Status of the Motion Master items in the automation framework's upstream report

[markosankovic]

This is the status of the Motion Master items (MM1 to MM14) of the upstream report from the automation framework team. The Motion Master fixes are in v6.0.0-alpha.91: https://github.com/synapticon/motion-master/releases/tag/v6.0.0-alpha.91

Fixed

• MM2. An out-of-range write is wrapped and written. A value that does not fit the object's data type is now refused with 400. The check is against the data type, not against the ESI's minimum and maximum. A parameter write that does not fit the data type is wrapped and written #75

• MM5. SDO objects in a monitoring slow every other call. Mailbox work on different drives now runs at the same time. The monitoring refresher reads different drives in parallel, and after each read it leaves the drive alone for as long as the read took. On an Integro bench, with the script from the report:

  • The refresher's share of the drive fell from 85% to 49%.
  • getDeviceStates with the monitoring went from 5.1 ms to 3.9 ms (median).
  • An SDO readParameter with the monitoring went from 8.9 ms to 7.1 ms, which equals the time without the monitoring.
  • The cost: on an Integro, the SDO objects of a monitoring update about every 18 ms instead of every 10 ms.

  A lock per drive for mailbox work instead of one lock for the whole bus #78 SDO objects in a monitoring keep a drive's mailbox busy, and one thread refreshes every drive #79

• MM6. The trigger-error caveat fits develop, not v5.6. Fixed as a documentation change. The caveat now says that on v5.6 the drive ramps down on 0x6085 whatever 0x605A says, and that from v5.7 the reaction follows 0x605A. The v5.7 behaviour is not backported to v5.6. The trigger-error caveat describes the v5.7 fault reaction, not v5.6 #82

• MM9. On a drive with a Safe Motion Module, 0x2001 reads 0 in OP. Fixed, both parts:

  • On a SOMANET device, a PDO entry with index 0x2001 or 0x2007 is treated as padding, and Motion Master logs a warning that names the device. So 0x2001 is read and written over SDO in SafeOp and OP. Padding that names 0x2001 in a SOMANET safety PDO is bound to Commutation angle offset #81
  • A device below SafeOp, or with the AL error bit set, no longer gets its input values overwritten while another device exchanges process data. The decode overwrites the input cells of a drive below SafeOp #80

• MM13. A failed FoE read does not say whether the drive refused or stayed silent. A refused FoE transfer now names the drive's FoE error code, for example 0x8001 (not found). This needed a patch in SOEM. A refused FoE transfer loses the drive's FoE error code #83

Changed in a different way than asked

• MM4. A power-cycled drive cannot reach PreOp until a re-scan. A scan stays the recovery, because it is the right one when the whole bus loses power. What changed:

  • Each device in GET and POST /api/devices/state now has reachability: answering, addressLost or notAnswering.
  • A state request leaves out a device that does not answer at once, instead of after the timeout.
  • GET /api/scan tells whether a new scan is needed, and the Console shows a "Scan needed" notice.

  A power-cycled drive cannot reach PreOp until the whole bus is scanned #77 A drive that lost its station address reports a stale state, and PreOp fails only after the timeout #91 Report when the bus needs a new scan #93

The cause is in the drive firmware

• MM1. Below OP, process data is served frozen as live. The wire captures show that the drive sends frozen input bytes in SafeOp, and Motion Master passes on what arrives. Motion Master will not switch to SDO below OP, because ETG.1000.6 requires valid inputs in SafeOp. Firmware issues: Circulo somanet_software#5807 (cause found, it also explains the working counter of 1), Integro somanet_software#5808. In Motion Master, the expected working counter now counts the outputs of a device in SafeOp too. Below OP, the process image serves values that no longer change as synced #73 The expected working counter counts a device in SafeOp as inputs only #74
• MM3. Every drop from OP to PreOp latches CyclicHb. The Integro faults whether Motion Master stops the cyclic frames first or keeps them running. A Circulo 7 with the same firmware version faults in neither case. Firmware issue: somanet_software#5806. A SOMANET Integro latches CyclicHb on every drop from OP #76

Not planned

• MM7. A command lease. Motion Master stays lean, and coordination between callers is the caller's job.
• MM8. An unlock operation for protected files. Won't fix.
• MM11. writeDevicePdoMapping and fixed PDOs. Writing a mapping correctly is the caller's job, and Motion Master avoids doing things in the background.
• MM12. The SDO abort reaches a caller only as text. A design decision for now. It can change later.

Not a Motion Master item

• MM14. OBLAC Drives does not warn when a setting leaves the encoder's range. This concerns OBLAC Drives and is not handled here.

Also in this release

When an FoE transfer fails without an answer from the drive, Motion Master now aborts the transfer on the drive. Before, a SOMANET bootloader stayed stuck after a read that took too long, and it answered every file request with mailbox error 0x0008 until a power cycle. File reads now wait 6 s with 3 attempts, and a write waits 10 s for each packet. #97