PM1022 lathe CNC Conversion

[dautry]

Yeah I’ve just heard back from linearmotionbearing2008 regarding a 20mm doublenut ballscrew for the y axis. When I was talking about machined ballscrew I was talking about the ends machined by him linearmotionbearing2008 . Precision Mathews sold a 1127 6-8 years ago. I acquired it used. I have a PM25 thats been converted to cnc . I also have a PM45 CNC that they put out for a short while. You had to run it really slow or it would miss steps. I think they probably caught a lot of flack from customers. I got it used at the same time as the 1127 lathe. I’m almost done with it as I upgraded to DMM sevos and Centroid Acorn operating system. I’ve also built a belt drive, 1.8kw servo for the spindle motor and PDB. I’m approximately half complete on building a counter balance for the z axis. I’m not letting myself start actually working on the cnc conversion for the lathe until the PM45 mill is done. I do know parts from China take awhile to get. So I needed to research and get them ordered. I checked on C5 ballscrews, 3000 was eye opening. Nice chat. Keep after it.
Sincerely,
D. Autry

Thanks for the nice words. It's taken a little over a year so far but coming along nicely.

I had good luck with linearmotionbearings2008, no experience with zyltech. I didn't think ball screws were machined? I'm familiar with rolled and ground, but not machined. The ground look real nice but cost buco bucks unless you find a deal on used.

CL_MotoTech mentioned converting his 1128 lathe. I wonder if that is similar to yours? It could be a typo for an 1127, or a typo for an 1228. It would be nice to have a lathe slightly larger than a 10" but for me space is crazy tight.

Will you start a thread on your build?

Thanks.

Hugh[/QUOTE]

[RTP_Burnsville]

Hugh,

Thanks for the project updates, still following along... You are doing a great job....

Robert

[currinh]

As I mentioned the next task is to build an enclosure for the lathe. I went with wood 'cuz it's easiest for me and, since I'm not using flood coolant (just mist) it should work well. I've spent a couple of weeks putting this together and I like the result. Not up to cabinet quality but none of my woodwork is.



It fits pretty well with two fold back door for access. I cut 0.065" aluminum brackets for the door joints. To assemble I used JB Weld in the joints and to hold the brackets solidly. It attaches with three screws at the headstock and two bolts at the tailstock. If needed, it comes off easily.

The doors won't close with the compound in place, but I plan to just use a riser block for the quick change tool post. If needed I can put the compound in place and leave the doors open. I used sheet metal at the headstock, it needs as much space as possible to install/remove chucks.



The doors fold back nicely giving good access.



It will go up against a wall which will stop the doors from going this far.



There's about 6" behind the enclosure. I may put the power supply and controller here, but that would make it harder to remove the enclosure. There is a shelf under the machine I build for computer and controller. But it would fit so nicely here. We'll see when we get to the controller. That is likely next, while putting 4 or 5 coats of poly finish on the box.



It is coming along.

Hugh

[currinh]

I'm slowly moving along with the lathe conversion. But there is progress to show.

I built up the electronics, mostly. It's on a piece of plywood and I hope to move this to an electronics box once everything is working.



It has a MESA 7i76 board at its heart. This feeds the X and Z Gecko stepper drives. They are supplied from the 48v supply. At the bottom are three Index Pulse Boards from CNC4PC. These are for an A/B/Index spindle encoder. There is also a relay board that isn't wired in yet. I have a 48v to 24v buck transformer that I need to mount. It will provide Field Power to the MESA board and juice to the relays.

I've had a heck of a time buying computers for the lathe and mill. I figured this would be the easy part and ordered MESA 5i25 boards which are PCI. I know computers with parallel ports are hard to find but assumed any old computer would have a PCI slot. I ordered a refurbished DEL through ebay. (first was dead, second was a tower [too big], but the third at least started.) But when I tried to install the PCI board I found a PCIe slot. I come to find out PCI slots are now rare, I'm just behind the times. I found a newly manufactured mini-itx J1800I-A board from China so ordered two. I've used this board before and it works fine for LinuxCNC. I bought cases with power supplies, memory, and CDROM drives. After the requisite month delivery time from China, I tried them. One seemed OK but the second would not install linux, so back it went. The first one installed and seemed to work till I started to configure LinuxCNC. During the config it stops responding to the keyboard and mouse, frozen. I next ordered a refurbished Gigabyte GA-E350N mini itx. It was DOA. It would display the BIOS splash screen but would not let me into the BIOS settings. It would also not install from a CD or USB stick. It's going back. Finally an old used Intel D425KT board from EBay. It's here and seems to be working. All these mini-itx boards have a parallel port also, but it's the PCI slot that I need.

