J3455, mid 50k servo jitter and mid 60k base thread ok?

More
19 May 2020 00:02 #168211 by jhandel
I have tried everything under the sun with Preempt_rt to get this little J3455 mini-pc (amzn.to/2ZbzSsi) to hit "decent" numbers.. So now I have to wonder, is the machine just not right?

The best numbers I can old are around 50k for servo and 60k for base... I've tried cpu isolation, a ton of other kernel boot settings, custom builds of the kernel, multiple images, etc... Those numbers are with Preempt RT or RTAI...

Before I throw up my hands has anyone else had luck with J3455 procs? is a mid 50k number ok? I am looking to build an EtherCat servo based 5 axis machine and I'll want to push it pretty decently..

Sadly I can't seem to do anything with the fans or the bios (super locked down bios) so it could be that my problem is in parts of the bios that is nuttered.. But before I assume that I figured I would ask if that "feels" normal.

Thanks all

PS: Its not a total waste, I need to update my very aging Mach 4 machine so there is that :-).. And I can still dev Linux CNC UIs on it.

Please Log in or Create an account to join the conversation.

More
19 May 2020 00:32 #168215 by cmorley
are you using Mesa hardware or parallel ports?

Please Log in or Create an account to join the conversation.

More
19 May 2020 00:36 #168217 by PCW
He said Ethercat, that should probably be OK with 60 usec latency

Please Log in or Create an account to join the conversation.

More
20 May 2020 13:50 #168382 by jhandel
So at what point is Servo thread too slow? and about how much IO & Axises can you "fit" into a given amount of latency before there is a risk of bad results...

Just trying to decide if 50k jitter is going to be sufficient for 5 axis, tool changing, etc...

Thanks

Please Log in or Create an account to join the conversation.

More
20 May 2020 14:08 #168383 by tommylight
That is still usable even for software stepping/parallel port, with somewhat limited velocity.
For hardware stepping it can be much worse and still work properly.
Latency has nothing to do with how many axis or IO are in use.

Please Log in or Create an account to join the conversation.

More
20 May 2020 14:15 - 20 May 2020 14:16 #168384 by PCW
The amount of data is not very important as its all very small relative to the data
transmission rate. The effect of latency on axis motion depends on the control
method chosen and whether the hardware has the ability to retime read and write
operations or filter/slew limit velocity or position commands

Assuming a simple position mode system with no re-timing and no filtering, 50 usec of jitter
would cause a 5% error in position updates. For example at say 120 IPM = .002" per mS,
5% = 10 uinch position command noise.

A velocity mode system is less affected by jitter (and unaffected during non accelerated motion )
Last edit: 20 May 2020 14:16 by PCW.
The following user(s) said Thank You: tommylight

Please Log in or Create an account to join the conversation.

More
20 May 2020 16:55 #168389 by jhandel
So jitter is more about positional accuracy at any given moment in time and less about "bandwidth" to read/write... Got it.. Thanks

Please Log in or Create an account to join the conversation.

More
31 Aug 2020 14:23 #180166 by PeterKey
Hello!
Sorry, my reply may be late.
I have a J3455 desktop with Linux Mint LMDE3 and LinuxCNC 2.8. Using an integrated video had a unsatisfactory latency. The discrete videocard lead to make a latency about 15000 despite of any software load.

Please Log in or Create an account to join the conversation.

Time to create page: 0.092 seconds
Powered by Kunena Forum