[siam]

For several different reasons, I decided several months ago to build a CNC lathe from scratch. The design I've settled on is very similar to a lathe built by DumpsterCNC several years back; I consider it such an elegant solution that I'm somewhat surprised that no one else has followed his lead and built similar lathes. I need to mention that I currently live in Thailand which provides both advantages and disadvantages when building your own CNC lathe: skilled labor is rather inexpensive compared to my home country, the USA, so having the tubular steel frame professionally welded shouldn't be cost prohibitive. I should also be able to have other parts, such as the spindle and bearing housings, inexpensively machined. The main disadvantage is that finding parts is always a challenge.

Getting back on pointe,...A few specs:

Overall approx. dimensions: 56" X 29" wide X 20" high.
bed: 1 meter
max diameter which can be turned: 14"
fully enclosed - allows for fluid cooling & more importantly to me, keeps my "shop" clean.

Here are a few CAD drawings of what I plan to build:

Attachment 292020

Attachment 292022

Attachment 292024

Attachment 292026

I few weeks back I was visiting family back in the US, and was able to purchase a few parts I knew would be hard to find in Thailand, & brought them back with me in my checked luggage (2.5HP Leeson motor with motor driver, stepper motor drivers 4 amps & 48v, timing belt & pulleys),...yes, my luggage was opened & inspected.

I'll be posting much more as this build progresses, but I'm interested in feedback on this rather unique design,...especially from anyone whom has ever tried this approach before.

Mike

[hanermo]

I have done lathes.
Some experiences.

Never direct-drive steppers.
Steppers are too slow for threading, in the pullback, and too inaccurate. They dont accelerate fast enough, in the pullback.

The index sensor will not be accurate.
Only encoders give a sufficiently crisp, sharp, signal.
The other options will be fuzzy, and the angular relationship wont work at all speeds.

Most motion control cards dont support threading correctly.

You want about 5x more resolution than the minimum feature size you plan to machine.

Small pulleys are not rigid enough for positioning under load.
They always lag, and this is made worse with steppers, as they act like springs with microstepping.

The design is classic .. but I think you would get better results with a vertical on the backturret instead of 45 degree angle.

Most motion controllers wont support lathes well.
Problem areas:
Homing, probing, rear turret if using 2 tool, offsets, css, threading.

My solutions:
Bigger screws. Better mounts. AC brushless servos. HTD8/30 pulleys. Servo spindle. CSMIO motion controller.
You cannot make it too heavy - You literally cannot make it heavy enough.
These are the right solutions, given that you are making a 14" lathe.

Fwiw..
I use size 35 mm linear guides, for something like this.
They are 4x more rigid than eg 25 mm guides.
Strength is not the issue, but rigidity very much is.

[rocketflier]

Should be a fun build. Another I followed, perhaps some ideas:
http://www.cnczone.com/forums/vertic...51688-cnc.html
And before you say the pictures are missing:
https://plus.google.com/photos/11506...55557749407665

[siam]

Thanks for all the suggestions Hanermo,...you make some good points. I especially value your suggestion on the need to use a resolver for accurate spindle positioning for threading, as I do hope to be able to cut threads with this machine at some future time.

[siam]

Thanks for the link to the photos Rocketflier,...I had read through the Slant bed post you mentioned but could never find the photos. I do like the massive frame & bed on this build, and as Hanermo mentioned, when building a lathe, more mass is always better, however, in my case I need to make my lathe as light as practical. Reason: I currently live in a rented house which has no real shop...I use one of the rooms in the house as my workshop. It would simply not be practical to move a 500lb (or more) lathe into my current environment. When I buy my own house (hopefully within the next few years) I will need to move my entire shop, so keeping everything as mobile as is practicable is a big plus; to that end, I will build a lathe bench, specifically designed for this lathe, which has casters on all four legs, making it easy to move around the shop.

A few other considerations that went into the design of this lathe:

It will be used to support my hobbies in a non-commercial workload, meaning that I can cut & thread as slowly as needed to attain the desired precision. At least under my initial light-load bench tests, the speed control (from Geminicontrols) provided smooth, very low RPM control of the 2.5HP DC drive motor, with excellent load compensation. More testing is needed, but I'm certainly pleased with the initial results.

To compensate for the lack of metal mass, I've used more "ribs" & larger square tubes then is actually needed.

