Unexpected realtime delay and genserkins

Hi!

I have prototype of roborics arm with 3 axis based on steppers

7i92 + 7i76x2

Wheezy, LCNC 2.7.12
Max jitter: BP 32000, SP 23000

Homing works propertly

But when i try G0 command, I have "Unexpected realtime delay" error

What can I do?

Thanks

You might be able to ignore the alarm, but it indicates that your PC is not meeting the somewhat modest latency demands of your config.
forum.linuxcnc.org/18-computer/19022-faq-the-latency-problem

Is it bad latency with hardware stepgen`s on 7i76?
Max jitter: Base period 32000, Servo period 23000

Pentium Core Duo, chipset H110
Grub config: nomodeset, isolcpus=1

Well, start be deleting the base thread from your config.

I have no idea why you have a 50000ns base thread loaded, your config isn't using it for anything.

Once you are rid of that your realtime alarms will likely go away.

In your hal file on line 3 delete
base_period_nsec=[EMCMOT]BASE_PERIOD

If that cures the realtime alarms, you might find that you can increase the servo thread speed back up to the more usuall 1,000,000, without any problems.

I already deleted base_period_nsec and traj_period_nsec parameters

It works with servo period only more 3,500,000 when short moving by G0 command

But long moving ends with errors: joint following error, sserial error 13 and unexpected realtime delay

Something makes the delay

I would also remove the isolcpus statement, With Preempt-RT lowering the overall CPU speed (which isocpus does)
will often make Ethernet latency worse. The latency test results will show gross latency problems but
gives little information on the Ethernet latency of the system.

Also if you have Intel Ethernet interface you very likely need to turn off IRQ coalescing
This is done with ethtool:

sudo ethtool -C [Ethernet device name] rx-usecs 0

To determine the Ethernet+LinuxCNC hal latency of the system you need to monitor these parameters:

motion.servo.last-period
servo-thread.tmax

100*(servo-thread.tmax/motion.servo.last-period) is the worst case percent of total servo
thread period used by LinuxCNC and communications. For our Ethernet cards, its best that this be
less than 50%

I try, but

dj@debian22:~$ sudo ethtool -C eth0 rx-usecs 0
[sudo] password for dj:
Cannot get device coalesce settings: Operation not supported

How correct this option?

Thanks!

That probably means that eth0 is not a Intel MAC (so the driver does not have that option nor is it needed)
What does dmesg say about your Ethernet chip?

dmesg | grep -i eth

[ 1.557538] r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
[ 1.562748] r8169 0000:01:00.0: eth0: RTL8168b/8111b at 0xf8246000, 4c:cc:6a:b0:6d:90, XID 14100800 IRQ 137
[ 1.562750] r8169 0000:01:00.0: eth0: jumbo features [frames: 4080 bytes, tx checksumming: ko]
[ 17.171030] Bluetooth: BNEP (Ethernet Emulation) ver 1.3
[ 18.901063] r8169 0000:01:00.0: eth0: link down
[ 18.901091] r8169 0000:01:00.0: eth0: link down
[ 18.901161] ADDRCONF(NETDEV_UP): eth0: link is not ready
[ 22.229996] r8169 0000:01:00.0: eth0: link up
[ 22.230048] ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
[ 32.365302] eth0: no IPv6 routers present

Right, so that's a Realtek MAC so it doesn't have (or need) the IRQ coalesce option set

Also your hal file has the wrong thread order(order of addf's), it should be
7i92 read
everything else
7i92 write

What actual PC hardware are you running?