Advanced Search

 [justify]Hello, I am looking for some advice here while still learning about Linuxcnc.I have robot arm with 6 custom drives and I would like to integrate it more with Linuxcnc and use kinematics module. Currently it is driven in joint space with mesa 6i25 via step/dir interface. From what I saw there are multiple ways to get joint positions into Linuxcnc. Have to tell here that encoders are in some way absolute as they use batteries and on power on send current position over uart and from that point and further on just behave as incremental and send A and B pulses (these pulses are added in software to first read out position). Position read is handled in drives. I am thinking about best way to transfer 6 encoder readings to Linuxcnc. Have to check with gear ratio but data from encoders would be at least 24 bits(with multiple turns) and lets fix it to be 4 bytes per encoder. Another thing is that this communication should be isolated from drives as step/dir are also isolated from drives.

My questions are :


1. Do I need this enc data only for handling robot joints limits and for first start up position read (as all loops are closed into drives)?

2. Are these encoders data needed every 1mS for servo tread (in respect that I want to move robot in Cartesian space)?

3. What would be constraints on timing and synchronization of these 6 encoder readings?
  
   My idea is to use one additional mcu (stm32) to do "fast" communication with 6 drives (via uart, spi, i2c or can but most probably uart(RS485) with speeds in range from 1Mbps to 2Mbps ), do acquisition of 6 encoder channels and send it Linuxcnc. Mcu could issue 6 read commands "simultaneously"  on 6 UARTs, wait for all responses (ISR with control logic in drives works at 16KHz so worst case ~62uS delayed response concerning priorities), pack data and send to Linuxcnc.

4.  Which one would be best/easiest to use? Mesa smart serial, Mesa PktUART, pc serial port with custom protocol, ethernet -> UDP or something else?

- PC serial port seems to be slow with max speed of 1115200 - 128000bps, have to do isolation

- Mesa smart serial, as I see no isolation (if I want to use it just with bare 6i25 + RS422 chip, perhaps there are some add on boards which offer isolation? ) and concerning post from 29 Aug 2021 18:33 #219118
"
Replied by PCW on topic Mesa Smart Serial Interface Questions (expanding and/or encoder card options)

PWM would be possible on sserial as would encoders.
    encoders use a lot of bits so it's not too practical to do multiple
    encoders on a remote sserial card due to the current 96 bit
    packet size limit (this has been expanded to 224 bits but no current
    sserial remotes use more that 96 in and 96 out)
    "
    If I use stm32 for acquisition of all enc data, I would need 6 encoder * 32 bits -> 192 bits in total (possibly little more), is there new firmware for packet size of 224bits already available for 6i25?  If so where I can find it?  If I don't want to use stm32 then I need fpga firmware (or mesa board) with at least 6 smart serial blocks but in this case each one of 6 drives would have to cummunicate over smart serial, not sure here if drives could follow smart serial timings due other higher priority tasks (as read on forum after sending request mesa serial is expecting reply in some time ~20uS(not sure in which phase discovery or...)      
    - stmbl code available
    - arduino code availale (work in progress as stated)

- Mesa PktUART, as I read available speeds can go very high around 10Mbps, have to provide isolation, Rx buffer size of 1024 bytes, would implementation be straight-forward or there could be some issues?
  I would still plan to use it with stm32 which collects 6 position data (24bytes) and send it via PktUART. Only issue here would be number of available UARTs on stm32 but possibly this could be resolved with remapping RX/TX pins.
  I suppose concerning fpga space, firmware for PktUART could be made along with 6 step gen for 6i25?

- Ethernet -> UDP seems to me as good option as it provides isolation and could be handled with available stm32 + W5500 and
  seems that there is some source code already available

 www.forum.linuxcnc.org/38-general-linuxc...or-linuxcnc?start=10

 forum.linuxcnc.org/24-hal-components/287...rduino-over-ethernet
 
Thanks[/justify]

