[peteeng]

Hi All - I see that Volumill has "active chip thickness control" so its out there! Peter

[deadlykitten]

There is no reason the geometric chip load can't be used as a input
Volumill has "active chip thickness control"

the cam software can generate a toolpath for constant chip thickness, but using chip load control inside a cnc is a different thing

for a cnc, is easy&common to achieve constant load (cl), but much harder&rare to achieve constant chip load (ccl) :
... cl : reaction occurs after several tool rotations
... ccl : for an endmill z6, the reaction needs to occur 6 times within a single tool rotation, and each cut needs to generate it's own peek, and this works for ae << tool radius ; otherwise, for larger ae, then at least 1 chip is being cut, thus there is more than 1 flute making contact with the material ( please see below *1 )

in other words, for same scenario, ccl requires a much more developed cnc that will cut only at ae << tool radius

i hope it makes sense

but does such a thing exist ? yes, it does

example 1 : cl & ccl is the same thing for tools with z1, like turning operations
example 2 : cl & ccl is the same thing for toos with z>1, that have each cutting edge cutting simultaneosly, thus their chip load diagrams are identical or overlapped, and this is drilling, plunging, or turning simultaneously with 2 turrets, etc

see the analogy ? now let's focus on real examples, that puts ccl away from cl

example 3 : okuma dynamic load control, that can adjust feed multiple times within a single spindle revolution; it uses the basic cnc system capability, for cuts that generate individual load for each insert, and is designed to reduce costs for expensive inserts in exotic materials; in other words, you will no longer be worried about toolholder runnout

example 4 : mapping infeed threading patterns

(*1) there is math for such things, like having, for example, 2.3, or 1.79 teeth in cut, thus when performing calculations, is possible to consider a decimal number of teeth in respect to spindle phase

cam targets geometrical chip thickness, while the cnc targets load paterns and may not always solve for constant chip load

[mactec54]

That fanuc article is from 2012 and still considered valid methodology although the exact feature naming has changed. It is still a trajectory planning algorithm and has nothing to do with the drives.

---------

Features
Whereas the rate-of-change of velocity is referred to as Acceleration, the rate-of-change of acceleration is technically
known as Jerk. Jerk is therefore the third derivative of Position, with Acceleration the second and Velocity the first.
Wherever a part program causes a significant change in cutter-path direction, such as in a transition from a straight line
to curve, any axis acceleration that is required to maintain a constant cutter-path vector feedrate through that transition
may cause vibration or shock to be induced into the machine structure.
The Jerk Control function can be used to reduce shock and any resultant position error from sudden axis acceleration.
Jerk Control temporarily modifies the cutter-path vector feedrate by applying a smoothing effect progressively modifying
the rate-of-change of acceleration (Jerk).
Without using Jerk Control, Bell-Shaped Acceleration alone would apply a harsher, constant rate-of-change, acceleration
profile. The permissible rate-of-change of acceleration (Jerk) is set by CNC parameters for each machine axis; these
limits are usually set in advance by the machine-tool-builder. Any velocity adjustments applied by the Jerk Control
function are calculated as true positional commands and hence incur only minimal acceleration or deceleration delays
(Acc/Dec Before Interpolation).
This function is a basic function in FANUC Series 0i-MF Plus (Type 0, 1 and 3) and 0i-LF Plus.
Benefits
• Reduction of machining errors caused by shock and vibration induced into the machine structure that result of
sudden changes in axis acceleration

----------

Page 262 of their current 2020 feature catalogue (more specifically relating to jerk control COMBINED with G64 motions, but its all the same at the root. https://www.fanucamerica.com/docs/de...-catalogue.pdf

Likewise siemens who outline the same method. And Delta. And LNC. And SGZH. And Heidenhain. And DDCS. And Centroid. And Beckhoff. In fact, I have yet to find a single control that states they do anything differently besides the hobby controls which do nothing at all.... Including Eding.

Eding has no jerk control methodology, it makes no reference to anything other than constant acceleration in the manual. It is also a open loop pulse driven system with no communication with the drive - not even encoder feedback. "Make your servo settings sloppy and hope for the best" is not jerk control, it's NO control.



You need to do way better in your trolling, this is getting old. I only reply to make sure people don't take your information as some sort advice, but by now they all get it, so I'm done.