Tailstock: For my personal use, the tailstock will always be manually positioned & operated; therefore, no need to be placed on linear rails & bearings. Instead, design a manually clamped-on carriage that attaches to the existing rails.

The majority of planned work will be in the 3" to 8" diameter range. On rare occasions I may need to turn a 14" diameter piece of metal, but I can always make slow, light cuts with lots of coolant.

Mach 3 for motion control. I already use Mach 3 for the CNC mill I built several years ago, and I'm fairly familiar with it; therefore a shorter learning curve.

48 Volt, 10 amp, DC power supply. Stepper drivers from MassMind.org rated 50v at 4 amps.

Chuck mount: D1-4 Camlock

Spindle: custom built. mounted in two back-to-back angular contact ball bearings nearest the chuck, and one standard ball bearing on the opposite end.

[handlewanker]

Hi, just came across this thread and some time back I thought about the problem of synchroing the spindle with the leadscrew for screw cutting with CNC, and as I'm new to CNC it appeared to be a too hard basket situation, as in why go to CNC turning if screw cutting is so complicated.

Unless screw cutting is the prime object of making a lathe , any lathe, CNC or manual, straight turning is just an X and Y movement at all times.......and an A movement if you have a turret tool changer etc.

So, if screw cutting is to be done simply, what would be the problem of adding change wheels to the spindle and leadscrew, as in a manual lathe, but control the saddle and cross slide movement with the steppers, servos...... whatever.

OK, so you need to manually change the gears or throw a lever or two at the start to get the pitch when required somewhere along the way, usually when the job is finished with the turning part, but as screw cutting is about 1/50th of the time spent in turning, to my way of thinking, the CNC mode for synchroing the spindle and leadscrew is a complicated one.

In a manual lathe the headstock spindle drives the leadscrew when screw cutting, so if the lathe is CNC'd with a geared spindle and leadscrew and screw cutting is needed, the stepper, servo motor....whatever....driving the leadscrew just freewheels.

It would be a simple matter to have a toothed clutch and solenoid on the gear train to disengage the gears so that high speed turning under CNC control can occur unobstructed.

It could mean that the cheap small lathes on EBAY could do CNC turning with geared spindles and leadscrew without having to go to great lengths with encoders on the spindle for the odd amount of screw cutting that might occur occasionally, but would need to be present for that operation.

I doubt that anyone would want to create an odd pitch size like .70mm pitch.......change wheels will do it anyway if you have enough of them with the right tooth count.

It's just a thought......can anyone see a problem with such a set-up?......I'd be interested to find out, as I can't see any.
Ian.

[Hezz]

Hi, just came across this thread and some time back I thought about the problem of synchroing the spindle with the leadscrew for screw cutting with CNC, and as I'm new to CNC it appeared to be a too hard basket situation, as in why go to CNC turning if screw cutting is so complicated.

Unless screw cutting is the prime object of making a lathe , any lathe, CNC or manual, straight turning is just an X and Y movement at all times.......and an A movement if you have a turret tool changer etc.

So, if screw cutting is to be done simply, what would be the problem of adding change wheels to the spindle and leadscrew, as in a manual lathe, but control the saddle and cross slide movement with the steppers, servos...... whatever.

OK, so you need to manually change the gears or throw a lever or two at the start to get the pitch when required somewhere along the way, usually when the job is finished with the turning part, but as screw cutting is about 1/50th of the time spent in turning, to my way of thinking, the CNC mode for synchroing the spindle and leadscrew is a complicated one.

In a manual lathe the headstock spindle drives the leadscrew when screw cutting, so if the lathe is CNC'd with a geared spindle and leadscrew and screw cutting is needed, the stepper, servo motor....whatever....driving the leadscrew just freewheels.

It would be a simple matter to have a toothed clutch and solenoid on the gear train to disengage the gears so that high speed turning under CNC control can occur unobstructed.

It could mean that the cheap small lathes on EBAY could do CNC turning with geared spindles and leadscrew without having to go to great lengths with encoders on the spindle for the odd amount of screw cutting that might occur occasionally, but would need to be present for that operation.

I doubt that anyone would want to create an odd pitch size like .70mm pitch.......change wheels will do it anyway if you have enough of them with the right tooth count.

It's just a thought......can anyone see a problem with such a set-up?......I'd be interested to find out, as I can't see any.
Ian.

