Backlash free rotary table

I want to build a rotary table that can handle continuous milling!
Do you think my drawing would work?
The upper gear belt pulley can rotate freely on the axle, and is connected to the axle via the spring, and is preloaded.

Attached


IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617135811210.jpg)

Nope... The axle is eccentric and the gear rotates on the axle so the holes would make it worse! You need a counterweight that rotates with the axle and has the mass opposite the direction of the gear.

Right, looked on different scheme. Drills must be in eccentric itself

I like the NEMA based reducer Zoidberg! I am planning to do something similar for a NEMA 23 input motor. I am also looking at the cam followers (needle bearing) on VBX since they have one that is 13mm OD. The only problem is they are $10 each and my goal with the project is to use stuff I already have in the shop
The counter balance is a necessity if you are going to run at high speeds. I could feel the wobble when running mine with the drill.
It looks like my post with all of the pictures got deleted. I assume due to the DNS issues my site has been having. So I have attached the important files to this message.
cad_dwg.zip is the Autocad 2000 3d drawing file of the reducer.
hypocycloid_script.zip contains the Python files I used to generate the profiles. You should be able to run these on any platform, you will just need a Python interpreter.
Here is the updated video that shows the internal operation of the reducer:
"http://www.youtube.com/watch?v=SH46bpe1cNA"]YouTube - Hypocycloid reducer, 90:1, Internal view
The original video showing the reducer running at 2000rpm input:
"http://www.youtube.com/watch?v=Ye8NtIZkixI"]YouTube - Hypocycloid reducer

Attached


IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617141211228.JPG)


IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617141211229.GIF)


IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617141311230.JPG)

Attached

I started to think about a nice thing! Even if the gearbox has backlash, you can just build 2 of them and run both of them from the same servo/stepper, but with one of them as a slave preloaded with 1 Nm or so (on the input).
Then you take the outputs and connect them to the rotary spindle with gearbelts, and get 90 Nm preload!
Like my previous idea:
IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617141511231.jpg)

There are also 688RS2 bearings AKA "roller mini" with 16mm OD. They sell in handy 20 pcs pack.

I made a rhinoscript based on your code One nice thing is that it interpolates the curve from the points!

I made work ZincBoy's script, now Im trying to use yours in rhinoscript using the same parameters as this image:
(entered the parameters in the same order)
IMAGE(http://img83.imageshack.us/img83/2940/dxfexamplei.gif)
and I get this in rhinoscript, cool image but its not what Im looking for
IMAGE(http://img23.imageshack.us/img23/7738/hypo.jpg)
What are the parameters you used to generate this image?
IMAGE(http://img142.imageshack.us/img142/9136/attachmentmi.png)
Thanks!

I found what I was doing wrong, ZincBoy's script uses the pitch value, Zoidberg's uses the Pin circle diameter value (value=1.6 in this example) and then calculates p. This is solved by editing the 1st lines as follows:
Code:
rem b = Rhino.GetReal ("Pin circle diameter")<br /> p = Rhino.GetReal ("Pitch")<br /> d = Rhino.GetReal ("Pin diameter")<br /> e = Rhino.GetReal ("Eccentricity")<br /> n = Rhino.GetInteger ("Number of teeth")<br /> rem p = (b/2)/n

Now I get this:
IMAGE(http://img16.imageshack.us/img16/155/hypo2.jpg)
I just need to find a way to make the circles around the pin centers and the center/excentric centers
Thanks!

How about this concept?
I'm sure it has been done before, but if the two blue pulleys were preloaded toward each other, and allowed to float, then there should be zero backlash, and decent stiffness as you have one straight tensioned leg on each belt. For more reduction, two stages could be used. The pulleys are just extra wide, standard pulleys.
The trick would be coming up with a floating tensioner that would stay in place. I suppose that three belts could be used, the center one twice as wide as the outer two, canceling out the torque caused by the offset between the belts.
Keith

Attached


IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617141511232.jpg)


IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617141611233.jpg)

What do you think about this? I extended the shaft trough the output so you can put 2 gearboxes inline with torsional preload between them! Then you don't need to have completley backlash free gearboxes!
Also i included the XL gearbelt pulley in the output!

