[QuinnSjoblom]

Thinking about a Y axis cnc lathe build (with live tools) and theres some challenges im running into for the general layout and/or programming. I already have a functional mill turn that is running on mach3 with custom macros and edited posts. I know how to address all the challenges of switching between milling and turning within one program as well as handling tool offsets, but the main issue with this new build is X/Y control that is not 90 degrees. When i look at most y axis lathe layouts, the x and y axis are 45 degrees to eachother and i can see many reasons why they do it that way. This layout just works out conveniently when it comes to rigidity and clearance, as well as the fact that turning forces on static tools are not directed inline with the y axis ballscrew, rather they are at 45 degrees which makes the forces much more manageable for the y axis servo. Ive attached a pic of a haas st20y showing this wedge configuration. Pretty much all y axis lathe designs ive looked at use this same general layout.


So theres 2 options. I build a machine with this common layout and come up with a programming solution for the 45 degree x/y, or i design a machine with 90 degree x/y. At this point i have no idea how to make it work with mach3/mach4 with 45 x/y, but im curious if anyone has any ideas. I know there is actually a feature in mach3 that allows this, but its very buggy and has many issue. Really wouldnt consider it an option. The other possibility is handling it in the post processor. I might be able to configure a post processor that outputs Y axis moves as a combination of x and y, but this method will also have plenty of its own issues. Arcs will need disabled, touching off mill tools will be very tricky, jogging y will not move 90 degrees to x, just endless issues. I really dont know if theres a solution ill be happy with.

As for the designing the machine with 90 degree x/y, i do have a couple ideas for that, but it will be mechanically less convenient than the standard layout. Picture a configuration like the pic i posted, but with the Z axis rails tipped back 45 degrees instead of on the same plane as the sub spindle rails. This would open up the 45 degree wedge to 90 degrees, but the "saddle" would become pretty huge and it would add a ton of mass to the already very heavy Z axis. Im open to any ideas for this project and im curious if anyone knows of y axis lathes out there that use a 90 degree x/y configuration

[Jim Dawson]

An interesting challenge.

Maybe it's time to consider a different motion controller, a real one. My prefered controllers are Galil Motion Control products. It is possible to interface these with Mach3 using the Galil plugin and to call routines in the controller itself. Not my prefered way of doing it, but possible. You would still be held hostage to the limitations of Mach3. To get around those limitations I just wrote my own software, I can make a system do anything I want. I'll give you the source code if you want to consider that route.

The Galil products give you the capability of coordinated motion on up to 8 axes, and can gear two or more axes together at any ratio.

As far as the mechanics of the system, I really don't have an opinion.

[QuinnSjoblom]

A new control software is definitely an option. Im surely not an experienced programmer so i would have a lot to learn if switching. Ive started to become somewhat comfortable with mach3 macros, but it would probably make more sense to learn a less limiting platform before i start spending real money on machines. This lathe as well as the mill i plan to build will not be cheap, probably 10 to 15k minimum each. Epoxy granite base, decent quality ground screws, servos, etc. The mill is going to basically be a copy of a robodrill. I really like the design of the tool changer. Easy to build and lightning fast tool changes.

Ill take a look at Galil and see if its something i would even be capable of learning to use for more complicated setups. The nice thing about Mach is the endless amount of posts and videos that ive been able to learn from.

[QuinnSjoblom]

it looks like camsoft pc based control is what i need to be looking at for Galil motion control, correct?

[Jim Dawson]

it looks like camsoft pc based control is what i need to be looking at for Galil motion control, correct?

Camsoft software is a good option to consider. Very flexible and takes advantage of the full power of the Galil controller. I've not used it, but I have heard good things about it.

[hanermo]

I suggest building it with a vertical y axis.
It will be vastly simpler, and can be re-used if stuff happens in life.

I have scratch built a VMC with a huge work envelope, 1600 mm wide table, 2400 kg in steel.

