Shimano Nexus/Alfine Inter-8 hub gear disassembled / reverse-engineered

On a beautiful day in 2011, when I visited my favorite Berlin bike builder and dealer on the Schöneberg Island, I noticed that he used the shell of a Shimano Inter-8 gear hub as an ash tray. When I asked him about the piece, he told me that one of his customers or friends had serviced the gear hub and not reassembled it correctly, making the ball bearings smash when he used it again.

Left is the reduction stage, right the 4-gear stage
(including carrier shared by both stages).

I asked for the innards of the hub which Conrad still had below a shelf in a corner of the shop. He gave them to me and this week I finally got around to look at them more closely.
I already knew that this hub is logically composed of two stages of planetary gears. One has two gears, the other four. Both can be switched independently to give eight gears and the ratios of the cogs are tuned in such a way that the eight gears have roughly equal spacing and no gear overlap. (So logically they work like the 3×8 gears of a 24 gear derailer system, only that all the gears are shiftable in sequence.) After disassembling the hub, I counted the teeth of all the cogs and used the counts to calculate the gear ratios. I was happy to find that I just got the same results that are published on several places of the Interwebs. For those who can't wait to see them, here's my calculations. All others can first read on to find out how this stuff works. Actually, those who've never read anything about planetary gears, go start with my introductory blog post on planetary gears to learn some of terminology. Then come back to find that I am experimenting with different terminology here (after all, this is a work in progress and I need to figure out what works best). For now, let's say “cog” to the little toothed round devices that turn inside the gear hub to translate speeds. Then we can use the word “gear” for the resulting ratios of speed, just as you ordinarily do when you say “first gear, second gear” and so on.

Now, here's a summary of the hub's architecture: since both stages share the same planet carrier, the carrier serves as the one and only power transfer between the two stages. The carrier is therefore output of the first stage and input of the second stage, which conveniently serves the fact that the first stage reduces speed, while the second stage increases speed. In both stages the sun cogs are locked and the ring cog serves as the other moving part. (In the second stage, there are three sun cogs locked via switchable one-way ratchets.) One-way ratchets (aka freewheels) are used in many places in the hub. Thanks to the freewheels several gears can be switched “active” at the same time which has two key advantages. One, when shifting gears, the mechanism only needs to activate or inactivate the higher gear, while the lower will automatically be inactive (freewheel) without being actuated by the shifting mechanism. The second advantage is that during shifting (or when there's a problem with the shifting cable), there is never a no-gear stage, since the lower gear will always be active as fall-back in case that the higher gear doesn't engage properly.

Reduction stage with two planet sets.

A first surprise I found when looking at the hub innards was that the reduction stage has two planet sets to create just one reduction ratio. I had previously assumed that there was only one set, that is, one sun cog, a couple of planet cogs all engaging with the sun cog, and one ring engaging with all the planets. But in reality, the three planets are two cogs in one: a smaller cog that engages the sun and a little bigger one, that engages the ring. I should actually have thought of that because with a single planet set, it is quite hard to achieve an 89% gear ratio as this stage of the hub does. (The planet cogs would need to be to tiny, since their size approaches zero as the ratio nears 100%.)

The power from the chain and cog is transferred to the ring cog of the first stage via a freewheel. To bypass the first stage (as in gears 5 to 8), a clutch connects the incoming motion directly to the planet carrier. Since the ring gear always turns faster than the carrier, it will be then freewheeling with respect to the input motion. (Inversely, when the clutch is disengaged and the ring is driven by the bike chain, then the carrier will turn with the translated speed, so it's important that the clutch is completely disengaged, since the gears would otherwise jam.) From experience during about 10'000 km riding a similar hub (and by what I've heard from others), switching this clutch is the most difficult and noticeable switch in the hub. It's the one between fourth and fifth gear. Also the one that first becomes edgy when the shifter cable is misadjusted. (Happens rarely and is easy to fix.)

Finally, let's look at the second stage which as I said is driven by the carrier and transfers its motion to the hub shell (and thus the wheel (and thus the road (and thus the earth, which makes is turn))) via the ring gear or (if in direct drive) via the carrier. Both ring and carrier connect to the hub shell via a freewheel so that the faster one will be driving while the other will be in fallback mode. Interestingly, direct gear (1 and 5) is achieved by locking none of the sun wheels which makes them turn freely, making the planets turn freely and the ring not being driven, so the carriers wins the race to drive the wheel. One last factoid: the ring cog sits on the smallest of the planet sets (with the biggest sun cog) which when locked makes the biggest gear (4 and 8). Locking the smallest sun cog, gives the smallest (non-direct) gear (2 and 6). The jump from direct drive to this gear is the biggest in the entire hub (22% compared to just 14% from gear 3 to 4). I think the reason for this is that the sun cog can't become much smaller than it already is because the ratched mechanism for locking it still needs to fit inside. (And the solid hub axle is still inside the ratched mechanism.)
I actually disassembled the hub one step further than the manual explains. I only figured this out by accident after removing one more stop ring (shown on the right in the very first picture) when suddenly the shafts holding the planets of the second stage fell out of the carrier. When I tried to make them all fall out, the shafts holding the first stage's planets also fell out from the same holes! :-D After taking out the first stage planets, the ring cog from the second stage could be removed since it was only held by the planets and then the planets from the second stage could be removed. The sun cogs are still inside the carrier. I can see and feel another stop ring inside them, but didn't try to take it out. I instead just counted the sun cogs teeth by marking one tooth and then turning the cog until the marked tooth came up again. It's really interesting how some parts are purposely held by stop rings while others (such as the second stage ring gear) are just floating on other parts (second stage planets) and held sideways by again other parts (first stage planets). I am a bit curious whether I could put that part back together again. Maybe some day... ;-)
Here's my calculations again for those who I made curious. I've already got a follow-up post in my head in which I'll explain the formula that the spreadsheet is using.

Sources:

• Shimano service manual (PDF)
• Same manual in German (pictures are better)
• Inter-8 how it works with tables and pictures. (German)
• List of gear ratios for popular hub gears