Attached


IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617141611234.jpg)


IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617141711235.jpg)

Hi Mactec

The only problem is your Idea does not work, or everyone would already be doing it like this

"I have traveled the length and breadth of this country and talked with the best people, and I can assure you that data processing is a fad that won't last out the year."
The editor in charge of business books for Prentice Hall, 1957
"If man were meant to fly he would have been born with wings"
Attributed to Milton Wright, father of Wilbur and Orvillle and Bishop of the Church of the United Brethren in Christ, about 1903, but probably dating back to the time of Leonardo Da Vinci
Cheers

Hi Mactec
"I have traveled the length and breadth of this country and talked with the best people, and I can assure you that data processing is a fad that won't last out the year."
The editor in charge of business books for Prentice Hall, 1957
"If man were meant to fly he would have been born with wings"
Attributed to Milton Wright, father of Wilbur and Orvillle and Bishop of the Church of the United Brethren in Christ, about 1903, but probably dating back to the time of Leonardo Da Vinci

I guess you can BS as much as you like but your posts tell the real story

I guess you can BS as much as you like but your posts tell the real story

And what is the 'real story'? I am curious.
For me, the real story is that technology advances. Once it was hard to bore out a cannon to get a usable smooth bore. Today we have a precision machine tool industry. But is that industry to be frozen at the current state of development - or can we improve further?
Be prepared to cannibalise your own best-selling products, because if you don't, someone else will.
Standard high-tech business credo.
Direct-couple encoder-controlled lathes for precision threading? Of course.
Direct-coupled encoder-controlled rigid tapping? Of course.
Spindle orientation for 5 axis machining? Of course.
Actually, I believe there are products on the market today which do this already ...
Cheers

Wow, I feel like a fly on a giant's dinner plate..........some questions have been answered, but I'm still in the dark as regards to the viability of sensing the movement on the final drive output.
In the design I am pushing across my graphics tablet, the output shaft is 35mm diam which is the bore diam of the bearings, so an encoder, that I assume is available and to my limited knowledge achievable, must be a disc with a bore of 35mm and an OD of whatever is practical to make and etch/scratch/ or otherwise mark 7,000 + lines on it so that as the spindle rotates it can generate a pulse count to enable the controller to control the servo motor for the count needed.
The servo motor will be coupled to the spindle by the friction drive mechanism with a reduction ratio of 1:20.
In other words, if you were the navigator in a car driving down a road, you would instruct the driver to stop when 10 telephone poles have passed your window, and base on the fact that the poles are spaced approximately 100 metres apart, when the tenth pole is reached the car is stopped, having arrived exactly 1 KM from your starting point which is the destination you were instructed to go to......the fact that you were not looking out of the windscreen is irrelevant as the pole count is more important than the actual direction you were going in.......the road was a winding curvy one anyway and the driver is responsible for accelerating and braking to achieve the travel speed to get from starting point to destination while allowing you to count the number of poles accurately.
To go a step further, my next destination is located at 800 metres from my starting point and so I instruct the driver to go into reverse and stop when I count back 2 poles.
If I can handle that simple navigation problem so can a more sophisticated controller for a CNC 4th axis.
I assumed in all cases that the line count is what decides where you are at no matter how many times you go forward or backwards, and it is also assumed that if the encoder is accurate enough and the read head capable enough to read the lines as they pass back and forth across the read head, then by all accounts you will not lose position within the tolerance of the encoder line count.
Is it possible to etch or otherwise mark 7,000 + lines on a disc of say 100mm diam, which will give you a circumference of 314.2mm and a line spacing radially of .045mm at the edge of the 100mm diam disc, and then read the lines as they move pass an optical read head at no more than 100 rpm?
The encoder will generate the pulses as it passes the read head in either direction, but can the controller decide if the count is plus or Minus......the lines going back or forth will still generate a pulse no matter which direction the disc is revolving in......the logic of this is a puzzle to me.
How does the controller know if the disc is rotating forward or backwards?
If this is a simple read head ability to read the oncoming lines or the outgoing lines, whatever, then the problem of direction differentiality is simple, and we have a 4th axis or a screw cutting leadscrew or whatever you need to have.
I could imagine that if the lines were grouped in a sequence that gave you a trigger for the next line you could decide which way the disc moved by when that trigger occurred.....bumpitty bump or bump bumpitty....LOL
If the pole count was analogue, it would be colour coded to give the direction of travel....red white black indicated forward travel, and black white red gave reversing etc, you'd know by the sequence of passing colours which way you were going......how this is done digitally I do not know, maybe twin discs with twin read heads.......my digital calliper handles that quite easily.
The next issue is cost for an encoder/read head for a 7,000 + line count.
Ian.

