Britbike forum Forums Projects started by Britbike forum members Members bike projects BSA 6-Spring Clutch

Note added 17 May 2018: The 6-spring clutch was used in tens of thousands of BSAs produced for over a decade starting in the late 1940s indicating that it is a fundamentally sound design. Despite that, after a half-century of abuse, wear, and incorrect rebuilding, today it has a bad reputation among some people who advocate "upgrading" to a Triumph 4-spring clutch. However, if you follow the instructions in this thread to return it to as-new performance you will have a 6-spring clutch that doesn't slip under power, doesn't drag when disengaged, allows neutral to be easily selected when stopped, and requires no more force on the lever than is required by a modern bike with a hydraulic clutch.
--------

This began as a thread in the BSA forum where some additional background information can be found:

Briefly, what initiated these threads was that the clutches on my BB and Catalina Gold Stars required much more force to operate than that of my Special Competition. Both of the former machines were built by the same man and I had not had the primary case off of either since acquiring the BB a year ago and the Catalina six months after that. Since the Catalina was already on a hoist I started with its clutch. However, at that point I found there is very little in print or on the web about 6-spring clutches, and that some of the information that does exist is incorrect.

Note that BSA produced a few variations of the six-spring clutch so, in addition to wear, it's possible components currently in a clutch have been mixed over the years. Also, to avoid possible confusion, in what follows I've used terminology from the BSA Parts manual rather than common U.S. descriptions (e.g. Clutch Actuating Cap instead of pressure plate). I've also capitalized the names of components to make them stand out.

What follows will be roughly in the order of Disassembly, Measurements and Refurbishing, Upgrades, Reassembly, Special Tools, and Adjustment. I expect it will take about five or six posts before reaching the end.

The Gold Star manual calls for setting the spring tension by having one thread showing above the lock nut. When I opened the primary case I found the builder had set the tension at roughly that value although he had drilled the nuts for lock wire instead of using double-nuts to maintain the setting.


While wear can add to the problem, the primary reason for a stiff clutch is that BSA's recommended spring adjustment appears to be based on a worst-case scenario of making sure the clutch wouldn't slip even if accelerating up a steep hill while carrying a heavy pillion passenger. Since this situation doesn't apply to how I use my Gold Stars, after refurbishing the components and incorporating a few upgrades I can adjust my springs with lower tension without worry of slipping.

After removing the Nuts, Springs, and Spring Cups the Clutch Actuating Cap comes directly off allowing the five pairs of Clutch Driven and Clutch Driving Plates to be fished out.



I placed the Actuating Cap and the ten Plates on a surface plate and found all of them to be flat. Had any of them not been flat, or had there been insufficient friction material left on the Driving Plate inserts, I have a number of others in a box so I could have found replacements. Absent that, I would have removed as much of the warping as possible with a press and/or hammer and then made a jig to mount plates on my lathe to skim them.

To remove the Mainshaft Nut requires a tool to lock the Driving Plate Centre to the Chainwheel. Although the 6-spring clutch requires a different locking tool than any of my other clutches, luckily I have a large stock of used plates so it was easy to make one.


I made the handle on this tool long enough to catch on the footpeg so I could concentrate my force on the nut.

Note: you have to be careful when pulling the Chainwheel off if you plan to reuse the present bearings because some of the tiny balls might come free from the cage and roll away on the floor unless you have placed a small pan or rag to catch them.


Once you've removed the Chainwheel you likely will find wear grooves in the slots like those in the following two photographs.



Ridges in the slots in the Chainwheel, as well as any in the Driving Plate Centre, would cause difficulty in disengaging the Driven and Driving Plates so they need to be eliminated. An air file is the fastest way to smooth the faces of the slots while keeping them flat although it could be done with a hand file. In addition to making the slots smooth so as not to "catch" on the tangs, any material projecting toward the center of the clutch also has to be removed since it also could "catch" on the edges of the plates. The Driving Plate Centre seems to be less prone to such wear but it needs to be inspected as well.

If the tangs on the Driven and Driving Plates look like the following they need to be smoothed because they can cause difficulty with clutch operation. Again, an air file is a fast way to eliminate the burrs and to square up the faces that engage the slots.


[to be continued]

After removing the Clutch Driving Plate Centre I used a 1-3/4" length of 1-1/2" ID scrap Al tube held with the Mainshaft Nut (reversed back-to-front) to press the Clutch Plate against the Clutch Sleeve. I found the total indicated runout (TIR) at the edge of the Clutch Plate to be a rather large 0.037". Ideally, this should be 0.000" so the cause needs to be located and eliminated (or reduced).



To check if the excessive runout was caused by a defective or worn Clutch Sleeve I measured the runout of the lands on the Sleeve and found it to be several thou. This may sound small but because the lands are at such a short distance from the center of rotation it will result in much larger TIR at the edge of the Clutch Plate.



The necessary puller for the Clutch Sleeve has a female 1-3/16 26 tpi thread. After removing the sleeve, the next photograph shows that the tops of the lands of the Clutch Sleeve look pretty good. However, visual inspection isn't sufficient to know whether or not the imperfections are large enough to cause, or contribute to, the 0.037" runout I measured 3" away at the edge of the 6"-diameter Clutch Plate. Since the Clutch Plate is pressed against the lands of the Clutch Sleeve any burrs (or manufacturing defects) on these lands will tilt the Clutch Plate and cause the Clutch Plate to wobble.



At this point I checked the TIR of the end of the mainshaft and found it to be ~0.004. That is, the end of the shaft is bent by ~0.002". However, in addition to this the Clutch Sleeve itself might have problems so I needed to accurately check it as well.



I put a good mainshaft in the 4-jaw chuck of the lathe and adjusted it to get the TIR to below 0.0002"



After pressing the Clutch Sleeve on the shaft I measured the lands and found them to be the same height to better than 0.0005". So, this part doesn't contribute to the excessive runout of the Clutch Plate.



The next photograph shows the portion of the Clutch Plate that is clamped against the Sleeve. Since the area between the tangs is clamped against the tangs of the Sleeve, if it is not parallel to the face of the Plate it will cause runout at the edge of the Plate.



