Converting a benchtop mill to CNC

[pensiveape]

[xfer42]

I did it with a Grizzly G0463. The only "kit" I bought for it was the ballscrews/ballnuts for the mill. The rest I pieced together.

Servos, ballscrews and ballnuts.
LinuxCNC Controller (opensource/free).
Gecko 320x drives (forgot where... some place that had the lowest price).

Later I ripped the DC motor and gears out and used a 3 phase motor ($49 from ebay +$50 shipping), belt drive 1:1 (timing gears and belts from sdp-si.com) and VFD (from ebay)

Got the ballscrews and ballnuts from CNC fusion. Not sure if they are around anymore. All they did was cut, grind and thread ball screws down for your mill.
Servo Motors w/encoders from ebay ($69 each). This was actually the first purchase, because it was a huge unknown as far as availability, price and size.

Almost all of the GCode I run through it is manually written. This was the first mill I ever touched, btw. No previous experience with them.

Couple of vids of it running:

[xfer42]

More details on the parts (as far as I can remember, in order of purchase):

Servos
3x MCG servo motors with 5000 count Renco encoders 60v, 23lb/in (~$69 each). Ebay
- I started with the servos because I didnt know if I was going to be able to find these for an affordable price.
- Steppers and their drives are much cheaper, so I figured start with this to determine if Im going to go down this route.
- Old link to pic on ebay of what I purchased: http://vi.vipr.ebaydesc.com/ws/eBayI...1&secureDesc=0


Power Supplies
3x 48v 8.3A power supplies ($59.99 each). I should have gone 60v+, but these work fine. Ebay
- The motors are rated for 60v, so they can handle more than 60v of pulse width modulation.
- I didnt know how hot they would get while holding their position (dithering), so I played it safe at 48v.
- These are the generic rectangular power supplies you see on ebay.


Drives
3x Gecko G320X servo drive. $115 each. Practical Micro Design (ebay is about same price or less now)
- Tested the servos once I got these to see if I could get sufficient resolution.
- I was surprised that I could step the DC motor 6 steps within a degree. Insane resolution, but thats unloaded. Theoretically, better than .0001 in resolution. My goal is to get .001 once all together and dialed in.


Parallel port optoisolator
1x ???? ($47). www.cnc4pc.com
- They dont make mine anymore. They have smaller, cheaper, better ones now.
- This is so the servo drives are electrically isolated from the computer and they dont pull any current from the parallel port.

Ballscrews and Ballnuts
1x "Small Mill CNC kit #2 - Ballscrews" ($389.00). DONTGOHEREwww.cncfusion.comDONTGOHERE (defunct, dont go there now. Looks like a bad site)
- This came with 5TPI Thomson ball screws cut, ground and threaded for X3 mill. Still needed mounts for it.

Mill
1x Grizzly G0463 Small Mill ($1250 + $99 shipping = $1349.00)
- From grizzly


Misc:
- Empty computer case to put everything in (The hardest, most time consuming part of the whole project!)
- Cables. Used Shielded CAT6 + Speaker wire. The CAT6 is used for encoder signals and power. Speaker wire is used for servo motor.
- GX30 10 pin aviation connectors from China (ebay, ~$10 each).
- Endmills, boring bar, clamp set, oil
- Just started buying 6061 aluminum stock. Didnt know what all I needed, but I knew I needed an assortment to start with.


Mounts:
- Once I got tooling and stock, I started making the motor mounts. This was a bit tedious since I needed to tear down the mill, write down measurements, then reassemble it to make the parts.
- I learned the holes drilled into the cast iron were far from precise. They obviously put the parts together, then drilled it by eye. So kind of a pain to transfer those holes into specific locations on paper. The solution was to best guess and oversize the holes for wiggle room.
- For Z axis, I took a guess that the motor could handle the Z with a 2:1 reduction. Works great.