And what is the 'real story'? I am curious.
For me, the real story is that technology advances. Once it was hard to bore out a cannon to get a usable smooth bore. Today we have a precision machine tool industry. But is that industry to be frozen at the current state of development - or can we improve further?
Be prepared to cannibalise your own best-selling products, because if you don't, someone else will.
Standard high-tech business credo.
Direct-couple encoder-controlled lathes for precision threading? Of course.
Direct-coupled encoder-controlled rigid tapping? Of course.
Spindle orientation for 5 axis machining? Of course.
Actually, I believe there are products on the market today which do this already ...
Cheers

Easy guys, let's not get carried away here! We can discuss this without jabs. And we can do so without meaningless platitudes - this isn't a campaign orcommunity rally! We can talk all we want about beaking the limits of present technology; and it's so easy to do so from the armchair.
I think I already brought up the use of direct drive servos; and yes - they are used for metalworking machinery. The rub is that access to this stuff is not cheap. Take the encoder. We can readily purchase a 14-bit encoder... great, until you consider direct drive motors such as NSK's MegaTorque have a 2.6 MILLION step encoder; 160 times the precision of the 14-bit encoder. Where do we buy THAT, easily, or cheaply? You can buy a surplus MegaTorque and matching drive, but you're now in the $2000+ category. Heck my twp homebrew routers didn't cost that much! It would be hilarious to bolt something like that, with a repeatability of 2 arc-SECONDS, onto my machine with an accuracy of .002"/ft. (not including flex.)
I have a few items that I've experimented with for driving a 4th axis. I do have a belt-driven 4th axis bought on eBay a while ago that has a 6:1 reduction. It's fine for most stuff really; though the plastic shaft bushing and large timing pulley ring - held to flats on the spindle with set screws and used to fix the shaft to the housing - lead me to believe that this is best suited for wood and very light intermittent aluminum milling. But it was cheap enough and served its purpose. Also the hollow shaft has a 5C taper so those collets bolt directly on; no need for a chuck! I will rebuild this with ball bearings and better timing pulley mounting.
I have a few planetary gearheads as well. The one I considered is a 30:1 reduction, with an accuracy under 8 arc-minutes. Seems to be good enough; that would be 6000 steps/rev using a stepper and full step. Already has needle-roller bearing supporting the output shaft and 1/4" shaft clamp for the stepper. I could mount a chuck directly to the shaft with a home-made chuck plate and be done with it; but I won't have a through-shaft.
My final option, and the one I think now I'll be using, is a new-old-stock Harmonic Drive gear assembly. The component set is already mounted to a housing. The input shaft (which has a 20mm though bore!) is supported by two deep-groove radial bearings, and the output plate is supported by a heavy-duty crossed roller bearing (which BTW has a dynamic load rating of over 3000lbs.) It also is 100:1 ratio, zero-backlash, and a positional accuracy of 1 arc-minute or less. All for under $300, shipped (from Korea!) You can't even buy the crossed-roller bearing for that price! Everything is completely sealed fromt and back. I will get for now a block of aluminum to make a housing, that will allow me to position the stepper motor (or servo when funds allow) on either side of the housing, so I can use the rotary on-the-flat or vertically. Really even if the motor was on the same side as the chuck I don't anticipate it getting in the way, but I'd have to shield the motor from chips and possibly coolant.
So there's really nothing earth-shattering here. Heck, harmonic drive gearheads have been around for 50 years or so. There's no harm in discovery, if you are financially capable of doing so and have the time. But time is the one thing you can never ever get back. Not saying that such pursuits are a waste of time; but they can come at a irretrievable cost.

Hi Ian

I feel like a fly on a giant's dinner plate

How about a nice juicy steak instead of a fly? Pass the mustard. :-)
7,000 optical lines on a disk is well within industry capabilities today. But a disk of 100 mm diameter with a thru-bore of 35 mm - that is harder, as the photo-lith optics have to be much bigger. Yes, I am sure it is available, but the price would be much higher. It would be from an 'old technology' company, and would come complete with precision bearings, armoured casing, etc. The encoders I am thinking of use YOUR bearings - that is why they are much less expensive.
I had better make it clear here: the big thru-hole optical ones with bearings and casing use a very similar technology to make the optical element; the extra expense comes from the bearings and housing. You can also get big magnetic ones from other suppliers - quite expensive too!
When you buy a DC Brush servo motor with integral encoder from any of the mainstream suppliers (eg Baldor, Reliant, EG&G, etc) , what you get is one of the encoders I am suggesting. An encoder wheel is mounted on the rear end of the motor shaft, and the read head is mounted on the bearing support plate. A plastic cover clips over the lot. Typically the encoder will be a HEDS unit, originally uniquely from Hewlett Packard, but they spun their electronics section off to form Avago many years ago. Today, they have competition (of course).

