Backlash free rotary table

[lerman]

Hi Rich, I wonder how the "practically backlash free" aspect comes into play, seeing as how it still has gears meshing with gears that require close fitting to make flank contact continuous to remove any play.

It looks like there is a differential action with one less tooth to the other side of the split planetary gears, so effecting the reduction, but I can't see how the reduction in backlash occurs, and for a CNC rotary table, this is the most important part.....the actual reduction is of secondary importance.

Some illumination would be handy here.
Ian.

The answer comes from Featured Technologies — Gear Bearings: increased capacity and performance for superior gear drives

=======
They provide smooth and accurate control with rifle-true anti-backlash, meaning that the inner portion of a gear spins (like a rifle bullet) as it moves axially via spring action. The spinning motion continues until the backlash is taken out. But, the rifling angle is a mechanical locking angle so the gear cannot back-drive. This produces a planetary transmission with zero backlash.
=======

Ken

[handlewanker]

Hi, it doesn't equate to "rifle true" as in the speed of a bullet spinning to maintain axial trueness in a forward motion, if'n that is what you are saying.

Or does it mean that the inner gear teeth are being pushed against the corresponding gear teeth while going forward..........what happens when you go in reverse......you get backlash from the transition of tooth flank coming away from tooth flank and passing through a non contact zone which not having contact means you get lost motion until the reverse flanks of the gear teeth make contact and take up the drive.

What is misunderstood is that in all cases of applied pressure you require an equal amount of reverse pressure to offset lost motion from elasticity, called preload, otherwise you have to make up this loss of pressure by waiting for the pressure zones to come to equality, iE taking up the pressure before motion can be achieved.

Put simply, all materials have elasticity......preload anticipates the elasticity of the object it is applied to.......remove the forward force and the reverse force has to equal the applied force but in reverse direction, and you get lost motion until the force is overcome.....that is the backlash we define as the lost motion.

You could mesh a pair of gears, forced together with enough force so that there is metal to metal contact between the tooth flanks, and with enough pressure against the flanks (both sides) to overcome any lost motion from lack of flank contact due to lube film thinning or inaccuracy in manufacture, and given enough power generated by a stepper motor achieve almost perfect backlashless rotation, but the stepper motor could miss a beat from the load it must now carry, and in time the wear on the tooth flanks will "run in" the gears and so you end up with lost preload and backlash once again.

I think there must be another feature in the design, (not apparent to me), apart from just close fitting of the gears, that gives the "practically backlash" free aspect to the design.
Ian.

[RICHARD ZASTROW]

Yes there is some backlash. I said "practically backlash free". If it can focus the Hubble, it's close enough for me.LOL

About 25 years ago I had the good fortune of meeting a German/Czech gentleman, Waldemer Elsdorfer, who provided the rotary tables to NASA for satellite tracking. I was shopping for a 60" table with a +/-4arc second accuracy. He said that's "junk". He wouldn't build anything that loose. His worst commercial tables were 4 arc second total tolerance envelope.

I asked how accurate he can make them. His reply was one two-hundredths of an arc second if he was allowed 20 seconds to let the electronics settle. I asked how he measured that level of accuracy. He said he mounted a telescope on the table and focused it on a known star. Then moved the table to a position where the star should be at a certain time. Any error shows up in the scopes reticule at the prescribed time.

I had to take him at his word, his office was full of autographed pictures of him with the astronauts from before the shift to the shuttle.

Dick Z

[handlewanker]

Hi Rich, for an everyday use in CNC mode how much would such a highly accurate rotary table set you back?

I don't think the wait of 20 seconds would come into the CNC machining when all you want is to get back to zero without missing a bit on the way.

The majority of the time you cannot afford to have any hesitation at the reversal of the table otherwise you'd get a series of indentations.....even though they were very accurately positioned....LOL.

I have a great interest in the Eccentrically Cycloidal drive as posted by Bill Todd, and I feel that the rolling motion of the eccentrically driven drive roller without friction from sliding, while at the same time maintaining a constant contact with both forward and reverse slopes of the lobes, has great promise for a relatively simple solution.

