Advanced Search

ProblemAs a newbie to Mesa cards, I spent a whole night trying to connect to a 7I76EU card (IP 10.10.10.10, jumpers W2=OFF, W3=OFF). mesaflash kept throwing errors: „Unsupported ethernet device 7I76EU” or „No 7I76EU board found”, even though ping to 10.10.10.10 worked and LEDs CR4 (FIELD POWER) and CR14 (LOGIC POWER) were on. CR11 (FIELD I/O FAULT) was lit, and CR13 (FPGA /DONE) sometimes lit up, hinting at a firmware issue. I wasn’t sure whether to use 7I76EU or 7I76E in commands, as 7I76EU is less common.What I checked

• Power: Connected 24V DC to TB1 (VIN) and TB2 (FIELD POWER), following: YouTube: Mesa 7I76E Power . Jumper W1=ON. Did not use P4/P5 (5V) as the card has built-in regulators.
• Network: Used a standard Ethernet cable (non-crossover), connected directly to PC. Initially tried IP 192.168.1.121 (W2=ON, W3=OFF), but switched to 10.10.10.10 (W2=OFF, W3=OFF), setting up the network per: Forum: Mesa Ethernet Setup . PC IP was 10.10.10.11/24, firewall disabled (sudo ufw disable).
• mesaflash attempts: Commands like sudo mesaflash --device 7i76eu --addr 10.10.10.10 --readhmid failed. I noticed the error mentioned „7I76EU” even when I used ETHER or 7i76e, suggesting the card was detected but not supported by mesaflash.
• LinuxCNC: Using Pncconf, I configured for 7I76E (IP 10.10.10.10). Changing INI to BOARD=7I76EU caused LinuxCNC to fail, but BOARD=7I76E worked, and CR11 turned off. On TB2, a multimeter showed DIR+5V signals changing with axis movement, and shorting IN17 to +24V activated the input in HAL. The card was working!

SolutionThe issue was an outdated mesaflash version that didn’t support 7I76EU. Updating from GitHub fixed it. Steps:

1. Check version:
   bash
   Copy

   mesaflash --version

   My version was too old.
2. Install dependencies (Debian Bookworm):
   bash
   Copy

   sudo apt update sudo apt install libpci-dev libmd-dev pkg-config build-essential git

3. Download and compile mesaflash:
   bash
   Copy

   git clone https://github.com/LinuxCNC/mesaflash.git cd mesaflash make sudo make install

4. Verify version:
   bash
   Copy

   mesaflash --version

   Showed: 3.5.13.
5. Test communication:
   bash
   Copy

   mesaflash --addr 10.10.10.10 --device ETHER --readhmid

   After fixing a cable mix-up, it returned card details: BoardName: MESA7I76, HostMot2 firmware, 5 StepGen, 3 I/O ports.

Notes

• Ensure the PC is in the 10.10.10.x/24 subnet and firewall is disabled.
• Jumpers W2=OFF, W3=OFF set IP 10.10.10.10; W2=ON, W3=OFF sets 192.168.1.121.
• LinuxCNC with BOARD=7I76E works because 7I76EU is compatible with 7I76E in older versions.
• CR13 (FPGA /DONE) stopped lighting up after LinuxCNC started correctly, but updating mesaflash was key for proper diagnostics.

ThanksThanks to advice from LinuxCNC.org , YouTube videos, and Grok’s guidance, I solved the issue. Hope this helps other beginners!
Show in sidebar

Hello gentlemen !

Im a Newby and want to run a CNC-mill with a Mesa 7i76E and a Rpi 5.
I have  read the thread and think a RPI 5  with 8 GB will be  the right choice for
bookworm and the mesa. The 16 GB RPI5  won't work or is there any kind of news ?
The freason for the question is that i have to buy a RPI but which one is the best opinion and why ?
Can you help me ?

Thanks   a lot
Ultrakurz
 

It could be a wiring/noise issue but typically this is because of the relatively
instantaneous velocity estimation (typically updated every ms), which tends
to accentuate the encoder inaccuracies. This fast update is important if a feedback
loop is used but a nuisance for a human readable display.