How does the controller know if the disc is rotating forward or backwards?

The encoder puts out two trains or channels of pulses (on two wires), usually called A & B. A single encoder disk with two read heads, with the second offset slightly. (This is stock standard acros the whole industry.) It is easiest to explain if we take the leading edge of the A channel pulses as the timing reference.
Now, does the leading edge of the pulses on the B channel precede or follow those on the A channel? If they lead we might say the encoder wheel is going forwards; if they trail it is going backwards. All the rest is bits of electronics and/or micro-processor controller. The bottom line is that we can sense both direction and rotation by using those two channels.
Some encoders also can put out an Index Pulse. This is a short pulse on a third wire, and can be used as an alignment reference. It is currently used by Mach3 CNC control software to sync the rotation of a lathe chuck to the movement of the leadscrew to give you threading. Some people never get this going - they forget to enable the use of the sync pulse in Mach3! In Mach4 it is rumoured that the threading may be able to use the A/B pulses for greater precision - which would be very nice.
Yes, your digital caliper does exactly the same thing.
Everything else is as you describe.
Cheers

Hi Roger, in an attempt to KISS.......while I count the poles to get from A to B, I know which direction the car is moving in by the fact that at the start the driver was instructed to go forward.....so the count begins.
At some point the car stops.......the count stops at that pole.......the driver is instructed to go in reverse.......I start to subtract the pole numbers BECAUSE I KNOW WE ARE GOING BACKWARDS.
Going forward again I add to the pole count because I know we are going forwards etc etc for every move, and this continuous for the whole of eternity, presumably until I fall asleep and so lose the pole count, but if I don't sleep the pole count will be exact to the number.
Could this be simplified for an encoder position interpretation by having a 7,000 + lined disc with lines etched on glass etc and a micro processer do the sums for each time the servo is activated either in the forward or reverse direction to count the pulses and either subtract or add them to the position status?
In other words, doing the pole count means I only have to count the poles passing the window and not the direction they come from.
To my reasoning this is simply a count circuit with an additional function that did the +- sums each time the driver went forward or reversed.
Would this then just break down to a G code (or A code?) that moved a servo motor in a particular direction until a line count had been reached and a micro processor to keep score of the line count, either more or less, as the code made the servo motor go forward or reverse?
If the servo motor strategy was the key to directional reasoning for the micro processor to act in the line count, I think it would be a simple circuit board with just an optical line counter read head and a microprocessor if that is all it takes.
That would have to be attached to the final drive as the output is what is being acted on not the input.
I was quite impressed with the planetary gear head idea, (gears without backlash.... I thought this was a simple sun and planet assembly) and have been led to believe that it is totally backlash free, but at a cost as it's intended for very highly sophisticated equipment that justifies the cost, unless sourced on a piece of used equipment at a NASA junk sale.
As far as I can recall, it was stated earlier in the thread that anything with gears in mesh requires clearance to run smoothly, and also has cyclic variations, so I can't come to terms with anything that has gears in the design and not think backlash, however small.....when the quest is for the Holy Grail, a plastic wine cup is not a compromise solution.
Ian.

Hi 6911
With respect, $161 is not all that low-cost!
But your assumption is wrong. These encoders are designed for precision applications. Yes, you can get up to 40,000 quadrature edges per rev today. You suggest they are not repeatable: I don't understand this claim and challenge it. Find me a recent reference which supports it...

The single magnet encoders being considered in this thread are not accurate, and their datasheets state a minimum of +-0.2 degrees of error, along with several counts of noise and hysterisis. This is a fundamental property of trying to sense the angular position of a single 50 cent magnet.
Here is a typical magnetic encoder datasheet (in fact, a very expensive renishaw magnetic encoder) that explicitly states its poor accuracy: http://rls.si/media/pdf/en/65be32569...f065fc2e18.pdf
I think its fair to assume that a part is not designed for precision applications if nonlinearity/accuracy is entirely ommitted from the datasheet, as is common for many magnetic encoders.
Other types of encoders are designed for precision applications, but low-cost and high count only comes in single-magnet form.
A glass scale encoder generally has low counts because each line must be individually etched/printed: US Digital | Products » E5 Optical Kit Encoder
The genuinely accurate encoders are essentially linear scales wrapped into a circle, and are a whole different category in terms of cost.
RESR rotary (angle) encoder system

Could this be simplified for an encoder position interpretation by having a 7,000 + lined disc with lines etched on glass etc and a micro processer do the sums for each time the servo is activated either in the forward or reverse direction to count the pulses and either subtract or add them to the position status?