Timing gears and belts:
- Z axis: Not sure what I bought, just made sure my combo was 2:1. https://sdp-si.com/


Stuff added later:
-Boxed the mill in. Kids were getting aluminum chips in their hair, and all over them. Drove the wife nuts. This severely cut down the mess in the garage and aluminum chips around the house.
-3 phase 1hp motor, 1:1 belt drive. This cut down the noise. Not a whole lot of benefit. (Motor was from Ebay, Timing gears and belts were from sdp-si.com. I had to bore out the center to fit the motor and the mill).
-EPDM rubber sheet to protect the back. Ebay, I think.
-Arduino controller to manage the ERR/RESET pin on drivers. This is just so it has blinking lights, and I can visually tell the status of the drives and which one faulted.

First run, walk around:


Testing backlash setting in LinuxCNC:


Later added an Arduino controller to the Gecko drives so it would tell me which axis has a fault.
Previously, all ERR/RESET lines were tied together in parallel, so if one faulted, they all faulted and stopped.
This was perfect at first, but later I wanted to know which axis threw a fault after tuning the drives.
First part of the vid just shows it can tell you what axis faulted. The second part is a test for noise. The arduino is super sensitive and it was reporting faults due to EMI. I fixed it in software by reading the pins 1000 times on each pass, and counting the 1s and 0s. If there were more 1s than 0s, then a 1 was returned and vice versa. Kind of a software debouncer/filter. There was no way I was going to rebuild this thing with more shielding. It can scan all 3 drives in less than a ms, so not a biggie.

[ar15barrels]

Couple of vids of it running:

How come you didn't just mill the threads?

[Mesa Defense]

Looks like a good time

[xfer42]

How come you didn't just mill the threads?

Because its more difficult for me to break a large tap
Honestly, I didnt know I could until you mentioned it. Never thought about using a cutter and spiraling down. Now I need to tool up for that...

[ar15barrels]

Because its more difficult for me to break a large tap
Honestly, I didnt know I could until you mentioned it. Never thought about using a cutter and spiraling down. Now I need to tool up for that...

It's called thread milling.
They sell cutters that are pitch specific and cut a bunch of threads at once.



You actually only need to make a few rotations to rough in and finish.

You can also use a single point cutter and spiral all the way down.
There should be lots of canned cycles for this.



You have the ultimate control of pitch diameter once you start thread milling.
Just tweak your cutter comp and you can fine tune threads for an almost lapped fit.

[pensiveape]

Wow, thank you xfer42!!! for your helpful and detailed response.

I am a little overwhelmed by how much I don't know. I don't have experience with robotics, electronics, or even more than a casual proficiency in computers... but I'm willing to learn!

My questions may seem pretty stupid, but I honestly don't even know where to start to learn the fundamentals.

Is there a reason why you picked the Grizzly G0463 over other benchtop mills?
-The two machines I was looking at was the Grizzly G0704 and the Precision Mathews PM25. I was told the PM25 has an upgraded motor over the g0704. I'm not sure what makes a belt driven motor better.

Why did you pick linuxcnc as a controller over mach3?
-There is another controller called "centroid acorn", but it seems to be less popular than the above two. Did you have a criteria that helped you choose linuxcnc from the above options?

You mentioned needing an "Parallel port optoisolator" to electrically isolate the computer from the motors.
-I vaguely recall seeing a conversion video where someone mentions that you could use an ethernet cable to do the same, as opposed to parallel port or USB (I think this was in regards to mach3). Is this the case with LinuxCNC too?

How do I determine which drivers go along with which stepper motors/servos?
-For stepper motors, Nema 23 and Nema 34, seem to be the most commonly used...
-Ideally I'd like to purchase the best performing parts I can, within reason. How do I determine the quality and performance of the stepper/servo motor and the subsequent driver?
--are there accuracy metrics?

Suppose I'd like to upgrade my mill later on from stepper motors, to servos...What exactly would I have to change? Is it just the motors and the drivers, or is it more complicated?


I considered purchasing a pre-converted/assembled benchtop CNC from https://procutcnc.com/, but I found a conversion kit they were offering on amazon, which had consistent negative reviews, and ultimately im deciding against it. While it will take much longer to learn and convert my mill, I think ultimately it will pay off.

Thanks for your help. I realize that it's probably too much to ask to get into the minutia of my dumb questions, but if you can, please guide me in the right direction of where I would go to learn the conversion process, from the most fundamental aspects and on.