It has infinite possibilities to drastically reduce any backlash by making the drive as a very large reduction.....so reducing the final backlash figure at the table to insignificent proportions.

The drive per se should be backlash free due to the fact that the roller is in constant contact with the surface it rolls on and does not retract from the rolling surface when it reverses.

This is tantamount to rolling a roller on a flat surface back and forth......and it is obvious that the roller would not leave the surface when it is rolled in the reverse direction....therefore it has no backlash in it's primary motion......the drive forward and back does not apply to the principle.
Ian.

[RICHARD ZASTROW]

Those super accurate rotary tables I mentioned (just as an FYI thing) are not for DIY. In fact, I don't believe there are many top corporations that could spend the money it requires.

That's a project that few governments can undertake. Fortunately, that was at a time the USA government could get enough of our tax money together to send manned rockets to the moon.

Unlike the users of CNC rotary tables, the NASA rotaries only had to give direct readings of a positioning device. My understanding of the readout was from multiple Farrand scales mounted directly to the table or it's shaft. The input from the scales was mapped. In operation, the multiple scales were then "averaged".

Yes, a bit pricey. Not something I will undertake soon.LOL

Dick Z

[handlewanker]

Grrrrrrrr....maybe that's one of the reasons the NASA guys had to quit early, exponential cost budget.

Maybe they could have got the same results with the wrapped wire drive....no backlash whatsoever, especially as the table don't have to go round and round like the CNC tables do.....this may seem facetious, but what is the difference between a drive that has "practically" no backlash, like in the NASA tables, and the EL Cheapo method which is 100% no backlash for a thousandth the cost.......mega bucks for a start.....somebody knows somebody in the Space Bureau...ah haa?

Come to think of it, I seem to remember the Russian "space" pencil that is able to write in a pure vacuum in any position and has an infinite shelf life....the NASA space ball point pen cost several hundred thousand dollars to develop, several thousand dollars to make and was only good for working in a vacuum in a totally gravity free environment.....and they couldn't get the cap off to use it....LOL.
Ian.

[aystarik]

back to cycloid drives, I've put 1st stage animation together, enjoy:"http://www.youtube.com/watch?v=oussNksQ6UM"]http://www.youtube.com/watch?v=oussNksQ6UM

[Desertrunner]

Its clear from your animation that the lobs rub against the pins. It also supports the claim that the backlash could be close to if not zero.
Tony

[vegipete]

If the pins are fixed, yes they would rub on the lobes. But the pins could be bearings, or mounted on bearings, to roll freely. And if the pin mounts were slightly eccentric, then they could be tightened down against the central lobed shape to take out backlash.

The little fiddley toy sample of one of these that I have from Sumitomo has two identical lobed rotors, phased 180 degrees apart.

Also, while Aystarik's animation very nicely shows the motion, it omits the reduction output that can be seen, for example, on the Wikipedia page, which appears to be another source of backlash.

======================
Saw the following on a web page:

Characteristics of Cycloidal Drives
Cyclo drive provides an efficient and durable transmission system with a shock load capacity of more than 500%.
[snip]

Disadvantages of Cycloidal Drives
There are several disadvantages also. They exhibit speed and torque ripple because of their methods of transmitting motion. The eccentric motion inside the drive translates into ripple and is evident at slow speeds. This is problematic in industries requiring smooth motion. They cannot effectively achieve ratios below 50:1. They have low torsional stiffness. They have higher initial cost and higher operating maintenance cost.

[handlewanker]

Hi all, I had the same feeling that the cycloidal "gear rotor" type drive was "rythmic" in it's transmission path, seeing as the path traced by the roller to lobe is a hypocycloid and not a linear progression as you would have with an involute gear form driving a rack or ring gear.

I favour the eccentrically cycloidal drive as posted by Bill Todd, with the multi phased stacked lobes being driven by multi phased stacked rollers, therefore each one of the roller lobes in the stack makes contact in a phased manner with each one of the corresponding phased lobes of the ring gear or rack stack whichever is being used, and is therefore not rythmic in it's rotation.

It cannot be rythmic if'n for argument sake there were 10 stacked lobes on the roller each phased at 1/10 of the roller's circumference, therfore each roller lobe in it's stack is making a phases contact point with each of the driven ring lobes at equal intervals of the rotors cycle.