The way this is typically fixed is by low pass filtering the encoder velocity signal
before displaying it. If this has already been done, you might increase the time
constant of the filter

Hey everyone,
I’ve been setting up a CNC mill using LinuxCNC with a Mesa 7i76e board, and I’ve run into a bit of a headache. The spindle RPM readings in the LinuxCNC GUI are fluctuating a lot even when the spindle speed seems stable when measured with an external tachometer. I'm using an encoder on the spindle for feedback, but I’m not sure if the issue is from signal noise, encoder settings, or something else in the HAL config.Has anyone else had similar issues with inconsistent RPM readings? What would you recommend checking first encoder wiring, debounce filtering, or maybe even firmware updates for the Mesa board? Any guidance would be a huge help, I’m still getting comfortable with HAL and custom configs.Thanks in advance!

References:

www.theengineeringprojects.com/2023/06/i...l-with-linuxcnc.html
www.cnczone.com/forums/linuxcnc-formerly-emc2-/
store.mesanet.com/index.php?route=product/product&product_id=58

theslawek wrote:

BINGO, adding the halui line did the trick. Thank you PCW and spumco, you guys are amazing (and lighting fast responders).
 

Thanks for reporting back, and glad to hear it's sorted out.

BTW, the "=>" and "<=" in the line are just characters to help the user visualize the pin to signal connection.  They are not required, and are ignored by LCNC.  I personally find single-line HAL entries nearly impossible to read, but breaking them up in to multiple lines does make for a much longer HAL file.

For example, the single line:
net machine-is-on hm2_7i76e.0.stepgen.04.enable halui.machine.is-on

is the same to LCNC as:
net machine-is-on  <= halui.machine.is-on
net machine-is-on  => hm2_7i76e.0.stepgen.04.enable

After my last post I was trying "machine-is-on" in the lines WITHOUT hm2. I tried your replacement and the result is M3 and M4 no longer do anything. No error.

#net spindle1-enable => hm2_7i76e.0.stepgen.04.enable
net machine-is-on hm2_7i76e.0.stepgen.04.enable

Even tried with the same "=>" and still same behaviour.

Do I have to do something with my regular spindle hal entries for it to play nice?
net spindle-enable => hm2_7i76e.0.7i76.0.0.spinena

PCW wrote:

Simply replacing:

net spindle.1.on hm2_7i76e.0.stepgen.04.enable

with

net machine-is-on  hm2_7i76e.0.stepgen.04.enable

Should fix the issue.
 

PCW - I didn't see the halui signal machine-is-on in his HAL file; maybe I missed it.

OP - if that signal is already in your half file, do what PCW posted.

If it's not already there, you can tweak it to this:

net machine-is-on <= halui.machine.is-on
net machine-is-on => hm2_7i76e.0.stepgen.04.enable

I can confirm this fixed my stepgen spindle after PCW suggested it to me last year.

Simply replacing:

net spindle.1.on hm2_7i76e.0.stepgen.04.enable

with

net machine-is-on  hm2_7i76e.0.stepgen.04.enable

Should fix the issue.

theslawek wrote:

1. Should I expect M5 to stop all spindles, or do I have to get used to using M5 $-1?
   • Is it possible to get M4 to work? It doesn't do anything right now.

M5 - I recall some discussion on Github about what should be the default behavior for multi-spindle M5 commands.  I found a commit that appears to change M5 to 'stop all spindles', and any of the $ modifiers are for individual spindles.
github.com/LinuxCNC/linuxcnc/commit/01a0...d65486193ead51027c29

Probably want to do some testing yourself.

M4: I think what's going on is that step/dir spindles require an explicit reverse speed value in the INI file, despite what the manual indicates.
linuxcnc.org/docs/devel/html/config/ini-...:ini:sec:spindle-num

adding
[SPINDLE_1]
MAX_REVERSE_VELOCITY = 550

Might do the trick.

If this doesn't work, the next thing is for you to get used to using the 'halshow' feature in LCNC.  This is a simple but powerful tool for debugging since it lets you see what LCNC is doing behind the scenes.  In your case, you want to know why M4 isn't working so you should display all the HAL pins which affect the spindle - including the PID pins.

When you open halshow there's a folder tree on the left where you can click on all the pins (real physical pins, or logical pins) in LCNC.  Once you select those they will appear on the right in the 'watchlist'.  On/off (bit) pins will look sort of like red/yellow LED's, and other pin types will display the value.
What you want is to display all the pins which form the internal logic chain that starts with and M4 command and ends with the mesa stepgen outputting signals to the drive.  Figuring out all the pins and how they're logically interconnected is not always straightforward... but it also didn't cost you buckets of money.