[xfer42]

Sorry for the delay. Ill have to respond in chunks. Every time I try to start, something else comes up.

Is there a reason why you picked the Grizzly G0463 over other benchtop mills?

The G0463 is the Sieg X3 mill.
My requirements:
Needs to have enough Z travel to bore out the receiver extension (buffer). Thats it. If it can do that, it can do everything else I wanted.

I liked the X2 and X3 setups because the examples I saw controlled the head, and not just the quill. Also, the round column mills looked like a huge pain every time the head moved.
There were ballscrews readily available for the X2 and X3.
X2 did not have enough Z travel, so that narrowed it down to only the X3 (G0463).
The belt driven part:
Gears are loud and theres slack between the gears. Belt driven is much quieter, and the spindle is smoother. I switched from brushed DC w/gears to induction w/timing belt because I needed more power and I was trying to get the best finish on aluminum.

Why did you pick linuxcnc as a controller over mach3?

1. I could download it immediately, load it on a VM at work and home and play with it way before I was ready.
2. I was comfortable with Linux. I work in it daily doing python and perl dev.
3. Its an OS and app tailored for the task, not just an app. (RTAI kernel extension).
4. Its opensource. Constantly updated by enthusiasts and experts.
5. Its free.
6. Windows sucks for any automation. It will update, go to sleep, reboot when you dont want it to.

No solid reason really. Mach3 is probably easier for most people. Just a guess.

You mentioned needing an "Parallel port optoisolator" to electrically isolate the computer from the motors.
-I vaguely recall seeing a conversion video where someone mentions that you could use an ethernet cable to do the same, as opposed to parallel port or USB (I think this was in regards to mach3). Is this the case with LinuxCNC too?

You can take any two IO pins on the parallel port, plug it into a drive and make it move. One pin is DIR, the other is STEP. Toggle STEP 010101 and it moves. Flip DIR from 0 to 1 (or vise versa) and it goes in the other direction. Perfect for a test, but you dont want to keep them that way.

If you dont have a parallel port, then your options (AFAIK, but Its not like I stay up to date on this stuff) are:
1. Get a parallel port.
PC_Parallel_Port -> breakout_board/optoisolator -> drive -> motor

2. Use one of those USB/Ethernet smooth steppers.
PC_Ethernet_port -> SmoothStepper -> drive -> motor

Those USB to parallel port cables will not work. They dont come up as LPT1,2,etc. They come up as a printer. Then, there are buffering and timing issues that make it incompatible. Timed pulses can be sent out to the parallel port, but not the USB. The USB will buffer it, then send them all out at once. Thats where the Smooth Stepper comes in. The PC can tell another micro controller via USB/Ethernet to send a set of timed pulses out to the drive.

I dont believe LinuxCNC will work with a smooth stepper. I never looked into those.

[mwr]

Thread milling

I did not watch the videos (the reason the threads were not milled may lie within) , you obviously have experience with machining - my guess the reason the threads were not cut; was due to the inability to have three axis simultaneous movement? Coupled with the lack of cutter compensation, may make it very difficult to achieve he correct PD size.

Just my .02 worth

[xfer42]

How do I determine which drivers go along with which stepper motors/servos?

You either get stepper drives, or servo drives. They have relatively high current ratings in tiny packages, so you dont have to size them for desktop/hobby stuff. I think the Gecko G320X can do 80v@20A. Thats more than 2HP of servo, which is enough for a powerful spindle.

-For stepper motors, Nema 23 and Nema 34, seem to be the most commonly used...
-Ideally I'd like to purchase the best performing parts I can, within reason. How do I determine the quality and performance of the stepper/servo motor and the subsequent driver?
--are there accuracy metrics?

They are simple devices and I dont think motor "quality" is a factor when used as hobby or desktop. If it were running 24/7 for an expected 10+ years, it may be a factor. And you would definately pay a premium for the proven performance. Even still, it will be bearings that die (or brushes, but people like to stay away from brushes for continuous, long running things since they wear).

Suppose I'd like to upgrade my mill later on from stepper motors, to servos...What exactly would I have to change? Is it just the motors and the drivers, or is it more complicated?