(There is a reason most of the forum has him on ignore,) I wonder what happened and why you are replying then, I guess the logarithm is broken (nuts)

You should not talk about controls that you have never used or installed and have No hands-on experience with, it just shows your ignorance in your posts


The Fanuc completely changed their whole control platform, and none of what you have been posting is current or reverent, your Google search is only as good as what you find without any knowledge or education in the subject, you can pretend to know, but it shows in your Google posts you have no clue on this subject, anyone can do a Google search and post what you have

You seem fixated with G64 making it out to be more than what it is, in the real world nobody uses it. unless they have a very old machine this helps the Hobby user some, to smooth out the bumps in the road.

Eding does have Look-ahead corner control and works very well, again your lack of of use and knowledge, shows your postings are incorrect.

[datac]

Hi All - I'm interested in how machine controllers, trajectory planners etc and CAM programs control jerk.

In an earlier post, I made a comment that I thought GRBL had applied S-Curve / Jerk components in their control, but I was wrong. It was TinyG.

Some detail I find on-line speak of 3rd order, yet others talk about 6th order... whatever that means. Because it appears to be a community development, perhaps there is a lot to learn within their community.

I knew somewhere a long time ago, there was a really decent video demonstrating the very same machine using various controls dialed in as best as possible, then the difference with TinyG's Jerk control. Yes, it was night and day with the S-Curve motion.

This was surprising to me given TinyG would be considered a hobby control board and just over $100. It is the only cheap control I've found talking about S-Curve motion at all.

Unfortunately, I think the interface(s), though there are many to choose from, to control boards like the TinyG are still horrible.

https://tenmilesquare.com/resources/...ion-explained/

https://github.com/synthetos/TinyG/wiki/What-Is-TinyG

[ihavenofish]

In an earlier post, I made a comment that I thought GRBL had applied S-Curve / Jerk components in their control, but I was wrong. It was TinyG.

Some detail I find on-line speak of 3rd order, yet others talk about 6th order... whatever that means. Because it appears to be a community development, perhaps there is a lot to learn within their community.

I knew somewhere a long time ago, there was a really decent video demonstrating the very same machine using various controls dialed in as best as possible, then the difference with TinyG's Jerk control. Yes, it was night and day with the S-Curve motion.

This was surprising to me given TinyG would be considered a hobby control board and just over $100. It is the only cheap control I've found talking about S-Curve motion at all.

Unfortunately, I think the interface(s), though there are many to choose from, to control boards like the TinyG are still horrible.

https://tenmilesquare.com/resources/...ion-explained/

https://github.com/synthetos/TinyG/wiki/What-Is-TinyG

Tiny G is (was) low performance due to the hardware, but the software itself is really interesting.

[datac]

Tiny G is (was) low performance due to the hardware, but the software itself is really interesting.

And one should expect exactly what for a whopping $129.00 ? Pretty sure the same guys who designed and programmed the trajectories for TinyG could have used hardware that costed $10,000 or more for no other reason than to make some guys less critical. Probably not in their business plan.

For the thousands (I'm guessing) who have the opportunity to make cool stuff on a machine for under $200, I'd bet they are not as miserable bunch as some may think.

Nonetheless, the ability to poke around inside TinyG's dev group might just appeal to the OP and his search for answers regarding "jerk", to which I sure thought he meant CNC control, not the psychological world of human behavior.

[ihavenofish]

And one should expect exactly what for a whopping $129.00 ? Pretty sure the same guys who designed and programmed the trajectories for TinyG could have used hardware that costed $10,000 or more for no other reason than to make some guys less critical. Probably not in their business plan.

For the thousands (I'm guessing) who have the opportunity to make cool stuff on a machine for under $200, I'd bet they are not as miserable bunch as some may think.

Wasn't meant to be a knock, just a description of why it's in the "hobby" level despite the software being quite evolved. (although in fairness the mesa 7i96 is $119 but needs a PC of course) Someone, somewhere (cant find the link) ported it to arm a few years back. Some people are good with hardware, others are good at software to control said hardware, and then there's that special little group who knows how to implement math. Getting them all in the same room and paid to work continuously is the trick it seems

Nonetheless, the ability to poke around inside TinyG's dev group might just appeal to the OP and his search for answers regarding "jerk", to which I sure thought he meant CNC control, not the psychological world of human behavior.

*cough*

[dazp1976]

your Google search is only as good as what you find without any knowledge or education in the subject

Can't stand google search sometimes.
It's always making old cr*p look up to date.

[peteeng]

Hi Datac - Thanks for mentioning TinyG. I've had a dig around and coupled with linuxcnc I think tinyg has the backbone for what I want to do, having no experience with Linux I'm on a learning curve there as well. It may even scale for the 6 axis robot I want to build. ta Peter

https://tenmilesquare.com/resources/...ion-explained/

TinyG even considers snap, crackle and pop that's the 6th derivative of position....