And a cnc lathe, currently 3 axis, x,z,c.
Spindle c axis with 1:3 8/30 mm HTD8 belt drive on the C, spindle, AC 220V brushless, 10.000 counts, 2.5 kW continuous servo spindle.
The servo gives 30 Nm torque at 1:3 for 90 Nm at the spindle.
This is wonderful !

A HAAS ST10 has 102 Nm peak around 1200 rpm, quite a bit less at start. 11 kW.
Maybe 30 Nm for the HAAS St10, sub 10 rpm, 3 times less.
And it costs 60.000 €.

It works again.
The new lathe saddle, scratch built, 25/5 mm ballscrew, (16 mm acme originally, then 5/8 roton ballscrew), made the first bore today, for a 72 teeth alu pulley for the VMC, 15/5 mm HTD, 24 mm bore nominal, and 23.99 was just right.
This was version 2 of x axis refit.
It took about 400 work hours, clearances for the ballnut are less than 1 mm.
Running on 20 mm Hiwin linear bearings, and 200 mm working length, for a 12x24 lathe originally.
A 10.000 count AC servo, direct drive, brushless, 5 mm rise screw, 0.5 micron theoretical resolution, practical incremental resolution about 0.5 microns.

Originally a Chester craftsman, a very heavy industrial-type small lathe sold by Chester in england, and similar in the US.

My 3-axis CNC lathe runs on Mach3, with a CSMIO-IP-S industrial controller.
The new saddle assy has 2 toolposts, plus a mount for the 4-tool horizontal toolchanger (industrial 30 mm tooling, 3 phase), and a mount for the 7 tool turret (servo, industrial).
About 79 cm in total length saddle.

I expect it to be 6 axis eventually/shortly with live tooling.
Spindle sync for threading is always perfect, with 10.000 ac brushless encoders, and 750W ac brushless servos.
4 Mhz sync rate with the controller.
Z axis ballscrew is 32/5 mm, thrust is about 1000 kgf iirc, around 10 Nm, direct drive.

I can drill 20 mm drill in tool steel with a 6 mm pilot, just, and use fro to push it, and stall the spindle.
Nothing breaks, as the servos stall around 0.01-0.1 mm out-of-position mechanically, in around 0.2 ms theory and 1-2 ms practically as the mechanicals wind up.

--
The very good and heavy servo spindle mount is about 60 kg in mass and 2 cm thick steel plates.

But the very powerful HTD 8 / 30 mm belt drive is not rigid at all.
Sure it can drive 100 kW of power, plus, ..
But the positioning rigidity is no-where near good enough for C axis work, as I hoped.

It is excellent for indexing, making holes at nn degrees, with a 30.000/360 about 0.01 degree accuracy.
The servo is perfectly accurate (led readouts).
But the belt has a lot of stretch/compliance under load.

I am thinking of going onto a planetary gear for the spindle, 1000 rpm outputs (plenty for me) are common.

[hanermo]

The controller is really important.

The cheap chinese controllers won´t do well with multi-axis.

The better intelligent controllers for mach3/4 -- and for linuxcnc -- also all have their own issues.
And independent controllers like the UCxx stuff, and competitors.

I don´t have some of the better newer US intelligent controllers, to test.
I think they are probably very good and very reliable, approaching the CSLabs CSMIO-IP-S but have no idea how well they work in the endless edge cases multi-axis controllers should deal with.

The controller must deal with homing, mpg, fro/sso, probing and limits and threading and rigid tapping and CSS in hw.
Ideally with some digital and analog io in hw for spindle load, encoder tracking, and tool breakage detection.

[QuinnSjoblom]

yep, with my current mill turn machine that i built, i learned a lot about accuracy, rigidity, and speed of a turning spindle/c axis. As you said, timing belts dont have the rigidity needed for a c axis, but unfortunately neither does the servo. Even with a more rigid power transmission, the servo is still not going to give it a rigid hold during index milling unless its geared way down. Im using a dmm 1.8kw direct driven for 3k rpm max and its definitely not capable of holding the spindle rigidly for index milling. I built a pneumatic disk brake for mine and it works great. Also with a brake i was able to use a 5mgt belt which has less backlash than 8mgt.