I've installed the spiral springs on the lathe to protect the Z screw. Also put some threaded holes in the cross slide to hopefully mount way wipers and a leather cover to keep chips away from the X ball screw.





It is now alive, I have X and Z motions. Before much use I need to make the cross slide covers and an X belt cover. But it is coming along.

Thanks.

[currinh]

I need your insight and advise.

I have my PM1022 converted. Still needs work but it's functional enough to check out the motions. The X axis has one to two thousands backlash. This seems good for a single ball nut axis. I can work with that.

However, I'm seeing some strange and not good results for the Z axis. The backlash varies from about 0.005" to 0.010". Taking these measurement I'm jogging an inch or so back and forth by 0.10" steps. Readings are good going negative, towards the spindle. But jogging positive it will show 0.010" backlash, then 0.005", then 0.010", bouncing back and forth. It's in the ballscrew rather than the bearings, the bearings have maybe 0.001" backlash.

In fact, jogging by 0.01" I get these backlash readings in thousands over 0.2" (TEST 1) and another 0.2" (TEST 2).



It seems to cycle over 0.2". Strange. But the pitch of the screw is 5mm, or 0.197". So less strange noting the cycle repeats with the pitch.

This is a "DFU2005 Double Ball nuts ball screw" from LinearMotionBearings2008.com. Low cost ($79) ball screw I'm sure you've all seen.



Looking at the ballnut assembly, there are two half circle spacers separating the two ballnuts. The nuts are screwed together against these spacers and kept in place by two tabs (one screw each). OK, one half circle spacer is tight and the other is loose. As the ballnut is rotated the same spacer is tight and the other is loose, it's not rotation dependant. I noticed during assembly that the ball screw is not straight, wobbles some 1/8" over its 3' length.

Has anyone had similar problems? Any solutions?

I could make new spacers and/or fiddle with shims. I don't know if that would help. Alternately I could buy a new inexpensive ballscrew for about $80 plus shipping. I don't think these, or the one I have, are rated. This would help if the screw I have is faulty, but wouldn't if all these are this poor quality. Or I could get a nicer C5 screw for $240 (Amazon).

Would you 1) fiddle with the existing ballnut, 2) order the same ball screw figuring this particular one is bad, 3) move up to a C5 screw? What would you do? Anyone have good sources for this 2005 ballscrew/nut?

Thank you for any insight and/or advise.

[maxspongebob]

I would remeasure the backlash of Z all along the screw. Is it the same regardless of position on the screw? If so, the slop is in the nut, just undo the screws holding the nuts together and readjust the rotation of the second nut to minimize the backlash. Then put a hose clamp on the nuts to hold them together. If the backlash varies along it's length, then you probably have a bad screw.

An alternative to a double nut is 2 individual nuts adjusted to oppose each other. You could use the same mounting holes that you already have with one on the inside of the carriage and one on the outside. Use shims to clock the outside nut to the correct position.

[currinh]

I doubt it's the screw itself. Since the two nuts don't seem to align well the problem is likely there. I'll measure backlash along the screw tomorrow though. Good idea. If it still looks like the nuts, I'll spend a couple of hours looking for a fix as you suggest. I wonder if anyone has replaced the spacers between the nuts with spring washers (bellevilles)?

Changing to two separate nuts would require some redesign getting the spiral springs to work. Quite doable but a pain and some fabrication. If it comes to that I'll likely get a new double nut.

Thanks for the help, I'll let you know what I find.

[currinh]

OK. I found the problem and life if good. It was simple.

But first I checked along the bed in several locations. I saw the same strange results everywhere. Then I took the shims out of the double ball screw. Remember, one was tight and the other loose. I found that the thickness of the shims varies by up to 0.017". That sure seems like a lot. I fiddled with just holding the two nuts in contact and saw the same problems.

But then I found a few loose screws in the apron assembly. They either came loose or weren't tightened during final assembly. I tightened those and the problem went away. I feel so stupid.

I now have about 0.003" backlash in Z and 0.002" in X. Those are in the ballpark of what I'd expect and now corrected in the controller. Back to working on the encoder.

Thanks.

[currinh]

Good progress and I think (hope), all the major components are in place and working. I’ve finished the encoder and it works.