Most CNC lathes don't have a leadscrew in the same sense that a manual lathe does. All the high quality industrial CNC's use ground ballscrews because of their low backlash. The timing for threading is not done with a mechanical synchronization but both the spindle and the other axis are synced in the control software and it is very accurate with high quality servos. The synchronizing is actually done by the control software. It will sent one pulse to one servo and then the other servo at many thousands times per second. It seems to me that once the control software has been programmed it is far easier to do threading on a cnc lathe. At least that is my experience having worked on both manual and CNC lathes. The software controlled synching is more accurate because there is a lot less backlash then there is through all of the gears on a manual machine.

All you do in the programming code for a threading cycle is call up the right code and put in the coordinates and thread pitch and major and minor diameter. The designers of the control software have done all of the hard part in the background. I have seen parts machined where the threads had to start and stop at an exact angle relative to some datum on the part and if it is off by more than a degree the parts were rejected. It would be time consuming to get that kind of accuracy with a manual lathe on a thread form.

[handlewanker]

Hi, I don't think high quality CNC commercial lathes of an industrial nature were the topic of the thread even if they do use the methods for CNC screw cutting you mentioned..

. I quite understand the process required with CNC turning and screw cutting.......not the using of it though, and the "leadscrew", that pushes the saddle along the bed, which for any CNC work would need to be a ball screw, and as it has a pitch is still a leadscrew even if it has balls instead of a solid nut with threads.

BTW......a leadscrew on a lathe is called that because it generates the pitch required to be cut as a division of it's own pitch when connected to the headstock spindle with a gear train.

What I proposed, and wanted to find out the problems arising, is to have a CNC lathe and direct couple the leadscrew and spindle with gears instead of an encoder etc .

You would then not have to change anything virtually in the spindle drive train, apart from converting the rest of the lathe from Acme threads to ball screws for the saddle and crosslide as would definitely be required to do CNC turning...etc etc.

This does not include having linear rails on the bed or crosslide, but can be a retrofit as with the removal of a compound rest there will be bags of room to fit linears as required.

It's not conventional in the CNC sense, but simple......where are the problems I asked.
Ian.

[siam]

Ian, I believe your mechanical drive of the lead screw for threading purposes has merit and could be accomplished, however, for my design, it would involve far more work and require the addition of many more parts as compared to the relative simplicity of adding a resolver to the spindle or motor shaft, and allowing the software to do the rest.

Mike

[siam]

I have a small amount of experience with designing and building welded, thin-wall tubular steel structures, and I've learned that no matter what techniques are employed, it's nearly impossible to eliminate all of the heat induced twisting and warping of the finished product. So how to build the welded cage and obtain alignment between the spindle centerline and the main rails?? I came up with a solution which I believe solves this problem. Rather than attempt to straighten the finished welded cage, I will build a separate, much simpler weldment, onto which the rails will be mounted. This structure is so simple that it should be possible to straighten after welding, if needed, and to surface grind the rail mounting surface nice & flat, if required. This rail bed will attach to the main framework via ten 1" X 1/4" steel tabs welded to the rail bed and bolted to the main framework. This design will allow the rails to be aligned to the spindle centerline with a high degree of accuracy as it allows for X & Y adjustments over the entire rail length. I have some reservations about using a bolted-on lathe bed, but I believe it's the best solution to this particular problem.

Below is a CAD drawings of the Rail Bed and a drawing showing the steel "tabs"

Attachment 292090 Attachment 292092

Mike

[hanermo]

Threading:
I have spent about 2000 hours at it, so far, and my experiences can be pretty much taken for "true".

I have tried the cnc4pc index sensors, pass-through sensors, reflective sensors.
None of these work, correctly.

The signal is fuzzy, and does not trigger cleanly.
Thus, it is *impossible* to get good threads.

We conclusively proved this, with Sergey from Centipede, by logging the results into memory at 533 Mhz, wtih microsecond resolution.
About 4 years back, with a special kernel-driver running at hw speeds and logging every pulse into main memory.
Aftre this, Sergey designed and built the index-trigger modules he sold with the centipede.

The SS has all sorts of bugs, and on top of this, mach3 has some bugs in threading as well.
Some we documented.
Examples:
Waifor() while waiting to trigger on the threading movement.