I found the face of the Clutch Plate to be flat on a surface plate so it provides a good reference surface to use for machining the area that contacts the Sleeve. I clamped the Clutch Plate to a face plate on the lathe, centered the Plate, and turned the burrs from the tangs as well as the important area between the tangs that mate against the Clutch Sleeve. I had to remove something like 0.003" from this area to get it flat.



After machining the center of the Clutch Plate I reinstalled it and the Clutch Sleeve on the gearbox main shaft and measured the TIR at the edge of the Plate in each of the 6 positions. I then stamped the position that gave the minimum runout (0.010") using the keyway in the shaft as the reference as well as marked this position with red paint.



Although the red paint may dissolve away with time, it should ensure that I at least install the plate correctly when I reassemble the clutch this time. Although not necessary, I then mounted the Plate with the opposite face out, machined away the burrs projecting from the tangs, and used a Scotchbrite pad to clean up the face.

Again, the end of the mainshaft is bent so even if the Sleeve and Clutch Plate are in perfect condition there will be runout at the edge of the plate. This means I have three choices at this point:

1) Bolt the clutch back together as-is.

2) Put the Plate on my mill with a 0.010" shim at the appropriate edge and machine the center with that slight tilt to compensate for the bent mainshaft. If this were done the entire clutch would rotate offset by 0.005" from the center line while remaining parallel to the rotational axis.

3) Rebuild the SCT gearbox using an unbent mainshaft.

Of these three, my initial pick was choice #4. This 1962 Catalina should have the one-year-only ASCT gearbox. As luck would have it, many years ago I bought just such a gearbox to use on my '57 Spitfire until/if the correct SCT2 turned up (which it eventually did). Twice since then I've offered the ASCT for sale. After all, what were the odds I'd ever buy a '62 Catalina and, if so, that it wouldn't have the correct gearbox? If nothing else, the fact I now "need" it confirms my practice of never selling anything.

I took the ASCT down from the shelf and measured the runout of its shaft. Unfortunately, it was a huge 0.0.017" which I quickly identified as coming from the bearing. I could push the shaft up and down to the sound of clicking. I already have a stock of new gearbox bearings so I ordered the necessary oil seal and gasket. Depending on when various parts shipments from around the world converge at my house I'll either rebuild the ASCT now and install it, or accept the 0.010" runout in the SCT for the time being.

While rebuilding the ASCT and installing it in the Catalina is the right thing to do (as would be rebuilding the SCT with a good mainshaft), the fact the bearing in the ASCT was so trashed means I could find other serious problems once I open it up. So, I'm leaving my options open as to how I will proceed once parts needed for upgrading the clutch -- to be described later -- start being delivered. No matter what, mis-machining a perfectly good Clutch Plate to partially compensate for a bent mainshaft isn't an option I would pick.

[to be continued]

maybe have a go at straightening the bent shaft?. find point of max bend between centres on the lathe, mark this point, put shaft on V blocks and press straight? worth a go.
Excellent so far, the eccentricity check was a good test, no clutch would work properly with that amount of wobble.

That's a good suggestion. The reason it's not on my list of options is I'm blessed with a wealth of gearbox (and other) parts that were accumulated by a hoarder after my own heart.

To unbend the mainshaft would require dismantling the gearbox so if I were to do that I first would look to see if my hoard contains a good mainshaft to use instead. But, even then, I might as well dismantle the correct ASCT, repair it, and swap it for the SCT. There's also a combination fix to consider, of swapping the internals of the ASCT for those of the SCT while the latter is still in the bike. But that assumes I find the needle and ball bearings in the SCT housing to be in good shape, as well as crossing my fingers and hoping the seal is as well.

If I had only one option (e.g. unbend the shaft) my decision on how to proceed would be easy. "Unfortunately," I'm forced to weigh multiple options. Poor, poor, pitiful me...

The necessary oil seal arrived yesterday, I already have a small stock of the various gearbox bearings, and I have a 30 ton press, so all options are currently on the table.

Proceeding with the clutch, it's easy to assume all 6 tangs of each of the Driven and Driving Plates engage with the mating surfaces of the Chainwheel and Driving Plate Centre. However, thanks to wear and original manufacturing tolerances, full engagement doesn't happen.

The least worn Driven Plate in my horde is also by far the most rusty, but it still nicely illustrates the point about engagement. At the left of the next photograph I have the Driven Plate in the lowest position in the Driving Plate Centre and rotated as far to our left as possible where it is making contact with the Centre. However, simply raising the Plate to the top position it no longer makes contact. This happens because one or more of the other tangs has made contact with the Centre first and stopped the Plate from rotating any further to our left. This means this vertical post of the Centre is either tapered toward the top due to wear, bent slightly to our left, or was manufactured this way.


As long as enough tangs of the five Driven Plates make contact to carry the load asked of them this lack of full engagement of all 6 tangs isn't necessarily a problem, but it was worth pointing out (the same issue with engagement is the case with the Driving Plates and the Chainwheel).

Turning next to the springs, I modified a unit sold for testing valve springs by substituting a 60 psi gauge because it is more appropriate for clutch springs than is the 300 psi the unit was supplied with. I also installed the gauge "backwards" so, in addition to testing clutch springs in a vise, I could easily read the force when using it to push on components such as the end of the clutch actuating lever. The piston has an area of 1 sq.in. so the gauge reads directly in pounds of force. The gauge I used has an accuracy of +/-3% of full range, i.e. +/-1.8 psi. Although that is fine for these purposes, greater relative accuracy can be obtained for determining spring constants if the same pressure value is used for all and the decrease in spring length at that fixed pressure measured (rather than vice versa). This is how I made my measurement, by compressing each spring with 50 pounds of force.




One of the springs from my clutch is on the left. Although the free length of all 6 are nearly the same, between 1.04" and 1.07", I measured the spring constants to be 204, 218, 230, 232, 240, and 246 lbs./in.. This ~20% variation will affect the operation of the clutch. I found a set of six like the one on the right that have the same ID and OD as the one on the left, but they are shorter at ~0.92" and significantly weaker at ~140 lbs./in. Since these would fit in the clutch it points out the necessity of making sure all the components in a clutch are actually the ones that should be there.

At this point I ordered six new 66-3800 springs from British Cycle Supply since even if they turned out not to have identical spring constants they would let me mix and match with the ones I have to produce a set with the smallest variation. The springs arrived, all were between 0.996" and 1.034" in length, and all had spring constants of 211 +/-7 pounds/inch. This variation of +/-3.3% is three times smaller than the current springs so they'll work fine.

