[embraced]

Hi all,

After getting discouraged on a recent mill build (due to several factors), I've set my sights on making a second router to distract myself from the mill.

The first router was a R&P machine driven by AC servos and a smooth stepper. In the process of building it, I learned an incredible amount - this new build will build upon that knowledge.
(early) pics of said machine are below...


My main complains with that machine was the lack of stability with the Servos (due to several reasons), flimsy legs, R&P transmission and the awful problems I experienced with mach3/smoothstepper.



The new machine (the topic of this thread) will be much different:

600x1200*150 work area (or there about...)
Hybrid servo driven (4.5N.m closed loop steppers on X and Y, 2.0N.m on Z)
2510 double-nut ballscrews on X and Y (Y axis is dual-driven)
1605 double-nut ballscrew on Z
100x100x6 hollow-section steel used for entire frame
Welded construction
Gantry will be 150x100x6 hollow section
Linear profile rails on all axes (20mm on X and Y, 25mm on Z)
Laser-cut brackets
Aluminium profile table top
Possible sand/epoxy sand fill
Controller will NOT be mach3. I will probably get a Chinese stand-alone controller (the ones with the screens and controls on one unit)
Spindle has yet to be decided on - I'd like to get an ATC spindle with a ISO20/ISO25/BT30 nose. Having a fixed collet chuck is truly a productivity killer.


Current weight (pre-sand/epoxy fill) is estimated by Solidworks to be 400kg.


There are some details in the photos that have not been detailed yet, notably:
How the Z axis ballscrew is coupled to the motor (will be belt-driven, probably with an HTD5M pulley)
The ballscrew mounts on the Y axis, and how they attach to the frame
Lack of fasteners all-round
Most of the renders show a 600x900 machine - i've since decided to extend the Y axis to allow for 1250mm of Y travel.


I would like to hear everyone's thoughts on the proposed machine. The build will be documented here soon, given that the design doesn't change dramatically as a result of this thread.

[pippin88]

Strongly consider a square section for the gantry (e.g. 150x150, or best yet 200x200), as it will be much stiffer.
When not cutting the primary force is gravity (thus down), however when cutting you'll have twisting forces on the gantry.

Do it yourself CNC router: Design Considerations, the Gantry
That page has a diagram which sort of indicates what I'm talking about.

Have a look at LinuxCNC. I'm not aware of any serious Chinese control software. Got any links?
The options I'm aware of are (in very rough increasing order of expense): LinuxCNC, Mach3, PlanetCNC, KFlop (various ways including running Mach3 front end), Flashcut

[embraced]

Thanks for the link regarding the gantry.

I have a KFLOP + Kanalog right now for my existing milling machine, and while it's an excellent motion controller, it falls behind with the UI. The KMotion CNC app is just not that great.

For the chinese controller, these are the things i'm talking about:

http://www.sotengcnc.com/cncsystemsl...s-list-67.html
SZGH-DK1000MDc-4 four axis engraving CNC system
http://www.adtechen.com/products/ADT...ol-System.html
GSK 218M

All of these controllers (except the last one) are under $900. They all have screens and controllers, eliminating the need for the PC - a quality I do like. They also have sub-panels as an option, as well as an MPG for $100.

[pippin88]

800 x 480, 8 inch colorful TFT LCD display - not for me. Eventually I'm going to upgrade my 17" and 19" displays to larger widescreens.

$900 would buy you a perfectly good second hand pc + Mesanet hardware step generation (e.g. 5i25) for use with LinuxCNC. Well documented, customisable.
Versus known proprietary controller.
The only Chinese CNC related software I've used is Moshidraw (for lasers) and it is hard to believe how bad it is, despite years of development.

[embraced]

I understand that the resolution might be low, but I am really not interested in a PC-based controller.

These controllers are in use in thousands of existing machines. Comparing awful Chinese software (i've used Moshidraw too) to a hardware controller is not really relavant. I have used the machine in the second link (the SZGH one) and it was fine. Easy to use, cheap and reliable.

[wizard]

Thanks for the link regarding the gantry.

I have a KFLOP + Kanalog right now for my existing milling machine, and while it's an excellent motion controller, it falls behind with the UI. The KMotion CNC app is just not that great.

And the products you listed below are? The machines may perform well as a controller but the interface looks like a throw back to the 1970's. I did look at each link and frankly they don't look too bad for the price if you are expecting a retro controller.

For the chinese controller, these are the things i'm talking about:

http://www.sotengcnc.com/cncsystemsl...s-list-67.html
SZGH-DK1000MDc-4 four axis engraving CNC system
http://www.adtechen.com/products/ADT...ol-System.html
GSK 218M

The last unit seems to be the most impressive but I have to wonder about documentation and the ability integrate it into your machine.

All of these controllers (except the last one) are under $900. They all have screens and controllers, eliminating the need for the PC - a quality I do like.

I'd be very interested to know why you don't like to have a PC as a base for the machine controller. Going PC eliminates a lot of issues common with proprietary controllers.

They also have sub-panels as an option, as well as an MPG for $100.

I guess I don't see the point. I have no doubt that the hardware linked will function to an extent but it just seems to be the wrong direction to be going. At least in my mind it is.

[wizard]

I understand that the resolution might be low, but I am really not interested in a PC-based controller.

I'm honestly wondering why. Most guys in this forum thread are specifically going this route to avoid proprietary controllers.

These controllers are in use in thousands of existing machines. Comparing awful Chinese software (i've used Moshidraw too) to a hardware controller is not really relavant. I have used the machine in the second link (the SZGH one) and it was fine. Easy to use, cheap and reliable.