This is where coming to terms with HAL is necessary:

All this stuff like this below still makes no sense to me as to what its purpose is.
net spindle1-vel-cmd-rps        <=  spindle.1.speed-out-rps

'spindle.1.speed-out-rps' is a logical (not physical) pin inside LCNC.  When LCNC starts up and reads the HAL file, there are a number of default functions that are automatically loaded. One of these functions is called 'motion.comp', and a description of what it does and all the related pins which are created is located here:
linuxcnc.org/docs/devel/html/man/man9/motion.9.html

If you read through that page, what you find is a list of logical pins which are created when LCNC starts.  Once those pins are created, you can do 'stuff' with them by connecting them together.  I won't go in to specifics right this minute, but any pin which is connected to spindle.1.speed-out-rps will receive a value that is equal to the commanded rotary-axis spindle speed in revolutions per second.

Hopefully this will click with you soon.

BUT

Assuming you didn't mess things up significantly during your editing, I think there may (still) be at least one bug in Mesact or PNCconf for step/dir spindles.

See where spindle.1.on is connected to hm2_7i76e.0.stepgen.04.enable through signal spindle1-enable?

This connection will result in the stepgen shutting off instantly when the spindle is turned off - no controlled deceleration.  This result in a massive BANG! when the motor drive stops dead, and can lead to over-voltage trips in the drive or broken things.

This should have been fixed in LCNC 2.9, but it appears to still be an issue:
github.com/LinuxCNC/linuxcnc/issues/2748

This bug shouldn't affect M4/reverse not working, but I bet your spindle stops HARD right now and doesn't decel gracefully.  

Hi Spumco, I appreciate your dedication to helping me accomplish this.

I ran into a hurdle immediately with MesaCT 2.1.7. When set to 7i76E, or any other card for that matter, the Spindle -> StepGen tab just has a message of "not functioning yet". I even cloned the lastest master from GitHub and the same message is present.

I was about to press Post on this reply, but then I thought, why don't I try PNCConf in case that's what you meant. PNCConf seems to have what you indicated. I did the changes in the INI, but I'm hung up on the HAL part.

Here are the substitutions I made from the PNCConf HAL into the clone of my good xyza config. (A parts were removed)

1. pid.s is pid.s1 (including in the loadrt and do-pid-calcs parts)
2. SPINDLE_0 to SPINDLE_1
3. changed all ".s" entries to ".s1"
4. entries such as "net spindle-vel-cmd-rps" changed to "net spindle1-vel-cmd-rps"

I've loaded the config many times and addressed my oversights.

I'm stuck with this error:

twopass: Error in file ./hallib/main.hal:
    Pin 'spindle.1.speed-ut-rps' does not exist

Which I think relates to this block. I'm not sure how to go from here. Are these related to the GUI?

net spindle1-vel-cmd-rps        <=  spindle.1.speed-out-rps
net spindle1-vel-cmd-rps-abs    <=  spindle.1.speed-out-rps-abs
net spindle1-vel-cmd-rpm        <=  spindle.1.speed-out
net spindle1-vel-cmd-rpm-abs    <=  spindle.1.speed-out-abs
net spindle1-enable             <=  spindle.1.on
net spindle1-cw                 <=  spindle.1.forward
net spindle1-ccw                <=  spindle.1.reverse
net spindle1-brake              <=  spindle.1.brake
net spindle1-revs               =>  spindle.1.revs
net spindle1-at-speed           =>  spindle.1.at-speed
net spindle1-vel-fb-rps         =>  spindle.1.speed-in
net spindle1-index-enable      <=>  spindle.1.index-enable

My latest INI and HALs attached.

I'm not really sure exactly what is does, but this is where it is used in my hal file.
It looks like it is set to -10800 because 5400 to -5400 is -10800.

# ---Analog output signals/setup---
setp hm2_7i76e.0.7i83.0.1.analogout5-scalemax [SPINDLE_0]OUTPUT_SCALE
setp hm2_7i76e.0.7i83.0.1.analogout5-minlim [SPINDLE_0]OUTPUT_MIN_LIMIT
setp hm2_7i76e.0.7i83.0.1.analogout5-maxlim [SPINDLE_0]OUTPUT_MAX_LIMIT
net spindle-output => hm2_7i76e.0.7i83.0.1.analogout5

Hi Spumco,

I can't find any official documentation about caxis.comp, just many forum posts about it from you. Is this something new not in the documentation yet?

