Advanced Search

I’d like to replace my weak HP EliteDesk G2 on which I currently run LinuxCNC. In my setup, the PC drives a Mesa 7i95t over ethernet, and displays a LinuxCNC UI on a monitor. Form factor is important since the PC is only used for LinuxCNC.I’m looking at various options for ~$800 range mini PCs, but am struggling to understand how they will perform in a realtime environment. Some of the options are:

• Beelink EQR7, cpu: AMD Ryzen5 7535U
• Beelink SER9 PRO+, cpu: AMD Ryzen7 H 255
• GEEKOM GT13 Pro, cpu: Intel Core i9-13900HK
• ASUS NUC 14 Pro, cpu: Intel Ultra 7 155H
• MINISFORUM DeskMini UM870, cpu: AMD Ryzen 7 8745H

Could someone please help shed some light on how to evaluate whether these models will perform well running the LinuxCNC realtime environment?

Thanks!

38? -3 here.
What made it work?

Don't know why they don't allow switching between csp and csv.
You switch to homing mode and back, so you do switch modes on the fly.
Had a guy in here that switched modes, although he switched between two cia402 components,
one was Dominik's for CSP and one was mine for PV mode (mostly a quick hack github.com/MetalMusings/cia402pv)

Searched and found many drives that allow dynamic switching while in op-enabled.
That's not a proof it is universally allowed, at least it isn't forbidden.

The current component is quite limited in that respect. It could support CSP, CSV, PV, PP, Homing, maybe also Torque Profile mode.
The main difference is which pins are active as inputs and outputs I guess. Plus the feedback position/velocity/ferror

 

what linuxcnc version do you have ?

in older versions you need to use: motion.spindle-speed-out

Looks like you've lost your network. what does ip a say?
Look to make sure you have an IP address on your network side 
can you ping your router?

Yes I did get your point but glossed over it. The joint becomes enabled and that is passed to cia402 to enable the Ethercat drive. the state machine turns it on safely and  manages errors/faults
The drives measure in pulses per rev and we use pulses per mm so it converts between scales too. (I think that's right, been a long time)
Looking at the current comp github.com/dbraun1981/hal-cia402/blob/main/cia402.comp#L142 
The velocity/position mode is only set at startup (in write-all), Is there a reason for that? Could we change modes when in an idle state? (which I will have a state for) Does not seem it would matter much to the drive

I have not started yet. I wanted to publish a couple of videos first. Got my Ethercat instalation one done this morning.

I've tried to update the system, but I get this:

As I can understand, there's something wrong (it seems almost...impossible to not find anything. Windows maybe, but Linux.....): too old version (2.7.14)???.
Anyway, I'm a little bit demoralized...
2 years ago I made the last job (a double sided PCB) and everything was perfect, then shutdown the system.
I can't understand what happened.
Before keeping going on with making you waste your time (I don't want it), as per your experience my problems can be related to:
1) software (linuxcnc) failure?
2) Mesa board failure?
3) MB failure?

In case 1), if I install the last linuxcnc release, will it be any chance to retrieve my current setup and customization (hoping it can fix the current problem)? 4 years ago I've spent almost 4-5 months to set linuxcnc (and with your needfull support, by myself I can't do anything....) and customize my axis with the function I normally use, so lose everything is something I don't want to....

Anyway, please tell me your opinion, I'll follow your suggestions.

The same thing happens when people switch between csp and pv modes (rotary axis to spindle motor and back),
they have problem with the rotary position feedback. Could there be a general way? Save the position, and later
create an offset to add to the position feedback. Sounds like that could work.

The point I wanted to make about cia402 state machine was that the device decides the state,
the cia402 component can only react and ask for a change of state. 
 

The cia402 component really only uses two states, disabled and op enabled, controlled by the enable pin.
Either transition to and stay in disabled mode, or transition to and stay in op enabled mode.
When in op enabled state, you can switch between the different operation modes. Anyway.

Is there no message in the syslog when you have a mismatch between pdos? I would have expected something there.
 

The sp_scurve code is based on S-curve (7-segment) trajectory planning mathematics, which uses jerk-limited motion profiles derived from the kinematics equations of motion.

Mathematical Foundation
The core math is based on third-order polynomial motion equations with controlled jerk (rate of change of acceleration):


Position: P(t) = P₀ + V₀·t + ½·A₀·t² + ⅙·J·t³
Velocity: V(t) = V₀ + A₀·t + ½·J·t²
Acceleration: A(t) = A₀ + J·t
These equations appear explicitly throughout the code (e.g., lines sp_scurve.c:187-197, 842-844, 1089-1091).