I've been working on an EtherCAT build for a couple of months now, and wanted to avoid posting here, but it seems I have hit a wall.
For some background: I am completely new to LinuxCNC but wanted to build an EtherCAT AXYZ cnc with Cia402 drives. I got an EK1100, EL1008, and EL2008, and my brother gave me a ClearPath-EC Teknic Cia402 (CoE) drive. I had a raspberry pi 5 available, so I am testing my configuration on that right now. I used the Pre-packaged rasp-pi install (Debian 12) with LinuxCNC from the linuxcnc website, and had very little trouble following Rod's installation from repositories. From there I had a lot of trouble, but got a configuration based off of the sim.axis config working with the EtherCAT slaves, and was able to test I/O with no trouble. 
Now, as I am attempting to incorporate the drive, but I can't get it into the OP state. The EtherCAT master recognizes it, and I can define pdos, but sdo 0x2000 and 0x2011 seem to just not exist, which I currently think has to do with the problem. I'm still having trouble wrapping my head around this, so let me know if there's anything else you need to help me. I would really appreciate the help!

teknic.com/files/downloads/ClearPath-EC_User_Manual.pdf

From your hal fileRemove the / from the first 10

What does
ip a
return?

Here are the .ini and .hal files.
The static IP of the RPi is 10.10.10.100 - I can ping the Mesa card from the terminal.
   

I am not sure if this is specific to qt_dragon or not.

However, when I try to rerun a tool because it broke or I need to adjust comps, the process makes me load and set tool length for each tool before it in the nc code.

I'm using versa probe and the Qt auto tool setter routine.

Is there any way to avoid changing each tool before rerunning the desired tool? Often times it's easier and quicker to just repost the single tool path as a separate program to avoid changing all the tools. I'd like to avoid that, if possible.

I'm using an RPi5 and a PCIe HAT with another NIC and some extra USB ports. Seems to work fine. It's the waveshare 27541. The setup needs the 5V 5A power supply to avoid voltage drops and hard resets. The 5V 3A power supply was not sufficient. I am using the official power supply in the control box as I had issues with other solutions. I'd also suggest a cooler for the cpu if you go this route.

From your error report:Usually means you have set the IP address on PC same as on Mesa card, the PC must have different last number for IP address, so
10.10.10.11 is OK for PC if Mesa is set at 10.10.10.10

What can you tell me about this program?

# --- Load the userspace for probe error manager ---
loadusr -Wn probe_error_mgr python3 probe_error_mgr.py

I have been running my machine in a kind of sketchy estop version, where LinuxCNC doesn´t really correlate with my physical estop hardware (which is running seperately).
Currently i am finally trying to implement feedback of my physical estop hardware (pilz estop relais that checks various inputs) into linuxCNC
Original hardware used two momentary switches connected to the main pilz relais to switch the main power on/-off. Once the machine is running i have a estop-ok signal 24V connected to my mesa inputs. 
Currently i have serious problems to synchronize GUI interface, linuxCNC and the main hardware. I have tried so many different versions of estop routing i am now totally confused what would be the best approach and like to have some directions for a new setup from scratch.
I would love to have it work like that:
- i connect my machine to power and the PC and mesahardware boots up, Estop Hardware is supplied 24V (works perfectly)
- once GUI is loaded (Gmoccapy) i press machine-on (which is currently not available to be pressed because the GUI is in estop)
- machine-on sends a 24V oneshot through mesa outputs to pilz e-stop relais, which switches on the rest of the hardware. (legacy function from the original setup)
- after a few seconds the estop hardware delivers a e-stop ok signal to linuxCNC which finally pulls the GUI estop button. halui estop is activated -> false

Main problems i have encountered so far are:
- halui estop is always actived. No matter what kind of reset functionality i have used in my main hal. 
- i can never use the GUI machine-is on button to actually start my machine. But i need to do so to recieve a estop ok signal from my hardware. So it needs to be usable before. Currently i use a jumper cable to start my machine and would love to get away with not have to implement a physical switch somewhere as the install with a big touchscreen doesn´t have anywhere to mount one.

Some direction would be much appreciated, thanks. 

I'm on 2.9.3. I'll update and see what I get.

If you call toolchange with parameter, eg.

O<toolchange>  call [1][#<toolno>]

then #2 has value of second parameter. #1 to #30  are used like that.
If you need the number of loaded tool, just use #5400.
Be aware that probing result is with current active coordinate system setting. If Z is has an offset...
Some people release from toolsetter after contact.

Thank you for all the replies, Appreciate the help though I will look through the links!
In the meantime I have a pi coming, thinking I can use usb-ethernet for non-realtime comms, though I prefer less components.

Good advice, I tend to forget there's a GUI for stuff like that.

What I did: In xfce you can go System|Synaptic Package Manager, search for Linuxcnc, mark linuxcnc-uspace. In Menu -> Package Force Version->you can select an older version -> apply.