[ryan99alero]

I'm sure thats an oxymoron saying quality and affordable in the same sentence. I'm in the final stages of designing a home built CNC router in CADD before actually building. Rough Dimensions are 1500mmx1500mmx550mm. Already purchased LinEngineering 5718L steppers and had planned on installing some of there encoders either on the stepper's or on the other end of the linear rail. X axis will be dual driven so I'd like a controller that supports gantry correction. Also purchased 4KW Spindle with Invert Drive. From what I've found my best bet is the CS-Labs CSMIO/IP-S controller. Haven't purchased Stepper Driver yet but was looking at either leadshine or Lin Engineering as well. Wanted to get some Gecko's but I don't think they support closed loop for steppers.

Was curious if anyone had any other recommendations. I'm sure I don't need step feedback but from my understanding the whole system will operate that much better if it knows where its precisely at at any given moment. Currently building the whole thing so I can build some molds for carbon fiber parts for building a custom drone I'm working on. Was given a price around 3k for 8 molds so decided it was time to build a cnc machine. Besides I like finding projects to do with the kids for educational. We built a 3D additive Printer that will print a piece 12"x12"x24" tall. So of course we need a subtractive 3D printer... lol

I'm sure I left out some vital information so I apologize in advance. I'd appreciate any help I can get.

[ger21]

The only real closed loop stepper options are KMotion or LinuxCNC.
Almost nobody runs steppers with a closed loop controller. It's just not necessary.

The IP-S only works with Mach3, Mach4, and CS Labs new SIM CNC.
Mach3 is obsolete, and I've heard that CS Labs Mach4 support is poor.

You might want to look at Centroid's Acorn controller.


You need to work on that design, as imo it needs some work.

1) The gantry sides are too small, and too tall. The gantry beam will sway side to side a LOT.

2) Ditch the bottom bearings on the gantry sides. The only reason they'd be remotely helpful, is to help stabilize the too tall sides. Otherwise, they double the cost, with little benefit.

[PCW_MESA]

The only real closed loop stepper options are KMotion or LinuxCNC.
Almost nobody runs steppers with a closed loop controller. It's just not necessary.

There are some circumstances where this can be of benefit, for example, a number of LinuxCNC users have combined
linear scales with feedback to steppers on mills and gotten measurably better accuracy. This does not make sense on a router
or similar less rigid/massive machine however due to the relatively low feedback bandwidth.

[ryan99alero]

a

[ryan99alero]

The only real closed loop stepper options are KMotion or LinuxCNC.
Almost nobody runs steppers with a closed loop controller. It's just not necessary.

The IP-S only works with Mach3, Mach4, and CS Labs new SIM CNC.
Mach3 is obsolete, and I've heard that CS Labs Mach4 support is poor.

You might want to look at Centroid's Acorn controller.


You need to work on that design, as imo it needs some work.

1) The gantry sides are too small, and too tall. The gantry beam will sway side to side a LOT.

2) Ditch the bottom bearings on the gantry sides. The only reason they'd be remotely helpful, is to help stabilize the too tall sides. Otherwise, they double the cost, with little benefit.

I'll try and find some specs on what is required for load force and cnc machines. I already purchased the bearing rails so I'll keep them. Fusion360 says with 500lbs of lateral force on the upper top edge the whole unit would deflect a max of .008193". The Y axis Gantry braces are steel 1.5" x 6" x 25". I ran several load simulations in all directions with 500lbs and it passed all.

You have a recommendation of what lateral load force I'd need. I figured the tools would snap way before I ever got to 500lbs. I was wanting to stick with using closed loop as I also plan on designing a 2 axis head so I'll have a 5 axis machine. But if you're saying using open loop with steppers and no encoders the machine will function pretty much 98% the same as far as accuracy, simplicity of cut and speed as a closed loop system would then that would save me some money. Just a lot of the articles I've read says go closed loop if you doing more than super simple 2 axis cuts. Machine will do mostly Carbon Fiber and Wood with some aluminum. The height I have it at is for some wood projects I'm wanting to do.

[peteeng]

Hi Ryan - Apply the FE load to the spindle collet area. You will need about 10N/0.001mm deflection at the tool for a stiff machine 5N/um minimum. How are you modelling the bearings? They are compliant so you need to figure that out for an accurate model.

I think you need a reality check, if you push that column sideways with your hand it will deflect 0.008" even if its 1/2" steel?? I also think the column bearing spacing will have to be wider for a 1500mm wide gantry. I'm sure you won't need feedback if the motors are sized right but they are better if you have the budget. But maybe better to spend it somewhere else on machine.. Peter

Just reread and they are 1.5" thick serious metal !! seriously heavy for this machine Aluminium would do but keep at the numbers... Did your reality check and all good!!

[ger21]

But if you're saying using open loop with steppers and no encoders the machine will function pretty much 98% the same as far as accuracy,

If you don't try to run faster than the motors are capable of, it will be accurate 100% of the time. There are tens of thousands of machine running just fine with open loop stepper systems.


I guess I didn't realize the scale of that thing. I've never seen anyone use gantry sides that thick.
With a 550mm Z axis, my concern would be twisting of the extrusion.
Any time you'll be cutting near the table, that tall Z axis will be a long lever, which could easily twist the gantry beam
Since you are already using heavy steel sides, a tube steel beam would be a much better choice.

