[ihavenofish]

Hi TTT - Have a look at the Buildbotics controller. It has S curve control. S curve control controls velocity profiles which control accelerations, not jerk. I'm not sure it controls servos ask them. They have been prompt to reply to my questions. I have not yet used one of their controllers but want to in a future machine.

Hi UHNF - thanks for the conversation it has prompted me to research and clarify my thoughts via a restless night. Where I'm coming from is that jerk is mainly created by the path so if you have G code that consists of straights and arcs you have many inflection points and huge jerk at every intersection. If you have g code create from 3rd order splines they are smooth so jerk is minimised. That would be how high speed roughing paths are done these days. This sidesteps the issue of a controller having to introduce transition curves into the code on the fly. S curve or parabolic curve control is about the velocity profile not the path. As the controller inherits the path from CAM (gcode) it can't change it. All it can do is manage the velocity. Since acceleration around curves is (vel squared / radius) by slowing slightly you reduce accelerations greatly. Inertial loads created by "rough" accelerations is what shakes machines. That's what the trajectory planner does. It looks at the max accel setting calculates the radius reduces the V to achieve the max accel setting and goes thru the curve (I think that's what happens) at no time is jerk calculated or measured in this approach. And since machining is good at a constant velocity in point to point rapids the accel is capped via its setting.

The overlay to all of this is that the toolpath is calculated as a "smooth" curve (lines arcs or splines) then at some point its broken into gcode which is a polyline ie a series of point to point moves all of which have inflections, then the motion /trajectory planner plots a path and the mechanical/electronical system smears this back to a pseudo continuous path. Its fraught with various issues all along the way.

So if you want very smooth toolpaths you have to look at the algorithms creating the Gcode as this is where the rubber meets the road in terms of jerk I feel. If you have "smooth" toolpaths then s curve velocity control will work very well. Regards Peter

https://www.jpe-innovations.com/prec...otion-profile/

Ya lost me. This is basic 7th grade math. One last time.

S curve acceleration is (IS) jerk.

position = M = static position
velocity = M/s = linear position = infinite acceleration
acceleration = M/s/s = linear velocity = S curve position = infinite jerk
jerk = M/s/s/s = linear acceleration = S curve velocity


It's literally called jerk. Thus, jerk "control" is controlling the rate at which the machine accelerates in the same way acceleration controls there rate velocity changes, and velocity controls the rate the machine moves. S curve acceleration means 3rd order velocity profiles, and 3rd order motion means jerk. S curve acceleration = jerk control. That is the literal definition.

It is strictly a trajectory planner dynamic and has absolutely nothing to do the motors or encoder counts in the same way acieration has nothing to do with encoder counts. You don't want the drive controlling jerk any more than you want the drive controlling acceleration. Obviously more encoder counts and more PID bandwidth in the drive and the control give more ability to track the given path, but this does not in any way make up for jerk control, and in fact could just make it worse. Blind on drive filtering is not jerk control. Fully controlled systems could implement jerk control in coordination with the drive and have the control interact with it to maintain the path, but the systems we are talking about (except siemens) are not fully controlled in that manner, many of them do not even read the encoder for feedback.

Control makers at the top end I'm sue have all manner of creative solutions to implement more advance jerk control that could interact with their own proprietary drives, but they still don't change the definition of jerk. (like how mactec seems to believe jerk means some sort of mechanical physical "jerky" motion, which it is not, even though the manifestation could a physical jerky motion).

I'm dropping this topic here as it has zero to do with this thread anymore other than it being desirable for tom faster than they typical hobbyist's machine.

[ihavenofish]

Hi Mactec - The way I read this is that the controller is reading the G-code and interpolating the code with a spline then re coding that spline with a finer, smoother code. That's OK but this should have happened in the original code. Any reworking of code means the resulting code could be off the intent of the original geometry. By the way the G in G code stands for geometry Code I think.... Using velocity profile control (S,P,T or other) is an implied jerk control not direct jerk control... It does control acceleration. Peter

Nano AI is not jerk control, it is advanced continuous path smoothing like G64 on most machines. It just in theory does it better. Siemens has a similar thing new brother machines have several smoothing modes, allowing to to pick speed vs precision. LinuxCNC's G64 smothing is rather good actually, it just lacks look ahead so it gets bogged down it tiny line segments at high speed. G64 as I've said before negates a lot of the need for jerk control in continuous motion (except the start and end). As you say, it is kinda an indirect jerk control as it effectively does control the acceleration profile via arc motion. There might be some instances of jerk left in the path, but they are nowhere near as violent.