This will not work, because the entire point of an encoder is to measure unexpected movement. Machining forces will backdrive the servo (especially if you are running a small reduction like 20:1) and you will have no idea what direction the table is turning.
I would not consider building an encoder to be a viable solution.

How about a nice juicy steak instead of a fly? Pass the mustard. :-)

Mustard on steak?! Sacrilege!

Hi Ian

Could this be simplified for an encoder position interpretation by having a 7,000 + lined disc with lines etched on glass etc and a micro processer do the sums for each time the servo is activated either in the forward or reverse direction to count the pulses and either subtract or add them to the position status?

In essence, yes to all of this. In detail: that is what Mach3 and a suitable servo driver unit do. I will spell this out simply: forgive me if some or all of this is already known to you.
Suppose I want to mark out a square of side 1 with my CNC, starting at 0,0 . I write a program which consists of four instructions:
g0 x1 y0
g0 x1 y1
g0 x0 y1
g0 x0 y0
Mach3 and a stock servo driver unit do everything else.
Now you could have the SW and HW do this in several ways, but it has become sort-of accepted that the interface will model a stepper motor. Let's assume for now that a single 'step' represents a distance of 0.01 (whatever units).
Mach3 issues the necessary Step & Dir signals to draw the square. For the first g-code instruction it will issue 100 step pulses to the X driver, with the Dir signal set for positive travel. The X driver box will make the X motor go around to do this, and the X driver itself will keep track of the encoder feedback. If the X driver works, all is well. If there is a problem (you ran into a brick wall?), then it will reply with a fault signal and the whole system will stop.
That means that Mach3 runs in an ideal world, only worrying about a single conversion factor between your dimensions and encoder pulses (for each axis). If your system uses stepper motors, than the X Driver simply creates the necessary signals on the stepper motor wires. That's simple, but stepper motors only go so big. beyond that you go to DC Brush motors, DC Brushless motors, and AC drives.
For these other motors, there is a microprocessor in the driver electronics which does all the housekeeping to make sure that the real world complies. That is where the encoder comes in: the micro checks that the right number of encoder pulses comes back. Where is the encoder mounted? It doesn't care. What if the motor overshoots, or you want to change direction? The micro tracks the A&B pulses so it knows which direction the motor is turning. It can accelerate the motor, and it can deccelerate the motor.
How many encoder pulses will be required for a travel of 1.000? That is determined by the system calibration. It might be something like 1,384 encoder pulses: it doesn't have to be a nice number. So Mach3 will really issue 1,384 pulses to travel 1.000, but you don't worry about that.
So to summarise all of that: 'a simple circuit board with just an optical line counter read head and a microprocessor if that is all it takes' as you wrote.
Cheers
Roger

Hi 6911

The single magnet encoders being considered in this thread are not accurate,

Look, I am sorry, but this bears no relation to what we are discussing. I know about those single-pole magnetic encoders, and they are simply not relevant here. We are talking about optical encoders here.

A glass scale encoder generally has low counts because each line must be individually etched/printed

If you call 10,000 lines a low count, that's fine by me.
They are etched all together using modern photo-lith processes which can get down to under 20 nanometres in a memory chip. In an 10k line optical encoder the lines are at a spacing of about 14.6 microns. Relative to the state of the art in photo-lith, that is huge, and easy. (We know it is easy, becasue the encoder wheels are not all that expensive.)

The genuinely accurate encoders are essentially linear scales wrapped into a circle, and are a whole different category in terms of cost.

Yes, I know about the Renishaw encoders, and I have priced them too. Lovely stuff, but way overboard for most machines!

This will not work, because the entire point of an encoder is to measure unexpected movement. Machining forces will backdrive the servo (especially if you are running a small reduction like 20:1) and you will have no idea what direction the table is turning.

I can see you and I may have to agree to disagree here.
First of all, the whole point of the encoder is to tell the controlling power driver where the motor or shaft has got too. The word 'unexpected' is not applicable. The encoder provides feedback which is a crucial part of the servo control loop.
Second, the encoder senses any movement, not just 'backdrive', so the controlling power driver can correct for it. The encoder not only senses movement but also the direction of the movement.
Third, this sytem works just fine in tens of thousands of machines around the world. This includes little hobby machines, large routers, converted Bridgeports, and going up from there to big production machines. If the machine is not running on steppers, it uses encoders for feedback. For most low to medium cost machines, this means optical encoders.
Cheers

Mustard on steak?! Sacrilege!