If installing a new race in the Chainwheel it is especially important to check the clearance between the bearing race and the Clutch Plate. If the race makes contact with any part of the Clutch Plate it hasn't been installed correctly and the race needs to be pushed further in. For my clutch the clearance is roughly 0.040" between it and the ridge in the Clutch Plate marked with the red arrow, the portion that comes closest to making contact with the race.


While not having the race touch the Clutch Plate is necessary, it's important to have the center of the race located as precisely as possible in the same plane as the sprocket teeth since only then will there be no sideways torque on the Chainwheel. Actually, it's the centerline of the two rows of balls in the bearing that need to be precisely centered in the Chainwheel, but it's good not to have the race too far off from that. On my Chainwheel the bearing is properly centered to 0.002", which is within the uncertainty in the series of the individual tolerances and measurements that have to be made to determine this.

Since I will be working on at least two more 6-spring clutches in the months ahead I took the time to machine a gauge for checking the location of the race, which needs to project 0.084" above the back face of the Chainwheel. This is in the next photograph, along with 0.003" and 0.004" shims I cut from shim stock positioned where they would go if the bearing were too high or too low:


I bought two ea. bearings, inner, and outer races from Draganfly [Update: read next post to see why I do not recommend these bearings] to use their upgraded bearings in my Catalina and BB Gold Stars. Dimensions of these races, the Clutch Sleeve, and the other relevant clutch components are below for easy reference. When ranges are given these are the low and high values I measured at different positions. I spent a reasonable amount of time with calibrated instruments to obtain these values so I expect both the relative and absolute accuracies are within +/-0.0001" on dimensions that are given to a ten-thousandth:

OD of my Clutch Sleeve: 1.3745-1.3748"

Inner bearing race:
-- ID of BSA race: 1.3763-1.3765"
-- ID of Draganfly race #1: 1.3744-1.3745"
-- ID of Draganfly race #2: 1.3735-1.3747"
-- Width BSA race: 0.5308-0.5319"
-- Width of Draganfly race #1: 0.5390-0.5392"
-- Width of Draganfly race #2: 0.5360"-0.5370"

Chainwheel bearing race
-- ID of BSA race: 1.8992"
-- ID of Draganfly race #1: 1.8942-1.8948"
-- ID of Draganfly race #2: 1.8945-1.8948"
-- OD of BSA race: 2.2532-2.2540"
-- OD of Draganfly race #1: 2.2546-2.2548"
-- OD of Draganfly race #2: 2.2552-2.5554"
-- Width of BSA race: 0.4098-0.4107"
-- Width of Draganfly race #1: 0.4084-0.4087"
-- Width of Draganfly race #2: 0.4083-0.4085"

Note that the ID of the Draganfly Chainwheel bearing races are significantly smaller (~0.0.005") than that of the (worn) BSA race. With bearings tenths matter, let alone thousandths, so unless the Draganfly inner race tracks compensate this could be a problem.

Length of Clutch Springs: 1" (nominal; range approx. 1"-1.03")
Length of Spring Cups: 0.5" (nominal; range approx. 0.503"-0.507")
Diameter of balls in BSA and Draganfly bearings: 3/16"
Length of pushrod: 11-3/8"

[to be continued]

In my last post I wrote that the ID of the Draganfly Chainwheel races were ~0.005" smaller than that of the BSA, but whether or not that was a problem depended on the inner race. Well, unfortunately, it's worse than a simple issue of too little clearance (which could have been solved by honing the outer race).

As for whether or not the design of the Draganfly bearing is "improved," I can see no functional advantage to their design that has one race holding the balls rather than two separate ones. So, even if it were a good bearing, it wouldn't be "improved."

But, aside from the layout, their cage is aluminum which is not appropriate for an application that has the top and bottom of the cage rubbing against steel, as well as each of the holes rubbing against the rapidly rotating steel balls. However, worse than the choice of construction materials is that the holes in the cage are drilled for two tracks 1/4" apart but the inner race was mis-manufactured with the tracks ~1/64" too far apart. This can be seen in the next photograph where I have marked the sides of the tracks with purple to make them easier to see and aligned the left track with the holes on the left of the cage.


As can be seen, with the cage centered over left track the right track is too far to the right. As a result of the mismatch, the cage forces the balls up onto the shoulder of the tracks resulting in reduced or negative clearance, excess friction, and increased wear of the balls and races due to the poor lubrication at the elevated temperature.

But, unfortunately, it's even worse. You might have noticed what looks like furrows plowed into the race where the balls roll. They actually are furrows. From the regular spacing of the grooves it appears the inner race was made on a CNC lathe rather than ground with a stone, and the tooling marks are such that the rms roughness (Ra) I measured is ~10x worse than it is allowed to be, i.e. it's 78 micro-inches vs. the 8 micro-in. or less that it needs to be for a bearing race.



The roughness of the outer race Draganfly sold me isn't quite as bad, but it is still ~3x too high at 26 micro-in. In service the roughness of the inner race would chew up the balls fairly quickly, leaving the owner who was not equipped with microscopes and surface roughness instruments wondering what he did wrong.

In case anyone is still left wondering at this point, I do not recommend people buy Draganfly's "upgraded" race. So, does anyone know who still has a stock of genuine NOS BSA items?

Here is what I believe the whole essence of this forum is about.
A learned discourse based on practical experience posted for the benefit for us all. Thank you, MM.
As I have said to another on Britbike, who I hold in the highest regard..... you'll do( for a Seppo !)

Looking at the collection of parts in your photos they all look very worn.I admire your work to help others to show potential problems associated with these clutches.I have only had one good working 6 spring clutch and that was fitted to a B33 that had been stored for a very long time and was mainly unmolested and abused, the rest of the 6 springers I have had made very good frisbees. Personally I would bite the bullet and fit a four spring clutch with correct adaptor, I have four spring clutches fitted to my WM20 & my old ZM20 both were fit and forget with a very light clutch lever pull.. Good luck in you quest.. Dave

The hoard came from a guy who never threw anything away. But, the examples I showed aren't representative of what I have, but rather are worst-case items I picked to best illustrate the effects of wear.