Key Mathematical Components
7-Segment S-Curve Profile - The trajectory is divided into 7 phases (S0-S6/n0-n6):

S0: Increasing jerk (acceleration ramps up)
S1: Constant acceleration
S2: Decreasing jerk (acceleration ramps down to 0)
S3: Constant velocity (cruise)
S4: Decreasing jerk (deceleration begins)
S5: Constant deceleration
S6: Increasing jerk (deceleration ramps to 0)
Cubic Equation Solving - The solve_cubic() function at lines 80-184 solves cubic equations using Cardano's formula and trigonometric method for finding roots.

There is a Newton-Raphson Method - The solute() function at lines 54-68 uses Newton-Raphson iteration to solve polynomial equations, but it is not in use, Cardano's formula is computationally superior.

Stopping Distance Calculation - Based on the relationship Amax² = V·J + 0.5·a² (line 949), derived from integrating the jerk-limited motion equations.

Attribution
The code credits go to 杨阳 (Yang Yang) as the original author, he also referenced a Chinese technical document at doc88.com/p-9119146814737.html for the S-curve formulation.

Thnk you , for fast respond
Today I try again with your advice I and D change to 0 and clear all paremeter in calibration box
again its not help
now
X axis
P 17 , I = 0 , D = 0
deadband same result can't touch
I try re config FF0 FF1 FF2 but not any change
now FF0 = 0 FF1 = 0.04 FF2 = 0

but there is still a slight vibration/noise at standstill that I cannot get rid of yet.

FYI : DELTA ASDA-A3 750w

Thank you again for your help.
Asis,

Good job my friend!
Can you provide any paper or info of the math background you used for it?

I bought a similar grease gun. You can get cheaper ones off ebay if you search for Jiabara threaded style for 70-80 gram tubes. 
The tubes are a clever accordion that screws onto the grease gun.
I did not like the nozzle/rigid stem. It locks on really tight. So maybe the real deal works better.
If you have old LM blocks with brittle plastic, you risk the chance of pulling the grease fitting out of the plastic threads. 
Be aware that some manufacturers use grease fittings with an odd pitch of M6x0.75! In case you need to modify something.

I have a machine that recommends 2 different greases. NSK Grs AS2 for the LM blocks and LR3 for the ball screws. AS2 is probably fine for both. 

I'm thinking of adding a centralized lubrication system to my current project instead of trying to access each block under the way covers. 
I have used Bijur pump and metering units adapted to ball screws and LM guides in the past using vactra no2.
I'm interested in finding out what others are doing that is more of a modern system. 
I know Haas went through a couple of redesigns with their remote lube/grease systems.
 

Did you try lowering the acceleration?
The default of 750mm/s/s is to much for some machines.

github.com/grandixximo/linuxcnc/tree/9d

This is phase 0 completed, according to my plan, this is extremely experimental work in progress, there are known issues at the moment, no arcs no blends, no scurve, as illustrated by the plan you can find at this post, which I got not much feedback..

forum.linuxcnc.org/38-general-linuxcnc-q...ons?start=640#341856

Next phase it is IMO too hard to be vibe coded, but I will try in my spare time.

The issue with the original code was it was developed with a machine that homes to index so it does not work cleanly with home to sensor. The real problem  is that during device homing, Linuxcnc throws following errors because it has not commanded movement. What generally needs to happen is the position feedback is synced with commanded position so Linuxcnc is happy. This is why most working solutions also provide their own cia402.comp

Yes, I think homing can be standardized for cia402 devices and the new home comp is there now to support that. So much of homing.c becomes irrelavent with device homing so the homecomp allows control of what is needed. I've had a couple of goes at it. The first was obsoleted by a rewrite of the homecomp paradigm.

The cia402 state machine is actually very simple. very linear with the exception of error and safety steps to prior states. The flowchart here is a good resource
www.kebamerica.com/blog/comprehensive-gu...a-402-drive-profile/

Anyway, back on task. It occurred to me that I had included some standard PDO's that are not reported by Ethercat pdos. I will revisit that. 

 

Thanks for your reply @andypugh

I copied and pasted your suggestion into my HAL file, replacing the original erroneous text.

Unfortunately when launching Linuxcnc, an error is shown up as attached in the .png file.

Revised HAL file also attached.

Regards

Phil