[HobbyCNCer]

I recently picked up a used Harbor Freight 12x36 lathe. I think it's pretty lightly used and it came with a lot of tooling and accessories including DRO(old Sargon setup). Because I like to tinker and keep myself busy with interesting projects I started thinking about converting it to CNC. A friend heard about it and I then ended up with two large stepper motors that he never could find a good use for: Anaheim Automation 42D212: http://www.anaheimautomation.com/man...ec%20Sheet.pdf

I then started trying to figure out how I could drive them. I found these inexpensive driver board kits a got a couple of them: http://www.hobbyengineering.com/H1259.html
I assembled one and connected it up to one of the motors and actually used an old computer ATX power supply and used the 3.3v line (motor shows a requirement of 3.6v) from it, that's supposed to be able to supply 20A, for the main power source.

It actually seems to work quite well. I've tried the driver board so far in manual mode which just continually turns the motor with a speed proportional to the position of an onboard pot and it also has a direction reversal switch. It seems to run well under all of the speeds with an apparent limit at the upper end where the motor will begin to jitter and then just stop moving and hum. I'm assuming it's reaching it's steps/second limit? The torque seems very strong as well. I can stop it by hand with effort at the high speed, but in low speed I cannot.

I'd like to get some more experienced opinions on what lies ahead. These driver boards can be controlled directly from a parallel port, though I may end up building an opto-isolated circuit for each just in case.
I'm also sure I need to upgrade the lead screw and/or rack drive for the two axes on the machine. I figured that if I direct drive each axis, the cross-feed will give 0.0006" per step and the x-axis will be a horrid 0.0032" per step. Even with a gear/belt reduction on the x-axis that's still very course. I've started looking around at other racks/pinions I could replace on it to help get that number down. Though what about the back-lash?
And from looking around on this board it looks like a ball lead-screw would be a good option for reducing the back-lash on the cross-feed.
I haven't looked too much into spindle speed detection but that's something I'm considering as well.

I think at this point since I'm decently ahead in hardware(motors and drivers) that it seems to make since to go ahead and try it.

-Wayne

PS: this is my first post...quite a community here! And sorry for the username...I thought I was being clever until I signed in and saw there's commercial stuff out there with nearly the same name...

[RotarySMP]

Welcome.

I have a CNC'd chinese 7x12.

A Nema 42 is mega overkill for a 12x36 lathe. You probably only need a Nema 23 on X and a Nema 34 on Z. You can probably sell those NEMA 42's on Ebay and earn enough to buy some nice motors for the Lathe. If you sell the DRO as well you could probably buy two gecko's to drive them.

You should be running at the maximum voltage this board allows (+35V), even so you won't get close to the motors true potential.

Hook it up bipolar to a Gecko G201 and a 70V PSU and you will get near its maximum, but it will be far more torque at low speeds than you need or can use, and won't get to very high speeds at all.

Computers are quite good at maths. Don't worry about weird feed per step sizes. With those big slow Nema 42's you won't wont to gear down.

Forget feeding Z as a CNC through the Rack.

Backlash on a lathe is not as much of a problem as on a CNC mill, and you can live with it. It really matters if you turn curves which passes through a deflection point (turning a barrel for example). You will also find it hard to turn really precise bores, such as bearing seats if the backlash is not consistant.

The longer you play around with CNC machines, the more you see that the correct solution is anti backlash ground ballscrews. Yes you can make other solutions work, but this is the easiest, best solution. Ebay is the best source.

What software are you using? The spindle index setup will depend on this, as TurboCNC, Mach and EMC all use different encoder/indexer set ups.

[Mike Nash]

Don't raise the power supply voltage for the motor without adding a couple of resistors between each winding center tap of the motor and the V+ terminals on those boards. They are not current regulated chopper drives, just digital switches. If you raise the voltage to say 24V you'll need 2 each 3.34 ohm 125 watt resistors per motor. That's with no room for error on the resistor wattage. And at a standill, one or both of those resistors will be radiating 125 watts of heat. And even then your max speed will be pretty slow indeed, although better than with the 3.3 volt supply.

On the other hand, if you try it as is and those big motors give you the speed and torque you need, then go for it! But like Mark, I vote for the Geckos, big motors or small.

[asuratman]

What software are you using? The spindle index setup will depend on this, as TurboCNC, Mach and EMC all use different encoder/indexer set ups.

