Advanced Search

Have you done all the suggested BIOS changes to reduce latency?

(disabling all power management, speed-step, hyperthreading, C state > C1 etc etc)

If you only get the error when enabling some motion it may be an EMI issue
Is the 7I96S's frame ground connected to chassis ground?

Also can you report the result of:

sudo chrt 99 ping -i .001 -q -c 60000 10.10.10.10

( This assumes the card IP address is 10.10.10.10 )

The test will take 1 minute



 

the SPI speed depends also on the FPGA-Clock, the SPI-Slave is programmed in verilog and need some clock cycles.

So you need a higher FPGA-Clock or someone have to program a plugin based on the buildin spi slave, but this can not be ported between the different FPGA's (gowin, ice40, ...) and the pins are not freely choosable

I need your help because I'm completely lost.

as at the beginning of this topic, i had an “hm2 error finishing read” problem.
i took the pc to feed a 7i96s i never have the “finishing read” problem.
i bought a new computer with much better latency and constant ping.
I plug in my lathe and it displays “hm2 error finishing read” all time when i ask change tool on my turret and somethime when i enable the drive (F2)...

before :


it used to work fine but now the turret runs for 1 second with this error... and I haven't changed a thing since.

below is the histogram of the new pc

 
 
 

Oh, so it really is as simple as that. capture variables, do math, then set rotation.

Thanks!

meister wrote:

 

Can't recommend this set-up enough. For those who aren't particularly interested in engineering hardware and want a Snap-On Ready Kit, it just won't get any better than that. That'll do 50meg, prolly capable of more. Oliver did it all for grab  github.com/multigcs/rio-tangbob

Flash the bitstream over, plug-in a shielded 100mb cat6, and be done!!

Ok, I found a previous post of yours:
forum.linuxcnc.org/show-your-stuff/32103...nc-tool-table#296627

I have unblocked your new 'egemens' account but unfortunately your old posts cannot be linked to it.

Please tell me if which one of your two new accounts you want to keep and I will delete the other.
Thanks

Zayoo wrote:

I can confirm that generic_spi communicate with board.

@MirkoCNC
And not just headers, but also SPI lines. They should be as short as possible. I tested on my bord at 4Mhz no errors at 4.5Mhz wrong data and to many errors.
 

Something must be awfully wrong somewhere if constraint to only 4meg. I can do 50Mhz no problem with a dedicated PLL clock and differential data lines (terminated at only one end). And more or less 25Mhz over cheap/inductive dupon jumper wires... 

You don't need to restart LinuxCNC before changing the stepgen scale
you just have to make sure motion.motion-enabled is false when changing it.

I would like some advice please on how to proceed with the computer I am using to run LinuxCNC.  I have just installed the latest version of LinuxCNC (2.9.3 Debian12 Bookworm Preempt -RT) and have run the Stress Test for an hour - the figures are not encouraging: 85953ns max jitter.  I know this is way too high, and if I were to use this computer I would have to make some changes, or it might be that no matter what I do it will not be suitable for software stepping.  If so I might have to save-up for a Mesa Board and do the stepping there.  The problem is that  I have no real feel for what scale of improvement various changes might have. 

The computer is a Dell Optiplex 745 with an Intel(R) core (TM)2 CPU 4300@1.8GHz.  It has 6GB of DDR2 SDRAM and is using the on-board graphics card.  I know the graphics card is the first place I should look to make an improvement, and I have the opportunity to buy an AMD Radeon 450 with 4GB GDDR5 - but it is not that cheap. I have a simple 3 axis cnc mill with just 3 Nema 23 stepper motors, but I would like to upgrade it at some future point.

My first question is: in very general figures, is it realistic to think that a new graphics card might reduce the jitter figure by, say 60000ns, or is this too optimistic - is it even worth a try?  If any new graphics card could not make this scale of improvement it might be better not to buy the Radeon and put the money towards a Mesa board.

My second question is: if I were to buy the Mesa board would I still need to upgrade the graphics card anyway, or could I stick with the on-board card?

Finally,(preparing my self for the worst !), I have been looking at the range of Mesa boards available and it is a little confusing for a novice.  The advice on this site seems to be a year or 2 old, and so I would be most grateful if you could recommend the most suitable board for me.  I cannot imagine I will ever require more than 5 axis - I probably will stick with 3, but I might double up the motors on Y.  I need the ability to control the spindle (1500Kw 24000rpm currently using a Chinese VFD), and the mist (just a compressed air feed).  
Grateful for any pointers
Mitch