[Jim Dawson]

I'm crazier than both of you guys I'm going to use a poly-V belt for my C axis drive. Should be close to 0 backlash, but I'm not going to use the motor encoder because I already have an encoder on the spindle. The motor will be a DMM 1.8KW servo but will be geared down about 3:1, it doesn't need to turn at spindle speeds, that's what the spindle motor is for. I'll just clutch it in and out as needed, and I already have a disk brake on the spindle. And I have an 8 axis Galil card sitting in my computer right now and already have the software written for the C axis addition. Just need the time to work on it. I really wish I had a Y axis, but I haven't figured out how to make that mod on my machine.

[QuinnSjoblom]

in your case, i think you have the best solution. with encoder on your spindle and a disk brake, v belt will for sure do fine. Id say your servo and gearing is pretty overkill for just being a c axis, but definitely wont be a problem and if you already have it, no reason to spend more money on a smaller servo. If i had a separate turning motor, i'd probably use a 400 watt servo geared down 10:1 or so for the c axis. One thing is for sure though Jim, with your setup using the 1.8kw and minimal gearing, you will be able to index your c axis a full turn in a few milliseconds lol, as i can do with mine. Although I only went that large because it doubles as my turning motor. Having the encoder directly on spindle is really the way to go.

Through lots of data collecting, ive found all the downsides to having encoder on servo, then driving spindle with timing belt. First issue is, the encoders on these servos are not as accurate as you might think. Extremely good resolution and repeatability, but accuracy of a specific location, not so much. My dmm servo will very accurately repeat to any of the 65,536 steps, but they are not split proportionately. if the servo is at encoder zero and you command a 180 degree move so the servo is now at 32,768, it will not be exactly 180 degrees, More like within a tenth of a degree and dmm even states this in the specs. What ever the error is, it will repeat very accurately, so if you can make use of screw mapping, that error can be compensated for. For driving a ballscrew like most servos do, a tenth degree of error is pretty much insignificant. With a 5mm pitch screw, that would be about 1 micron or half a tenth of linear error. So no problem for ballscrews, but a tenth degree of error on a c axis is not acceptable in my opinion. If geared down, that error is divided, but for a 1:1 drive like mine to keep rpm, the error stays. In mach 3 there was no way to apply screw mapping to this problem so i came up with my own solution using a macro. The macro only works to correct errors for index positioning, but thats all i use my c axis for anyway. This macro i wrote is combined with my axis lock macro for the disk brake that my post processor calls after every c axis move. Before the disk brake locks, the macro calls in the c axis dro, lets say its at 90 degrees for example. Through many measurements i was able to determine that if i want a true 90 degrees from zero, i need to command 89.9102 degrees to compensate for the encoder error. So the macro looks for multiple scenarios after it calls in dro, 0,90,180,270. Once it knows which position its sitting at (currently with error), it can then apply the corrected value i have determined. So it calls in dro and gets 90, it then commands the code G0 C89.9102, then locks the brake, continues with program. That solves the encoder error for indexing, but doesnt help if you want precise continuous movement of c axis.

