5S21 and 5S42 are Seiko's attempt in the 1990s to give us sweeping quartz. The second hand sweeps smoother than a Hi-Beat Grand Seiko and appears as smooth as Spring Drive or Accutron while still being quartz.

The original paper 《低刻音薄型長寿命掛時計用ムーブメントの開発》—馬場康治 describes the working principle of this movement. It's fun to read. AI can OCR and translate it for you if you can't read Japanese. It turned out the smooth second hand movement is a happy side-effect of the original intention to develop a silent and long battery life clock movement.

Honestly, readers interested in how a 5S21 movement works can simply read the above paper and maybe the technical guide of 5S21 and that'd have been sufficient. Here, I am going to inspect into an example of 5S21 I received that would give us more insight into the mechanism, as with many other posts on this site, when things don't go as smooth (heh) as they should.

It all started with this rather handsome looking 5S21-7A30 I received. When I dropped a new SR920SW in, it was running. By the way, the same battery model as Seiko's contemporary smooth quartz movement VH31. Both are 4Hz too.

The movement was clean enough. I was just curious and performed a service anyway.

Once I have removed the dial, I was surprised to find the bottom of the viscous rotor unit cracked. No dampening fluid lack on this side of the movement observed though. I thought this was ok and cracked on (ahem) with the disassembly.

Ok, there is indeed a leak (red circle, top-right corner). Just that the thick dampening fluid found themselves comfortable staying near the viscous rotor unit instead of spreading throughout the whole movement. Had they found a way into the hairspring wheel I would have given up then and there as I have no idea how to clean the hairspring wheel without breaking it.

At the time, I had no idea how this would affect the running of this movement. See, the reason the second hand appears to sweep smoothly is because when the step motor rotor turns 4 times a second, the energy is stored into the hairspring for a gradual release regulated by the viscous rotor unit. It didn't seem obvious but both the intermediate wheels mesh with the fourth (second hand) wheel so the second hand turns at the rate of the slower wheel of the two.

The more energy stored, the viscous rotor would turn faster, and vice versa. Assuming the viscous rotor works as designed, there would be enough friction in the viscous rotor to strike a balance with the 4 ticks per second and the second hand would move in a comfortably sweeping motion.

If for any reason, the slope of the straight line representing viscous drag is to change (e.g. dampening fluid leak), then the range where fourth wheel speed varies would have changed. For example, let's say it's a horizontal straight line i.e. no variation in dampening moment in relation to torsional movement of coil spring, then the rotary speed of fourth wheel would vary across a wider range due to a faster release of torsional movement of coil spring.

As a result, the movement of the second hand would resemble a VH31, where it's 4 sharp ticks instead of a continuous motion.

Interestingly, looking at the original paper, the reason for this isn't to have a smooth second hand sweep, but to reduce the jagged movements of the wheels, which is noisy. We get smooth second hand sweep as a side-effect.

So what happens to this 5S21-7A30 unit? There's clearly a leak at the viscous rotor unit and a substantial loss in dampening fluid. According to the above, it should appear to tick at 4Hz like a VH31, as if the viscous rotor doesn't exist.

Oh yes it does. The effect is more pronounced with naked eye, where the speed variation isn't sufficiently dampened by the viscous rotor unit, so the second hand would move with high velocity and slow down quickly until the next step rotor tick.

Comparing this and the perfect sweep on a good 5S21 the difference is as clear as day. Human brain is just too good at registering the stuttering.

But when put side-by-side with an SP68 (left) and VH31 (right), 5S21 (middle) still has a smoother sweep, as the viscous rotor unit still provides a level of resistance rather than an instantaneous unwind of the hairspring when there isn't a viscous rotor unit.

In order to add more friction to the gear train, I re-lubricated the movement side pivots with Moebius 8200 on top of 9010 that's already there on top and bottom. Best case it addes enough variable friction it fixes the issues. Worst case the difference in friction between top and bottom pinions twists the pinions broke, we'll see.

5S21 is a fun watch movement. It's designed to be easy to service and if the viscous rotor unit wasn't broken and I didn't bend the second hand at one point, I am sure this 5S21-7A30 would look amazing after a service.

The good news is, even then, this watch is still functional and looks handsome.