I wrote to Draganfly about the problems with their bearings, and included a link to my previous post for them to have more information and to see for themselves the issue with misalignment and surface roughness. This is the full text of their response:

Thank you for your email and comments concerning the bearings.

We are sorry that you are not happy with the quality of the parts. These are the best that we are able to supply and do not have any replacements. The 65-3910 clutch ball cage we have made by a trusted and reputable engineering company. We sell approximately 80 each year and we are not aware of any concerns raised.

We are happy to issue a refund upon receipt of the parts.

I have to wonder how many of the complaints about the functioning of 6-spring clutches are from people who rebuilt theirs using aftermarket components made by Draganfly's "trusted and reputable" company. However, despite Draganfly's assurance that they aren't aware of any concerns with these bearings I'll continue to trust my own eyes and instruments.

The issue isn't so much the ~$60 each I paid for them, or the $10 in shipping costs I'm going to lose out of this, it's that the unacceptable "quality" of these aftermarket parts wasted my time trying to install them, and they have set back my rebuild by at least a few weeks as I now had to start over to find bearings that actually will work.


p.s. Although built to a price, and subject to wear and abuse, I see no fundamental design flaw in the 6-spring clutch. Just the fact alone that for over a decade BSA installed 6-spring clutches on tens, or even hundreds, of thousands of machines including Gold Stars indicates there's no reason they can't be made to work well. However, a half-century of wear, abuse and neglect, coupled with faulty aftermarket parts, means there are things that must be paid attention to. With this thread I'm trying to pay attention to all those things.

I suppose its all relative, the old 6 spring freewheeling bearing looks like a far superior item than the later 3 spring cush drive offering which is a bit of a joke by comparison. The bearing in question has fairly light duty, maybe thats how the RAF Draganfly offering gets away with it.
The pictures are very illuminating,I wonder how hard the tracked races are? You would expect a ground finish.

A picture really is worth 1000 words. I could have quoted dimensions for the race and cage to the nearest 0.0001", and reported how many micro-inches of roughness there was, but all one has to do is simply look at those pictures to know there's a problem. Of course, I only have the two I bought from them to measure so maybe the other 80 they sold this year were properly made...

My tester only goes up to 60 Rockwell, but the OD surface of the larger race is at least that hard so that's one thing they did right (although I didn't measure the inner race nor the hardness of any of the balls). Of course, the harder the furrowed race is, the faster it will grind away the balls, so the fact it might be properly hardened in no way compensates for the mis-machining.

It's possible the inner race is roughed out on a CNC lathe and then sent to grinding as the next step before hardening, and that my two missed the grinding step. Of course, missing that step would be inconsistent with them having been "made by a trusted and reputable engineering company."

Turning to some good news, I inspected my old inner race and it's in good shape with no sign of Brinelling so I can reuse it if needed. I have two NOS outer, and one inner, races coming to me from separate sources, and NOS bearings coming to me from British Cycle Supply. I pressed the Draganfly Chainwheel race out this afternoon and have both sets ready to box up and mail back to them to refund the ~$120 I paid for that junk. Less the $10 in postage, time wasted with these bearings, and delay caused by having to find proper bearings. Not that I'm complaining...

On the recommendation of others I bought the components for SRM's clutch pushrod upgrade. Thankfully, I had better luck with the quality of this than I had with Draganfly's bearing "upgrade." (Note: unfortunately, I discovered a few months later that the quality wasn't what I thought it was when I wrote the previous sentence).

----------Note added three months after writing this post:
Measurements I made on another SRM pushrod assembly found the roughness of the two bearing surfaces to be 3-4x higher than 20 microinches that Torrington specifies as the maximum allowed for this type of thrust bearing and, ahem, roughly the same as that of the Draganfly bearing race. The component that bolts into the pressure plate is 65-70 microinches (perpendicular to the grinding marks) and that of the pusher is greater at 75-80. This can be seen in the next photograph:

Although the location of this bearing in the center of a spinning disk means lubrication will be minimal, degrading the lifetime no matter what, it still should benefit from a few minutes spent polishing the surfaces with a fine India stone.
-------- end note ----

To install the SRM Plate Adjuster requires removing the present "button" in the Clutch Actuating Cap. I clamped the Cap to the table and milled all but a few thou. from the top of the "button" and then tapped it through with a punch.



I then used the 1/4" hole to center the Actuating Cap in the lathe. Centered this way, the outer rim was Concentric with this hole to within ~0.01". This will be fine since the 6 vertical "fingers" of the Driving Plate Centre passing through slots in the Actuating Cap are all that centers the Cap when it's part of the clutch mechanism. Although the Actuating Cap is somewhat flat in the center it still is a stamping so, rather than simply drill the hole larger and tap for the 3/8-24 thread of the SRM Plate Adjuster, I first skimmed the mounting surface to make it truly flat as well as parallel with the clamping surface. I had to remove ~0.006" to accomplish this.



After drilling and tapping it 3/8-24 UNF I screwed the Plate Adjuster tight using the broached hex hole SRM conveniently provided for doing this and attached a stainless lock nut on the outside, using red Locktite for both Fasteners. However, later I milled away most of the external nut to be sure it wouldn't interfere with the clutch cover.



The stock push rod is 11-3/8" long. Subtracting the lengths of the SRM parts it should be shortened to 9-7/8". However, since I could remove more material much easier than I could put it back I cut it to 10" and tested it. The 10" length seems good but I'll check it again once I have the clutch fully assembled. The pushrod has hardened ends that are slightly dimpled to match the balls at either end. After cutting the rod only one end will be hardened so it is important that end goes against the actuating arm.



In addition to reducing the length of the rod I straightened and polished it as well as cleaned out the bore in the mainshaft using a gallery brush and solvent. The decrease in friction after doing this was quite noticeable so it was well worth doing.

I had hoped to use the original race in the Chainwheel but a magnifying glass revealed Brinelling at several locations on one of the tracks. After I removed the race from the Chainwheel I photographed the worst of it at two magnifications in a microscope.


Although this is the worst damage, two other places in the same track are almost as bad. There was no choice but to replace it.