French Dijon whole-grain mustard, please!
If you haven't tried it ...

Hi Ian

the line spacing for 8,000 lines on a 70mm diam would be .028mm at the edge of the disc, so a very fine pointed diamond scratcher would be required.

That's 28 microns between lines at the edges. To get that reliably, you would need a diamond point starting at about 5 microns, and you would need the surface flatness and smoothness of the disk to be well under 5 microns. The result, compared to photo-lithography, would be poor at the best. The flex in the disk would be a problem for a start. Bearing slop while engraving would also be very coarse and problematic.
The old idea of a ruling engine for making gratings was great in its time, but a high performance ruling engine able to do what you want would be a 10 year undertaking in itself. Unless you could buy an obsolete one discarded from National Standards. These days it's all etched chromium. Somewhere in my collection I have a discarded chromium on glass linear scale, given to me many many years ago by a friend at our National Standards labs. It was already obsolete.
Cheers
PS: you would probably also need to run the ruling engine in a temperature-stabilised lab - held to about +/-1 C.

We used the Cone Drive "zero" backlash double enveloping worm gear units in the past. We had two problems with them, both self inflicted. We applied too much force to them which compressed the belville washers on the worm. Also didn't give enough thought to the cost, a bit pricey.LOL
Dick Z

Hmmmm, so a diamond pointed scriber won't do the job....LOL......and my garage has a temperature differential of whatever the ambient decides to be.....drat.
Maybe I'm attacking the problem from the wrong angle.........why attempt to make the encoder when all I want to do is move a spindle forwards and backwards in very small increments.
Perhaps the worm drive should be looked at with a view of making it backlash free......looking outside of the box for a solution.
Every time I examine the worm drive I gravitate to a geared set-up with multi compounded gears sets to get the resolution, and that is simple to achieve as it's been done thousands of time before.
Gears are my favourite form of movement transmission as they roll and do not slide, and ever since the evolution of the windmill and water wheel with wooden pegged gears, the toothed wheel has always been the fall back solution.
Even the solid state digital time piece, which killed off the Swiss watch industry overnight, has gone back to geared movements with silicon chip movers.
So perhaps the answer is staring us in the face.....it doesn't pay to ignore history.
I quite like the idea of the encoder reading the poles as they pass, but what if it misses one pole......gears don't go out of mesh or miss a tooth.......the jury will stay out until a solution is found.
I've come to the realisation too that in a worm drive the worm/wormwheel combination is under a constant pre-load condition as it drives forwards, and that means it must also be under a pre-load condition when it's driven back or you'll get a "taking up of the slack" as the combo comes under load each time it goes forward or backwards.
Having a constant load present when no load is apparent will just increase the wear rate with the soft materials of the combo.
You could make the worm and wormwheel from hardened steel relying on the oil wedge to keep the two materials from wearing each other out, especially as the drive is a slow one, and in my opinion this might give a closer contact with less backlash than the conventional materials.
Having hard materials slide on hard materials is not new......the overhead valve gear on most cars, where camshafts press directly on the vale ends do so with the hardened camshaft lobes sliding on the valve head with a cap on top, separated by a splash oil film.

But that is a sliding solution, whereas the geared method is a rolling one, which is more ideal as it can be done with relatively little hi tech input using practically all off the shelf items.
The more I look at gears the more I think it's in the realm of KISS and an encoder is very hi tech to say the least.
Ian.

We used the Cone Drive "zero" backlash double enveloping worm gear units in the past. We had two problems with them, both self inflicted. We applied too much force to them which compressed the belville washers on the worm. Also didn't give enough thought to the cost, a bit pricey.LOL
Dick Z

Hi Dick, double cone drive?......does that mean the worm is made in the form of two cones and one is pressed against the worm wheel tooth flank to pull the other coned worm against the opposite flank?
Wow, for the forces involved in a 4th axis drive and the light loads this might be the simple answer.......but I would make all of the components from hardened steel with a heavy gear oil to do the load carrying/transmitting to keep the steel surfaces apart.
Or.......I think the worm halves could be made from bronze as they can be considered sacrificial and easier to make and replace.....the worm wheel would be hardened and ground steel to keep it in shape and form.
Provided the preload Bellville washers on the worm were not too hard to make wear a problem I think the drive could be viable for the 4th I need......backlash would occur if the drive was overloaded, but it has distinct possibilities, and the backlash would only be a problem when the cut was in a roughing mode which would not occur when a finishing lighter cut was being made.
The worms could be cut in the average lathe very easily.......easier to replace the worms than the wormwheel.
Ian.

