Somehow the tracking result look good when observing visually after jitter tweaking, but when I look at individual bones like the arms etc. they have weird behavior on one axis, in this case the Z-axis.

This does not happen if I just export an unfiltered file.

The reason I want to remove jitter in IPI instead of Blender (in this case), is because removing jitter on a parent bone when the child bone is already good, the child bone will get jittery after parent bone cleanup. #frustrating

Another weird thing is that usually on one side of the body it is the top of the jittery curve that is correct, and on the other side it is the bottom.

Is this just a "noob" problem?

...

Kind of hard to see from just an F-Curve snapshot what the actual appearance is, is it possible to use a screen recorder and upload the video to YouTube, or similar service, then post the link here?

You can also send the project files to iPi through a Cloud service and they will have a look, it may be an issue they are unaware of, or simple operator error.

What camera set up is being used? Are you adding any trajectory filtering before export? I can't see getting less jitter without using that smoothing filter before export though.

You will most likely need to use a bit higher JR setting on all parts, especially if the action in the motion is slower, or less active, with the TF on 4-5 before exporting, or you will get a bit of jitter remaining, but if using default settings on the JR or TF, some jitter will show after JR in the exported file.

What version of iPi Studio is being used also, I am guessing the latest version, with up to date video drivers for the graphics card being used?

Blender usually imports the parent from iPi fine, it's when you re-target to another character things get more wonky.

I pretty much use only a 6 cam PS Eye set up, but I have one Kinect 360 for testing that side, or if I need a quick transition motion I sometimes use it, but not often.

...

Yes, Snapz is right. A project file would be more helpful as illustration of your problems.
If you do not want to share in public then send a link to our tech support email.

Thanks for reply guys!

Well, before I upload, let's add some information.

The problem happens on every recording and stays within IPI beofre export and is not related to Blender (Blender only shows the curves). The Z-axis shown above is the axis that run along the upper arm bone on both sides. The forearms can also have this problem, but less so.
In the curve snapshot you can see that the lower segments are flat, means these are the problem areas and they lack tracking data.

I use a single Kinect V2, and have the latest IPsoft version and GTX970 drivers. I used cameras before but the setup took too much time and space.

Now, the thing is that the main track is good, flawless save for a bit of noise, and there is no refitting or anything needed, it just works. Also the Refinement and trajectory filter works fine. It is ONLY when jitter removal is used that the problem appears and that is even if arms/legs are set to Zero in the JR options. The locked deviation from the track appears to be 20-30 degrees of rotation. I have not fiddled with any other buttons than the minimum as shown in tutorials.

I am retty sure there is something wrong with the JR algorithm as it creates "dips" in the data along with the smoothing. But haven't anyone else encountered this?

...

Best to let iPi have a look, if you uploaded the files to them yet, you may be correct.

I do know this issue does happen with Kinects, or PS Eyes when imported onto 3ds Max Biped, always an upper arm bone spinning on "X" rotation that isn't in the iPi studio animation.

I corrected it inside 3ds Max with a biped initialization elbow setting change, but doing that also affected some arm trajectories a bit, though that is easier to layer out than the twisting upper arm bone... I don't use Blender for animation editing, but you say it's before import to Blender anyway.

Hope you can get it worked out.

...