Since I also would be rebuilding 6-spring clutches in the BB, Spitfire, and possibly the Special Competition in the future I decided it was worth the time to make some special tools. The next photograph shows drifts for removing and replacing the race in the Chainwheel. The piece at the left supports the Chainwheel from the inside in my 30T press while the drift in the center pushes the old race out from the back side. The tool at the right supports the Chainwheel from the back side while the drift in the center pushes the new race in from the inside.


I machined the center portion of the piece at the right lower than the rim by 0.084" so that when the drift pushes the race in it automatically ends up in the correct position. From left to right these are made from 3-1/4"OD x 2-3/4"ID x 1-3/4" long tubing, 2-1/4"OD x 1-1/4" long rod, and 3"OD x 2" long rod.

The inner race is close to a fit on the Clutch Sleeve so the two drifts shown below deal with this. The drift on the left pushes the race on and the one at the right pushes it off.


These are made from 2"OD x 1-1/2"ID x 1" long tubing and 2"OD x 1-1/2" ID x 1-1/2" long tubing.

Unfortunately, at this point I hadn't discovered the defective "quality" of the Draganfly replacement bearing and races so I installed their outer race in the Chainwheel. I later removed it with the drift in the next to last photograph to ship back to England for a refund and as of this writing I am waiting for a proper NOS BSA race to arrive so I can install it. The time I spent making the special drifts and fixtures wasn't wasted.

[to be continued]

Hi MM,

When you have the length of the pushrod finalised,
The cut end should be re hardened
I'm sure you know how
The pushrod is made from silver steel, known as "drill rod" in USA

John

I'll harden and temper it because it's easy, but it shouldn't be necessary because thanks to the bearing the SRM Pusher won't rotate against the end of the rod.

As an update, the Draganfly bearings are now on their way back to England with $22.50 in postage attached to the box. That, plus the L3.95 I paid to have them shipped here means this little debacle cost me ~$27.50 (I'm my own shipping department and obviously can't cut as good a deal as they can with postage). But, hey, it's only money, right? And time. That's ~3 hours of my life that I won't get back, and a several week delay in getting the Catalina's clutch working again. But, looking at the glass half full, I'll regard the $27.50 as tuition for being re-taught the important lesson that, where aftermarket parts are concerned, caveat emptor.

I have noticed the lowest prices on items I get from the States to UK comes with Pitney Bowes marked on the envelope, its lower by a considerable margin, 40 to 50%. Last package came in at $5 and had electronics board inside.

Having a postal machine allows someone to buy postage in bulk and get a discount, as does registering with the post office to do Customs and print forms on line. All of that takes time, of course, which would be worth it if I did this frequently.

After assembling everything the Actuating Cap will need to be adjusted to lift evenly, and slight unevenness and roughness of the top surface will interfere to some extent with accomplishing this. Even though the slightly bent mainshaft will contribute 0.010" runout at the edge, that can be compensated for by indexing the Chainwheel to it because the lift will be perfectly vertical with respect to the clutch plates when the properly-indexed runout is 0.010". So, since there's no reason to squander even a thou. of lift from this clutch, I made a 0.004" facing cut on the outer ~1/4" of the surface of the Actuating Cap for later use with a dial test indicator when adjusting the Cap.



The various holes punched in the Actuating Cap were pretty rough and a test fitting showed a few of the slots dragged a little on the "arms" of the Driving Plate Centre.



I marked one of the slots and an arm with paint so I could install it in the same orientation each time and hand fitted it to the Centre using a file on the slots. This made a noticeable improvement in the smoothness of engagement and disengagement.

The NOS outer race I found thanks to a PM from chaterlea25 arrived showing its age in the form of a dull finish, so I put it in the lathe and polished it with 0.2 micrometer/8 microinch abrasive, the same as a modern roughness specification for a race. However, the actual measured roughness was 16 microinches (also the tracks of the inner race).

To polish the tracks of the inner race I made a jig from 1-1/2" Al so I could spin it in the lathe.


The turned-down portion is slightly shorter than the width of the race allowing the bolt and "washer" to clamp the race to the jig.

After polishing the inner (top) and outer (bottom) races they looked like the following:


I didn't take the time to make an accurate measurement of the ID of the outer race prior to pressing it into the Chainwheel, but I did check it before and afterwards with calipers. Since the difference was small the accuracy should be pretty good. In any case, the ID of the race decreased by exactly 0.0050" according to the calipers (resolution of last digit +/-0.0005").

With the bearing situation under control the steps to reassemble the clutch are:

-- Press outer race into the Chainwheel
-- Install the two sets of bearings and races on the inner race
One has to be careful here because the orientation of the both sets of bearings and cages in the race is very important. Looking closely at the next photograph the two sides of the cages aren't identical, with one side closed and the other partially open. The top image shows the bearings correctly oriented, and the bottom incorrectly oriented.


If you install the two sets of bearings in the race with the closed sides of their cages side-by-side it will force the balls further apart than the tracks in the race. This will force the balls part way up the side of the track and result in zero or negative clearance when inserted in the Chainwheel race. They have to be installed with the open side of one race next to the closed side of the other, or with the open sides facing each other and the closed sides to the outside of the assembly allowing the cage edges to serve as a shield against any large pieces of debris.

-- Insert Clutch Sleeve in Clutch Plate, making sure they are indexed as previously marked to minimize runout
-- Install inner race on Clutch Sleeve
-- Insert the Sleeve/Plate/bearing assembly in Chainwheel
-- Place assembly on mainshaft
-- Install the Driving Plate Centre
-- Install the Locking Washer and Mainshaft Nut and torque the nut to 65 ft-lb.
The lock washer in my clutch has the rim slightly raised on one side. Be sure to install this with the raised side up to make it easier to bend into position later.
-- Bend the Locking Washer
Note -- do not do this until making the final assembly; see below. When it is time to bend the washer I found a 90o screwdriver helped along with an adjustable wrench to pry the edge of the washer up enough to grab it with Channellock pliers to complete the bend.

At this point the Chainwheel spins freely, i.e . it travels more than a full turn with a fairly gentle flip of the sprocket teeth. I also measured the amount by which the Chainwheel moves straight in and out on the races. On my clutch this movement was approximately 0.04". Play in the bearings was enough to allow the outer edge to rock by 0.090".