In the purely cycloidal drive with a multi lobed roller rotating inside a lobed ring gear with one lobe less, it will as it is shown exhibit a rythmical path, which for CNC rotary table applications is a non starter.....purely for a reduction drive yes......but CNC...No.

As was shown by Bill Todd in the video of the eccentrically cycloidal drive, the roller rotates around the ring gear lobes in a rolling manner and therefore is not sliding.

Perhaps more light can be thrown on this drive in case there is a hidden factor that nullifies the advantages.
Ian.

[Desertrunner]

Ian the cycloidal rotor you mentioned (Bill Todd), if I remember correctly had both centres anchoured together, if that so I think you will find it won't work. I tried to find a commerical unit with the same method but there we none aviable. I would like to be proven wrong but if you check Sumitomo multi reduction drives you will see what I mean.
Tony

[handlewanker]

Hi, nope, the eccentrically cycloidal drive is the one I am referring to....I seem to have lost the link I had with the demo of the advantages of the drive as compared to a regular worm and worm wheel.....it looks like a series of discs that rotate on a ring gear also comprised of a series of lobes......does that make any sense?

In one form it looks like a spiral worm running on a similar larger spiral gear, and one rotation of the worm drives one lobe of the larger gearwheel.....I'll check back amongst the posts as I think it is the answer to the worm drive problem.

Found the link....post #304 on this thread by Bill Todd, with a further link back to the EC gearing video on rotary worm drives.

The reason I like this gearing is it would seem it could be made simply by making the "gears" in the form as they show it with phased discs stacked side by side.....the ratio shown in the video was 1:5, possibly it could go lower to 1:3 maybe....the top end is without limit......a rack in fact.

The gear form rolls around the lobes and this in my opinion is without friction and also without backlash.
Ian.

[aystarik]

it omits the reduction output

here you go...

[simonrafferty]

My little contribution to the Hypocycloid theme.

I have modified the Rhino Script Zoidberg Posted to work with Rhino4. On my copy at least, it did not have the ATan2 function - so I coded this long hand.
I also added a clearance value to leave a gap between the pins & rotor. Although 0 is desirable for minimum backlash, it needs a little to run smoothly.

Lastly, it now draws circles for the pins/rollers and for the cam in the centre of the rotor.

Option Explicit
'Hypocycloid cam generator
'Based on script by Zoidberg on the CNCZone Forum
'In turn based on a Python script by Alex Lait

'Modified to work with Rhino4 (which doesn't have ATan2 function) and 
'Addition of pin and cam circles by Simon Rafferty www.x-eng.co.uk
'I also added a clearance value - as reduction will not run if rotor is touching pins all the way round.
'0.1mm seems to be a good compromise between backlash and smooth running

'Credit To:
'        Zoidberg
'        http://www.cnczone.com/forums/linear_rotary_motion/72261-backlash_free_rotary_table-9.html#post602329

'        Alex Lait
'        http://www.zincland.com/hypocycloid

'        Formulas To describe a hypocycloid cam
'        http://gears.ru/transmis/zaprogramata/2.139.pdf

'        Insperational thread On CNCzone
'        http://www.cnczone.com/forums/showthread.php?t=72261

'        Documenting And updating the sdxf library
'        http://www.kellbot.com/sdxf-python-library-For-dxf/

'        Formulas For calculating the pressure angle And finding the limit circles
'        http://imtuoradea.ro/auo.fmte/files-2007/MECATRONICA_files/Anamaria_Dascalescu_1.pdf

'Suggestions:
'        - Eccentricity should Not be more than the roller radius.  Generally about 0.25 x Pin radius For a smooth result

' Try with values:
' 160, 15, 3, 14, 0.2


Sub Hypocycloid()
 Dim p, b, d, e, n,c
 
 b = Rhino.GetReal ("Pin circle diameter")
 d = Rhino.GetReal ("Pin diameter")
 e = Rhino.GetReal ("Eccentricity")
 n = Rhino.GetInteger ("Number of teeth")  
 c = Rhino.GetReal ("Clearance between rotor & pins")  
 p = (b/2)/n
  