[peteeng]

Hi IHNF - There is no S (sin) curve acceleration control. If it is stated like that it is incorrect. Sine, Parabola or Trapezoid are velocity profiles not acceleration profiles. These velocity profiles control accelerations by limiting the velocity around corners or by limiting absolute acceleration in point to point moves directly to the set value. These control jerk indirectly. Controlling jerk directly from Gcode is not simple. Cheers Peter

[ihavenofish]

Hi IHNF - There is no S (sin) curve acceleration control. If it is stated like that it is incorrect. Sine, Parabola or Trapezoid are velocity profiles not acceleration profiles. These velocity profiles control accelerations by limiting the velocity around corners or by limiting absolute acceleration in point to point moves directly to the set value. These control jerk indirectly. Controlling jerk directly from Gcode is not simple. Cheers Peter

The profile applies to whatever you are graphing.

I feel like I'm in a political discussion where one side has decided to invent new meanings for all common words.

[peteeng]

OK lets call it a day - Peter

[mactec54]

OK lets call it a day - Peter

PM me your Email and I will send you what one of my controls have and how Jerk control can work in a control, it's never perfect because you have a constant acceleration / deceleration you are on the right path though. you should not bother with ihavenofish, he has lots of posts from Google, and has no hands-on real-world experience what's so ever.

I'm not sure why anyone in the hobby business, have any thoughts of Jerk control as 70% of all industry cnc controls, don't have any Jerk control, go to a industrial machinery show and watch the machine demo's, you will see the machines that have Jerk control and the one's that do not, it's very easy to see

Most Hobby users are using steppers every movement is a Jerk and no amount of Jerk control will ever change that, so the best a Hobby builder can do is to use a quality servo motor with as high PPR Encoder they can find or afford, this just by itself is enough for most even in the industry, most machine manufactures don't use anything less than 20Bit Encoders today, only a few years ago they only were using a max of 12Bit Encoders,

How do I know Encoders Help /work I tested this, starting from 8Bit up to 20Bit Encoders, all using the same control and the same load, this was done on a CNC Mill using only the Best Yaskawa Servo motors the Jerk did not start to get better until 14Bit Encoder from this point is where you start to see a difference, 16Bit was smooth motion and would be good enough for most, 17Bit 18Bit and 20Bit being the smoothest machine movement at the time, I have since used 32Bit with the most basic control and had very smooth motion at low and high speed, so anyone that says high Encoder resolution does not make any different, have never tried it to know, that using the most basic control you can get smooth motion

Here is a snip from one of my controls how they control Jerk Motion

The Basic Motion class uses a trapezoidal profile in which a constant acceleration and a constant deceleration are
employed. This class is useful for users who do not require advanced motion profiles and instead need a simple
motion API.

The Jerk Motion class allows the user to choose one of three motion profiles: trapezoidal, S-curve, and jerk-limited.
The trapezoidal profile is identical to the trapezoidal profile of Basic Motion. Compared to the S-curve and jerk limited profiles, it has the disadvantage of having infinite acceleration jerks and deceleration jerks.

The S-curve profile employs a constant acceleration jerk and a constant deceleration jerk. This means that the acceleration and deceleration profiles will be wedge-shaped/triangular. The velocity profile will follow an "S" shape.

[mactec54]

Ya lost me. This is basic 7th grade math. One last time.

S curve acceleration is (IS) jerk.

position = M = static position
velocity = M/s = linear position = infinite acceleration
acceleration = M/s/s = linear velocity = S curve position = infinite jerk
jerk = M/s/s/s = linear acceleration = S curve velocity


It's literally called jerk. Thus, jerk "control" is controlling the rate at which the machine accelerates in the same way acceleration controls there rate velocity changes, and velocity controls the rate the machine moves. S curve acceleration means 3rd order velocity profiles, and 3rd order motion means jerk. S curve acceleration = jerk control. That is the literal definition.

