An improved Crayford or not?

[Stub Mandrel]

Coincidentally John Wall has recently passed on and I'm making a Crayford focuser.

It mystifies me why the Crayford uses four ball races instead of three, as three plus the adjustment spindle would constrain the tube perfectly in three dimensions. using four, the tube is probably only located by three no-ideally located points (unless it distorts) unless all four races are perfectly positioned.

A pair of widely separated bearings in line with another single bearing at 120 degrees and half way between the other two along the tube would seem ideal.

Can anyone fault my logic?

[Chriske]

Four  bearings will hold the focusing tube in line with the optical train. Three cannot do that.  If it would be possible with just three it would already been done.
And indeed these four bearings should be perfectly placed in the focuserbody. But in case one bearing should start to 'float' no worries three will guide the focuser tube without going 'off direction'.

[hughgilhespie]

A perfect focusser will only have one degree of freedom and will only allow in and out movement. You might want to be able to adjust the rotation to achieve a particular camera angle but the rotation angle will normally be locked. So, from kinematics, you need to constrain 5 degrees of freedom and a minimum of 5 constraints will be needed. If you are not worried about rotation then 4 constraints will do as in 3 bearings and the friction rod that drives the tube.

Regards,

Hugh

[lux eterna]

Hi Neil,

I guess in theory you are correct - 3 bearings would be enogh. But the third bearing would have to take twice the load compared to the other two. And it takes just a tiny bit of flexure at that position to induce a much bigger play at the draw tub end. With 4 (perfectly aligned) bearings the tube ends will have a minimum of flexure.

When I inspected my own Crayford, I was surprised to see how close (in axial position) to each other the two pairs were located, which allows for quite a bit of flexure at the tube end. I have also noticed that the star pattern in all my subs have a small offset from one sub to the next, even if no dithering was done. Tightening the focuser adjustment screws and the three 2" tube locking thumb screws did not help. But that is history now, since I introduced this focuser flex killer :

Of course I have to release the tripod screw each time I refocus, but I trade that for having minimum flexure during long exposures.

Ragnar

 

[Jessun]

I like the thinking. (Not knowing what I'm talking about really since I'm no engineer)

My Feathertouches are in storage at the moment so I can't examine them. They work. I used to have Moonlite's but pretty much threw them away in anger. There were bearings all over the place and the draw tube wobbled like crazy from the get go.

/Jesper

[hughgilhespie]

Interesting. John Wall's original drawing show that his design was based on kinematic principles.

https://en.wikipedia.org/wiki/Crayford_focuser#/media/File:Crayford_2.jpg

His arrangement of 4 bearings approximates a vee-block for fixing the x and y positions of the focus tube but allows movement along z axis.

Regards,

Hugh

[Stub Mandrel]

Sorry, but I'm going to defend my position!

 

13 hours ago, Chriske said:

Four  bearings will hold the focusing tube in line with the optical train. Three cannot do that.  If it would be possible with just three it would already been done.

Sorry, three points are all that is required. any three points that can contact the tube will fix it in only one position.

In contrast, with four points they all need to be perfectly positioned to work.

It's exactly analogous to why a tripod will sit securely on an uneven surface but table legs need to be exactly the same length to sit well even on a flat floor.

11 minutes ago, lux eterna said:

I guess in theory you are correct - 3 bearings would be enogh. But the third bearing would have to take twice the load compared to the other two. And it takes just a tiny bit of flexure at that position to induce a much bigger play at the draw tub end. With 4 (perfectly aligned) bearings the tube ends will have a minimum of flexure.

The loads are negligible in terms of what even tiny bearings can sustain, and will be distributed evenly. With four one bearing will always be less loaded than the others unless the system is perfect and the load is normal to the plane through all four contact points.

 

5 minutes ago, hughgilhespie said:

Interesting. John Wall's original drawing show that his design was based on kinematic principles.

https://en.wikipedia.org/wiki/Crayford_focuser#/media/File:Crayford_2.jpg

My contention is that this is wrong.

I have two sets of precision v-blocks. They were costly and have to be precision ground to support a cylinder perfectly. I could hack out a three-point support for a tube that word hold it solidly aligned in a few minutes.

John Wall's use of four bearings was necessitated by placing them (in the original design) on opposing pairs either side of the mounting plate.

This carried though to all later designs because it works, not because it is the optimal solution.

1 hour ago, hughgilhespie said:

A perfect focusser will only have one degree of freedom and will only allow in and out movement. You might want to be able to adjust the rotation to achieve a particular camera angle but the rotation angle will normally be locked. So, from kinematics, you need to constrain 5 degrees of freedom and a minimum of 5 constraints will be needed. If you are not worried about rotation then 4 constraints will do as in 3 bearings and the friction rod that drives the tube.

Agreed; the tube is free to move in and out, so (as we can ignore rotation) kinematics only requires four points to constrain the tube.

