the middle more slowly. For the same offset and amplitude, a 21,600 vph movement such as the 6R35 or the NH35 reads a third higher than a 28,800 vph one. Sellita's 0.8 ms on a 4 Hz SW200-1 corresponds to about 1.1 ms on a 3 Hz movement, and its 0.6 ms to 0.8 ms. Our beat rate converter gives the frequency of any movement you are unsure of.

Neither effect changes the watch, only the number. If a reading is borderline, take it again at full wind before deciding anything.

What to do with the number

Taken dial up, an hour or more after a full wind, on a 28,800 vph movement:

For a 21,600 vph movement, multiply the boundaries by four thirds: about 0.8 and 1.1 ms in place of 0.6 and 0.8.

Beat error is not what makes a watch gain or lose. A watch can sit at 0.0 ms and run 20 seconds a day fast, or at 1 ms and keep time to a few seconds. A large beat error shows up in starting instead: the movement may stall low in its reserve or fail to restart, one of the causes in our answer on why an automatic watch stops overnight.

Correcting it means turning the hairspring collet or moving a stud carrier such as the 6R35's stud support. Ask for it by name, because "please regulate it" buys a rate adjustment only, as our answer on whether a watch can be regulated explains. It is usually done within a full service, priced in our answer on what a watch service costs.

The mistake that makes it worse

The beat adjuster sits next to the regulator, which makes it tempting to use on a rate problem. A thread on OmegaForums from May 2023 shows where that leads. The owner of an Omega Cal. 564 found it running "very fast +300 with a beat error of 0.6". With the regulator at its limit and the watch still about 150 seconds a day fast, they moved the beat lever until the rate came down to +4, and "of course the beat error has increased to 7.3ms."

The cause was magnetism. Two days later the owner wrote that "after de-magnetising it tonight I could regulate the HS to -1 to +3 in different positions with a BE of 0 to 0.3." A watch gaining minutes a day is the classic sign of a magnetised hairspring, covered in our answer on whether magnets can damage a watch. Neither adjuster can fix it. The order of work is demagnetise, then read beat error, then think about rate, and never trade one against the other.

When these limits stop applying

• The watch is vintage or from another maker. Sellita's figures describe modern 4 Hz Swiss calibers. Slower movements read higher for the same fault, and another maker may set its own figure. Seiko publishes none, and the ETA technical communications we read (2892-A2, 2801-2 family) carry no timing table at all.
• The watch is a chronometer. The seven COSC criteria are all rate criteria, with no beat error among them. Sellita's own Chronomètre grade carries the same 0.6 ms as its Top grade.
• A new watch reads above 0.8 ms. Sellita's limits are applied per delivered lot: 95% of the pieces must be inside them, so up to 1 in 20 can leave the factory outside. It is a fair warranty question if the watch also stops or keeps poor time; a reading alone is a weak case.
• The watch has no swinging balance. Quartz watches and Spring Drive have no tick and tock to compare.

Read beat error alongside the other two numbers on the screen. Our answer on what a good amplitude is gives the makers' limits for the second, and our guide to timing your own watch puts all three together. On beat error alone, the makers' answer is short: 0.6 ms if you want the best grade, 0.8 ms to pass, at full wind and dial up.