Russtuff has done a YouTube series on his G602 lathe conversion. In Video 11, Spindle Encoder, he shows his encoder. Very nicely done so I shamelessly copied his scheme. Being me I had to modify it somewhat (and I couldn’t find his drawings). But basically the same.

As russtuff did, I cut an encoder disk that fits onto the outboard end of the spindle. It fits outside the existing gear on the spindle. That increases the stack length by 0.08” which caused no problem. Mine has 36 slots with one being longer for an index pulse.



I decided to use three sensors where russtuff used only two. This gives A, B & Index pulse trains to the controller. With A and B its possible to determine which way the spindle is turning. It also gives twice the resolution. It also makes the set-up in LinuxCNC easy. The sensors are arranged as:



The A and B sensors are mounted on the same plane with their sensing point on line with the short slots. The index sensor is on an angle to catch only the single long slot. I screwed up from the first sketch on. The A sensor is at the center of a short slot as shown. The B sensor was centered between slots (old). In hind sight this gives 180 degrees between A and B. When A is on B is off and vise versa. That won’t work as they need to be out of phase by 90 degrees, not 180. I eventually moved B to be on the edge of a slot when A is centered (new). But we’re getting ahead of ourselves.

I built up the bracket of aluminum. Reasonably straight forward job.



The M3 taped holes are for the sensors with the cut out for corresponding wires. The pic below shows the sensors mounted.



I mounted the bracket with sensors at the tailstock using long M4 screws. I fiddled with it to get good signals on all three sensors. Then marked one hole while holding it in place. Not too hard once I found the disconnected sensor wire. Drill and tap this first hole. Once in place and working I marked, drilled and tapped the second hole.





I used an “Index Pulse Board” from CNC4PC on each sensor. Russtuff used the same boards. I believe it cleans up the sensor signal giving a nice square wave. They work well. It’s nice to have the LEDs that show on/off for debugging.

Well, the readout in LinuxCNC was getting some kind of signal but it was gibberish. That’s not right. I started checking with LinuxCNC’s HALshow and HALscope. HALshow told me the signals were getting through to the controller and being correctly recognized. I finally noticed there was no middle points where A and B were both on or both off. Hmm. After learning a minuscule bit about HALscope I was able to get a trace of the A and B signals. Yep, there were 180 degrees out of phase. Lights went on and I reviewed the original design.

I moved the B sensor over to be 90 degrees out of phase. That involved reaming and filling the original mounting holes with press fit, and loctited, aluminum pins. I did use the new, now functional, cnc lathe to make the pins. WooHoo. After filling the original holes, I drilled and tapped the new ones.

Back on the lathe for testing. HALscope showed the signals to be correct. It looked like the sampling rate was too low to see consistency of steps or exact phase. I couldn’t figure out how to increase the sample rate. Looked OK as far as I could tell. The readout jumps around maybe 10% but a lot better than the first test.

I tried to cut a thread and was told the spindle wasn’t turning. Stupid machine, you have an encoder signal showing it is. But no, LinuxCNC needs an M3 S? command. I don’t think of this as none of my machines have spindle control from the computer. Stupid controller.

But after giving a start spindle command threading worked as expected. Here’s the first thread from this lathe (cnc or manual).



It’s a miracle and a 1/4-20 nut fits nicely.

Now on to cross slide way wipers, covers to protect the X slide and a belt cover. That’ll wrap it up but for a lot of cleaning up, electronics box, wiring, etc. But there’s light at the end of the tunnel.

Thank for coming along.

[CL_MotoTech]

That looks really great! Congratulations on the mile stone of threading. Impressive work you have done here, and building your own encoder kind of blows my mind.

[currinh]

CL_MotoTech:

Thanks. It's taken a bunch of time so far, and still quite a bit of clean up to go.

The encoder wasn't that difficult. It would be harder to get 100 or 200 division, but 36 isn't too difficult. The CNC4PC cards and sensors make the connection to the controller easy. Then it's just designing a bracket to hold the sensors in the right locations. But, as above, it's easy to mess up the 90 degree phase between A and B (at least for me). However, my speed varies some which must be due to errors in sensor placement? If it continues to work for threading though I'm good.

Thanks for the kind words.

[currinh]

Almost a wrap!

I cut out aluminum pieces to go on the cross slide for way wipers and way protection. For this I used a piece of sacrificial plywood for backing.