[Chriske]

Do not understand yet how you would solve the arrangement with just 3 ball bearings. Without the aid of friction pads of course.
Do you have an idea, example, drawing...?

I'd be interested to make one and test it...

[Stub Mandrel]

34 minutes ago, Chriske said:

Do not understand yet how you would solve the arrangement with just 3 ball bearings. Without the aid of friction pads of course.
Do you have an idea, example, drawing...?

I'd be interested to make one and test it...

Very simple principle, simply replace one pair of ball races with a single one aligned half way between the other two.

This drawing only shows most parts in outline, despite appearances the angular positions of the bearings are symmetrical:

[Chriske]

Oh.. that one...!

Made one years ago but quickly abandoned the idea. Problem is when using very heavy eyepieces it often goes out of alignment when focusing to far outward.
To avoid that problem you could support whatever is in the focuser(large eyepiece, camera,...), but it is far easier to add that fourth bearing.
A second solution I came up with is to enlarge the distance between the two bearings (at the same side of the tube). But the focuser got to long. Not a good idea because I like the low profile focuser style.
I also tried to add more tension with the focusing rod on the tube, but so doing you lose the butterly smoothness of the focuser, And that's what I like so much with Crayford's.
Again if it would be a better solution, manufacturers would make it that way.
But try it, maybe you will find a solution to the problem. Keep us informed...

[Stub Mandrel]

1 hour ago, Chriske said:

Oh.. that one...!

Made one years ago but quickly abandoned the idea. Problem is when using very heavy eyepieces it often goes out of alignment when focusing to far outward.
To avoid that problem you could support whatever is in the focuser(large eyepiece, camera,...), but it is far easier to add that fourth bearing.
A second solution I came up with is to enlarge the distance between the two bearings (at the same side of the tube). But the focuser got to long. Not a good idea because I like the low profile focuser style.
I also tried to add more tension with the focusing rod on the tube, but so doing you lose the butterly smoothness of the focuser, And that's what I like so much with Crayford's.
Again if it would be a better solution, manufacturers would make it that way.
But try it, maybe you will find a solution to the problem. Keep us informed...

Thanks, it's useful to know if the theoretical solution doesn't perform.

Alternatives include more bearings, with eccentric mountings so they can be aligned for collimation and smooth movement, and using bearings instead of the 'focusing rod' to locate the tube, so the focuser simply provides movement, not location.

One arrangement would be six bearings at each end with at least one of each trio adjustable*, and mount the focuser spindle in bearings AND sprung.

 

* removing a short section of thread from M3 screws would allow sufficient latitude of adjustment as long as the screw holes are accurately placed.

[hughgilhespie]

Hi Neil,

Thinking about this......

If you have 4 bearings in a 'classic' Crayford layout of course that gives you 4 points of contact. But, because they are arranged as 2 'Vee' pairs,  separated by a certain distance, they will ALL be in contact with the focus tube - at least if you assume that the focus tube is rigid. At each pair, the tube is held down by gravity (plus some force from your tension spring) The tube will go down until BOTH bearings in the Vee are in contact. Same at the other Vee. The focus tube then will move in or out along a line that passes through the effective midpoints of the vees. Not necessarily in line with the optical axis but nevertheless it is constrained to move in a straight line.

Regards,

Hugh

[Stub Mandrel]

14 minutes ago, hughgilhespie said:

Hi Neil,

Thinking about this......

If you have 4 bearings in a 'classic' Crayford layout of course that gives you 4 points of contact. But, because they are arranged as 2 'Vee' pairs,  separated by a certain distance, they will ALL be in contact with the focus tube - at least if you assume that the focus tube is rigid. At each pair, the tube is held down by gravity (plus some force from your tension spring) The tube will go down until BOTH bearings in the Vee are in contact. Same at the other Vee. The focus tube then will move in or out along a line that passes through the effective midpoints of the vees. Not necessarily in line with the optical axis but nevertheless it is constrained to move in a straight line.

Regards,

Hugh

In theory, only if the two vees are perfectly aligned.

Imagine one of the vees is rotated a couple of degrees with respect to the other; the tube will lie askew and only contact three of the bearings.

For perfect four-point contact both vees need to be exactly aligned.

It seems that in practice the alignment is either good enough that flex in the drawtube/housing is either sufficient to allow four-point contact OR the tube only contacts the bearings sufficiently well to rotate them all, but only three bear the load.

[Ravenous]

20 hours ago, Stub Mandrel said:

Imagine one of the vees is rotated a couple of degrees with respect to the other; the tube will lie askew and only contact three of the bearings.

I'm having a hard time seeing this. I'm sure the standard Crayford is sufficient; based on the kinematic principles Hugh mentioned earlier.

You can constrain a plane with three contact points. (It'll be free to slide in two dimensions and rotate in one)

You can constrain a sphere with three contact points (It'll be free to rotate in three axes)

To constrain a cylinder, you need four contact points.  (It'll be free to slide in one axis and rotate around one direction.)

