Yaskawa incremental encoder and Mesa 7i48

    I recently had a terrible time trying to implement an older (90's) Yaskawa servo motor with "incremental" encoder UTOP1. It would run rough in one direction and smooth in the other with an AMC brushless drive and 7i77. I opened the encoder cover and found odd looking tracks (see 1st and 3rd encoder band on picture). I'm curious on the purpose of it? It's clearly not a standard incremental evenly spaced grating. The additional 3 sections that repeat 4x I assume are the commutation hall signals and the inner 2 are the index. Maybe open yours and see if the tracks are odd looking?  

If they are anything like the ones i had, and they sure do look and sound so, they have ABC not ABZ signals, all are incremental outputs, but not linear at all, the rate changes depending on motor angle.
Open hal scope and monitor all 3 pairs, use something to consistently turn the motor, slowly, see the width of each line change at a certain point.
I really hope they are not the same, as those are useless for anything except Yaskawa drives.

I spent quite a few hours yesterday watching A, B and C lines (as well as rawcounts value) in HalScope and I found a nice pattern there. I have created excel spreadsheet where I tried to draw those graphs with cell boarders and also wrote the rawcounts values at which those transitions take place (basically only transitions are shown, everything in middle is ommited).
docs.google.com/spreadsheets/d/1M04yZmUE...2&rtpof=true&sd=true

I have plotted full revolution of motor shaft, there are 24 transitions (25th is the same as 1st). The pattern repeats 4 times (that is 8 pole motor - 4 electrical revolutions), so pattern is divided in 6 parts of similar length.I should note that difference of rawcounts between 1st and 25th transition is slightly more than expected for 8192 CPR encoder, but I was having realtime errors as I tried to use TightVNC to connect to the LinuxCNC PC from my laptop so that I could see the screen (most importantly - numbers there) while standing near the motor and turning it. But it was so slow on updating screen that I ended running back and forth to the PC anyway.

One more thing - there were several occasions where I noticed strange spikes in C graphs. At first I thought that those are glitches etc, but at one time when I saw it again I tried to catch that place again in HalScope. So I got 3 repetitions and all 3 attempts show a spike there so that makes me think that something similar to index pulse is integrated in C channel (I have attached all three log files for this spike, it is between rawcounts 33658 and 33659, I changed the extension from csv to txt, because adding csv files is not allowed).I did save log files for almost all other transitions as well if that helps (forgot to save 2 or 3 due to lack of concentration as it was around midnight).

Is there any chance to get anything meaningful out of this? I would like to think that this pattern can be used similarly to Hall sensor feedback.

And this thread alone shows that I am not the only one who would had liked to use Yaskawa motors (I have one more machine that will need replacement of at least 2 drives and one previous retrofit that also is not working anymore because of drive faults and would need to be reanimated).

There is some information regarding C-Channel in this document:
www.dropbox.com/scl/fi/f43c4x9juh0dsgcp9...y3ktvoa1buw&e=1&dl=0

Relevant sections:

 
 
 

 

I wonder if a HAL component could be written to interpret this startup data?
It would rather depend on the rate that it comes in at. It is probably too fast to sample in a servo thread, so can't be directly read from the mesa driver pins.

It is said in there that it takes 40 us to send that message and if I understand correctly it sends 15 pulses. so something around 3 us per pulse which is 300-400 kHz range?