It is strictly a trajectory planner dynamic and has absolutely nothing to do the motors or encoder counts in the same way acieration has nothing to do with encoder counts. You don't want the drive controlling jerk any more than you want the drive controlling acceleration. Obviously more encoder counts and more PID bandwidth in the drive and the control give more ability to track the given path, but this does not in any way make up for jerk control, and in fact could just make it worse. Blind on drive filtering is not jerk control. Fully controlled systems could implement jerk control in coordination with the drive and have the control interact with it to maintain the path, but the systems we are talking about (except siemens) are not fully controlled in that manner, many of them do not even read the encoder for feedback.

Control makers at the top end I'm sue have all manner of creative solutions to implement more advance jerk control that could interact with their own proprietary drives, but they still don't change the definition of jerk. (like how mactec seems to believe jerk means some sort of mechanical physical "jerky" motion, which it is not, even though the manifestation could a physical jerky motion).

I'm dropping this topic here as it has zero to do with this thread anymore other than it being desirable for tom faster than they typical hobbyist's machine.

You should stop copying your content from Google, you have no real-world Control Building experience what's so ever

Acceleration / Deceleration is what the Jerk Factor is all about and how the control can handle it, most controls can't do much for Jerk control, so you look to what can give you smoother motion, and this is Servo Motor with high Encoder Resolution= Smooth Motion

[peteeng]

Hi All - I have started a thread "Jerk control for cnc jerks" here. My intent is not to argue about technicalities but to get an understanding from members and their systems of how current CAM and control systems deal with 1) toolpath creation 2) motion control (including jerk) 3) results obtained 4) issues uncovered or to be solved. I have a maths intern looking for a project so maybe something lies in this area for him and its an area of interest for me Peter

https://www.cnczone.com/forums/mecha...30780-cam.html

[TeeTomTerrific]

I think I've figured out the transmission chains placement and mounting now. That leaves me with placement and mounting of the proximity sensors yet to do. All my extrusions and custom parts were ordered on Monday and are now in production. I'm shipping by sea to save a good bit of shipping cost. I found out they will land in the port of Houston, TX so I'm going to drive to pick them up when they arrive in a couple months. Shipping and port fees really drove my cost up, but the entire cost for everything is still probably saving me between 40% to 50% from what this would cost me to purchase everything in the US. I only have a few minor miscellaneous parts to get domestically, (not including the controller board and other electronics). I'll be able to focus on the electronics while waiting everything else to arrive so expect more questions regarding that subject soon.

[ihavenofish]

nice. are you having custom extrusions done then?

George (xzero) had a couple done here in canada. (my old client made the dies). The main "issue" was precision (twist I guess being the biggest offender) because george was not machining the extrusions flat where the rails mount. Wound up costing a fortune and they went through many iterations before approval. In they end they they were awesome, but it was a lot more involved than simply sending a drawing and getting extrusions back a few weeks later.

How are you dealing with this?.

[TeeTomTerrific]

No everything is stock extrusions from a company called Langle. I found 3D drawings of everything they make at https://b2b.partcommunity.com/community/ and downloaded a 30150 profile that was exactly what I was looking for for the table top. They got sent my email address that I had downloaded an extrusion 3D drawing. They in tur contacted me asking me if I had a project I needed quoted and it sort of took off from there a few weeks later when I had a good list of what I needed. Price wise they approximately 1/3 of what Misumi charges. Misumi has been my go to extrusion supplier. If I was in a hurry I would buy from them but I'm not so I'm willing to wait for a boat. The custom parts are just CNC machined from metric 6061 plate. The U-channel will be bent from 3mm aluminum sheet stock which they suggested because they didn't have any U-channel extrusions. It has a pretty generous 5mm outside radius in the bends. Wherever I could I picked the thickest walled most rigid extrusions. This table is going to be a beast, probably close to 400 Kg, or 800 lbs at least from the 500 Kg shipping weight they calculated. I also arranged for the company I' getting my ball screws, linear rails, motor, etc. to coordinate with Langle and all that is coming in the same crate as the extrusions saving about $800 in shipping cost. Those weigh 81 Kg.

However, If I wanted to do custom extrusions, I bet Langle would do it and at a reasonable price. That's always available for a possible future project.

