Controllers for 3D cutting
This is a quote from the F360 manufacturing forum seems to summarize my situation.
"Unless the 3d op toolpath is aligned with an X or Y axis, it’s going to be line segments. It’s a fundamental limitation of many controls and probably fusion."
I have a Shop Sabre model 23, 3 axis mill with WinCNC and am unhappy with the 3D cutting times. I have spoken with Shop Sabre customer support and WinCNC support several times each. After more than a dozen trials air cutting my conclusion on the machine is that Fusion 3D parallel (raster) give the smallest tool path size and shortest actual cutting time. I have tried most every 3D option and only morph spiral comes close. I have also adjusted tolerances and smoothing to reduce file size. Actually, 2D adaptive clearing is faster than 3D adaptive clearing on my machine. The 3D parallel appears to simply raster with Y (in my case) being held constant. The actual cutting time with gcode of 180 ipm is usually 80-90 ipm actual. That is 45-60% of fusion estimated machining time with file sizes of 700 kb. 2D actual cutting with gcode of 180 ipm is usually 170 ipm, 95% of the fusion ipm estimate. This is with a HSD 2.2 hp spindle. It is annoying to watch a machine this size cutting at 90 ipm.
Although the machine can rapid 300+ ipm, vortex recommendation with their tooling is 180 ipm max with the 2.2 hp spindle. No sense upgrading the spindle if the ceiling of 3D is 100 ipm.
Question – Is there logic to the controller affecting the cutting time? Are there controllers that are better at performance for 3D tool paths? It is interesting to me that none of the files cut so far have any lost steps, loss of detail or failed. The on screen running gcode steps are in sync with the machine movement and code scroll stops when the spindle rises to return home. I am not sure if the code scroll indicates the controller not “feeding” drives in real time or not.
Until a couple years ago I had access to a ShopBot PRS48 and Buddy. With those machines I was cutting 280 ipm actual. Of course, 4.5 hp, servos and 440 v. My shop now is limited to 220 single phase besides not having the space or budget for machines that size. Probably adds to my being annoyed with 90 ipm.
Steve. (apology for the dissertation, but wanted to give enough information.)
Re: Controllers for 3D cutting
Quote:
Originally Posted by
SteveS
"Unless the 3d op toolpath is aligned with an X or Y axis, it’s going to be line segments. It’s a fundamental limitation of many controls and probably fusion."
When you boil the G code down to the controller level, all moves are converted to line segments. The controller only understands steps or encoder pulses, it has no idea what an inch or a millimeter is. Arcs are also converted to line segments at the controller level, there is a lot of trig happening there. Arcs are converted to a series of X/Y/Z coordinates measured in steps or encoder pulses. So a simple G2 move that is one line of G code, might be 1000 or more lines of segmented moves at the controller level.
Fusion outputs line segments only in 2D adaptive, except for the leadin/out moves, not sure about 3D adaptive because I have not used it. I'm not sure about any other CAM software. One thing I have noticed about Fusion is that it does not create the best tool paths and seems to waste a lot of time on moves that make no sense. There are some settings that you can tweak to reduce the tool lifting and that type of thing, but I have not figured out any way to get it to work from one side to the other in a logical sequence without unneeded rapid moves from one side of the work to the other.
On my lathe, many times I hand edit the G code to optimize the tool path. Fusion seems to waste a lot of time with unneeded moves. But the lathe G code files are very small and are easy to edit. Hand editing a 200,000 line mill G code file would be next to impossible and would take more time than you would save in cutting time.
Quote:
Question – Is there logic to the controller affecting the cutting time?
It could, but if there were any significant delays in processing the commands you would notice a jerky motion in the movements. I would expect to see a difference in the controller performance between a million dollar machining center and a home shop machine. In the high end controllers there is some optimization and trajectory planning done on the G code as it's processed with a deep look ahead. You are not going to have this in a low end controller.
Quote:
Are there controllers that are better at performance for 3D tool paths? It is interesting to me that none of the files cut so far have any lost steps, loss of detail or failed.