My goal is to make my CNC router use the rotary axis and simulate a lathe. I've actually already done this before moving to my Mesa 7i76e but using an external device with a dial ( amazon link ) to control the speed. I'd like to have that programatic control over the speed for obvious reasons. I actually have a YouTube video on what I did then. 

I don't need dynamic change between velocity and position controls. I'm more than happy to restart LinuxCNC with new config when changing operations.

I understand now, silly me.
Ok found I had to use setname & found the Linux version here
www.forum.linuxcnc.org/27-driver-boards/...board?start=10#65737

and I can communicate.

Thanks again

Here's an example:

 

It's a bit tricky because you do need to use muxed encoders for all
encoders (because the 7I89 uses physically muxed encoders)

 

Hi All

My intent is to merge some vhd files to get a 7i76e + 7i78 + 7i89 

I saw a post here (from PCW I believe) sayign the count in the 4th column was the number of those modules, so for the stepgens I want (6x or more) I've been putting x06, and same for the encoders (1x spindle + 6x on the DB25 headers), just increasing the built in Qcount. I believe I don't need muxed ones (I have single ended A/B/Z glass scales), finally I then make sure the total number of entries in the whole array was the same as the standard hostmot2 7i76e examples. No ide if this logic is sound or not... 

(HM2DPLLTag,x"00",ClockLowTag, x"01",HM2DPLLBaseRateAddr&PadT,HM2DPLLNumRegs,x"00",HM2DPLLMPBitMask), (WatchDogTag,x"00",ClockLowTag, x"01",WatchDogTimeAddr&PadT, WatchDogNumRegs,x"00",WatchDogMPBitMask),
(IOPortTag,x"00",ClockLowTag, x"03",PortAddr&PadT,IOPortNumRegs,x"00",IOPortMPBitMask),
(QcountTa    x"02",   ClockLowTag,x"06",QcounterAddr&PadT,QCounterNumRegs,x"00",QCounterMPBitMask),  
(StepGenTag,x"02",ClockLowTag, x"06", StepGenRateAddr&PadT,StepGenNumRegs,x"00",StepGenMPBitMask)
(MuxedQcountTag, MQCRev,ClockLowTag, x"01",MuxedQcounterAddr&PadT,MuxedQCounterNumRegs,x"00", MuxedQCounterMPBitMask),
(MuxedQCountSelTag, x"00",ClockLowTag, x"01", NullAddr&PadT,x"00",x"00",x"00000000"),
(SSerialTag,x"00",ClockLowTag,x"01",SSerialCommandAddr&PadT,SSerialNumRegs,x"10",SSerialMPBitMask),  
(LEDTag,x"00",ClockLowTag,x"01",LEDAddr&PadT,LEDNumRegs,x"00",LEDMPBitMask),
(NullTag,x"00",NullTag,x"00",NullAddr&PadT,x"00",x"00",x"00000000"),
(NullTag, x"00", NullTag,x"00",NullAddr&PadT,x"00",    
more null
more null

until I get to the same line number as the examples so same number of modules. 

I think I've made a dud bit file and written it to my 7i76e at some stage, the first time I tried to write it, mesaflash did the erasing, got halfway through the writing "WWWW..." thing and the terminal just hung. Ctrl + C'd the terminal, and I think the 7i76e stopped responding. 
I power cycled the card itself, and now mesaflash can see it and write known good bit files to it, but after I write new bit files and --reload, the --readhmid just returned "None" for all of the channels. 
I've since power cycled a few times and it seems to be writing standard bitfiles again by some miracle.

Can someone explain the header logic or point me to an explanation? 
Can I just keep changing the module count freely and the Xilinx tool will flag if it runs out of space? 
If the Xilinx tool compiles it okay can I be confident it won't brick the board?
How do I know if it's in fallback or normal boot? 
What do the other columns in the header do, do I need to set memory addresses where the modules start or block sizes or anything? I'm wondering if I could be increasing module count and then overflowing the following module or something. 
Does the total number of modules in that array need to be correct / a specific number / does it matter? 

I've skimmed the hostmot2 dataflow vhd file but fallen at the first hurdle: 





This line obviously reads that count but I can't see where NumberOfModules (function? attribute of the an object called "work"?) is defined so I'm lost. As you can tell I don't have a clue how this code works so I'm just a smidge lost  
I've had PCW post a few bitfiles here for me, for which I'm eternally grateful, but I feel bad so I'd prefer to acutally understand what's going on and be able to do it myself (and pass the favour on if it's not potentially harmful to someone elses setup!)

Thanks, 
Kealan