Since the four small pieces only have one screw hole each, I left holding tabs. These broke out well and cleaned up easily on a belt sander. The final pieces are as:



The pieces fit on the front and back of the cross-slide.



The holes were drilled and tapped using the mill while the lathe was apart. The lower blocks hold felt in place as way wipers. The upper is to hold leather to cover the ways.



The leather just slides over the front and back. The leather should hold up well to oil and chips for a few years. If it wears out it will be easy to replace. I left the front leather long figuring the weight would help pull it over the X pulley. After trying it I suppose it could be shorter but the length doesn’t hurt anything. The back side is the same but with a shorter piece of leather.

I also built a simple belt cover for the X timing belt drive.



The sheet metal I have here is fairly thick and I can’t easily bend it. I also don’t have a welder here. I don’t have a piece of aluminum thick enough to cut the cover from, and would hesitate to cut up that nice a piece for a simple belt cover. Makes one (me) think they (I) should get a 3D printer. It would not work well made entirely of wood. But then I thought of building the face of 0.08” aluminum sheet metal and the edges from plywood. No pictures, but I cut the inside of both the face and plywood on the mill. The two parts were bonded together (JB Weld) by aligning with the screw holes. Finally the outside was cut by holding the assembly to a sacrificial piece of wood using screws through the three holes. It worked well as is seen above. It looks a little weird to have the wood on the lathe, but it is just a belt cover to keep chips out. It’s also covered with the leather way cover.

Life is good. Almost time to learn how to run a CNC lathe.

But the last task, besides cleaning up the wiring, is to build a riser block for the tool post. I’ll try to get to a steel store this week for that large (large for me 2”x3”x5”) chunk-o-steel.

Thanks for coming along.

[CL_MotoTech]

You should be proud of your build. It is very well done!

I always try to include at least one bit of wood in my race cars. Wood is fantastic and is often overlooked. Natures composite, that's what I like to call it.

[currinh]

CL_MotoTech:

Thanks for the kind words.

Wood is pretty amazing. This application was simple but seems to be working well.

[currinh]

OK. I finished a tool post riser block.



Building this I cut a new shaft for the QCTP of 5/16” 4130 round bar. Not a hard job but I could see the compound flexing. Also, the threading tool didn’t want to hold position. I’d check the position, take a threading cut and recheck. It would shift .005” or .01”. This is devastating for a CNC lathe. One really needs to set up a tool table for at least basic tools. Once done the set up for a job is simply setting G54 along the bed, Z. Any slipping or shifting of the QCTP destroys the tool table and all tools need to be reset. I looked for lost steps and found none laying around. So, I’m convinced the QCTP is slipping. This was with the compound in place AND with the bare riser block.

I planned to build the riser block and put stops on, after some time, if it didn’t hold position. But the above proved positive locking is needed.

I had trouble finding a chunk of 2”x3”x5” steel for the riser. I called all over Phoenix, a large city, and could not find that stock. Nothing special, just a chunk of A36 hot rolled. (Would be nice to make from tool steel using a surface grinder. However, that much cost and effort isn’t justified on this inexpensive Chinese lathe, plus I don’t own a surface grinder.) I finally bit the bullet and paid Online Metals to ship me the stock. I think the low cost of material from Online Metals made up for shipping cost.

Once the stock arrived I squared it up on the mill with a 2” face mill. Not hard but took some time. (Everything I do takes excessive time) I drilled and counter sunk four holes in the corners over the cross slide T-slots. Also drilled/tapped two 1/4” x 5/16” lengths of stock for T-nuts. A 1/2-20 threaded hole on top for the QCTP shaft completed the basic riser.

The lathe has two T-slots along the cross slide and a blind hole to locate the compound slide. I carefully measured and put a bar and short post to locate the new riser.



The bar was machined to be a good fit in the T-slot. On it you can see two dowel pins that positively locate it to the riser block. Same with the post which locates in the blind hole. A 1/4” counter bore holds the short post. The post itself has the 1/4” length that fits into the counter sink and a larger length that just fits the blind hole. It fit quite well. There’s some interference but can be pulled into place with the four screws, like a light press. So far so good.

I made a stop for the top of the riser to keep the QCTP from rotating. Its to be held with a central M4 screw and two dowel pins in reamed hole.



I installed the QCTP and squared it as well as I could. Then put down a width of blue painters tape and used superglue to hold the stop in place. I pushed it against the QCTP as I could. I saw this trick on the Internet and decided it might work well here. And it worked very well. The superglue held the stop in place while I drilled and reamed holes in the riser block, also drill and tap the center hole. It was easy to pop the stop off using a utility knife. Clean up was quick and easy. I’ll use this “fixture” again.

