If you remember upthread I mentioned that one of the slightly crazy aspects of the (W) hubs is that the sun pinions are not only biased rightwards, but they must move under the action of this spring if you are to get all five gears. This means that failure of the biasing spring means failure of the hub. And biasing spring problems are not unknown. To rectify this fault you are expected to purchase the 'Atlas assembly'HSA675 to fit (W) five speed hubs, rear (LHS) uppermost
The spring, which would cost pennies by itself, is buried within this assembly which is riveted together with six chunky rivets. This wouldn't be so bad but (bafflingly) the Atlas assembly is only slightly cheaper than a complete internal is. Not so much an 'Atlas' assembly as an 'Alas' assembly...
Having several Atlas assemblies with duff/no springs (at least one came out of a faulty hub in pieces) I wondered if there was a cure. I contemplated trying to feed a new spring into the assembly, but reckoned that there was no way of doing it without distorting the new spring. Also, not only did I not know exactly what the new spring should look like, I didn't know if parts of the old spring were still inside in every case or not.
I decided that the thing would have to come apart, and that meant drilling the rivets out. There are six rivets, with large heads at the rear (LHS) and smaller ones at the front. Since one repair option is to convert the rivets to bolts, and there is probably more space (or could be) on the LHS for bolt heads, I decided to drill from the LHS; this might also make the rivets easier to drift out, since the atlas assembly will sit nicely over a bench vice one way up but not the other. It wasn't too difficult to do the drilling.
The rivets have shanks that are 4.4mm diameter and inside the assembly I found a spring that had either been made wrong from the start or had seen some weird loads (the possible nature of which I don't understand at all) and had become crushed.
I decided to try and reshape the spring;
The spring above appears to be made in wire which doesn't have full temper, in that I was able to reshape the spring without undue difficulty. On reflection I am thinking the 'deformation in service' hypothesis is probably correct; the spring diameter had evidently increased because the spring didn't have an integer number of turns (which spring designers tend to avoid) and the diameter was both uneven and larger than would comfortably fit in the recesses which the spring sits in.
The spring is (AFAICT) ~4 turns, ~0.9mm wire diameter, and about 19mm OD. The length.... I'm not so sure about. I shall try about 12mm length and compare the force on the planet gears with a 'known good' example. I shall come up with a plan to turn the rivets into threaded fasteners, and probably I shall have to grind pockets into part HSA676 to make room for the fastener heads.
But I am slightly optimistic that the Atlas assembly can be repaired this way.....we shall see if my optimism is misplaced or not.