-- Starting with a Driven Plate, insert the five pairs of Driven (metal) and Drive (friction) Plates
-- Install the pushrod (shortened, if being used with an SRM bearing upgrade)
On my clutch I made the initial fitting with a 10" pushrod but found that the 9-7/8" length I originally had calculated for the rod would be better so I trimmed 1/8" from it.

-- (Install the SRM Pusher and bearing if this upgrade has been incorporated)
Check the total motion of the pushrod. In my clutch it was 0.130". This is the maximum amount available to separate the plates, with compression of anything along the way (such as the clutch cable) reducing it.



-- install the Clutch Actuating Cap
-- install the six Spring Cups, Springs, and nuts (tight, to take up all slack)

The first time through the above steps is only a test fitting because having done all this work it's worth a few extra minutes to confirm all is well.

-- measure the runout at the edge of the Actuating Cap
I found 0.012" on my test fitting, essentially the same as the 0.010" I measured for the Clutch Plate alone on the slightly bent mainshaft, so all is well. This runout remained unchanged when I rotated the Driving Plate Centre while holding the Chainwheel fixed, which in turn rotates the metal Driven Plates, and when I rotated the Chainwheel while holding the Centre fixed. The fact the runout remained unchanged means both faces of the metal plates and the friction plates are parallel with each other. It's important to do this test even if using new aftermarket plates -- or especially when using aftermarket plates.

One last measurement I made before partially disassembling the clutch prior to the final assembly was the torque required to spin the Centre with the gearbox in neutral and with all the springs loose while I was holding the Chainwheel (~2.5-3 inch-pounds), and when the springs were tightened enough to drag the Chainwheel along (~1.5-2 inch-pounds) . The 3/8" locknut in the SRM Plate Adjuster nicely served for attaching the torque wrench. The difference of ~1 inch-pound is the torque required to drag the dry plates past each other. Presumably this small amount of drag would be even lower when the plates are coated with a film of 20W or ATF.

[to be continued]

While I do not normally frequent this particular section of Britbike I have been informed by a reliable source that I should come over a suitable slice of humble pie and offer restitution in the form of suitable amounts of beverages to numb the pain should the paths cross in the future.
There has been no complaints about the Dragan fly bearing, only praise from all who have fitted one but your attention has been duely noted and of course future suggestions for improvements must only be from first hand experience and not hearsay even if it come from those whoes opnions are highly regarded.

Not necessary at all. As learning experiences go, it was remarkably inexpensive. And, thanks to chaterlea25 pointing me to NOS items that were delivered fast, the clutch was back together a full day before the end-of-Thanksgiving-holiday "promise" that I made prior to the Draganfly diversion.

To paraphrase the late, great Chico Marx, "Who ya gonna believe, them or your own lying eyes?" I posted photos (and roughness measurements from a profilometer) so people could see for themselves the problems with these bearings (i.e. inappropriate material for the cage, tracks with wrong spacing, and wildly excessive roughness), rather than rely on what I might say.

Even if we speculate that I received a bad batch due to lack of quality control, the cage still would be Al and the bearing still would consist of two rows of 16 balls spaced 1/4" apart just as in the original bearing. That is, the bearing would be identical in function although with a cage that would wear much faster than bronze.

Since even if machined properly and with a bronze cage Draganfly's bearing offers no improvement in function, and since I already started speculating, I'll speculate further as to why "highly regarded" people might have praised them. If someone's races are worn out of spec, the fact the mis-made cage forces the balls up onto the shoulders of the tracks immediately cures the problem of excess clearance. Voila!, no need to press new races into place. With the "repaired" clutch safely closed up in the primary it's out of sight, out of mind, so the fact the Al race wears and the excess clearance returns won't be immediately apparent.

Good work MM, I find it helps to mark each plate at one tongue or notch with white correction fluid, so that each plate goes back in the same groove as well as the same order, helps a lot when the clutch has all settled down , especially if it has to come apart again at a later date, it will go back and function with least bother.

I wonder if the mainshaft bend could be slightly " Adjusted " by applying the final torque up force in a away that would help to straighten the shaft out. i.e. Place max run out towards rear wheel, set final torque with wrench at 6 o clock and push forward to torque and perhaps straighten.
Did you lose 0.005 " off the bearing ID after polishing? seems a lot.

Thanks very much for your comments. You may be a lot stronger than me because I can't imagine me being able to bend a shaft of that diameter without hydraulic assist.

I removed a negligible amount from the race when I polished it. The 0.005" decrease in diameter I measured with calipers was before and after pushing it into the Chainwheel. Even that seems like a lot, but I wanted to get on with the job of finishing the clutch so I didn't take the time to drag out the bore gauge and gauge blocks to make proper measurements.

A lot more measurements will come in the next post, including ones of use to anyone setting up a 6-spring clutch (as long as they own a torque wrench).

In an article about gearboxes Phil Irving said that shafts which are out of truth up to about .005" can be straightened, but beyond this, there is a distinct possibility of the case hardening cracking during the straightening process, even if it has not already done so.
This clutch thread is very interesting, MM.

i can imagine it with a scaffold pole/ fork stanchion ( its the West Highland Way), joking here, the drive bearing would not like this.
I had miss-read the bearing clearance change, I see now ,you lost 5 thou with crush.

Even if the shaft is case hardened that only matters on surfaces that wear, such as gears. The bend is near the end where only a tapered fitting is attached.
Thanks very much. I appreciate it.

I'm lazy. Unless I get really creative I find it much easier to cause more damage with something like a 30T press than I can do manually.


There are a lot of words in what follows in this post. But, since there is almost no information available on 6-spring clutches I wanted to supply as many details on my measurements as possible in case someone spots any errors. Also, pay close attention to the units because many of the numbers preceding them are similar in what follows, but 100 lbs. is not the same as 100 ft.lbs. nor as 100 lbs./inch.

Another possible source of confusion is I used torque wrenches on two different components. Although the readings of both are in units of torque, ft.lbs., I was able to convert one of them to units of force, lbs., thanks to a calibration/conversion experiment described below. That is, one torque wrench measures torque but through this conversion factor the other torque wrench determines force.

Finally, if your eyes start to glaze over from all the words, just jump to the final paragraph to learn what you need to know to take full advantage of these results.