The other issue with encoder on servo is belt variance and backlash. On a 5mgt belt, the backlash isnt so bad, but ive still mostly accounted for it in my error correcting macro. Before applying the corrected value to a position, the servo is first commanded 4 degrees CCW, then rotates CW to corrected position which i have found to consistently bottom out backlash in one direction. Belt variance is a whole different issue. A 5mgt belt does not have its teeth consistently spaced 5mm with absolute accuracy throughout. This is another variable that can be accounted for in the error correcting macro, but only if the belt is in a consistent position for indexing. If you are using the servo for turning and indexing, the belt will be in a different position each time you go from turning to indexing. In my case, with 72 tooth sprockets and 150 tooth belt, the belt only repeats to a consistent location on sprockets every 25 turns. Right now if i want to make use of my error correcting macro, i have to line up a mark on my sprocket with a mark on my belt to get to the calibrated belt position. After homing, i have to enter a 360 degree move to c axis repeatedly until i see the mark. I do have a way to make this happen automatically, but havent done it yet. either needs a 25 to 1 geared homing switch on servo, or a hole in belt to line up with an optical sensor that enables the homing switch.
Before i wrote my error correcting macro, my C axis was only accurate to about 0.2 degrees when adding up error of encoder and belt. With the macro i get about 0.04 degree accuracy.
Long story short though, with a good accurate ring encoder directly on spindle there is no need for any of this and you also have ability for very accurate continuous c axis motion. In mach3, i didnt have the option of putting an encoder on spindle to control c axis, the servo has to be controlled by its drive and mach3 can only send steps to drive, not accept encoder feedback (open loop).

My new machine is going to be setup very differently. still timing belt driven spindles with servos, But C axis indexing will be locked with curvic couplings or wedge lock from the side. Extremely rigid lock and extremely accurate without relying on an encoder for precise position. The servo and belt will get the position to within 0.2 degrees or so, then the pneumatic coupling will lock up and pull it into precise position. I have a few reasons for wanting to stick with servo driven spindles rather than motor driven with just servo for C axis. For one, It makes it very simple to have perfect synchronization between main and sub spindle. The 2 servo drives will simply be parallelled to one output from the motion controller. The 2 will be perfectly in sync at all times for turning and indexing unless i disable the sub with a relay. It also simplifies the mechanics to not have a separate turning motor and C axis.

[hanermo]

As Quinn said.
100% correct.

The belts and pulleys are not consistent throughout.
And HTD8 has much more error than HTD5.
Dunno why, but it is so.

As we both said, based on expensive experience.

A C axis implies milling with a rotating spindle.
Not just indexing.
Thats what I was hoping for, but not what I got.

My servo accuracy is excellent, unlike Quinns experience.
Optical encoders tend to be near-perfect or have no real major cyclical errors re: their resolution.
Due to how they are made, it is very hard for an optical encoder to have an error anywhere near 2 counts, or even 1 count.

But the belts and or pulleys do have significant errors for angular positioning.

[QuinnSjoblom]

so heres one rough layout i was messing with today. 90 degree X/Y. I also chose to position the lathe bed tilted forward 45 degrees so it would be at a 90 to the z rail surface on the backside. The lathe bed could be level, putting it at 135 degrees to the z surface on the back, but i figure that would be a much more difficult built to get 2 surfaces 135 degrees to each other true. With 90 degree, After casting the epoxy granite base i would probably start by scraping the Z rail beds coplanar to eachother, mount rails, use carriages on those rails with indicator to start dialing in the subspindle rail beds. If that was 135 degrees between surfaces instead of 90, i would have no idea how to do that. Also i think the forward 45 degree tilt of the bed could work out well for chip clearing. So anyway, heres just one possibility i see for a 90 degree X/Y axis lathe. X axis is 8 inches of travel, y axis is +/-2 inches, Z axis 20 inches.

[Jim Dawson]

Quinn,

Your post #10 really shows the limitations of Mach3. You should not have to map the error on any axis. I am a real fan of putting the feedback device on the load, be it a rotating (spindle) or linear load (table or carriage). But this requires closing the loop at the controller level, not possible with Mach3. Having the feedback device on the load automatically compensates for backlash within reason, but there is no substitute for a tight machine.

This is not to say that linear or rotary encoders are perfect, but pretty darn close. On linear moves the maximum error I get on my knee mill is +/- 2 microns using 1 micron resolution scales, even with 0.004'' backlash in my X ball screw. Circular interpolation is not good because of the X backlash, I can count on the circle being up to 0.002'' difference in the X and Y axis depending on the cutting speed. The servo just doesn't have time to correct for the error. That's going to get fixed in a couple of weeks, sending both ball screws out for rebuild.

