[newbie] Design considerations for home-built CNC

Hey guys,

First, I'm sorry for the noobish nature of this post. I know everyone gets annoyed by posts that are basically "help me, I'm too lazy to RTFM". I have tried to do some research into what I need, and I will present my thoughts below, but I would most definitely appreciate input from anyone who is kind enough to provide it. So, that said, here goes:

I am in the process of building a CNC electrochemical mill. Briefly, this is a device that uses a controlled corrosion process to remove material (so is a NON-CONTACT process, this is important a little bit later). Typically, such things are bulky and expensive and not really hobby-friendly. However, being an actual electrochemist, I know a bunch of tricks to translate the clunky machines down to a lightweight tabletop version, and I have done so successfully by repurposing a craigslist-find fireball V90, equipped with 260 oz-in steppers. Basically I replaced the router with a custom-machined (I have access to a great machine shop already) toolholder, and then designed and built some driving circuitry to control the voltage and current, etc. That system works, but it has some problems, the largest being that it is built with a pretty shitty frame to begin with. The V90, if you didn't know, is basically made of fiberboard and is held together with some long lag bolts. While this makes it easy to mass-produce and assemble, it is mostly designed for woodworking use, where the tolerances are not machine tolerances. I am basically running at the lower limit of the resolution of the machine, after having replaced the leadscrews to be 3/8-20. Realistically, my tolerances are on the order of a couple thousandths of an inch when running at useful speed. Some of this is probably my inexperience with linuxCNC, but certainly a large portion is the fact that it is an open loop (i.e. stepper-driven) machine running on a somewhat rickety frame. For iteration #2, I would like to do much better, now that I know the electrochemistry side of things is working as intended.

What I want ultimately is a servo-driven 4-axis machine, where one axis is a lathe axis and the other three axes are standard XYZ, but I want to be able to pattern the inside and outside of objects held in the lathe portion. To do this, I need to be able to set one axis to allow for internal vs external cutting, I will assume this is the Z axis. I will also need an axis to move the tool for the actual patterning, which I am assuming is X. I won't need to run both mill and lathe at the same time. So, my first question: It seems like all I need to do is have two separate configs: a mill (XYZ) and a lathe (XZ). If I have an encoder on the lathe spindle, I should also be able to do spindle-synchronized motion, which, assuming that my Gcode is good enough, should let me make arbitrary patterns in the work surface. Is this correct, or am I missing something crucial here? Alternatively, can I just set it up as a XYZA mill and achieve the same result? Are there CAD/CAM advantages to doing it one way over another? The functioning of the machine is more akin to a laser-cutter than a milling machine, if that makes a difference.

Right now, my plans consist of the following:

1. Build frame out of 80/20, gusseting all corners for maximum frame rigidity
2. Use granite surface plate as work surface (the surface doesn't need to be conductive with the way I have my workholding set up)
3. 3/8-20 lead screws with bronze nuts, unless I can find some 3/8-40 with bronze nuts. Double-up bronze nuts with adjustable fixture to provide anti-backlash capability. (I don't trust plastic nuts, even in this non-load-bearing application).
4. Servo-driven. Googling around, it seems like I can get pretty nice servos from Pittman motors that come with encoders from the factory. I was thinking of using the 14207S008-SP ( specs ) to drive the X and Y axes, and then either those same motors or motors with gearboxes to drive the Z and A/lathe axis. I think I would rather have a higher resolution encoder than a lower-res encoder + gearbox. Basic math suggests that for a cylinder of diameter 0.6", a 500-pulse encoder gives me a spatial resolution of around 0.004". A 2000-ish encoder brings me to 0.001" which is more acceptable to me, and there is no backlash to account for as with a gearbox. The linear speed of the z-axis isn't too important, nor is the rotational speed of the lathe axis, as for the synchronized motion, the patterning speed will certainly be gated by the X-axis speed, in fact slow rotation is probably preferable as it puts less strain on the X axis, and I don't have to worry about chip load. etc being influenced by spindle speed.

Keeping in mind that the cutting action is non-contact, is there any reason why this plan might not work? I will be the first to admit that I may not completely understand motor performance, but it seems from the performance figures that the "260 oz-in" rating is in fact a maximum rating, whereas the servo I linked above is rated at 50 oz-in continuous, and eyeballing the numbers, it seems like the servo motor is way more powerful over a broader range of speeds (the max torque fro the servo is around 400 oz-in, and decreases linerarly over the range of speeds, unlike the quasi-exponential looking decay for the stepper).

Finally, should I be looking elsewhere for servo motors? Pittman was the only company that sold the sizes that seemed to be in between hobby servos (uselessly small) and bridgeport-sized servos (no longer benchtop-compatible).

Again, I'm sorry for having (what seem to me, anyway) such basic questions, but I want to get it right the first time. I appreciate anyone's input, and if someone has suggestions for how I might implement things alternatively, I am definitely interested in hearing about it.

Thanks in advance!

Best,
Thomas

You are asking quite a bit for a servo system to keep within +- 1
count (+- 3 to 10 counts is more typical) so you either want to
gear down the motor (with the attendant backlash problems )
or use a higher resolution encoder
(note encoder resolution is lines or PPT *4)

Note that you can use a plain servo motor
and your own higher resolution encoder on the spindle

valensdiesel wrote:

Finally, should I be looking elsewhere for servo motors? Pittman was the only company that sold the sizes that seemed to be in between hobby servos (uselessly small) and bridgeport-sized servos (no longer benchtop-compatible).

Pittman also makes some really AWESOME brushless motors, I have some of their 4443 units with
Hall sensors and encoders built in. Not quite hobby-priced, but you may be able to find
them surplus, like I did. Also, Automation Tech (formerly Keling Inc) has some nice NEMA
size 23 motors in both brush and brushless that are pretty affordable.

Electrochemical machining sounds like slow movement once it starts a "burn" and
so very little motor power is needed.

Jon

Wow! Already you guys are awesome. Thanks for the links to alternative motors. Also I didn't realize that the software had rising/falling edge capability (presumably this is where the factor of 4 comes from?) If I am interpreting what you are saying correctly, then an encoder with a listed resolution of 2048 ( example ) can actually read out at 8192 counts per revolution, which helps to offset the 3-10 count error?

Thanks!

Thomas