[deadlykitten]

You mention its hard for a machine to follow a spline but at the Gcode level the machine does not know if its a spline or an arc

hello pete i delay some answers, trying a simple way to put it

common cad geometry is "point line circle elipse bezier", while for g-code is only "line circle"

"point" is not for cnc interest, thus there is no use to ever implement a g-code for point

as for " elipse bezier ", most topics are on why is not possible to g-code beziers ( there are less request for elipses ), and in reality few cnc controls have g-code for such

an obvious reason, is that complex shapes can be segmented, thus replaced with a chain of G01s, and this works for most parts, and on elder machines there is no G02/03, so even circles have had to be segmented; in other words, G01 is king well, king until serious problems appear, and i will cover this soon, because is a bit more chalenging

a less obvious reason, is that in order to easy implement a g-code for a shape, that shape needs to be parameterized in such a way, that equal distant parameters will generate egual segmentation for details, please check attached image 01

as a consequence, a shape can be offseted only if above condition is true; if you can not offset it, then you can not use the classical radius compensation ( thus no G42 / 41 ); please check attached image 02 : offseting an ellipse/bezier won't produce another ellipse/bezier, thus the original element degenerates gets screwed

---

as conclusion:
... shapes that degenerate require aproximation methods, so they will be discretized different in different cam softwares, or even same software may generate different results, and also their toolpath planing and radius compensation will be particular to how the cnc brand implements it, and still final motion may be discontinued and machined surfaces may be rough roughness, unless higher resolution encoders are there; beziers are much more tricky
... shapes that do not degenerate are cam & cnc friendly

[peteeng]

Hi DK - Not sure where your going with this. We do not need a Gcode for an ellipse to make an ellipse. same with bezier or any arbitrary shape. I create free form shapes regularly and cut them regularly no need for lines arcs or ellipses. The cam happily creates code for any shape.

Ultimately all gcode is point to point, I think you say this. That's the only way the machine can work. The CAD/CAM system turns solid or surface geometry into a toolpath which is a contiguous "line" which consists of lines, arcs or splines. Ideally this toolpath is nurbs as then it can be smoothed and manipulated to be curvature continuous within the cam program. Once this near continuous toolpath is calculated it is tessellated into Gcode. Some MC systems can "smooth" this via numerical analysis or they convert the gcode back to a continuum via interpolation then smooth, then re tessellate. Can be in real time or as a post process operation. In all of these events you end up with point to point gcode of some sort.

I appreciate your comment on shapes that degenerate I have had many a discussion with young engineers that use ellipses in their part design. Because the CAD can do it does not mean it should be done. I suggest you look up grasshopper toolpathing or any of the better CAM systems that explain how they work to get a deeper understanding of the maths or at least the logic. They have gone way past lines and arcs.... Peter

[deadlykitten]

hy pette

Ultimately all gcode is point to point, I think you say this.
In all of these events you end up with point to point gcode

on actual machines, those points become so close, that it interpolates at suffice accuracy considered real motion so is no longer point2point, but high resolution interpolation, and there are 3 types : line / arch / bezier

your questions on how to improve cad/cam side to achieve smooth motion, for point to point movement, are relevant for less capable systems

Ideally this toolpath is nurbs as then it can be smoothed and manipulated to be curvature continuous within the cam program

if i may, ideally for what ? is not possible/nor needed to always generate nurbs, and most cnc machines can not deal with nurbs, and those that do have nurbs g-code may behave bumpy

convert the gcode back to a continuum via interpolation then smooth

why should it smooth a continuum ? come on

Can be in real time or as a post process operation

when is it real time, and when is it postprocess ?

Hi DK - Not sure where your going with this

along the way, i felt like you hanged on idea on how an actual mcs works, and take that/some if it, to build kind of a feedback loop in your machine; my dear pete, is hard to go hunting, if you put jerk in cam system, and chip thicnkess into cnc as load input, thus you mix worlds of cam with cnc this creates a gap between what you say and what others are repling to you, making the conversation hard to reach common ground

your relate too much to cad/cam, always go back to reasoning why cad/cam is important, while trying to figure out a cnc; there are too many miss asumptions in your replies, and this shows a rush to conclusions ... give it time, things will settle surely

there is a key element missing, that you are revolving around, and as i see it, is how the cnc quantify smoothness, or does it quantifies smoothness ? kindly

[mactec54]

Hi DK - Not sure where your going with this. We do not need a Gcode for an ellipse to make an ellipse. same with bezier or any arbitrary shape. I create free form shapes regularly and cut them regularly no need for lines arcs or ellipses. The cam happily creates code for any shape.