The DMM servos are a bit lacking in the holding torque area, but with some tuning they can be tightened up, especially if you are using analog control. I tightened one up to where I could not feel any movement when the motor was locked, I was not able to achieve this in step & direction mode, but didn't have time to play with it further. After some communication with DMM tech support I now have a much better understanding of what all the setting do, the manual is a bit lacking in the torque mode explanations and in a couple of cases just flat wrong. Once I get another 1.8KW unit in house, I'll be putting it on the bench first to really understand the settings where I'm not pressured with time constraints.

The reason that I am going with a 1.8KW servo on the lathe C axis is because I have 4 other 1.8KW units on the machine and just want to keep them all the same in case of a failure, I can just swap out as needed. Right now I only have 2 products that the C axis would be used for, we do the second op on the Haas mill, but we could eliminate that step.

Right now my lathe spindle encoder is 1024 pulse (4096 counts/rev) which works fine for single point threading and rigid tapping, but when I add the C axis I'll most likely add a 90,000 count/rev encoder which will give a positioning resolution of 0.004°, and hopefully the actual positioning accuracy will be somewhat close to that. Do I need that kind of accuracy? No, but I'll try it just to see if I can do it. Million dollar machine accuracy on a budget.

The bottom line here is that the key to making all of this work is the controller. If you are going to build expensive machines, then it's worth installing the proper controls to make them work the way you want. You need to be able to close the loop at the controller and that eliminates Mach3/4. Galil, Dynomotion, or others are the way to go. But you are going to have to learn some programming and a new system. Based on your abilities with the Mach3 macros, I think you will have no problem learning a new system. It's all the same logic, just the syntax may be a little different.

[QuinnSjoblom]

Quinn,

Your post #10 really shows the limitations of Mach3. You should not have to map the error on any axis. I am a real fan of putting the feedback device on the load, be it a rotating (spindle) or linear load (table or carriage). But this requires closing the loop at the controller level, not possible with Mach3. Having the feedback device on the load automatically compensates for backlash within reason, but there is no substitute for a tight machine.

This is not to say that linear or rotary encoders are perfect, but pretty darn close. On linear moves the maximum error I get on my knee mill is +/- 2 microns using 1 micron resolution scales, even with 0.004'' backlash in my X ball screw. Circular interpolation is not good because of the X backlash, I can count on the circle being up to 0.002'' difference in the X and Y axis depending on the cutting speed. The servo just doesn't have time to correct for the error. That's going to get fixed in a couple of weeks, sending both ball screws out for rebuild.

The DMM servos are a bit lacking in the holding torque area, but with some tuning they can be tightened up, especially if you are using analog control. I tightened one up to where I could not feel any movement when the motor was locked, I was not able to achieve this in step & direction mode, but didn't have time to play with it further. After some communication with DMM tech support I now have a much better understanding of what all the setting do, the manual is a bit lacking in the torque mode explanations and in a couple of cases just flat wrong. Once I get another 1.8KW unit in house, I'll be putting it on the bench first to really understand the settings where I'm not pressured with time constraints.

The reason that I am going with a 1.8KW servo on the lathe C axis is because I have 4 other 1.8KW units on the machine and just want to keep them all the same in case of a failure, I can just swap out as needed. Right now I only have 2 products that the C axis would be used for, we do the second op on the Haas mill, but we could eliminate that step.

Right now my lathe spindle encoder is 1024 pulse (4096 counts/rev) which works fine for single point threading and rigid tapping, but when I add the C axis I'll most likely add a 90,000 count/rev encoder which will give a positioning resolution of 0.004°, and hopefully the actual positioning accuracy will be somewhat close to that. Do I need that kind of accuracy? No, but I'll try it just to see if I can do it. Million dollar machine accuracy on a budget.