The only good solution, imo, ime, is encoder-signal threading.
Note that unlike most I have tried many solutions, and none of them work, properly.
Thus I am not speaking from the basis of one hardware/sw solution, only.

The best solution, and the only one, afaik, is the CSMIO.
I know that the Hicon controller is supposed to support encoder threading, on mach4.
I have no idea what is the current state of the art, in general.

I sue a CSMIO-IPS, for my lathe.
It supports encoder-based threading, for both Mach3 and Mach4.

Hood, a very knowledgeable user, also uses CSMIO, the analog one CSMIO-IPA, for lathe.

I have just installed my new ac brushless servos on the lathe, finally, and they are not yet producing.

[hanermo]

This will work fine.

I have some reservations about using a bolted-on lathe bed, but I believe it's the best solution to this particular problem.
Mike

[PCW_MESA]

Linuxcnc has supported encoder/index based threading for many years and it works fine

[hanermo]

PCW_mesa...
Maybe yes, maybe no.

There is no evidence it works well.
Show me some leadscrews, done to a tolerance of 2 microns or less, over 400 mm.

Show me some threads done to aerospace 3a/b specs, with go-nogo gages passing it.

Close is useless.

The exact same problem exists with the machx stuff..
And I suspect I am one of the very few ones actually going to be able to deliver on high quality threads, in the near future.

I know some people who do industrial quality threads on industrial tools, hood is one.

Unless the threads have specs, and pass them, they are useless.
Root radius, crest radius, pitch diameter, and sizes must all be spot on. Within tolerance, iow.

In my case, I want to make perfect threads.
But my post refers more to everyday jobshops, who just want to make standard, good, threads to commercial specs using full profile inserts, in steel.

And yes, the customers will check them with go/no go gages, routinely.
And yes, its all industrial, and yes, the competition is industrial modern machines.

"Threads on" or "looks right" is useless, for our commercial stuff.

Have you fitted a machine with linuxcnc to do threads with encoders ? .. and are thus an authority on them ?
What servos did you use, and what interface cards ?
What is the step resolution ?
.. on what lathe, and what servos on z, x, and spindle ?

A LOT of stuff does not work correctly.
I am not doubting You. I am not being negative.

I am experienced, professional, and do industrial quality work for industrial money.
E.g.
A refit I do, will cost 20k and up.
It will have specs, and when I sell it, I will know in advance what specs it fills, what it does not, and which are "best-effort".

Often, these posts degenerate to useless comments, with little content, and little experience.

I was using the old emc in 2002, and have used production linux, with 70 servers we built and managed, for fortune 500 corporates over here, in mission critical use for years.
So .. I am not anti-linux, on the contrary.
I sell solutions, not theoretical specs.

Anything that works, may have a value.
Thus, anything actually working, and documented, and having specs, is of use, to me.

Thats also why often my posts are detailed, not very short, and tend to contain details.
This is an exact industry, with often-critical parameters.
Thus, Details Matter.

I have never seen any linux info re: thread quality.
Any links welcome.
I do know that most machx stuff does not produce industrial-quality threads, and I also know why, in many cases.

The costs are immaterial, until the quality is there. Imo, Ime.
Once the quality is there, and there are some specs, we can start to evaluate if something is expensive, cheap, better or worse.

[PCW_MESA]

Umm since you obviously dont have any recent linuxcnc experience why are you making any comments about it at all?

again more slowly:

Linuxcnc has supported encoder/index based threading for many years and it works fine

[The Blight]

"I have spent about 2000 hours at it, so far, and my experiences can be pretty much taken for "true"."

Anecdotal evidence doesn't count for much on the internet. Back up your claims with hard data the same way you demand data from the rest of the people posting here. Right now you are just coming across as arrogant.

It also helps to take into account where this machine is to be used. Hobby use doesn't have to comply to the same standards as industrial use. If you want to discuss encoder feedback and precision of professional machine tools, perhaps you could start up a topic on the subject and document it so that everyone can use this in future builds.

[handlewanker]

Hi, I think the poster of this thread had better chime in and state exactly what quality of work re- threading with CNC he would like to have and what would be an acceptable outcome for his needs.

It's quite fine to dwell on high tec industry standards and methods, but for certain classes of work good enough is good enough, otherwise you sink the ship by overloading with excess requirements.