I think not in the controllers that would normally be found in a home shop. That they work as well as they do has always been amazing to me.
Quote:
The on screen running gcode steps are in sync with the machine movement and code scroll stops when the spindle rises to return home.
The display should stop until the tool has moved to the position commanded by a particular line of G code, it won't execute the next line of code until the move is complete.
Quote:
I am not sure if the code scroll indicates the controller not “feeding” drives in real time or not.
It depends on the CNC software. Some software may take the G code line by line and process it in real time. Others may compile the entire G code file into something the controller can understand and drip feed it to the controller. The latter is how I do it, the entire G code file is processed, converted to controller compatible code, and stored in memory, and is fed to the controller as there is room in the command buffer. As the command code is being executed, there are pointers that select the correct line of G code on the screen so there is a visual representation of the program progress.
Re: Controllers for 3D cutting
When 3D cutting, the machine will almost always cut slower than the commanded feerate. The laws of physics won't allow it. The machine has to slow down to change direction, which results in lower speeds. The faster the machine can accelerate and decelerate, the faster it can change direction, and the higher the velocity will be. The speed of the Z axis is most often the source of the lower speeds.
I'm not familiar with WinCNC, but you should be able to change settings to make it faster, but it might come at the expense of loss of detail.
Imo, you chould be able to cut much faster than 180 ipm, regardless of what Vortex says.
Re: Controllers for 3D cutting
Hi Steve - What sort of 3D cutting are you doing? Does it need to be accurate? There's a lot going on between the point of creating the gcode and the actual real toolpath. All above points mentioned are valid. But you can tweak the controller settings to improve the outcome. The CAM system predicts the path time using the path length and the set speed. It does not know that the velocity planner will slow the tool down as it goes up and down dales and round corners. if you do not need accuracy eg you are doing sculptural work then you can relax the tolerances quite a bit. This makes for a wider roadway for the velocity planner to go down.... Cheers Peter
Re: Controllers for 3D cutting
Hi Steve - I looked at the wincnc manual. There are settings for vel, accel and it has S curve fitting which means it figures out tangential connection paths for various curves and straight blends. It also has a min vel setting for cutting. It also has a minimum rad vel setting...So between all of those you should be able to speed up your machine to a degree that gives you the accuracy you want. I'm sure the wincnc people can help sort it out also. Looks like a comprehensive system. Peter
to answer your question about better performing cam systems yes high end systems use "recursive" algorithms vs simple look ahead algorithms to control the velocity. In an recursive system the gcode is written, then it is played backwards or forwards and the acceleration and jerk are calculated then the path is changed to smooth things out. This may take several cycles to achieve a specified path within tolerance. This is like driving your car into an unknown sharp corner you won't get through it fast the first time, you probably will hit the brakes heavily. But the next time you will enter and leave different and start to smooth the corner out. Once you smooth the corner out you may even try to go faster, this is exactly what a top end system does prior to actually cutting something. Hobby systems look ahead once and when the delta accel is too high it hits the brakes. Then it speeds up again but there is no learning or feedback loop up front to smooth things out. If there's someone out there that writes these codes would be good for an insiders explanation...Cheers
http://www.wincnc.net/webfiles/CNC%2...nual_3.0r8.pdf
I also read the Fusion 360 stuff on toolpathing and there are tolerance and smoothing options within that. I suspect you may have a conflict between the F360 tolerances and the CAM tolerances slowing things down. If roughing then F360 has smoothing functions that will speed up various things by creating arcs vs polyline type paths. So read through the info and have a play with your settings... I use mecsoft and UCCNC and a UC100 so have a different set tools. I do notice that if I set curves to be broken into small (say 1mm) lengths then the tool can't go really fast... so have to play with that occasionally to speed things up. I run simulations in mecsoft with different settings and soon figure out how to shorten the run time... again regards Peter
Re: Controllers for 3D cutting
@Steves
In a CNC controller a lot depends by internal toolpath pipeline (Look Ahead, Interpolator, Smoothing, Jerk and A/D dynamic controls, etc) algorithms.
