A and C axis for a 5 axis milling machine

[RCaffin]

Hi Goemon

I suspect that you have misread the HAAS specs, which are typically 30 arc-seconds, not 30 arc-minutes.

Cheers
Roger

[dharmic]

Nice score. I like the hypocycloidic setup. It's stronger and easy (relatively) to DIY because the geometry is all quite friendly to round tooling on 3 axis machinery. Downside is there is some sliding action rather than, IIRC, the pure drop-on/lift-off engagement of teeth on a harmonic drive. Minimal, but it will eventually lead to wear.

118:1 might be a bit high reduction, but you can always drive it with a servo instead of a stepper and run the servo fast. IF the slow output causes you grief.

Do let us know how you get on with it, interested to see!

[handlewanker]

As I'm new to CNC I'll plead my ignorance and state that if you are using a 4th axis for pure simple rotational machining of a shape you won't be going much past 360 deg of rotation....as you would if it was a turning operation...….. but winding back to 0 deg at the end of the turn......as I see it.

However if you were doing a continuous turning type operation then backlash would become a problem.

For the first type of operation it would be practical to spring load the 4th axis spindle to make it always push forwards and in that way there can be no backlash provided the spring loading was not overcome by the cutter...……….OK, so the stepper has to also overcome the rotational force from the spring when it reverses , but it does give you no backlash for the machining operation.

For the finishing operation status the final cut, which is light, will not be backlash prone so a spring does not have to be hugely powerful

It does mean you would have to wind the spindle back to 0 deg..... the beginning...... at the end of a cut....every time........…..such a price to pay to be backlash free...….you have to do this with the XYZ slides anyway so for a 4th it would just be the same.
Ian.

[dharmic]

Different controllers do it differently. Some reset to zero, some take the shortest path - so if you're doing multiple operations which are repeated at indexed positions it will usually get to the last one of a set, then continue rotating in the same direction to the first of the next set.

Actually better for backlash issues because the axis never changes direction so the pressure is always on one side of the system and the play in it never gets encountered.

[handlewanker]

Different controllers do it differently. Some reset to zero, some take the shortest path - so if you're doing multiple operations which are repeated at indexed positions it will usually get to the last one of a set, then continue rotating in the same direction to the first of the next set.

Actually better for backlash issues because the axis never changes direction so the pressure is always on one side of the system and the play in it never gets encountered.

Hi, I see your point, but with any uncontrolled backlash it means you virtually have a "loose" spindle all the time and that equates to a backwards and forwards lost motion that can cause a cutter to dig in and pull the workpiece around such as will happen in a climb milling operation.

I doubt that a small amount of backlash on a small diam workpiece will affect a machining operation too much if it always goes one way, but for incremental positioning of the cutter on the workpiece then the larger the diam the worse the positioning becomes.

I can quite see why it is better to reset to zero and go forth again, but as it would be simpler for small batches to go back to 0 deg then the extra time factor is just a few seconds at the end of an operation.

A spring loaded spindle would be like having a weight on the end of a rope down a well…..the weight of the water in a bucket will always create a force on the winding mechanism in one direction.
Ian.

[dharmic]

Fair point. I could take my example, indexing only, and add a brake with a little drag on it then we're fine - the rotation will always be against the drag and the axis locked whenever cutter meets the meat.

But that is pretty useless for continuous ops, as you say the loosey goosey means the tool can drag the job any which way. Not to mention that continuous ops are likely to have rotational direction changes too. So backlash becomes an issue. A spring load would fix this provided the spring force is in excess of the cutting force. This brings its own issues of course, in terms of extra drag on the motor and wear in the system. Whether they are significant or not, who knows.

The purist in me says that the spring is a dodgy hack. Design the damned thing for minimal backlash and wear in the first place, then build it right, and let's have no more of this "just push on it til it stops wobbling" nonsense. But the purist in me is a jerk, and the pragmatist in me likes any idea that says "get the job done cheap, fast and well". In the end, the purist usually gives into the pragmatist.

[handlewanker]

This topic of a backlash free 4th axis has been thrashed to a million bits some time back.....my preference for that was to have a worm and worm wheel drive but with the worm in a spring loaded swing housing to get it to make metal to metal contact with the worm wheel teeth all the time.

Provided the worm teeth don't bottom in the worm wheel...….clearance needed....then you need to have a resilient drive to cater for the tight and loose spots on a not so new rotary table or 4th axis....whatever.