Are you talking about threading. Please explain how to do it? I mean software and hardware needed. What I know is using index pulse card from cnc4pc. But is there any different way to do it. If I use index pulse card from cnc4pc, I have to wait on order. I have extra encoder, I read in one of threads somebody used it. But I do not know how. I am building cnc lathe from scratch with GD cnc control.

[RotarySMP]

Mike, thanks for catching my Error, I thought those were choppering drives.

Asuratman, I used TurboCNC on my lathe, which only need a 1 per rev spindle index pulse. I provided this with a optical gate, but you could also do it with a hall effect sensor, a microswitch, the index line from an encoder etc. The Software doesn't care what method you use to create the pulse.

http://www.wrathall.com/Interests/CN...le_encoder.htm

Mach3 is a little more sophisticated, in that it can sync to multiple pulses per rotation. You could set it up with the same optical gate as mine, but replace the tabbed encoding disc with a multiple slotted disc, one slot of which should be 2x the thickness of the others, which is MACH3's way of detecting the phase.

EMC2 is set up for differential encoder, with the 1/rev indexing pulse plus only the A channel is used. You could also do this with 2 optical gates. One with a 1/rev tab like mine above, and second one having a multi slotted disc.

[hanermo]

I have a 1 micron resolution 12x and a +/- 3 micron 7x.

Now, first of all, everything is possible.
First is my recommendation, and at the end is why your current gear might not be the best bet, IMHO.

Your best sweet spot is likely to be using the cheap gecko 251 driver boards, a 48 v power supply, and some small nema 23 motors.

I have and use all the top end gear, including a hardware step generator (smoothstepper) and 10.000 / count encoder + 1 kw servos etc. I also have, unused, about 10 different old steppers and 4 different old driver boards.

And I still recommend the Nema 23/g 251/48v combo for you.

For a lathe, repeatability is the nr 1 issue (and resolution of course). You will be coming to the same spot lots of times to do a small incremental cut, and whether that cut is 0.02 mm or 0.04 mm has a huge effect on your results, as depending on tool shape it will grab, bind and jerk depending, depending.

Acme screws will work ok ... but ballscrews will work amazingly better. They do NOT need to be ground, but anti-backlash are better.

I just went through setting up a ballscrew on my big lathes x, using a 3/4 inch fixed-fixed ballscrew, and am extremely happy with the excellent results.

Fixed-fixed means it has dual ends fixed, and dual independent preloads (3 adjustable preload points)(4 angular contact bearings / thrust bearings). I needed to make 11 precision components to about 0.01 mm tolerance, and 4 special threads, about 80 hours work or 8 days in all.

I use a 1 micron digital dti, and get about 2 micron resolution (1 micron, but 1/7 is a 2 micron step or jump) with steppers.

For a lathe, you really want to gear the steppers down. Really. You should. As it is a Very Good Idea.

About 3:1 is good.
Timing belts (HTD is cheap and great, no backlash, as in under 0.001 mm) work well and cheaply for gearing.

Torque is not the concern. 3 Nm (2 will do) nema 23 small steppers are plenty at 30$ each.
Speed is not the concern. Lathes are fast. I run at way less than max speed (at max speed I could do end-end on extended travel crosslide in under 2 secs) using a 50 V commercial PSU (50 amps output lab-quality power supply).

For repeatable results, ie to make it work well, you need to build to about 10:1, ie a step should be about 10x better than what you need to produce.

For normal lathe work, you thus want to build for about 0.001 mm (Mine is 0.0003 mm, but my needs are higher-end industrial and I dont mind spending lots of hours at it, and some money.)

Big steppers:
Now, if You use Nema 42 steppers, your best speed will be less than 1/2 to 1/3 of the small ones. About 300-400 rpm at best compared to 900-1500 to 2000 rpm for the small ones.

Gearing:
Repeatability is a function of leverage and quality. You are not using top-end ground, guaranteed, hardened steel components. By using 3:1 gearing, your repeatability is vastly improved.

Resolution:
A stepper turns at 200 steps /turn. Using a 4 mm screw ... (acmes at around 3 mm rise for 12x size lathes, ballscrews about 4-6 mm rise typically for a 5/8 to 3/4 screw).

200 steps / turn / 4 mm / turn = 50 steps / mm or 0.02 mm.
It´s too coarse for a lathe. Gearing down 3:1 will give you 0.007 mm AND because of the extra torque microstepping will start to work for positioning, to about 1/3-1/4 step, giving your around 0.002 for a real-world repeatable result of better than 0.01 mm (just about).