The hardest part will be making new cables and rewiring. Swap out the drives (same inputs). Swap out the motors and run new cables. You dont even need to tell the controller that you are using servo/stepper. They are controlled exactly the same. You will need to tell the software how many pulses per rotation, how many rotations per in, etc.

I considered purchasing a pre-converted/assembled benchtop CNC from https://procutcnc.com/, but I found a conversion kit they were offering on amazon, which had consistent negative reviews, and ultimately im deciding against it. While it will take much longer to learn and convert my mill, I think ultimately it will pay off.

Only attempt to retrofit a mill if that in itself would be an extremely satisfying accomplishment for you. Desktop CNC mills are cheaper than ever now (all steppers though..boooo), and people have worked out the torque requirements for the setups.

I always wanted to do it, but I knew that I was committed once I made the big purchase (the $1200+ mill). Thats peanuts to a professional machinist, since it will payback. This was a hobby piece that was of no benefit to income, nor was it a tax deduction. A toy that the wife would most certainly complain about. Also, the wife and kids cant use it and it would take up more space in the garage. It was probably over $2000 for everything (rough guess for now), but I spread it across a year so it wouldn't sound off alarms with the wife.

I bought the servos first. Then I started buying stuff to make it work. Those were the unknowns, and cheaper than the mill since I never worked with servos and drives. Once I had suitable control hardware and was confident it could be done, I proceeded and bought the mill.

Im glad I did it. If I didn't do it, Id still be researching on retrofitting a mill to CNC. Its great tool to have.

[WinstonMonty]

I have a brand new harbor freight mini mill that I purchased last year to do this project, but got super busy at work. If you’re interested, let me know OP. I’d be willing to sell it at a discount to get it out of my garage.

[xfer42]

I did not watch the videos (the reason the threads were not milled may lie within) , you obviously have experience with machining -

I am no machinist. Just have a lot of hobbies. This mill is the first mill I ever touched.

my guess the reason the threads were not cut; was due to the inability to have three axis simultaneous movement?

All three axes can move at the same time. The vid of the lower shows it ramps down. Its only X & Z, but thats due to me manually writing it that way.

Coupled with the lack of cutter compensation, may make it very difficult to achieve he correct PD size.

Correct, no cutter compensation. I rarely use cutter compensation on simple objects, since it requires keeping an accurate tool inventory database in LinuxCNC and whatever CAD/CAM software. I prefer to write the gcode manually. I find it easier to just document how it will be machined manually as gcode in my spare time, then write subroutines for repetitive 2D operations.

The tool head was first attempted with Rhino -> RhinoCam. I didnt like the tool path, and it generated a boatload of GCode which I didnt want to even try to edit (project finished, but it was painful to watch. Beer was required). The manual GCode for the tool head is probably 100th the amount of text compared to the generated one from RhinoCam. This is due to G02, G03 (arcs) and subroutines. Basically, it was a lot more work than writing it manually. If I were doing it as a job, or mass production, then it would be worth the time. Once you have the measurements, its easier to start documenting the tool path rather than going down the 3D modeling, and trying to tell the computer how you want to machine it.

[jimmykan]

Wow... Manual G-coding is hardcore, or normal if you were from the 1980s.

My dad used to write G-code manually. Would crash a toolhead into a workpiece once in a while. Happened a lot less after he started using MasterCAM.

[xfer42]

Wow... Manual G-coding is hardcore, or normal if you were from the 1980s.

My dad used to write G-code manually. Would crash a toolhead into a workpiece once in a while. Happened a lot less after he started using MasterCAM.

Its probably due to simulation. LinuxCNC is free, so I have several instances of it in virtual machines. I prefer to simulate it in LinuxCNC because theres some code that is specific to it (subroutines). Plus it eliminates any possible interpolation discrepancies. Its easier to manually write the GCode if you know how it should be manually milled. If you are making something, and start with a sketch and 3D model, then its certainly easier to use MasterCAM or some other software to generate it. This was the last thing I made that I didnt even want to think about the Gcode. It was simple, but I didnt feel like figuring out the inside curves.

(last project):
18058080_10209980326847938_2422762444377251171_n.jpg
18056838_10209980225285399_231899569466317309_n.jpg