Below is a pic of the mounted riser block with stop and QCTP shaft.



Finally a pic of with the QCTP mounted.



I found a little bit of slop between the stop and the QCTP. A 0.005” piece of shim stock took that out.

I think the QCTP and riser could be taken off and reinstalled without loosing alignment. I haven’t used it since I put the alignment parts on, however it feels rock solid and I’m sure it’ll be a great improvement.

That should do it for all the hard parts. I still need to put the electronics in a box and clean up the wiring. It’s now usable so not sure how long that last step will take. I’ll report back then.

Till then, thanks for coming along. I hope you got something from the build.

[CL_MotoTech]

I admire your craftsmanship. Keep it up!

[nlancaster]

Love this, really thinking hard of buying the PM-1022V lathe and then doing a CNC conversion.

[currinh]

Love this, really thinking hard of buying the PM-1022V lathe and then doing a CNC conversion.

I though a long time and ended up with the PM1022. It would be nice to have a larger lathe but space (and $) prevented that. After deciding on a 10" lathe I went back and forth a lot between Grizzly and Precision Matthews. I can't find it now but Grizzly had a 10x22 lathe with (I think) a three phase motor and build in VFD. That was attractive and some have converted other Grizzly lathes to this configuration. But, I decided on the PM1022 for the "quick change" spindle nose and T-slots on the cross slide. These two would be very hard to change after purchase on a Grizzly. So far the motor and controller on the PM are OK. I really disliked the powered cross slide. Don't get me wrong, for a manual machine it would be wonderful. But for a CNC conversion I hate discarding the whole mechanism.

I haven't run it much yet, just finishing up the conversion. But so far I'm very happy with the PM1022 for a small Chinese lathe. So yes, go for it.

[currinh]

I finally installed the electronics in a proper box. Just kept the 1/2” plywood backboard and installed the whole thing in a plastic electrical box. It sounds so simple but took some effort.

The first picture below shows the box open. On top is the main power supply 48v DC. Just below this is a small power supply giving 24v DC Field Power to the MESA board. I installed four relays but currently none are hooked up. Then the MESA 7i76 board along with two Gecko stepper drivers on heatsinks. And finally three CNC4PC boards that clean up the encoder opto signals. A latching ON/OFF switch mounted on top of the box plus a couple of fuses.

I wanted to put the fan on the side. However, I didn’t find a great place so mounted in on the cover. Hard to see but there are many 1/2” holes on the floor of the box for venting.



And closed.



The DB25 cable to the computer comes out the left side. Otherwise wiring comes out the right. The top wire is for 110V AC wall plug. Next is to be for a 24v DC control signal for mist coolant, and below that cables for the X and Z stepper motors. Then 110v AC for the Z stepper fan and finally three cables in one connector for the encoder sensors.



The above pictures show it mounted on the wall behind the lathe. For general use it mounts on the chip enclosure at the back side of the lathe. Not wanting to move the lathe often I made the box movable, it can be taken up and out to be temporarily hung on the wall for maintenance.





The lathe is on casters which also have stabilizing feet. The picture below, from Amazon, shows one.



With the feet raised the lathe is easy to move around. But it’s very stable with the feet down, though a bit of effort to transit between the two. Even more of a pain is re-aligning the lathe after any move. Since I have way too little room to leave access to the back, this is my solution.

I’ve used the lathe a little and starting to understand its capabilities and limitations, plus how to use a CNC lathe. Still a long ways to go but so far I think it was a good thing to do. With backlash comp accuracy is likely in the +/- 0.001 range (I need to measure more carefully). That’s accuracy of motion, still need to apply technique to get good parts. So far I’m happy with it.

So I think that’s a wrap on this build. Thanks for coming along.

[currinh]

I had a PM request for drawings. I though I'd respond here also if anyone else is interested.

I have the drawings in *.pdf and would be glad to provide them. You would have to agree that the drawings would be for non-commercial use and, if distributed, they are to retain a notice that I am the original author. Note that only one conversion has been done using these plans so there are likely still errors in the drawings, user beware. I would like to hear if there are errors with enough detail to fix them. If this design isn't useful you may return them for a full refund (but since they are fee expect nothing back. :-) Just let me know that you agree as above and forward me your e-mail address.

Thanks.