How perfect is perfect in the real World......it cost a fortune to not even get close to good enough in most cases.

I spoke of a geared headstock/lead screw solution, purely as a passing comment, but the application comes from way back in the 50's........to be precise it was used on a Ravensburg face plate type lathe........face plate being 6 feet in diam.....I have photos of that lathe.

On that lathe the headstock and bed were two separate items not connected to each other mechanically for screw cutting, as the bed was mobile and could be mounted either across the faceplate or at right angles to it, depending on the job size etc.

To get synchronisation, the headstock sent out electrical pulses to a stepper motor on the bed that turned the lead screw, and the actual screw cutting pitches were done with change gears in the normal way.......that was long before computers or CNC were thought of.

BTW, I would have thought it would be simpler and also cheaper to buy a used lathe, machine off the Vee ways to leave a flat top surface and fit linear rails etc.....just thinking......all that welding will cause lots or headaches along the way..
Ian.

[siam]

This lathe is being built to support my hobbies in a non-commercial environment, so I have the latitude & luxury to cut & thread as slowly as needed to attain whatever desired precision I choose. Of course I'm aiming for the greatest accuracy possible, but within my capabilities & limited budget. Reading through Ian's & Hanermo's posts has given me some valuable insights into the threading challenges I face as this build progresses; certainly I need to learn more about both encoder & indexed threading. Fortunately for me, I doubt that I will ever need to produce the extremely high quality threads which Hanermo turns out on a daily basis. Right now & for the next few months, I'm more focused on getting the tubular steel framework welded with a minimum degree of warpage & finding a machine shop that will make the spindle & bearing housings; the choice between encoders or indexers, though good to plan for now, is presently way down the road.

[siam]

BTW, I would have thought it would be simpler and also cheaper to buy a used lathe, machine off the Vee ways to leave a flat top surface and fit linear rails etc.....just thinking......all that welding will cause lots or headaches along the way..
Ian.

The first & most important reason for not incorporating an existing cast steel bed into my design is the extra weight it would add,...I need to keep this build as lightweight as possible & practical.
The second reason for using an all tubular steel design is the sheer challenge of doing so; the engineering part of me just wants to prove to myself that a well designed tubular steel frame is a viable alternative to traditional castings.

[handlewanker]

Hi.....not too much of the "high quality on a daily basis" please, it makes me blush........last time I cut threads was a year ago.....not much turning now I'm retired.........and my old lathe is a 1930's Colchester Bantam.....all manual etc.

With a conventional manual lathe the idea is to munch as much metal off with each pass as the lathe can stand, time being money etc, and you have to stand and work a manual lathe hands on all the way.

With a CNC lathe, and a hobby one at that, you can apply the "death by many blows" practice, that is, you take off metal with lots of smaller passes and in doing so the frame can be much lighter......provided it's designed clever enough using steel tubing to resist the cutting forces, but it still involves a lot of bracing and welding too.

I mentioned the manual lathe as a donor because it already has a headstock with spindle and drive motor, only needing the saddle to be removed and linear ways fitted to the bed.........a new saddle for CNC machining is practically just a slab of flat metal with linear ways and bearing blocks attached to either side.

Just the headstock alone (and tail stock) would be a big time saver in the construction, but donor headstocks are rarer than hens teeth, or rocking horse poo poo.

So, what is a brand new lathe worth on EBAY, one with the headstock size you need for your build.......$1000 to $2,000?........I'm insinuating that you remove the headstock and tailstock (to save a lot of build time) and sell the rest to someone who wants to mess with building a lathe to their specs or using the bed as a CNC donor where weight is not a problem and selling the left overs for spare parts.

You get a headstock complete with drive motor and with a spindle and chuck, possibly with Camlock, and bearings designed to work at speed and pre-loading as for a lathe, and all the working parts heat treated and ground to specs you probably won't be able to replicate......ready to go.

A lighter construction would be a wood turning lathe, cast iron bed etc, not the steel tube type, but still relatively light, and having a flat bed top for mounting linear rails on, and no other parts like a saddle etc to get rid of, so are cost effective.

Wood turning lathes are very basic and designed to run at high speed.....also much cheaper initially....only big drawback is they don't have big holes, if any, through the spindle, and usually only have screwed spindle mounts for a chuck...... if needed........ but spindles can be re-engineered.
Ian.