--------------------------
EXECUTIVE SUMMARY OF THE INFORMATION IN THIS LENGTHY POST

A torque wrench (covering ~0-20 ft.lbs), modified 3/4" socket, and machinst's ruler are all that are required to set the springs on a 6-spring clutch to give a light action while still not slipping.
--------------------------

Since the bottom lever of the actuating arm is splined it could have been removed and incorrectly replaced on the shaft at some point in the bike's history. For maximum leverage it is important that it be replaced so that it is at 90o with respect to the rod. But, 90o when engaged or when disengaged?


Both positions have their adherents, and it doesn't make much difference in practice, but I set mine for 90o when the clutch is engaged, i.e. when I'm not gripping the clutch lever. Our grip is strongest when our fingers are closest to being in a fist, and the pressure on the actuating rod is highest when the clutch is disengaged, i.e. when the springs are compressed by the most. Both factors point to having the arm at 90o when the clutch is engaged because it's there that we want all the leverage we can get.

In an older post on the BSA A7-A10 Forum David Paddock wrote that using socket notched to fit over the clutch actuating lever and adjusting the springs for 5-6 ft.lbs. results in a light clutch that doesn't slip. He also says the force this applies to the clutch rod is 90 lbs. Since being able to measure the force externally via this torque is a great aid I milled a slot in a 3/4" socket to apply torque directly to the actuating arm. For this purpose an old fashioned analog torque wrench is better than a "click wrench."



Trevor in Australia (BSA_WM20) has written that if the clutch springs are tightened just enough to keep the clutch from slipping in 1st gear when the clutch is let out with the front wheel against a wall, it won't slip in operation. This prompted me to make a second tool from an 18 mm deep drive socket bored to 3/4" to slip over the Kickstarter Crank Spindle. Used with a 150 ft.lb. torque wrench, and with the engine locked, it will let me determine the torque on the Spindle required to cause the clutch plates to slip at any given setting of the clutch springs.


I drilled a hole in the socket and welded a 3/8" nut in the correct location to allow a bolt to lock the socket to the slot in the lug allowing torque to be applied directly to the clutch to measure where it starts to slip. Note that the previous tool uses torque to determine how much force it takes to push the clutch plates apart, while this tool measures how much torque it takes to cause the plates to slip against each other.

In an earlier post I determined the spring constants of the NOS springs I obtained from British Cycle Supply were 211 lbs./inch (=/-3%). That is, if one of those springs was compressed 1/2" it would take 105.5 lbs. of force to move the Actuating Cap. This is approximately the force mentioned earlier that is sufficient to keep the clutch from slipping. Since there are 6 springs, not 1, to achieve this value each needs to be compressed by 1/6 x 0.5" = 0.0833". The pitch of the 26 tpi studs on the Driving Plate Centre is 1/26 = 0.0385". This means each nut should be spun in by hand until it just touches the spring, and then a further 0.0833/0.0383 = 2.18 turns. Rounding up, 2-1/4 turns is a good place to start since that would give 109 lbs. of force total if all the springs were an identical 211 lbs./in. Again, though, this assumes the value of ~100 lbs. of clamping force is correct. As I show below this appears to be more clamping force than is actually required, although I will be making additional measurements in upcoming weeks to know better.

Keep in mind when fine tuning the nuts to achieve minimum runout of the Actuating Cap that each additional 1/4 turn of all of them increases the clamping force by a fairly significant 12 lbs. (11%). Because of this, back off one nut to partially raise a low point and at the same time tighten the opposing nut by the same amount to lower a high point. This will reduce the runout while keeping the total clamping force constant.

To recap, assuming approximately 100 lbs. of clamping force is somewhat over the minimum to keep the clutch from slipping, and assuming springs of the original BSA value of 211 lbs./inch, setting the springs such that the nuts just touch them, plus a further 2-1/4 turns would accomplish this.

I can think of five ways to consider for checking the results:

1. Once the clutch springs have been adjusted by trial and error, use a modified 0-60 lb. valve spring tester to push on the ball end of the handlebar clutch lever to determine how much force is required.

This is the least satisfactory way of determining anything because of unknown compression and friction in the clutch cable. However, once the necessary force is determined more directly, this measurement will be useful in deciding whether it's worth replacing the clutch cable.


2. Take the Inspection Cover off the gearbox and push on the end of the Clutch Push Rod Lever (which pushes directly on the Push Rod) with a 0-300 pound valve spring tester and directly measure the force required to lift the Actuating Cap.

The difficulty with this is if it does require ~100 lbs. that's a lot of force to muster in the first place, let alone near ground level while pushing sideways on a motorcycle that wants to tip over.


3. Push on the clutch cable end of the Clutch Operating Lever using a modified 0-60 pound valve spring tester and determine the force required for a given amount of movement.

This is more reasonable. Thanks to the 5:1 lever ratio of the Operating Lever it only requires 20 pounds on the end of this lever to apply 100 pounds to the pushrod. Jumping out of sequence I confirmed that 30 pounds on the end of the Operating Lever applies 150 lbs. of force to the pushrod.


4. See if 5-6 ft.lbs. of torque on the Clutch Push Rod Lever causes the plates to separate (if David Paddock's value is correct)

Although quantitative, something essential is missing; the point in travel where this torque should be measured. That is, did Paddock post the torque value when he saw the first motion of the Operating Lever, or when the plates were fully separated, or somewhere in between? Leaving aside that important question for the moment, to turn any such torque measurements into meaningful values of force applied to the Actuating Cap I clamped a spare clutch cover and the valve spring tester on the bed of my mill.


With this setup the torque on the Operating Lever causes the Push Rod Lever to push directly on the valve spring tester through the Al rod. The torque vs. force curve is beautifully linear over the full range up to the maximum I measured of 250 lbs., as Hooke's Law says it should be. The slope is 16.67 lbs. of force per 1 ft.lb. of torque applied to the shaft. Name the force you want (say, 100 lbs.) and this tells you the torque you should adjust the springs to achieve in order to result in that force (100/16.67 = 6.0 ft.lbs.). Since there is a solid metal connection from the torque wrench all the way to the Actuating Cap this method eliminates all uncontrolled variables. Well, nearly all. Wear of the Outer Cover caused by the shaft of the Operating Lever will have an effect at the smallest values of torque.