 Dim i, arrPoint, arrPoints(360)
 For i = 0 To 360 
  arrPoint = Array(CalcX(p,d+c,e,n,Rhino.ToRadians(i)), CalcY(p,d+c,e,n,Rhino.ToRadians(i)), 0)
  arrPoints(i) = arrPoint
 Next 
 Rhino.AddInterpCurve(arrPoints)

    
 Dim xtemp, ytemp, arrPlane(3), Plane
 For i = 0 To n+1
  xtemp = (p*n)*Cos(2*Rhino.pi/(n+1)*i)
  ytemp = (p*n)*Sin(2*Rhino.pi/(n+1)*i)
  arrPoint = Array(xtemp+e-d/2, ytemp, 0)
  arrPlane(0) = arrPoint
  arrPoint = Array(xtemp+e + d/2, ytemp, 0)
  arrPlane(1) = arrPoint
  arrPoint = Array(xtemp+e, ytemp + d/2, 0)
  arrPlane(2) = arrPoint
  
  Rhino.AddCircle3Pt arrPlane(0),arrPlane(1),arrPlane(2)
 Next  
 'Lastly, add cam
  arrPoint = Array(e-d/2, 0, 0)
  arrPlane(0) = arrPoint
  arrPoint = Array(e+d/2, 0, 0)
  arrPlane(1) = arrPoint
  arrPoint = Array(e,d/2, 0)
  arrPlane(2) = arrPoint
  'Origin
  Rhino.AddCircle3Pt arrPlane(0),arrPlane(1),arrPlane(2)

  arrPoint = Array(-d, 0, 0)
  arrPlane(0) = arrPoint
  arrPoint = Array(d, 0, 0)
  arrPlane(1) = arrPoint
  arrPoint = Array(e,d, 0)
  arrPlane(2) = arrPoint
  'Offset Cam
  Rhino.AddCircle3Pt arrPlane(0),arrPlane(1),arrPlane(2)
 
End Sub

Private Function CalcYP(a, e, n, p)
 CalcYP = ATan2(Sin(n*a)/(Cos(n*a)+(n*p)/(e*(n+1))), 1.0)
End Function

Private Function CalcX(p,d,e,n,a)
 CalcX = (n*p)*Cos(a)+e*Cos((n+1)*a)-d/2*Cos(CalcYP(a,e,n,p)+a)
End Function

Private Function CalcY(p,d,e,n,a)
 CalcY = (n*p)*Sin(a)+e*Sin((n+1)*a)-d/2*Sin(CalcYP(a,e,n,p)+a)
End Function

Private Function atan2(y,x)
Dim theta, pi: pi = 4*Atn(1)
If x <> 0 Then
theta = Atn(y/x)
If x < 0 Then theta = theta + pi
Else
If y < 0 Then theta = (3 * pi/2) Else theta = pi/2
End If
atan2 = theta
End Function

For those who have not used Rhinoscript before,
Go in to the Tools Menu, Select RhinoScript then Edit....
Paste the above code in to the window and save somewhere you can find.
Go to Tools, Rhinoscript, Load - and load the code you just saved
Go to tools, RhinoScript, Run
The command line will now prompt you for Prompt you for the first parameter. Enter the numbers - and it will draw the hypocycloid.
If you modify the code, you have to save it, then re-load it in Rhino before running.

Si

[handlewanker]

Wow, pretty awesome.......all we need now is a backlashless rotary table driven by a CNC coded drive to enable an end mill to make the hills and valleys without any flat spots on the peaks and troughs at the reversal points.

Presumably the cam plate would be cut with a wire cutter,laser, plasma, water jet.....whatever from hardened steel and just polished on the cam faces when finished cutting.
Ian.

[Mike Everman]

Very nice, Bill. You make that animation? I'm a big fan of all things epicyclic.

[simonrafferty]

In a previous life, I was involved with a company who made gearboxes with what they described as controlled backlash which were used in robotics.
They realised that when you are moving rapidly, you don't need as high a tolerance as slowly - so made the backlash controllable via hydraulic pressure.