Handlewanker,
The little table I built is a worm drive with about a 6 inch worm wheel. The wheel has 180 teeth and is driven with a 10 tpi screw. One rev of the stepper makes 2 degrees, so 1 step is .01 deg in full step. I run the stepper driver in 1/10 step mode which takes it to .001 degrees per step. Of course I could drag out the old fractional step drives that do 25,000 steps per rev.... But that's just toooooo much.

I quite like the idea of the encoder reading the poles as they pass, but what if it misses one pole

Since most every DC motor driven CNC successfully uses encoders to control the movement, I don't think that is a real big problem these days.

Perhaps the worm drive should be looked at with a view of making it backlash free

And many have been the attempts to do just that. None have been what you would call entirely successful yet. Mind you, some of the more expensive ones do come close. But maybe not on my wallet.
It would seem that the biggest advances in making worm&wheel combos has been the move to make both the worm and the wheel AFTER they have been mounted on their shafts. Otherwise you get tiny variations in radius. That does little harm to the rotation, but it plays merry hell with the backlash.

double cone drive

Just a fancy name for a shaped worm (conforms to wheel radius) and worm (teeth conform to worm radius). Nice, but hardly revolutionary. Better shock loading though.

You could make the worm and wormwheel from hardened steel

That is done today. The expensive ones are ground in an NC grinder after being hardened. There's a video on YouTube I think of that - or maybe it was on a vendor web site.
Dreams ... an NC grinder ...
Actually, mostly possible with some serious programming on a 4-axis mill ... but cover ALL the bearings and ways and screws!
Cheers

Ian (and others)
Barden (Schaeffler) have a number of good publications on bearings, including 'HANDLING AND MOUNTING PRECISION BALL BEARINGS' at
Barden Precision Ball Bearings *|*Publications *|* HANDLING AND MOUNTING PRECISION BALL BEARINGS
There's more there as well, such as
BEARING IN MIND: PRECISION BALL BEARING BASICS
BEARING FAILURE: CAUSES AND CURES (marvelous photos of dead bearings)
Super Precision Bearings Machine Tool Products
Double Direction Angular Contact Thrust Ball Bearings and Cylindrical Roller Bearings
A new Level of Performance in Angular Contact Spindle Bearings
All fascinating stuff.
Cheers

ConeDrive is the brand name, after the founder Samuel Cone mwho developed the manufacturing process for the drive at the time. The term for the drive is "double enveloping worm gear." And yes, first conceived by Leonardo DaVinci in the 1500s.
Hyperbolic and wormoid gear drives for rotary gear actuaros for automotive steering system and for aerospace flight controls.
Some other ones... The Spinea trochoidal gear reducer (still can't figure this one out.)
Reduction gears | SPINEA - Excellence in motion
KAMO reducer using steel balls. They claim pure rolling motion in their reducer.... Download the catalog to see the guts. They also produce roller pinion similar to Nexen.
Indexing Tables & Positioning Solutions - Precision Motion (COFIL) Ltd - Zero-Backlash Gearheads
KAMO SEIKO CORPORATION?- KAMO - [HPR]
There's some on eBay now! (must fight the urge to spend...)
Kamo Gear Head BR85UH 30g 11K4 Ratio 1 30 | eBay

Another thought; If you used hollow rollers (i.e tubes) , they would distort, instead of the outer rings.

Look up the formulas for epicyclical gearing. There are some +1 & -1's in the ratio calculations to compensate for planet rotations, I think. I have it here somewhere. ????
Dick Z

Look up the formulas for epicyclical gearing. There are some +1 & -1's in the ratio calculations to compensate for planet rotations, I think. I have it here somewhere. ????
Dick Z

I was trying to work it out from first pinciples (always doomed to failure ).
+/-1's sounds about right
Thanks Dick.

I think the ratio thing is explained under epicyclical gearing on wikipedia.
The patent # for gear bearings is US 6,626,792 B2 Sep. 30,2003
Dick Z

I made a new design of the traction drive with angled rollers and preload with belleville washers. The spacer is a plastic disc with slots for the planets to roll in and is placed between the input and output sun! Think this would work?

Attached


IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617141711236.jpg)

I made a new design of the traction drive with angled rollers and preload with belleville washers. The spacer is a plastic disc with slots for the planets to roll in and is placed between the input and output sun! Think this would work?

Good, very compact but...
With the tapers arranged like that there is a differential ratio across the taper, so there would be sliding friction as the roller turns. (That's why my rough design here (Backlash free rotary table) looks the way it does).