Still, it is necessary to determine at what point in the motion of the Operating Lever the torque reading should be taken. A common problem is the shaft of the Operating lever has worn an oblong opening in the Gearbox Outer Cover so the initial torque is "wasted" moving the shaft into a different position. After experimenting with this I decided that the torque value to use is when the end of the Operating Lever has moved 0.5 inches. One can hold a machinist's ruler on the cover with one hand while operating the torque wrench with the other. Done this way the error in measuring 1/2" of movement is acceptably small.

Note that because the clutch springs have been compressed an additional 1/2" this torque reading gives a higher value of force than the actual clamping force when the clutch is fully engaged. However, thanks to Hooke's Law, they are linearly related. Also, in the end few people care to know what the true clamping force is anyway. We just want a tool that lets us adjust our clutch and fit the primary cover with reasonable confidence we won't have to do it again because it will neither slip nor be too firm.

Fortuitously, my Special Competition Gold Star gives me an excellent starting point because its clutch doesn't slip despite requiring a fairly light grip. This bike requires a torque of 11 ft-lbs at 0.5" motion of the Operating Lever. Although I haven't determined if a different cable would make things even better, it requires 15 lbs. of force applied to the ball at the end of the clutch lever to push it to almost touch the handlebars. For comparison, the same measurement on three modern hydraulic clutches gave 8, 10 and 11 pounds, my Triumph 500 with a very old cable 16 lbs., and the BB Gold Star that I subjected an Australian to a month ago a Schwarzenegger-like 25 lbs.



This torque measurement eliminates whatever is going on in the clutch lever or cable. Again, my Special Competition GS requires 11 ft.lbs. at 0.5" while the BB GS requires 24 ft.lbs. This tells me the former's cable needs to be lubricated (because the force on the clutch lever should be no more than 11/24x25 = 11.5 lbs, not 15 lbs.), and the latter's clutch springs need to be loosened (because 24 ft.lbs is way more than needed to avoid slipping).

The leverage ratio of the Operating Lever/Push Rod Lever is 5:1 so movement of 0.5" of the Operating Lever forces the Push Rod out by 0.1", i.e. compresses the springs by this much additional. Since my clutch springs have a spring constant of 211 lbs./inch that's an additional 21.1 lbs. of force that each of the 6 springs exerts when the measurement is made at this position, for a total of 127 lbs. Since there are 16.67 lbs of force exerted for each 1 ft.lb. of torque, the 11 ft.lbs. at 0.5" corresponds to 183 lbs. of force at this extension. Subtracting the 127 lbs. due to the additional 0.1" compression of the springs this means the clamping force when the Special Competition's clutch is fully engaged is 56 lbs.

Referring back to the 5-6 ft.lbs. given by David Paddock, if he measured that at the first sign of movement of the Operating Lever it means he is suggesting a clamping force somewhat higher, at 83-100 lbs. However, when applying 5-6 ft.lbs. it almost seems as if it's not doing much more than taking up slack. That's why measuring the force at 0.5" extension is a more reliable approach. In any case, I'm going to set the Catalina's clutch the same as the Special Competition's, i.e. 11 ft.lbs. at 0.5".


5. See if ~100 ft.lbs. of torque applied to the Kickstarter Crank Spindle causes the clutch plates to slip when they are properly adjusted (if Paddock's value is correct).

Because of the nature of the slip/stick of friction and the lack of quantitative knowledge of how the oil or ATF will affect it, this measurement has the most uncertainty in it. Still, it was worth doing. As the next photograph shows I locked the engine with a sprag socket I had previously made, held in place with a crude clamp based on a bearing separator.


With the engine locked and the gearbox in neutral and at various settings of the clutch springs I measured the torque on the Kickstarter Crank Spindle required to make the clutch slip. As expected there were considerable variations from run-to-run. Despite this, near the 150 ft.lb. upper limit of the torque wrench I used on the Crank Spindle it appears that Operating Lever torques of ~12-13 ft.lbs. are needed to avoid slipping. Multiplying by 16.67 lbs./ft.lb = ~200-215 lbs. of force at ~0.2" displacement of the Lever keeps the clutch from slipping until the torque applied to the Spindle is 140-150 ft. lbs. Subtracting ~50 lbs. total because the measurement was made at ~0.2" displacement rather than 0" gives 150-165 lbs. of clamping force.

Backing the nuts out a little, and again with a lot of variation in the values I obtained, a torque of ~10 ft.lbs. = 167 lbs. force kept the clutch from slipping until the torque on the Spindle was ~100 ft.lbs. Subtracting the same 50 lbs. means 107 lbs. of clamping force kept the clutch from slipping until that torque was reached on the Spindle,

Unfortunately, the variation in values obtained by this technique are too large to make it generally useful. For example, in the previous paragraph I gave a value of 10 ft.lbs. (167 lbs. of force) corresponding to slipping at a torque of 100 ft.lbs. applied to the Spindle. However, one measurement at ll ft.lbs. (183 lbs. of force) found the clutch slipped at 60 ft.lbs., i.e. much less torque on the Spindle required even more clamping pressure to stop slipping.


EXECUTIVE SUMMARY OF THE INFORMATION IN THIS LENGTHY POST

A torque wrench (covering ~0-20 ft.lbs), modified 3/4" socket, and machinst's ruler are all that are required to set the springs on a 6-spring clutch to give a light action while still not slipping. Based on measurements made on one machine a value of 11 ft.lbs. at an Operating Arm movement of 0.5" seems appropriate. However, I will report results on the Catalina's clutch as soon as possible after I get its front hub back from Vintage Brake. Depending on what I find with it I will next adjust the clutch on the BB Gold Star and report those results as well.

[to be continued]

"tightened just enough to keep the clutch from slipping when trying to kick it over in first gear with the front wheel against a wall"
If the bike is in gear the kickstart is connected directly to the back wheel, even with the clutch pulled in the kickstart will not move unless the tyre slips against the floor

Hmm. Of course you are right. I hadn't caught that problem with the advice because my bike is raised on a lift and temporarily missing its front wheel so placing its wheel against a wall wasn't even an option.

Although the problem with the brick wall advice that prompted me to conduct that test doesn't affect any of the results, I'll edit the previous post. In any case, as I wrote there, the test itself (done in neutral with the engine locked) has too much variation in the results to be generally useful. However, within that range of variation, it did allow me to determine the actual clamping force at the onset of slipping.