Ultimately all gcode is point to point, I think you say this. That's the only way the machine can work. The CAD/CAM system turns solid or surface geometry into a toolpath which is a contiguous "line" which consists of lines, arcs or splines. Ideally this toolpath is nurbs as then it can be smoothed and manipulated to be curvature continuous within the cam program. Once this near continuous toolpath is calculated it is tessellated into Gcode. Some MC systems can "smooth" this via numerical analysis or they convert the gcode back to a continuum via interpolation then smooth, then re tessellate. Can be in real time or as a post process operation. In all of these events you end up with point to point gcode of some sort.

I appreciate your comment on shapes that degenerate I have had many a discussion with young engineers that use ellipses in their part design. Because the CAD can do it does not mean it should be done. I suggest you look up grasshopper toolpathing or any of the better CAM systems that explain how they work to get a deeper understanding of the maths or at least the logic. They have gone way past lines and arcs.... Peter

This may throw a wrench in the work, this STEP-NC has been around a while most won't even know about it, worth a look though, it does not use G-Code to cut a part.

Step-NC, STEP-NC Machine ® enables machining directly from a 3D model instead of the antiquated G-codes. https://b2b.partcommunity.com/commun...l/4863/STEP-NC

[mactec54]

There is a key element missing, that you are revolving around, and as i see it, is how the cnc quantify smoothness, or does it quantify smoothness? kindly

What is a big part of the machine smoothness journey, is the Servo Motor and Encoder Resolution.

[deadlykitten]

hy mactec there is okuma application for step-nc ( 3d ), and also it's conversational can auto generate operations for 2d contours; it requires some time consuming definitions ...

replacing the traditional way is a big thing, and there is at least the story of Gulfstream that no longer uses proprietary formats

also solidworks mbd is an aproach that considers to replace classical technical drawing with a 3d model and keep only critical dimensions and infos ( for example, how many cnc machines can deliver accuracy beyond what is needed for common parts, then all dimensions with normal tolerances can be eliminated )

few years ago i talked with someone that receives designs/3d models from abroad, and after that colaborates with many shops in the area in order to get his part manufactured, and then he assambles those, and he has to overcome many problems because of format compatibility, and non-sense or missing tolerances ... we discused a bit about standardized aproaches, and how some features of his models can not be machined ( because they are designed by persons with less experience in manufacturing )

in other words, if all is under your roof ( design -> product ), and you use 3d models, then there is a chance to boost your workflow

if it maters, i try to boost common things that do not require 3d models

What is a big part of the machine smoothness journey, is the Servo Motor and Encoder Resolution.

the big question mark for me, is how petteng said that he designs, builds and sells cnc machines, while having such doubts on jerk and other things ? i don't wanna sound rude, but i think he used simple systems for a long time, and now he is having a big bang

if you put a nurbs g-code into a hobby machine is a thing, in a normal one is another, if you activate adaptive feedrate is another ... latest days i started thinking he may need brief on diff/folowing and allowable error, inertia/mass evaluation, stuff related to the feedback loop ? i onestly dont know how to put it anymore, he is too much hanged in the cad-cam area / kindly

[mactec54]

hy mactec there is okuma application for step-nc ( 3d ), and also it's conversational can auto generate operations for 2d contours; it requires some time consuming definitions ...

replacing the traditional way is a big thing, and there is at least the story of Gulfstream that no longer uses proprietary formats

also solidworks mbd is an aproach that considers to replace classical technical drawing with a 3d model and keep only critical dimensions and infos ( for example, how many cnc machines can deliver accuracy beyond what is needed for common parts, then all dimensions with normal tolerances can be eliminated )

few years ago i talked with someone that receives designs/3d models from abroad, and after that colaborates with many shops in the area in order to get his part manufactured, and then he assambles those, and he has to overcome many problems because of format compatibility, and non-sense or missing tolerances ... we discused a bit about standardized aproaches, and how some features of his models can not be machined ( because they are designed by persons with less experience in manufacturing )

in other words, if all is under your roof ( design -> product ), and you use 3d models, then there is a chance to boost your workflow

if it maters, i try to boost common things that do not require 3d models



the big question mark for me, is how petteng said that he designs, builds and sells cnc machines, while having such doubts on jerk and other things ? i don't wanna sound rude, but i think he used simple systems for a long time, and now he is having a big bang

if you put a nurbs g-code into a hobby machine is a thing, in a normal one is another, if you activate adaptive feedrate is another ... latest days i started thinking he may need brief on diff/folowing and allowable error, inertia/mass evaluation, stuff related to the feedback loop ? i onestly dont know how to put it anymore, he is too much hanged in the cad-cam area / kindly