Yes i know, but i would need feedback from the output anyway
Another thing i have been thinking about is to be able to switch between 1:1 and 60-100:1 to be able to do turning and indexing on the same spindle/setup. But dont know how to to do it and to keep the indexing backlash free.

Another thing i have been thinking about is to be able to switch between 1:1 and 60-100:1 to be able to do turning and indexing on the same spindle/setup. But dont know how to to do it and to keep the indexing backlash free.

How about some sort of clutch to connect the input to the second stage planet, and a mechanism to disconnect the first stage planets from the outer ring? (i.e. remove the pre-load?)

Hmm, perhaps i could disconnect the grounded annulus from the housing so it could rotate free and lock the rotation of the planets!?
Check my sketch for another possible simple solution. The problem with that is it would have to be manual to be resonable.

Attached


IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617141811237.jpg)

Not to throw a wrench into the works, perhaps someone has already mentioned this, but I had used one of these a long time ago on a extremely high accuracy (±1 arc sec) air bearing rotary table:
http://www.onviollc.com/dotnetnuke/C...1/Default.aspx
It worked very well for the application, and I was wondering if it could be recreated in a home shop with low cost roller bearings or ball bearings.
Keith
NEATman

Actually i was reading about a DIY version of a DIY cycloidal reducer (for use with a telescope) yesterday! Interesting stuff, well worth to check into further!
http://www.telescopeshop.co.za/telescope_drives.htm

I made a test to see what happens with the gearing ratio when i lock/release certain parts of the assembly! I used gears to see that the gearing is correct.
When i lock the input annulus to the housing i get about 197:1 ratio.
When i release the input annulus i get about -5:1 ratio, but then i would need to lock the planet carriers orbital rotation to transmit any torque if i'm thinking correctly!?
This should be easy with 2 brakes, 1 for the Input annulus and one for the planet carrier that you switch between.
If i would release the input annulus and lock the planet rotation i should get 1:1
But that is a little harder to acomplish! Got any ideas?
Edit: I got an idea! What if you release the input annulus from the housing and connect it with the output annulus, the difference in gearing should lock up the hole assembly and give 1:1!?

Attached


IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617141911238.jpg)

I asked a friend if he wanted to make a hypocycloid reducer at work on the lathe... He want to build a gun turret for airsoft so just i made this. It's 49:1.
I will post pictures when it's done!

Attached


IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617141911239.jpg)

Here is the models and drawings for the gearbox. It's made for a NEMA 23 motor. The dimensions with parentheses is for measuring the cam and pins.
I have not worked out the tolerances and have no clue if it will work yet

Attached

Very nice Zoidberg, I just finished making my own 100:1 reducer based on your rhinoscript and a lot of help from ZincBoy via PM.
I downloaded your files, and Im importing them into SW to see how they fit. There is bearing at the input in the house thats not drawn right?
One non related question, how do you mate these parts to simulate the cam action in SW?
Here is a photo of my reducer in its previous incarnation:
IMAGE(http://img402.imageshack.us/img402/723/img4482x.jpg)
I made two attached 10 lobes cams, the funny thing about this setup is that the cams cancelled each other and the output remained fixed. The current one has a 10:9 lobes relation and it works just fine as 100:1 reducer
Thanks!

No bearings.. the eccentric is brass and sits directly on the stepper axle I use gear mate...

Thanks will try that.
Here is how you calculate the ratios (thanks to ZincBoy)
First Stage
a) # Pins 8
b) # Lobes 7
Second Stage
c) # Pins 7
d) # Lobes 6
0.020408163 Ratio = (b-a)/b*(c-d)/d
49:1 Inverse = 1/Ratio
Pablo

Ok nice I have just tried with simulation hehe

That formula is incorrect! I think it's missing the differential part of the gearing! :P
I'm writing a program. You enter max and min lobes and a target ratio. Then the program list all combinations (with error)... But it's not worth much without the correct formula If i get it to work correctly i will sort the list by error!

Attached


IMAGE(http://factorydaily.com/fdattachs/fdattachs6/112617142011240.JPG)

Acording to this page: http://www.darali.com/page17.html the formula it's OK but since I'm the pupil and not the teacher here I'll watch for your developments
Im quite happy I went from zero knowledge to a 100:1 reducer build in less than a week thanks to this thread, you and ZincBoy
Pablo

Add new comment

Images
More information
  • Files must be less than 2 MB.
  • Allowed file types: png gif jpg jpeg.
Documents
More information
  • Files must be less than 2 MB.
  • Allowed file types: zip rar.