Computer Migration loop tuning issues 5i25 to 7i92m

Hello Everyone,

I have an interesting issue and need some guidance.

My Fryer bed mill conversion has been working well, but I have been wanting to upgrade the computer for some time. The machine runs DMM servos in torque mode connected to a 7i77 and 5i25.

The current computer is Jetway J7F2WE1G2E 1.2Ghz Via C7 Fanless processor with 1GB of ram. This system runs very well, but has very little spare processor which causes slow update rate on the GUI while running the machine. This machine runs ~20us latency running 3 GLX gears, however the glx gears are only 2FPS each...


I purchased a 7i92m to migrate from the Mini ITX mobo w/ 5i25 pci card to a Zotac Zbox micro form computer running a Celeron N3160 processor at 1.6Ghz. This system can run 3 GLX gears and maintain ~60us latency. I migrated the profile from the old computer to the new computer and changed the hal file to suit the new interface and was up and running. I used it for a few weeks (making good parts) like this before I checked the tuning.

I did not expect the tuning to be perfect, but the machine sounded good so I went with it for a few weeks. Upon investigating, its Horrible!

Here is the scope from the original computer:


And here is the scope from the new computer at the same scale:


You can see the error is off the screen!

I have tried re-tuning the new computer and was not able to get the error into an acceptable range. It seems to always follow the commanded position. For an example, Original P gain was 6000. I got all the way up to 12000 and still had a very large error. If Im reading the values at the bottom of halscope correctly, 10 thou error.

One of the reason of moving to a new computer was to increase the servo thread speed. I set it for 2khz and started getting timing errors that were not present at 1khz. This made the error of the loop smoother but did not help it.

What would cause the error to not be able to be driven close to 0 on this new computer?

Im sure I missed some details but hope I have provided enough for someone to get me pointed in the correct direction.

-Adam

If this is a newer LinuxCNC installation You may need to disconnect the feedback-deriv pins on the PID component

FF1 seems off by quite a ways

Also watch out for possible analog issues with grounding, this would change with a 7I92
(since the analog ground would not be frame ground by default like it would be with a 5I25)

Yes. This is running 2.8.

Some more info, The old computer was running the 2.7 iso and the new system is running Mint 18.1 with preempt-rt kernel and 2.8.

So would it be wise to ground the 7i92 at the standoffs to the same frame as the 7i77? They are powered by the same 5V supply.

Andy,

I have no had a chance to reply, but PCW had the solution. Disconnecting the deriv pins fixed the error. I was able to get very close after this just by resetting FF1 to 0.

I am using inch. The current maximum error is about 0.001in and occurs when the axis enters or leaves the cruise phase of movement.

Any suggestions for how to fix this?



-Adam