Any CNC does all or some of this things (depends by complexity of CNC control) to obtain fluid movements and try to keep programmed feed following axes caracteristics, overall when advanced rought (trochoidal) are used.
Try to place a problematic G-Code file with evey axis caracteristics:
- Max axis Velocity
- Accelerations (ipm or mm/s2).
I will try to check with our CNC the results.
We are fall in all these behaviours during develop of 5 axis continous machining and fine tuned CNC algorithms to get best possibile from machine, so all technology is fall also in 3 axis configuration.
Example of 5 axis continous maching using 2000 look-ahead blocks, bezier smooting alogorithm, jerk (S-Curve) and Acc/Dec Dynamic Control:
http://www.youtube.com/watch?v=NLQyBfYOvj0
Re: Controllers for 3D cutting
Jim,
WinCNC is used on several brands of $40-$60,000+ commercial OEM CNC in the cabinet manufacturing and sign arena. Sold to individuals at $1,250 so not really entry level hobbyist.
If there was a controller that could be better at 3D, I would seriously consider it.
Steve.
Re: Controllers for 3D cutting
Rather than responding individually, I am going to cover most all here.
Peteeng - I have spent time on the phone twice with WinCNC engineers to optimize their INI file with my needs and some improvement was made.
Gerry - I changed the INI value (per guidance from WinCNC) for acceleration from 1,000 in increments to 3,000 and saw no appreciable difference, possibly 5% time reduction.
Gerry - Acceleration, that has been in my mind for some time and in the absence of any other good path I am going to replace the steppers to see (only money!!).
As well known, holding torque is the usual marketing hype and not all steppers have a curve of the torque at rpm. Current X & Y stepper are Nema 23 298 oz-in, rotor inertia 2.6 oz-in and 5a. The Z is a Nema 34 467 oz-in 4a. The driver cards are SureStep 7.5a per phase, 24-75 vdc and they offer steppers matching the driver card. So, I am probably going to go with their steppers. They offer Nema 23 287 oz-in, Rotor Inertia 2.6 oz-in and 5.6a, Nema 34 803 oz-in, rotor interia 14.8 and 6.3a. My understanding (?) is higher voltage will increase rapid speed and higher amperage will increase torque. Directional it seems better, but not sure of the percentage improvement to expect. Plus changing the same size steppers is elementary. The existing power supply is 48-53v 5a Meanwell, so going to upgrade that also.
I did get recommended and looked at ClearPath servos that took stepper command. Not really being able to how much that upgrade would help me for the cost it is on a far back burner.
I was also told that replacing the Nema 23 for Nema 34 may be directionally wrong due to the inertia needing to be overcome in the Nema 34 could be negative for short lines of movement. (17mm ballscrews)
Steve,
Re: Controllers for 3D cutting
I really don't think the controller is the problem. It's a limitation of the CAM programs. In Fusion for instance, many times I am asking myself why it just made a particular move, I'm sure other CAM programs are similar. Your controller will do what it's told. You might be able to tweak your operating parameters a bit to get higher accelerations and set the rapids in the CAM setup to the maximum that your machine will take, but at that point it can start causing other problems like missed steps and overshoot. Just changing the acceleration on my lathe made several seconds of difference in the run time of a particular part. But on the lathe I am running hundreds of parts in a run so it's easy to track changes in the run time when optimizing the G code and machine parameters. I use the lathe as an example because most of the time I'm running one off parts on the mill so no chance for optimization.
I understand that there are G-gode optimizing programs, but they are uber expensive. I'm sure that some of the high end controllers have optimization functionality built in, but you are most likely not going to find this in the $1200 range, I would guess you would have to add a couple of zeros to get that type of functionality. As an example, Rolls-Royce jet engine division sends their entire machine out to another company for programming to optimize the tool path for a particular part, when they get the machine back that machine runs that part for the rest of its life.