They used a planetary reduction where each of the ring gears were cut with a shallow taper. As you moved the rings up & down in relation to the planets (which were allowed to move radially a few thou), the tolerance and backlash changed. As you reduce this however, the frictional drag increased and at it's limit, the whole gearbox pretty much locked solid.
It meant that for most big movements, it was quite efficient and did not cause unnecessary wear, but when you needed precise positioning, it could be tightened and potentially lock.
This wouldn't be impossible to achieve on a DIY level, particularly if you cut your own gears. Just make the planets with say a 0.1 degree taper (bevel) running in an off the shelf internal ring gear?
You could possibly even achieve the same thing by tilting the axis of the sun gear?

My Jones & Shipman 4th Axis has a hydraulic lock - but otherwise it's just a worm/wheel construction. It's very good, backlash wise when unlocked - but solid as hell when you lock it. I did wonder about modulating the hydraulic pressure in proportion to the chip size or something?

Si

Si

[handlewanker]

Hi, for the average DIY machinist, cutting gears is quite a task....cutting gears and hardening them is quite a big task.......cutting gears and hardening and grinding them is way beyond the average DIY'er....if you don't harden and grind you may as well just stick with a bronze wormwheel and steel worm for all the accuracy you just won't get.
Ian.

[erosnicolau]

Hello! Nobody here since November 2012? That's too bad...
I've been reading all this thread, which provided invaluable information and knowledge to a newbie like me.
I noticed most of you guys need backlash-free solutions for rotary tables. I need them for fine angular control of a time-lapse photography camera motion controller. Since the HarmonicDrives are very expensive (and I need four 1:100 reductions for my robot, which in the best day would set me back more than 3000usd), I read all about your experiences and advices and the only solution I'm looking at, right now, is to build my own hypo-cycloidal reductions, based on the 1:100 model some of you guys already built. For now, the only "innovation" I'm introducing (nothing new, though) is using 5mm diameter deep groove ball bearings instead of the outer "teeth".
Here's a print-screen of the 3d model designed so far starting from your ideas (this is our third prototype under way)

I will be posting more news once I get this prototype into my hands.
Edit 1: all the yellow bits are bearings: 11+10=21 5mm radial, 2 thrust bearings and various other sizes radial bearings for support in various places
Edit 2: the first prototype (made of plexy-glass) had 100 arcminutes backlash LOL. the second prototype (made of Duramid) had 12 arcminutes backlash (we later discovered an X-Y distortion in the mill, which is now corrected). We hope to get to less than 5 arcminutes with this, third, prototype. Will keep you posted

[handlewanker]

Hi, engineering people need backlashless dividing heads and rotary tables for CNC work due to the need to go forward and backward frequently while cutting at the same time, and without missing a step.

I would have thought that you would only need to go one way to track the object against the rotation of the Earth etc, so a buffered drive would suffice for your needs.

If you understand CNC you will realise we cut in both directions, but without missing a step on the reversal.

Your design looks intriguing, if it works without as much backlash as a normal steel worm and bronze wormwheel rotary table, then you are going in the right direction.......maintaining the condition is the next target......anyone can make a bronze wormwheel drive that initially has almost nill backlash, but as the drive is a steel sliding on bronze force it will wear and so ......you are back to square 1.

I like the idea of the roller rolling on the inside of the ring "gear", and having ballraces for the rollers should eliminate any sliding friction and also wear.

I have toyed with the idea of a single roller and a ring gear, but as it needs a continuous contact at the roll over point on the ring gear crest, there would have to be a series of thin rollers stacked together and spaced around the pivot point of the roller, meshing with a similar stacked number of thin ring gears that are sequentially rotating to maintain a continuous drive.

In my imagination, the roller gear(sequentially stacked discs) and the ring gear (also sequentially stacked discs) would be wire cut from pre hardened steel discs that would only need polishing on the roller faces and no subsequent hardening or grinding.

The rings and roller discs are then fixed together to form the sequential roller/ring gear track.

This is how I would see a DIY attempt at a roller solution, as the wire cutting is very precise and working on pre hardened steel makes any after rework unnecessary.
Ian.