A worm and worm wheel type of 4th axis is almost good to go but the adjustment of the worm is the hold back factor as you need to cater for the tight spots too.

I have a small 100mm rotary table and the worm is adjusted by a cam housing.....at best it is almost there but for climb machining on the go the backlash would be catastrophic......it's designed to be used in the locked position

SO...…...I am in the process of making a 4th axis and will be using a spring loaded worm/worm wheel design to get rid of the backlash.....belt drive does not appeal to me etc etc, mainly because the reduction is too low at about 1:6 at max.

Basically the worm itself is contained in a swing housing with the usual thrust bearings and pivots at one end under spring pressure to keep it meshed with the worm wheel...…...that's it, not rocket science by any means.

You only need to have a few thou....10 at most.... of up and down movement to make the worm sit tightly in the worm wheel as all you are doing is to cater for the variations of the worm wheel fit to the worm and make it fit closely all the time without binding.

The pressure of the cut will be in a rotary direction and that means the teeth of the worm wheel will impact against the worm teeth and so will not be influenced by cutter pressure in any direction forcing the worm out of contact.....the spring will just keep it in contact.
Ian.

[RCaffin]

Mach3 has several options for handling the 360 degree rollover. Some of them are either terrible or unreliable. I spent some time verifying this.
If you keep rotating in one direction, eventually you run out of numerical range.
For further discussion on how to handle this safely, please see

Coordinate Spaces - a Guide
Coordinate system.

Disclosure: I wrote it in order to get it all sorted out in MY head.
Cheers
Roger
PS: I reckon the best solution re backlash is to use an HD so you have none. Biased.

[Goemon]

Nice score. I like the hypocycloidic setup. It's stronger and easy (relatively) to DIY because the geometry is all quite friendly to round tooling on 3 axis machinery. Downside is there is some sliding action rather than, IIRC, the pure drop-on/lift-off engagement of teeth on a harmonic drive. Minimal, but it will eventually lead to wear.

118:1 might be a bit high reduction, but you can always drive it with a servo instead of a stepper and run the servo fast. IF the slow output causes you grief.

Do let us know how you get on with it, interested to see!

Will do.

Hopefully, I won't have to worry about wear for a while as I got a new one and, it is rated for many times the force I will actually need. These things seem to have been designed to move some seriously heavy gear. I'll keep an eye on it though.

I have some 400w Yaskawa AC servos and drives that I will use if my steppers prove too slow. I think it will be ok for indexing based on my back of an envelope math.

There were two available and I kinda wish I would have bought both but somebody else snagged the other one while I was procrastinating.
I like that it has a built-in rotary table bearing and mounting surface. I am hoping that, with a few extra drill holes, I'll be able to directly mount an off-the-shelf 6" or 8" face plate and chuck.

[Goemon]

I only saw this after I bought my gear head (typical) but for those looking for ideas for making 4th and 5th axis heads (that rotate the spindle instead of the table), these industrial robot wrists seem to be available used for around $400:

https://www.ebay.com/itm/Yaskawa-Mot...item3d65a081f8

This is where they remove all those used harmonic drives and cyclo reducers from, so they come with two precision gear heads attached. I never liked the idea of building parts like that as a diy project but I could make one of these wrists work easily.

[dharmic]

Agreed on your best solution Roger. I don't know what range Mach3 uses - I was figuring on doubles or 64 bit integers which (to me) would suggest the sun would burn out before you ran off the end. A little disappointing to see this is not the case.

[Goemon]

Fair point. I could take my example, indexing only, and add a brake with a little drag on it then we're fine - the rotation will always be against the drag and the axis locked whenever cutter meets the meat.

But that is pretty useless for continuous ops, as you say the loosey goosey means the tool can drag the job any which way. Not to mention that continuous ops are likely to have rotational direction changes too. So backlash becomes an issue. A spring load would fix this provided the spring force is in excess of the cutting force. This brings its own issues of course, in terms of extra drag on the motor and wear in the system. Whether they are significant or not, who knows.

The purist in me says that the spring is a dodgy hack. Design the damned thing for minimal backlash and wear in the first place, then build it right, and let's have no more of this "just push on it til it stops wobbling" nonsense. But the purist in me is a jerk, and the pragmatist in me likes any idea that says "get the job done cheap, fast and well". In the end, the purist usually gives into the pragmatist.