The controllers that I use in my machines are around $2400, that is just the hardware, no software. And I'm sure that my systems are no more efficient than your WinCNC system. It's really all in the software, and I'm too lazy spend the 1000 hours or more that it would take write any optimization code to process the G code beyond simple trajectory planning.
Re: Controllers for 3D cutting
Quote:
Originally Posted by
Jim Dawson
I really don't think the controller is the problem. It's a limitation of the CAM programs. In Fusion for instance, many times I am asking myself why it just made a particular move, I'm sure other CAM programs are similar.
Jim,
I am amazed with Fusion CAM sometimes, I was milling a series of exactly the same circle pockets in a piece and it jumped around rather than going to the one just next to it.
Back in my work life I would go to shops that machined the dies for larger injection mold parts. The machines were larger than my backyard shed. I remember them having a "specialist" come to them from the machine OEM when all the design was complete to optimize everything. Cutting time was a week or more per half, cost me $300k or more. They were more concerned with time than me it seemed since they had a backlog for the CNC stations.
One big difference I believe is that most companies that make CNC machines for under $250,000 do not employ dedicated system engineers and only focus on generic CNC machines. They have engineers, but the synergy of the different functions from gcode to cutting is done from catalogs and not actual design specifications for end customer satisfaction. Engineering of the structure and mechanical movement being different. In my case and several others I know of, the machine OEM pushes complaints of performance (other than actual failures) at software, material or unrealistic customer expectation (even if they show pictures of such expectations on their PR gallery pages). In my case the tech support has no background in Fusion or any Autodesk product, so always "must be them".
Steve.
Re: Controllers for 3D cutting
Steve,
I think there is not a lot that can be done to correct the limitations of the various softwares involved, at least not at a price point that most of us can handle. We really can't complain too much about Fusion, free is a very good price. MasterCam might generate better tool paths, but at $20K per seat(+annual maintenance $$), it had better be good.
In your case it would be a big step up to go with servos rather than steppers. A set of 750 Watt servos would make your machine sing and are most likely overkill, but you can always turn the power down. 1000 IPM rapids would not be out of the question (subject to the RPM limitation of the ball screws) Then you would have the axis power to load your spindle to it's capacity, that would be the limiting factor. You could set your cutting speed at the maximum the particular tool bit would allow for the conditions.
I have been very happy with the DMM servos I installed on my lathe. The DMM servos will accept a step and direction input, and are actually a bit less money than the ClearPath units. Although there is some additional hardware needed so the final installed cost is about the same.
Re: Controllers for 3D cutting
Jim,
I think there is not a lot that can be done to correct the limitations of the various softwares involved, at least not at a price point that most of us can handle. [That is the crux of my question and situation. Is that my limiting factor and the rest only a 5% factor related to better 3D performance?]
We really can't complain too much about Fusion, free is a very good price. [Agreed, although I have been on the paid plan for couple years, not sure the paid gives me much for my use. The focus at Fusion (Autodesk) seems to be in other areas than speedy 3D. My best resource at this time.]
MasterCam might generate better tool paths, but at $20K per seat(+annual maintenance $$), it had better be good. [Between "might" and budget, that isn't going to happen. I may reach out to MasterCam to see if they will tool path one of my files just to see if the end performance is faster. Then I would look to farm out the Fusion CAD to have the CAM done. I'm sure someone would take on the side gig.]
In your case it would be a big step up to go with servos rather than steppers. A set of 750 Watt servos would make your machine sing and are most likely overkill, but you can always turn the power down. 1000 IPM rapids would not be out of the question (subject to the RPM limitation of the ball screws) [This is a change that I wish there was real world info on that pertained to 3D. For 2D work and larger (than mine) scale machines I understand the upside of servos, but at 400 sq-in normal work-pieces with thousand lines of short moves I am not confident.]
I have been very happy with the DMM servos I installed on my lathe. The DMM servos will accept a step and direction input, and are actually a bit less money than the ClearPath units. Although there is some additional hardware needed so the final installed cost is about the same. [I haven't looked at DMM, but clearpath does seem a quite basic upgrade to make.]