Yes, Okuma where one of the first to use and have the STEP-NC, available for their machines

There are a lot of Hobby users now that are using quality AC Servos with High Resolution Encoders, not much can change for the Stepper users though, to help the Jerk factor, this is where using a G64 can help to smooth things out a little, this is just a small band-aid for a much larger problem

[deadlykitten]

attached image shows a trapezoidal and an S curve ( that should perform theoretically smoother)

in other words, positioning in rapids on a performant cnc that follow trapez, will behave smoother than feeding nurbs on less performant systems / kindly

[peteeng]

Hi DK - Look at your raw gcode. there are no lines, arcs or splines. Just points. These are "mechanically" interpolated by the machine. ie there is no "line" between each point that the machine follows. To achieve very smooth motion you need smooth toolpaths and you need mechanical feedback. You can have poor toolpaths with a mechanical feedback, post processing and improve the motion. Or you can do all the "smoothing" in CAM. Its all a very big puzzle that everyone solves differently. We work in continuums because then you can do std mathematical functions to smooth paths, optimise paths etc. But at the end of that it has to be tessellated into points. Or You can work numerically (eg use quadrature maths) to work with the point data. All of these things are done in the most economical way. I don't mean $$$ economy I mean computational economy. DK I don't doubt anything. I have areas of interest that I investigate and deepen my knowledge, I also try to share some of that knowledge here. Thanks to you and many others that journey is interesting.

I'll mish mash some answers here:

Real time computation means its done while the CNC is working, Post processing is sorting something out before you run the machine. More points does not necessarily mean its "smoother". In fact most of the time more points make it bumpier. If you look at my logic diagram you wills see the advanced system has feedback loops. These are not in CAD-CAM but on the hardware side. Happy to discuss feedback loops you seem to be stalled on the trap vs S-curve stuff and I think your last comment got it back to front. S-curve will be smoother then trap not the other way round.

I used to off line program and "teach" robots toolpaths. Robots are quite a bit more complex then simple stepper driven 3 axes cnc's. Happy to discuss any of the things you mention you just need to discuss it.

Mactec- Ta for the step-nc looks interesting. Some 20 years ago Alibre tried to open up step to include the parametric data but the industry shut it down as the proprietary CAD companies didn't want users to be able to change CAD systems easily. But one day it may happen... Peter

[peteeng]

Morning All - To summarise where I'm at.
1) Discussion has sorted a few things in my head, thanks to all contributors
2) My next router Frankenrouter will use TinyG to give that a go. Plus I'll start looking at LinuxCNC maybe both all at once, learning curve overload
3) I'm now looking at encoders to see if I can loop them from the steppers to the MC. I use UCCNC not sure if it can do that yet... A side project for the intern... Peter

[deadlykitten]

hello again please, how a cam achieves smoothnes and how a cnc does it ?

We work in continuums because then you can do std mathematical functions to smooth paths, optimise paths

for most machines, nurbs is out of the question, toolpaths are line/arch, and most parts are simple

so i ask : is it possible for a normal non-nurbs machine to behave as smooth, or even smoother than one that runs nurbs ? how is smoothness quantified ?

More points does not necessarily mean its "smoother". In fact most of the time more points make it bumpier.

how can i find the right amount of points ?
is it the same for feed override 50% and override 350% ?
is it possible not to care about the amount of points, and always go for max settings ?

If you look at the logic diagram you wills see the advanced system has feedback loops. These are not in CAD-CAM but on the hardware side.

i have looked over it right when you posted it : please check attached response

S-curve will be smoother then trap not the other way round.

nonono, it will be smoother only if it is being programed so

considering acceleration/decc ramps, trap/s/combination(thus something in between trap and s may be used)/etc : for max performance, you need to reach an area beyond you can not go futher, then back off a bit; how do you reach it ?

usual factors : rigidity, inertia, cuting forces, material mass : how are those taken into consideration ?

Real time computation means its done while the CNC is working, Post processing is sorting something out before you run the machine.

feedback loop requires post process; no feedback loop means real time

how much nurbs(or whatever else) is it possible to push inside the machine ? when does a machine capable of post process, no longer be capable to postprocess ? how is this balanced ? kindly