The existing backlash free worm gears I saw used a double bevel design so preload could be applied in opposite directions at the same time. I.e. One shaft for each direction like this:

Worm and worm wheel running backlash free - OTT 2 half worm concept distributed by Allytech

It makes sense as backlash problems occur most with changes in direction.

The bigger issue for me is if it's worth time given the relatively low cost of quality used gear heads on ebay. Maybe you could bypass the worm gear and couple the rotary table with a shaft output harmonic drive or cyclo gear.

Is there any advantage to worm gears over harmonic or cyclo?

[RCaffin]

Very biased opinion: Harmonic Drives work and don't have any measurable backlash, while every other design is either weak (spring-loaded), is subject to serious wear issues (worm&wheel), or extremely expensive (and still has backlash). I spent a lot of time trying to find an alternative to an HD, and failed.

However, this is 2018 and robots have been in use for decades, so that there is now a thriving second hand marketplace for parts stripped out of recycled robot arms. This is become a very fine source of inexpensive HDs (via eBay).

Cheers
Roger

[dharmic]

Roger, you don't mention the hypocycloidal drives in your list there - did you study them with the others in your hunt back in the day?

[Goemon]

Roger, you don't mention the hypocycloidal drives in your list there - did you study them with the others in your hunt back in the day?

There is no reason to generalize about what types of gear mechanism have backlash because it is possible these days to design backlash free versions of gear mechanisms that otherwise have backlash. There are various worm drives and cyclo gear heads that either have zero backlash is near enough to make no difference. Nobody here has tested all the options. There is just too many.

It's a matter of common sense. The idea that harmonic drives are the only option for a high quality rotary axis is nonsense because the vast majority of high end 4th and 5th axis assemblies do not use harmonic drives.

After going through the research and shopping process, most of the industrial robots I saw use a mix of cycloidal and harmonic gears with some of the larger heavier duty robots using just cycloidal for their superior shock absorption, low vibration and high moment loads. If a drive is good (and precise) enough to go on a precision industrial robot, they are good enough for a diy CNC build.

The best way to look for a deal on a suitable drive is to first search for "backlash free" or "zero backlash" gear. Then, make a note of some of the brands that make them and then search for those brands on eBay. You can also try searching for "robot gears" as well as harmonic drives and cyclo gears. I found a bunch of deals on harmonic and cyclo drives that were not listed as harmonic or cyclo gears.

[dharmic]

I only asked because this is obviously a passionate subject for Roger, something he's spent a bunch of time on - regardless of how he seems to have fallen into the easy trap of "I have a solution that works really well for me now, so every other option is crap".

I'm looking at a two stage hypocycloid gearbox I made sitting on my desk and anyone who's read more than three of my posts know I lean toward them like Roger leans toward HDs. But he spent a lot more time than I did looking at stuff, I am interested in whether he looked at the hypocycloids and, if so, why he finds them less mentionworthy than the HDs. It ain't going to change my mind, but I'm curious.

[RCaffin]

Hi dharmic

True. However, I have read Ms Dascalescu's paper on generating the hypocycloidal profile (with clearance modifications) and tried to look at the weakness's in the approach. It seems to me (and I might be wrong) that the accuracy of the profile is crucial, and could be subject to wear and distortion. Also, the rollers required by the design dictate a certain minimum size to the the whole thing. I mentioned this limit many postings ago when I said that there were some other possibilities when you get above a certain size. For big drives, with precision grinding on hardened material, this is a possibility, but still with some backlash.

A gentleman at another website published some experiments at Hypocycloid Gear Reduction on making a hypocycloidal drive using plastic. Plastic is very forgiving compared to hardened steel. He made the parts 0.001" undersize: it worked but there was some backlash ("less than a degree"). This is to be expected owing to the design of the meshing teeth. They are not exactly spur gears.

There is a report on Hackaday of another attempt, which failed:
"a single gear is placed slightly off-axis inside a ring gear. On paper, it works. In reality, not so much.
[Dan] spent $3000 on a prototype hypocycloid gearbox that doesn’t turn without binding or jamming. The gear was made with incredible tolerances and top quality machining, but [Dan] has a very expensive paper weight sitting on his desk right now."
Perhaps this one was made 0.0001" oversize? Anyhow, the guy went on to investigate why, and found:
"In a hypocycloid there is no backlash because there's always lots of teeth touching. Also, the force is uneven. Torque on either the output shaft or the input shaft causes the shaft to bend off-axis, away from the point of contact between the inner gear and the housing. As soon as this happens the interior gear starts to tilt. A tilted gear is a jamming gear"

It seems that the idea is neat, but genuinely backlash-free versions are not easy to find. To be commercially viable, some backlash has to be permitted. Of course, if I am wrong, some URLs correcting me would be delightful.

