There is an art to this and it's not for the inexperienced builder. Also, many people don't have much time on a Bridgeport milling machine: and if someone else has a better way I'm all ears, even creaky old wooofers can learn new tricks (if there's a treat involved) –
If you had prints it would be a lot easier. Since these things are sixty years old, there should be some floating around. But if you don't have 'em forge ahead …
99 chances out of 100 the first surface they created when doing the original machining was the middle case split. So, If you welded on that casting, first thing I'd do would be to throw that case half on a flat surface and check that it is still flat. A surface plate is nice but float glass works almost as well and it is cheap. Run by your friendly neighborhood glass shop and have them cut you a 2' x 2' piece of 3/8” plate and sand the edges.
Throw the case half on it and see that it's flat. If it's only a little bit rickety then throw some fine valve grinding compound on the plate and lap it flat. May as well throw the other half on and do that one too while you're working up a sweat. Cases that don't leak are nice.
If your part is way warped, better talk to a guy who does a lot of aluminum welding about straightening it before you even start, not gonna go into that here.
So, you've lapped the case half and it's now flat as Twiggy's chest, grab the best indicator you've got and square up the head on the Bridgeport. Then do it again and make it better. If the table is older and worn down in the middle, consider picking up a piece of tooling plate for a sub-plate and facing it off. Where the cuts overlap you should see witness marks but take off no metal.
Oh wait. I forgot to mention stone the table first. Just a light foop-foop over the table, knock down any nicks or burrs.
Diversion into indicators : get the best one you can afford. No, get one you can't really afford. My favorite ever was an Alina that they don't Make any more, long lever, two inch dial that was built like a tank. ACCURATE !! You could read it from across the room. Each division was big. The needle didn't glitch all over. It didn't waffle around like a cheesy Central p.o.s. It actually saved time and grief because you ran that thing around a hole or down an edge and you could TRUST it. That's an important tool to have ; upon the indicator everything else depends, get a really good one and baby it. You'll be glad you did.
So now you've got a chance at keeping things square and flat and perpendicular and all that jazz. This is like top-level important. If this is wrong, nothing can come out right. EVERYTHING HAS TO BE RIGHT TO GET GOOD ACCURACY, AND THAT INCLUDES THE FLOOR IT SITS ON.
At this point I'd take a skim cut off the outside edge of the cam chest. That surface is supposed to be parallel to the centerline and by now it most likely isn't. Plus it's going to be your clamping surface. Take as little as possible but there's room in there to lose a few thou.
Now you can flip it back over onto the table or your subplate or maybe a set of 1-2-3 blocks. I love 1-2-3 blocks. For the amount of metal you are planning to remove, I wouldn't worry about supporting the transmission but that's up to you. Can't hurt, as long as you don't 'oversupport' it out of parallel.
Here's where you get to start guessing. Not wild-ass guesses, supposed to be educated guesses. Right?
It looks to me like the shafts in the tranny are vertically aligned. If you can just go off a trap door, skip this but if there's nothing there and you want to try to replicate what HD did, then I'd make a spud that fits the hole for the crank. Drop the case half on that (crank side facing up), snug it down lightly, then indicate across two of the cam bearing holes. You can get to those from the inside. Bang it round a little around the spud to get them aligned with the axis of the table. If you line the cases up crosswise against those holes, then chances are real good that you've got the cases lined up in X the same as the factory intended. Tighten that hummer down then run back to the tranny.
The advantage to doing this even if you don't have to is that when you come to check on the center distance between the tranny bearing bores, you won't have to be doing sines and cosines. I hate those. I like my numbers straight, not all kinky. Even if we are working on a tranny …
Pop on the other case half (you lapped it, right ? So it should be nice and flat and go on and off smoothly). Check the mounting surface for the trap door. Parallel ? Better be. I shoulda done this first. Move it up to the “before” step. Luckily we're doing this on paper. Did you ever hear that a good machinist is not one who never makes mistakes, it's one who can do a good job fixing his mistakes ?
If it isn't parallel, make it be. Take off as little as possible.
If you have a known-good door, slip it on (hope the dowels aren't too tight, you want it to be snug but still go on and off without hammering). Indicate the mainshaft in the trap door and indicate it in the cases.
If not (probably not, you just welded on that thing so most likely it's warped). Decide what you want to do. I would say that, since you have made all the mating surfaces good and flat and you welded the area right next to the mainshaft, if it's off then it should be fixed. You can bush it pretty easy, it's a steel insert cast into the cases, yes ? Seems to me you don't want to go really thin but a sixteenth wall thickness should be okay … if you've got the room, an eighth would be better.
I like a shoulder on bushes like this but maybe there is no room. And they are more work and take up space. Another way to lock it is to drill half in the bush and half in the parent metal then tap for a little setscrew. Loctite the screw in, chop the head off, bob's yer uncle that bush she ain't gonna go noplace.
The trap door is more likely to be correct than the cases, you didn't weld on it. Sleeving that bearing is going to be a bit of a pain in the ass but a crooked bearing is a bad deal and you already went to all this work … shafts that spin straight and true are good.
PARALLEL. That is the MOST IMPORTANT THING.
