[etherlab-users] Cyclic Synchronous Position mode

Graeme Foot Graeme.Foot at touchcut.com
Thu May 17 01:53:42 CEST 2018


> With my now modified cycle, I get a difference of about 4.45 cycles. As I now receive the current position before I send out the next target position (while it was the opposite way before), it makes sense it is one cycle more. But I compare the current position I receive at the begin of the cycle with the target position I send out right afterwards (I do not compare with the one generated later in the cycle and send out the next

> cycle!)

>

> Therefore, I think, it is still not 100% clear:

> - I would assume that in my now modified cycle (when following your

> example) I have 3 cycles delay, not 4.



Yeah, working out how many cycles a round trip takes always gives me a headache, and I usually forget a step, so don't take what I said as gospel.





> - SYNC0 offset does not affect the 0.x cycle at all. This however makes sense to me, as SYNC0 both affects the moment the target position is activated AND the moment the current position is captured. So modifying it will just move BOTH events together, making it invisible when comparing like this (of course, the real axis position will move relative to the real time).



Correct.





> - The reason for the 0.45 cycles is probably some internal delay in the drive between activating target position and capturing current position?



The drive has its own internal cycle rate.  The Yaskawa's I use run at 125us cycles (or was that 62.5us, I forget).  So your sync event will fall somewhere between these cycles with the target position being picked up for the drives next cycle, but the actual position returned is the one from the previous drive cycle.  You can usually find out your drives cycle rate in the documentation, or from its stated minimum supported cycle times.  (I've seen some that look like they run at 250us cycles.)  So that's some of the delay, the rest is possibly due to timings in the master side.





>

> <> What you need to do is find the Following Error PDO index, add that to <> your drive PDO data, and use it.

>

> I read this now (also 0x60f4:0). It is however not giving me the delay I am observing but the servo error due to axis regulation tuning. If I substract the following error from the current position and then do the compare of target position and modified current position I get very stable a delay of 4.456x cycles during constand speed movement and just slightly changing values during strong accleration. So I can measure the delay much more precisely.



I thought you were comparing target position to actual position to calculate your own following error.  If not, then is it just for curiosity / understanding as to what the delay is made up from?  Or is it so that you have some idea how long after you transmit a target position that it will take to physically be there?  If it's this one, then the total round trip delay doesn't matter, only when the hardware is where you expect it to be does.  This time will be something like:

t = t @ start of cycle  +  Sync time of slave  +  (up to) slave cycle time  +  master cycle time





>From your previous email:



> Is there anything else I need to keep an eye on when doing all the work after "master send"? Can I safely assume that all TxPDOs from the drive are still valid after the master send and I can read them after the master send using EC_READ_*? Or should I save them when I write the cached RxPDOs? In my test program, it works without caching though.



As Gavin said "overlapping PDOs" are off by default, so the TxPDO's (read data) are safe to read after writing the cached RxPDO's (write data).  Just make sure to don't overwrite any TxPDO read data.





As a bonus, just another little hacky trick I do...



I also call ecrt_domain_state() every cycle and ecrt_master_state() once every 10 cycles, outputting changes to the system log.  I don't actually perform any diagnostic functions beyond that using these (e.g. send commands to figure out what is not communicating and do something about it) as that would take too long.  But I do need to know if any of my active drives have stopped communicating and deal with it immediately (abort the coordinated motion, logically disable the drive, raise and EStop etc).



How I do this is:

- master receive

- domain process

- ecrt_domain_state()

- ecrt_master_state()

- call preSendPDOData() on each of my slaves

- write cached PDO values

- domain queue

- dc sync

- master send

- call postSendPDOData() on each of my slaves

- perform application calcs (writes to PDO data is cached)



preSendPDOData() for my drives caches the just read LastErrorCode (0x603f:00) value and overwrites it with the value of 0xFFFF.

postSendPDOData() for my drives replaces the LastErrorCode value with the cached value (from above) to return it to its just read state.



What this does is sends out a 0xFFFF value in the TxPDO (read) data every cycle.  If the drive is communicating the PDO value will be replaced with the drives current error code value.  If it is not communicating, the value will come back as 0xFFFF as it has not been replaced.  If my "application calc" code sees this value on a drive (along with any other non-zero error code) it will halt any coordinated motion the drive is involved with and perform any other cleanup operations my application requires.



>From the drive point of view, your drive should have its own watchdog to stop what it is doing if it looses communications.  But even if that is not configured / enabled there will be no new target position (assuming cyclic synchronous position mode) reaching the drive so it will stop moving anyway.  We also raise an electrical EStop condition if an active drive looses comms, so a drive in velocity mode will also be stopped when the power is dropped via the EStop.





Regards,

Graeme.




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.etherlab.org/pipermail/etherlab-users/attachments/20180516/398410d1/attachment-0001.html>


More information about the etherlab-users mailing list