So why does the HD work? I think it is because the design stays within the elastic limits of the materials used.

Cheers
Roger

[Goemon]

I only asked because this is obviously a passionate subject for Roger, something he's spent a bunch of time on - regardless of how he seems to have fallen into the easy trap of "I have a solution that works really well for me now, so every other option is crap".

I'm looking at a two stage hypocycloid gearbox I made sitting on my desk and anyone who's read more than three of my posts know I lean toward them like Roger leans toward HDs. But he spent a lot more time than I did looking at stuff, I am interested in whether he looked at the hypocycloids and, if so, why he finds them less mentionworthy than the HDs. It ain't going to change my mind, but I'm curious.

It will come down to the individual brand and model with any type. It is worth reading the small-print to see what the manufacturers consider to be low or zero backlash. A lot of harmonic drives I looked at (with zero backlash headlines) state "less than 3 arc minutes" or "less than 1 arc minute" depending on the required precision (and budget). I.e. In terms of backlash (or lost motion) harmonic drive makers advertise the same levels as other precision robot gears. Not better and, in some cases, slightly worse. E.g.:

High Precision Servo Mount Gearboxes | Harmonic Drive

The top option (for backlash) from this vendor is less than 1 arc minute - same as the Fanuc / Nabtesco cyclo gear heads. More than precise enough for any diy build here.

My guess is that the cheap harmonic 4th axis from China are probably not made so well and produce more lost motion.

I based my decision on the availability of affordable deals in a gear head that is large enough for my needs and rated the high enough loads. I would have been happy with any of the low / no backlash, high ratio, high load gear types.

[dharmic]

Cheers, Roger.

My own one is plastic and is sitting on the desk instead of being used because, like any fast FDM 3D printed thing, it has the precision and finesse of a rabid water buffalo mid-rampage. Still, it works as a demonstrator and the backlash feels like it's in keeping with the tolerance of the print. Yes, the design needs to be right and the manufacturing tightly toleranced but I still believe for a DIY high reduction gear, I'd much rather take on the challenge of a hypocycloid than a strain wave rotor. But that's me and the home build side of it. If I had a choice of the two to buy cheap-ish and they were both capable spec-wise of supporting my application, the HD would be my choice too.

:edit: Goemon, that it depends on the mfr and quality goes without saying. I keep having to remind myself that when people talk about harmonic drives they're usually talking about the class of strain wave rotor gearbox rather than products from the Harmonic Drive company who held the patent for many years.

[RCaffin]

Hi Goemon

A lot of harmonic drives I looked at (with zero backlash headlines) state "less than 3 arc minutes"
I have not found any of those.
Edit: The secret is to search for 'strain wave gear', not 'harmonic drive'. You will also find a horde of expensive market research reports you can buy if you have more money than brains.

What I did find was a Harmonic Drive PLANETARY gearbox which quoted precision and low backlash (standard: < 3-arc-min, optional: <1 arc-min), but a planetary gearbox is not a full HD gearbox at all. On the same page HD claimed for their genuine Harmonic Drives a (repeatability ±4 to ±10 arc-sec), which is far more realistic.

I haven't found any Chinese harmonic drives either, so any URLs to those would be appreciated. I need to update my education.

Cheers
Roger
EDIT:
A market research report commented:Only the high end vendors are able to provide harmonic drives that work, the other units become trash within days. Hum ... ! They also quoted the following companies as being in the field:
Harmonic Drive arc-seconds
Sumitomo Heavy Industries
Beijing CTKM Precision Gearbox
Leaderdrive quote <20 arc-sec and <10 arc-sec for various units: fair enough.
Zhejiang Laifu
Nidec-Shimpo don't seem to quote backlash figures (so draw your own conclusions).
Cone Drive

SWG Solutions offer 180 arc-sec transmission accuracy and 6 arc-MINUTES backlash on their harmonic drives. One has to ask what they are doing or whether they are even worth considering. I don't think so.

Strain Wave Gearing GBC Module | Akiyuki.jp has pictures of a LEGO implementation. One has to award full marks for style!

Cheers
Roger