The correct distance apart. This is also important but there is some room here. Gears are designed to have backlash. With a 14 1/2* pressure angle, the center distance change is about twice the number for the backlash. That is, if you have a pair of gears that are tight at 6“ centers and you want .003 - .005” backlash (that's about what I'd guess for Sporty gears) then you spread the center distance twice that - .006“ to .010”. So the real world center distance is only half as touchy as you think. 20* pressure angle is about one-and a-half times instead of two times (these are rough numbers, if you want to get right on the nipples grab a copy of Colvin & Stanley, Gear Engineering (I think, brain ain't what she used to be.) C&S is not full of arcane stuff, just useful formulas for figuring backlash, tooth thickness, o.d., p.d., worthwhile things. Machinery's Handbook has that stuff also - if you're going to work in metal, go buy a used five year old copy for cheap, everything you will ever need is in there but the print is pretty small for geezers.
In this case since the gearbox is driven by a chain, you really don't need to sweat the center distance from the crank. That's nice, gives you leeway in your bearing placement. And the shift mechanism is mounted to the trap door, not the cases, so no worry there. AND, because of that, I'd tend to make the trap door my defining locations. The shift shaft is a pretty sloppy fit in the tower, should work fine even if you move the gear locations by ten or even twenty thou.
If you gotta known good trap door, at this point just indicate the bores from the trap door then cut them into the cases. Viola, you're ready to rock. While you're in there, face off any thrust surfaces so that they are flat when you put a thrust washer up against them.
Speaking of which, apologies again if you already know this but for boring in welded aluminum (ugh, gooey stuff), I liked my square Criterion boring head and a high speed steel boring bar. Unless you have the diamond lapping thingy, you can't get carbide sharp enough for this kind of work. And if you have that slow-wheel diamond lapping thingy, I should be listening to you instead of the other way 'round. You could rough the hole out with a two-flute center-cutting end mill, or drill it first with a 1/4“ drill then the end mill. Those center-cutting end mills don't center cut as good as they say, a quick drilled hole up the middle is a big help. WD-40 is okay for this.
On the other hand, if you don't have a known-good trap door, you can still do the educated guess thing. Since the only critical thing here is parallelism and to a lesser extent shaft centers, you can be off quite a bit from HD location and it will still work exactly the same. You can measure what you do have and those numbers are probably crappy but will give you some starting info. Just remember that you if you get too far off you'll have to mess with the shift tower thingy, too. It's unlikely that you'd get that far off but just don't forget about it entirely.
These gears are DP (inch). You can be pretty sure of that because everything in a fifties HD is inch Even without the little metal gear tooth gages that roll with the teeth. Do all your numbers in inches.
Gears need backlash. We know that. In this size box they would probably go for .003 - .005” or more. These are 20* pressure angle (freeby info from me but you can check it with the tooth gages if you like) so the center distance change to get that would be about five to ten thou. If they were smart they went to the big side because a little extra lash doesn't hurt anything but binding gears in a somewhat feeble (the gears aren't 2“ wide on 2” shafts, they are gonna deflect under pressure) housing would be bad.
There are two ways to get backlash. You can spread the shafts or you can cut the teeth deeper. In general people choose to spread the shafts. It makes calculations easier. But sometimes they will cut the teeth deeper. It depends on the situation and the smarts of the designer. In one case you add a few thou to a single center distance. In the other you recalculate ten gears. If the designer worked in a shop he goes with Plan A. If he just came out of school and likes the pretty even numbers on a drawing and doesn't care who has to suffer for it, he goes with Plan B.
You can double-check this in a crappy way by sticking some gears on the shafts and throwing a set of verniers over the shafts. Hold them tight together. Subtract half the diameters of the two shafts and that's your no-backlash center distance. Add your ten and there ya go.
In some cases you can triple check your assumptions by throwing the verniers over the bearing bores then adding half the diameters. But in this case one bearing is a needle with no inner race so you're kinda screwed. And the bearings in the output side case are at the bottom of a hole For something rare and expensive you could make a little test fixture but maybe not needed for a Sportster …
Now that you know your guessed center distance, and knowing that the thing is chain-driven, you could put the bearing bores wherever you wanted, always remembering to keep the thing square to the shift tower.
So, if you are crazy and you really like the short-wheelbase Sportster, you could rotate that gearbox 30* I bet, cut out a half inch or more of the transmission housing, re-weld, and make a super-short wheelbase very Sporty Sportster. I wonder why they made that thing so long in the first place ?
About size, the fit is specified in the bearing book, you can look it up. I always try to go to the tight side 'cuz it's goin into aluminum which expands with heat but that's just a personal choice. Other people probably have a different opinion.
If the countershaft is just a teeny bit tight in the bearing, you could either leave it alone - figuring that with heat it will get better - or you can throw it in a lathe and run some extra-fine grit emery paper over it. Option #2 sounds bad but in real life works okay. You are not really taking enough off to make it out of round, you are just manually super-finishing it.
I will admit that I tend to make things too tight. For race bikes take all my numbers and double them. Race bikes don't like to be tight. They like to be just about ready to fall apart, cuz that's when they run best. But for a street bike, seems like smooth and quiet and long-lasting are good qualities. So if you like things looser, no argument from me. But KEEP IT PARALLEL !! That really does matter.
In clearancing the closed in bearing, you should be able to get a half thou reasonably easy. Use sharp tools with a little radius on the corner and go slow. No prob.
Do you have readouts ? That makes it a lot easier. They are not God, btw, don't believe them that you are actually going to hold .0002“, they don't, but it's better than indicators. And if you get the shafts parallel within a thou, it'll be great. Better than new.
Have at it, just be careful and think first, do second and it'll come out good.