The bottom line here is that the key to making all of this work is the controller. If you are going to build expensive machines, then it's worth installing the proper controls to make them work the way you want. You need to be able to close the loop at the controller and that eliminates Mach3/4. Galil, Dynomotion, or others are the way to go. But you are going to have to learn some programming and a new system. Based on your abilities with the Mach3 macros, I think you will have no problem learning a new system. It's all the same logic, just the syntax may be a little different.

You're absolutely right jim, closed loop is definitely the way to go and there's just no beating it. My only concern is my ability to learn a new control software and make use of all the features I need. With mach3 I already know how to handle all the challenges of a milling/turning machine and make everything work how i want. I was able to learn this because of endless posts and YouTube videos on mach3. I'm worried that learning a new system is going to be very difficult for me without the same vast resources for learning like there is for mach3. I am definitely going to look into it and if it seems feasible, I will absolutely switch to a more sophisticated closed loop system.

One thing I wanted to mention about the holding power of my dmm servo, i have it tuned to the point where it is absolutely stiff when I try to twist it with my hand, it is locked solid and will handle a very large input of torque. The problem arises when introducing an oscillating load, like an endmill as each flute takes a cut. It will handle milling forces no problem, but the chatter is terrible. Probably only oscillating a few micron, but it's enough to ruin surface finish. Once I added the disk brake, the surface finish was excellent when milling on parts in the turning spindle. Keep in mind though, this is direct driven to get enough rpm for turning. If geared down, it would be less of an issue.

[Jim Dawson]

You're absolutely right jim, closed loop is definitely the way to go and there's just no beating it. My only concern is my ability to learn a new control software and make use of all the features I need. With mach3 I already know how to handle all the challenges of a milling/turning machine and make everything work how i want. I was able to learn this because of endless posts and YouTube videos on mach3. I'm worried that learning a new system is going to be very difficult for me without the same vast resources for learning like there is for mach3. I am definitely going to look into it and if it seems feasible, I will absolutely switch to a more sophisticated closed loop system.

Your only limitation is in your head I have seen your work and have confidence in you. If you can understand the convoluted Mach3 macro systems and actually make it work correctly, then real motion control system programming will be a breeze. Much more straightforward, and you have full control over everything.

One thing I wanted to mention about the holding power of my dmm servo, i have it tuned to the point where it is absolutely stiff when I try to twist it with my hand, it is locked solid and will handle a very large input of torque. The problem arises when introducing an oscillating load, like an endmill as each flute takes a cut. It will handle milling forces no problem, but the chatter is terrible. Probably only oscillating a few micron, but it's enough to ruin surface finish. Once I added the disk brake, the surface finish was excellent when milling on parts in the turning spindle. Keep in mind though, this is direct driven to get enough rpm for turning. If geared down, it would be less of an issue.

Mass is your friend here. But that can be a problem in smaller machines. It does sound like you resolved the problem.

[QuinnSjoblom]

Ah good point. With a substantial lathe spindle, there's probably enough mass to dampen those oscillating milling forces. My current machine just has a Dunham 5c headstock, nice beefy iron chunk but not much mass in the rotating parts. The next machine I plan to build might have a larger main spindle, but not sure yet. What id really like to do is use my Dunham 5c headstock as the sub on my new machine. Already has the pneumatic closer, then use a larger spindle for main, big enough for 16c or 3j through capacity for feeding bigger bars through. Sub doesn't need big through capacity, I can use 5c step chucks on that end for part transfers with bigger diameters.

Some might scoff at this, but I've actually been looking on aliexpress at some decent lathe spindles and bt30 cartridges. Looks like some of it is decent stuff to me. About 1200 to 1500 for either one. They have specific ratings for balance, rpm, runout, etc. Also some reviews from people that bought them. There's a lot of cool machine building stuff on there. Ground ballscrews, genuine hiwin rails, big harmonic drives, big cross roller bearings, etc. There's obviously a lot of cheap crap on there, but some of it looks good unless they are just flat out lying about grades and ratings.