Edit: If you want it I can send you a link to download their catalog.

[ihavenofish]

Cool. I guess my question wasn't clear though. Are you machining the mounting faces for the linear rails?

[TeeTomTerrific]

Cool. I guess my question wasn't clear though. Are you machining the mounting faces for the linear rails?

I'm planning to using 30x30 extrusions that are solid 3 sides. I'll have to drill them to mount the rails. If there is any issue with warpage in the extrusions then I can fall back to getting bar stock machined. Hopefully that will not be necessary.

[ihavenofish]

Interesting. they wont be remotely flat (in machine tool spec terms), and they wont hold a thread well at that thickness especially in 6063. 6063 is cosmetic grade, takes anodising well and extrudes easily, but it not very strong. Fine as a machine frame as ultimate strength isn't important, but you need to use the T slot with steel t nuts to hold screws.

You also really want a registration shoulder or assembly will be very challenging.

[TeeTomTerrific]

Interesting. they wont be remotely flat (in machine tool spec terms), and they wont hold a thread well at that thickness especially in 6063. 6063 is cosmetic grade, takes anodising well and extrudes easily, but it not very strong. Fine as a machine frame as ultimate strength isn't important, but you need to use the T slot with steel t nuts to hold screws.

You also really want a registration shoulder or assembly will be very challenging.

They get through holes and will not be threaded. The bolts connect to t-nuts in the channel in the extrusion below the 3030. I also have two copies of an alignment tool being machined to make sure everything lines up as close to perfect as possible.

[ihavenofish]

Ooooh, ok, i misunderstood. Just an issue of flatness then which toy wont really know til you get it.

[TeeTomTerrific]

They get through holes and will not be threaded. The bolts connect to t-nuts in the channel in the extrusion below the 3030. I also have two copies of an alignment tool being machined to make sure everything lines up as close to perfect as possible.

Actually, one of those was an old drawing. The newer ones have a 10mm plate bolted to the extrusion to add rigidity and flatness. The Z axis will be machined though. I got some pricing to do that on X and Y, but it would have cost an extra 1K or more and I just couldn't justify it the way I could for the Z axis.

[ihavenofish]

Yeah machining from solid is not cheap right now. I did mine before the price went way up and even then it was very expensive.

Attachment 478698

[TeeTomTerrific]

Yeah machining from solid is not cheap right now. I did mine before the price went way up and even then it was very expensive.

NICE!!!

[TeeTomTerrific]

So I'm considering going with the UC300ETH-Max since it is only $40 more than the UC300ETH-UB1. I don't know if I would ever use all the extra inputs and outputs, but it seems like a good value. Along with that I think that I should probably also get the UCR201 USB wireless jog pendant.
The UB1 manual recommends a 24Volt 3Amp power supply and looking around for a DIN mounted power supply the Mean Well SDR-75-24 looks like a good fit. It's a 75 Watt, 3.2 Amp 24Volt power supply and I can feed it 220 from the 15Amp circuit I have available in the shop along with the servo drivers. However if there is a reason to go with a larger power supply, the same series is also available from 120 W all the way up to 960 W. I just don't have the experience to judge if there would be a need for a larger power supply.

As far as the 220 power I would appreciate some guidance on whether each drive should have a 2 Amp MCB, one 10 MCB for all 220 input or some other arrangement. Should inputs from sensors and such also have some sort of circuit protection like an MCB or fuses?

I realize putting all this together in a case will be the last step but I'm trying to identify all that will go in it including things like a DIN mount for the UB1-UD1 DIN Mount 220 terminal block, MCBs, Cooling Fans, etc. I'd like some suggestions for brands, sources and so on as well as anything else I should consider to make this a well built CNC controller system.

One other question. Does the Z axis normally have only one proximity sensor for homing and a touch plate to set the limit or is a limit switch in addition to the touch plate advisable? Your thoughts on a touch plate would also be appreciated. What do you think about the Avid Auto Z and Corner Finding Touch Plate? It's pricy! Is there something else I should consider instead?

I'm searching for all the things I will need to make this machine go, and go fast, and the folks following this thread have been very helpful so far so fire away with your comments so I can get this all figured out.

Thanks,

Tom