(For completeness, I should mention all three cases need a flexible constraint from "above" to prevent the object lifting.)

There's a mathematical argument to prove all this, but I last did this decades ago...

[Stub Mandrel]

1 hour ago, Ravenous said:

I'm having a hard time seeing this. I'm sure the standard Crayford is sufficient; based on the kinematic principles Hugh mentioned earlier.

You can constrain a plane with three contact points. (It'll be free to slide in two dimensions and rotate in one)

You can constrain a sphere with three contact points (It'll be free to rotate in three axes)

To constrain a cylinder, you need four contact points.  (It'll be free to slide in one axis and rotate around one direction.)

(For completeness, I should mention all three cases need a flexible constraint from "above" to prevent the object lifting.)

There's a mathematical argument to prove all this, but I last did this decades ago...

The constraint to stop lifting counts as one of the points.

A cylinder needs FIVE contact points  three to rest on (NOT four) like a tripod, one to stop it sliding back and fourth (not needed as we want it to slide) and one to stop it lifting.

Three bearings plus the focus spindle/roller/shaft = four points

The fourth bearing is (theoretically) redundant.

 

[Stub Mandrel]

OK a cylinder in space has six degrees of freedom:

up down

back and forth

left and right

rotating around each of these three axes

 

Put three points (the bearings) in a horizontal triangle narrower and shorter than the cylinder and rest the cylinder on them like a cradle, lying left-right..

Now place a constraint over the top (the focus spindle)

It can't move up and down, because of the three points below and the one above.

It can't move back and forth as the spindle stops it riding up the bearings.

It cant rotate around the vertical or front-back axes for the same reason.

It CAN slide left to right (the focusing motion) and it CAN rotate around the left-right axis (in practice the spindle is elongated and sits on a flat so it acts as TWO points and constrains this rotation).

Adding a fourth bearing underneath merely duplicates the function of the single bearing from a kinematic perspective, it does not add any constraint.

 

I imagine the reason four bearings are used has more to do with the narrow guide principle than kinematics.

[Ravenous]

Post deleted.  I give up!

[Stub Mandrel]

Well... I decided to follow the time honoured design with four 8mm bearings.

And you know what? The drawtube runs on them beautifully

[Chriske]

Can we have a peek..?

[Stub Mandrel]

See the thread here:

[Stub Mandrel]

After a three month+ hiatus I now have a fully functional Crayford, just lacks the focus knobs...

[Rusted]

I'd have gone for one, or a pair, of pivoted, lateral "seesaws" to allow the bearings to take up any slack across the diagonals.

Or, to put it more technically: The four legged stool on a cottage, brick floor syndrome. 

[Stub Mandrel]

3 minutes ago, Rusted said:

I'd have gone for one, or a pair, of pivoted, lateral "seesaws" to allow the bearings to take up any slack across the diagonals.

Or, to put it more technically: The four legged stool on a cottage, brick floor syndrome. 

If you do that, you may as well use three bearings as geometry the pivot replaces one pair of bearings with a single bearing.

[Rusted]

13 hours ago, Stub Mandrel said:

If you do that, you may as well use three bearings as geometry the pivot replaces one pair of bearings with a single bearing.

All me to disagree. The single bearing has no control [whatsoever] over lateral movement.

It can only resist pressure at right angles to its own axis.

With a 3 bearing system you are relying on a single pair for steering but really building a wayward "tricycle."

A tricycle has a deceptively narrow hinge line between surface contact points with very marginal stability and poor predictability.

Which is why they are so very unpopular for mass transport, racing cars and Crayfords.

Rocker arm [see-saw] bearing supports would remove any risk of diagonal rocking between contact points.

It becomes a precision, self aligning, linear bearing with a high degree of lateral constraint.

A fixed bearing construction relies entirely on precision construction to achieve it true axis of linear translation.

[Stub Mandrel]

1 hour ago, Rusted said:

s.

It becomes a precision, self aligning, linear bearing with a high degree of lateral constraint.

A fixed bearing construction relies entirely on precision construction to achieve it true axis of linear translation.

But if a pair of bearings can self-align by rotating around a pivot, their ability to constrain become the same as using a single bearing, that's why both a tricycle and a chassis with a single live axle (e.g. a traction engine) can sit on an uneven surface, but the pivoted axle provids stability not accuracy of positioning.

Think about a tube resting on four bearings, two on a pivoted arm. Nothing stops the tube moving side to side by tilting the pivot until you apply force from above. This is entirely analagous to the three bearing arrangement.

Once you add the focuser spindle, it is forced to a stable position, but no more accurately than the three bearing.

I would agree that it may being  a benefit of greater stability countering the problem @Chriske found, but it won't be more accurate.

In the end my tube keeps contact with all four bearings without having to be preloaded, so my machining was accurate enough, which is what matters in the end.