I will have to take your word for easy, cheap and reliable but you still have the same short comings of any other proprietary controller.

[ger21]

I'm curious why your switching from servos to closed loop steppers. I would think that you'd be giving up a lot of performance.

What kind of problem did you have with the servos you were using?

[embraced]

I suppose the main reason I'm avoiding a PC-based controller is that I want IP6x rated physical buttons to play with. I do not want to click on-screen buttons with a mouse.

As for why I'm switching from servos to CL steppers (or rather why I intend to): The servos I've got (600W Chinese AC servos) are very powerful, and combined with 8:1 reduction gearboxes I have on them, I have 15.28N.m. of torque and 611N of linear force from 0-59M/minute. So that was nice - but the stability was simply awful. I don't know whether they were out of tune or whether I had electrical noise, but they simply would not sit still. They whine when in an uncomfortable position for them - something which is fixed by physically pushing the axis until they quiet down.

I would like to switch to the CL steppers is stability. Super stable and they do not hunt when stationary.

Perhaps I would actually re-use the servos if I could get to the end of my issues with them.

[bhurts]

The next question is did you tune your servos or just assume they were good from the factory?

Ben

[blau_schuh]

Have you though about used Yaskawa or Mitsubishi AC servos? Probably can piece together the bits for cheap if you're willing to poke around on ebay a bit (got a bunch of mine from china). Took about an hour to get them up and running and at least the Mitsubishi ac servos I'm using self-tune.

I'm running 400W mitsu ac servos (+mesa +linux cnc) and they scream... Haven't tried to jog my x axis gantry with them yet, but these seem to have plenty of oomph. Was jogging my y axis at 800ipm+ in a test not too long ago.

[embraced]

The servos were not tuned from the factory. On first use they were extremely laggy - I could rotate the shaft by hand quite significantly before they would decide to respond. So I turned the gain up. This improved the response, but induced the hunting and noise. I do not know enough about tuning to know if I should tune the current loop, speed loop or position loop - all seemed to have an effect on the servos, even in step+direction mode.

I have not really thought about getting more servos for the current build - though I would love to get some drivers that were able to self-tune.

I have no doubt you could get 800IPM+ with those servos. I initially tested my machine at 2200IPM - something that's truly frightening on a relatively flimsy frame. It was impossible to stop even when creeping along I might add. Torque throughout the entire speed range is something I love about servos...

[blau_schuh]

Wow 2200ipm. Holy smokes!

I was limited to ~866ipm because the servo max rpm is around 2500rpm and i had 1:1 gearing. Had I tossed in a 1:3 pulley I'd be up around 2500ipm. Pretty sure the ballscrew couldn't handle that, probably not even wise at 866. I've had some DC servos that were over twice the size and would have had trouble stopping on a dime and reversing as quick as the ACs....pretty amazing.

I'm using the J2S series from Mitsubishi in step and direction mode. I'm told that in velocity mode they are even better.

[bhurts]

Do you have a Manuel for your servos and drives? Unless there complete trash you should be able to get rid of your hunting and noise issues. See if you can set a deadband. If so set it for 2-5 counts. That should fix the hunting. I know you don't care for the k motion gui but read there sections on motor tuning. It is a good guide to how it's done. If you have a pdf of the Manuel post it. I can honestly say I have no hunting and my motors are dead quiet and since I'm pretty slow I'm sure you'll get it.

Ben

[Tkamsker]

Hi have a Look at linuxcnc and some fpga Card ( i use General mechatronics ) then have a Look at gerainte devices vsd 160 in dual Mode. Abs now you can Choose any servo which will Not destroy you router lets say it is then a macanixal Problem ,..


Gesendet von iPad mit Tapatalk

[embraced]

Attached is the manual for the servo drivers - if someone were able to interpret this in a way I could understand it would be very, very much appreciated.

I think that the deadband function is what they call 'location finished range'. I will play with this function to see whether it has an effect on the hunting.

It still won't solve the terrible response though. I am in completely new territory when it comes to tuning the speed, position and current loops of servos...

As for the LinuxCNC + FPGA controller, it does look like a more robust solution than mach3. However, it doesn't give me what I would like the most - an IP6x array of physical buttons to play with, like the real controllers do.

[embraced]

For anyone interested, I received a quote for the stress relieving: $180AUD for stress relieving the gantry at 580-620 degrees celsius for 2 hours. Doing the same operation for the table and gantry ends up being $425AUD.

[Tkamsker]

Hi Embraced what do you mean with ip6 x Array of Buttons ? The General mechatronics deliver 4 Times 10 gpio pins i use gmoccapy which is 17 Buttons and an Keyboard Card on USB which gives me 50 keys so i am Quite happy i use Heavy Duty Buttons which cost around 20€ per piece But u can use them in wet conditions and Gloves ,.. (For testing i used 0,20€ Switches -


Gesendet von iPad mit Tapatalk

[Tkamsker]

Hi i Looked on page 10 if you have that Controller and servo just connect it to the fpga Card if Works exactely like my config. I am sure the General mechatronics Guys help you Out Thomas


Gesendet von iPad mit Tapatalk

[embraced]

IP stands for 'ingress protection' - first digit is protection against objects (fingers, dust etc), second digit is protection against water. IP6x means absolute protection against dust, but an undefined resistance against water.

It seems that the solution you suggest could be suitable, but I really am not interested in making an entire control panel. I already design and make electronic control gear for my day job, and I could not think of many things that I would want to do less in my own time!

Something like PoNET kbd48CNC would be more suitable - if I were to go that route (I may - but I haven't decided yet).