Absolute Encoder Questions / SSI via 5i25

PCW wrote:

Somehow the absolute --> relative bugs me but I guess there is really no major downside (other than a transient position read error would not get corrected)

I think it does. Consider the following sequence:
1000
1005
1010
1378 (oops!)
1020
1025

The deltas are 5, 5, 368, -358, 5.
The result of the computation is 1000, 1005, 1010, 1378, 1020, 1025.
It is very similar, in fact, to masking the lower N bits and or-ing in the new value. The difference is that in signed32 it auto-wraps into the 64-bit accumulator, (but not with variable bit)

A new patch (on master from the root of LinuxCNC -dev).
Now pushed to the master branch.
I was expecting to push this to the main distribution this weekend, but I have realised that it needs to cope with configurable data fields, and that has multiplied the complexity enormously.
So, this patch will let you run encoders now, but all you get is the encoder part, and the emulated index might not work.

I think PKM has got configurable index emulation working, but so much is due to change that I didn't see the benefit in merging it now.

Unlike my previous attempts, this one has actually been tried on hardware.

Now pushed to the master branch:
git.linuxcnc.org/gitweb?p=linuxcnc.git;a...b3201fa1a08657b9f69a
it should be available as part of the pre-built packages at
buildbot.linuxcnc.org

Please feed back any problems or feature requests.
BISS and Fanuc will share the same format.