[rodw]

I personally think that there are so many areas that inaccuracies can creep in in your design that worrying about encoders in your design is just adding additional complexity and cost that is not required. I would not be a fan of the extrusions for the gantry and worry about the flimsy gantry ends.

As far as the controller goes, you could do a lot worse than Linuxcnc. Pair it with a Mesa 7i76e 5 axis ethernet board with spindle control (or the cheaper Mesa 7i96 with fewer inputs) and you will have a high performance, gantry friendly system as LinuxCNC will square the gantry when it homes. I've been running my 1220mm x 1280mm plasma table without encoders with this setup and its really awesome. Its spot on accurate and I've never had accuracy issues even at up to 21 m/min. In my experience optimising for maximum speed and accelleration which plasma requires, if you are missing steps, you are exceeding the design parameters of the steppers.

I've never played with encoders but it is my understanding that you start missing steps whan you have no torque so the steppers can't catchup the missing steps.

[109jb]

I'm currently working on my 4th CNC conversion and on this one I'm using closed loop steppers. All the previous ones used open-loop steppers. The open loop work fine but the peace of mind not worrying about lost steps makes the closed loop worth it to me.

Now I know someone is going to say that a properly tuned open loop system will never lose steps but I disagree. If a stepper is used for a repetitive task where the parameters like required torque, rpm, etc never change then absolutely yes the open loop stepper should never lose steps. Think pick and place machine or something like that. Now on a CNC router, milling machine, lathe, etc, the parameters might not always be the same. I for example seldom make the same part more than once and do a lot of manual programming and MDI work. Drilling a 1/4 hole in aluminum is much different than drilling a 3/4" hole in steel. The loads vary and I have had my open loop machines lose steps and crash because of it. It doesn't happen often, but it does happen. Something as simple as a feed rate set too fast can do it. So is a feed rate too fast indicative of an improperly tuned machine, or is it s well tuned machine that "oops" I screwed up. I think the latter. Now I'm talking milling machine and metal working. If you are using a router for sign making then maybe you can get it to where your loads are always well within open loop stepper capabilities. My machines are also dovetail machines and even adjusting a gib too tight can add enough drag that a stepper will lose steps during a rapid.

As far as crashes, I remember one crash in particular where the chip stuck in the drilled hole and when the Z axis went to rapid up is stalled. The next move was a rapid XY and it broke the drill, ruined the part, and slipped the head out of tram. On my new machine that won't happen because the closed loop driver would have faulted, then sent a signal to my control software, which would have halted all motion immediately. This is the biggest benefit as far as I am concerned.

As far as accuracy the closed loop stepper DOES NOT improve accuracy. If anything a closed loop stepper is less accurate n a sense because it is allowed to be a certain amount off before an alarm.

I have also seen people comment that closed-loop on a stepper is useless because once a stepper loses steps you don't have the torque to get them back. Well I can't speak for all closed loop drivers, only the ones I have, but they have a digital display that shows how many encoder counts the motor is off from the commanded position. At slower speeds this always reads 0, but higher speeds and loads I can see it lagging by 20-30 encoder counts but it does catch up. This alone says that there is something that allows the closed loop stepper to catch up. My drives are Leadshine copies and have little documentation, but the leadshine manual states that high speed torque is 30% better than an open loop stepper and I would believe it. Even just the error tracking makes them beneficial and certainly not useless to me, but I do tend to think they perform better than open loop, although I'm kind of comparing apples to oranges since the motor sizes I have are very different (230-570 oz-in NEMA 23 open loop vs 1600 oz-in NEMA 34 closed loop).

I do think that AC or DC servos of similar physical size can be made to perform better, but most require significant tuning. I know some of the servo manufacturers are making it easier with auto-tuning software and such so this probably isn't as big of a deal, but there is no more tuning for the closed loop stepper when compared to an open loop stepper.

As far as tuning, on open loop steppers, I have reached a point where I set max velocity and accel to 60% of the stepper stall value for open loop steppers. I have enough confidence in the closed loop stepper, particularly its ability to signal a problem, that I've set these to 80% of what I found the stall values to be. So far I haven't received any errors during motion testing which has included some aggressive programs.

Bottom line is that I think the closed loop steppers have some real advantages over open loop. What those things are worth is up to the user, but certainly valuable enough to me that I don't have regrets going closed loop so far.

[Hood]

Modern day AC servos are almost plug and play.
I have older Allen Bradley DSD drives on a few of my machines (big lathe, plasma and Beaver mill), the Auto Tuning was a hit or miss and even then they required a bit of manual tuning even if it worked.

I have Telemcanique Lexiums on my Chiron and auto tuning was quite good on them and manual tuning required was minimal and likely most wouldn't bother,

I had a Samsung drive on the wee lathe and that didn't need anything at all, just hooked it up and it basically tuned itself and was exceptionally accurate. I hear the recent Chinese AC drives are just as good although I have no personal experience of them at this time.
Price wise I think the difference between closed loop steppers and AC servos is closing considerably but the performance advantage is massive with servos.