Screws:
Thicker is better, AND much smoother due to the way torque works on really small components. Friction effects (about 10-30 kg "stiction" or start-up riction on 12x lathes) has a huge effect on small screws, and a negligible effect on big screws. This means an acme screw driven lathe will have about 10 kg push needed before the cross-slide starts to move. With a ballscrew, this is about 1 kg. In effect you get 5-10x better resolution.
This means a 10 mm screw will twist and bounce around 0.01 - 0.02 mm.
A 12 mm screw about 0.01 mm.
A 18 mm (3/4 screw) about 0.0008 mm.

Acme screws are about 10-30% efficient. Assume 10%.
Ballscrews are about 90% efficient.

This means that a 3/4 screw, is about 18 times more stiff and repeatable than a top-quality 12 mm acme screw (which I just replaced).
Its about 5-6 times more efficient due to the ballscrew vs acme efficiency, and about 2-3 times more stiff due to 18 mm vs 12 mm. Total about 18 times.
The fixed-fixed preloaded mount makes it about 2x more stiff and rigid again, and truly low-backlash.

Conclusion:
To get truly great results for a cnc lathe, you need to build for the best results you can afford.
You can ignore the speed/power values, as they will not be limiting you.

A workaround for acme screws is to gear up with very large pulleys (4:1 or 5:1). The problem is the space needed, and making pulleys is somewhat technically difficult (about 8 on a scale of 4-ten) and big pulleys (72 teeth and up) are expensive. If you have 5 mm reamers, and a diving head/rotary table/spin indexer you can make your own HTD pulleys (but it is lots of hours, and concentricity is important).

BTW-
You dont need a preloaded ballscrew setup to start with.
A basic ballscrew, with a floating mount, and 2 thrust bearings will work fine.
I am still using (so far) a single un-preloaded ballnut.
I will be changing this, in time, but I wanted to let you know that many recommendations, while good and desirable in themselves) are not truly needed.

On a 12x:
[Kool-Aid test]
There is no question that industrial ground ballscrews, class c1 precision, would be much better. BUT the real-world results would ONLY improve positioning accuracy, from 0.02 mm, to about 0.002 mm.
They would NOT
- improve resolution from the 1 micron level (lathe cutting results top out at around 1 micron due to physics limitations of tooltips),
- nor the efficiency (already 90%),
- nor the speed (already limited by safety concerns)

Actual workpiece sizes are checked by micrometers.
5 identical pieces will be identical, within around 0.006-0.008 mm, currently.
Using ground c1 ballscrews, I would get around 0.003 mm.
The cost would be about 3000 € extra.

Where ground ballscrews would help, is in one-off workpieces. I would not need to measure or check workpiece sizes before doing final cuts.
IF I was doing single hi-spec hi-dollar workpieces I would have to buy the very best, to be able to deliver these workpieces in a timely manner. As it is, I am not, so this is currently not needed, for me.

CNC is more about repeatability than absolute precision, unless you do high-end work like metrology, optics, or mass-produced very high end metalwork (bearing journals, turbines etc).


Good luck,
hanermo

[blades]

Mach3 is a little more sophisticated, in that it can sync to multiple pulses per rotation. You could set it up with the same optical gate as mine, but replace the tabbed encoding disc with a multiple slotted disc, one slot of which should be 2x the thickness of the others, which is MACH3's way of detecting the phase.

Mark, is there a specific number of slots that is best to use with Mach?

Is 4 enough? Is 6 overkill?

Thanks for the info on the 1 slot being 2x wider than the others, I now have a new part to make for my 7 x 10 since I'm planning on using Mach.

- Bill

[RotarySMP]

Sorry, I have not used Mach, I just picked up on the multi slot thing at a time when I was investigating changing to Mach.

[blades]

What made you decide not to switch to Mach? Are you still using TurboCNC?

[RotarySMP]

The old P1 90MHz I had running TurboCNC died, so I looked at upgrading to Mach or EMC.

I have no space for a PC, as my machine shop is a corner of my inner city balcony, and when I ran the checks on my PIII 800MHz laptop, the singnal was not stable enough to run Mach. TurboCNC seems to run okay on it, so I stuck with that.

My only issue with TurboCNC is the tool offset management which is not separate from machine offsets, and don't work at all for me.