I'm going to look at the whole picture again and re-evaluate. Upgrading steppers is minimal cost and easy, will be a data point for mechanical potential. Outsourcing a file for MasterCam and maybe another CAM file would be another data point.
As of yesterday I have two fellow engineer retirees (none of us have the perfect background) that are working up a plan to instrument my machine with o-scopes and accelerometers for data collection to see what is going on. Don't anyone hold their breath.
Steve.
Re: Controllers for 3D cutting
@SteveS, for 3 axis toolpath I guess there are a good platea of CAD/CAM below your mentionated $20K.
I'm using BobCAD/CAM and produce very good toolpaths with license for 3 pro (including advanced rough/finiture aka trochoidal).
I've used for some time Fusion360, good in design part, optimal how post-processor engine (javascript and API SDK are beautiful), but yet a few poor in some machining controls and user hooks to control toolpath generation.
A CNC board can change a little program execution but don't make miracles, obviously.
However there are CNC boards that make true disasters in toolpath planning...
Re: Controllers for 3D cutting
Quote:
Originally Posted by
shineworld
@SteveS, for 3 axis toolpath I guess there are a good platea of CAD/CAM below your mentionated $20K. I'm using BobCAD/CAM and produce very good toolpaths with license for 3 pro (including advanced rough/finiture aka trochoidal). I've used for some time Fusion360, good in design part, optimal how post-processor engine (javascript and API SDK are beautiful), but yet a few poor in some machining controls and user hooks to control toolpath generation. A CNC board can change a little program execution but don't make miracles, obviously.
However there are CNC boards that make true disasters in toolpath planning...
I have had BobCad since version 17 in 2008, I'm currently at 31. (I still think 21 was the best.) Much of my current designs are assemblies and parametric which Fusion excels at, that in my mind those are not a BCC strength unless you add Solidworks. I'm also opposed to paying for any kind of support. A BCC strength to me and why I keep it is the basic design of 3D frames to export for use in assemblies, which I can do easier than in Fusion in some situations.
Steve.
Re: Controllers for 3D cutting
Quote:
Originally Posted by
SteveS
Jim,
As of yesterday I have two fellow engineer retirees (none of us have the perfect background) that are working up a plan to instrument my machine with o-scopes and accelerometers for data collection to see what is going on. Don't anyone hold their breath.
Steve.
Three retired engineers + a machine + test equipment sounds like a fun combination :)
I'll be waiting for the results.
Re: Controllers for 3D cutting
Quote:
Originally Posted by
Jim Dawson
Three retired engineers + a machine + test equipment sounds like a fun combination :)
I'll be waiting for the results.
Heh, what could go wrong... ;-) Reminds me of an old joke. Three engineers and an oscope walk into a bar...
Mark
Re: Controllers for 3D cutting
I have absolutely no experience with Fusion but I have used a few programs which allow you to specify a chordal deviation level that will result in an acceptable level of accuracy for the intended use of the part.The smaller the deviation-the greater the number of lines of code churned out by the program and the greater the fidelity to the original object.I took the point from the first post about the smallest toolpath file size and may be mistaken.What we haven't been told is the operating system being run on the computer that controls the machine and whether it has many other processes going on.
Re: Controllers for 3D cutting
Quote:
I have absolutely no experience with Fusion but I have used a few programs which allow you to specify a chordal deviation level that will result in an acceptable level of accuracy for the intended use of the part.
Fusion has tolerance and smoothing settings. I believe the first paragraph in this thread defines the problem:
Quote:
This is a quote from the F360 manufacturing forum seems to summarize my situation.
"Unless the 3d op toolpath is aligned with an X or Y axis, it’s going to be line segments. It’s a fundamental limitation of many controls and probably fusion."
Outside the 3 defined planes smoothing can't apply it's 'arc-line-arc-line' type path.
Others in this thread are familiar with Fusion but I suspect SteveS is the only person in this thread familiar with wincnc.
