[Etherlab-users] DC and oversampling with EL2262
dabbede at gmail.com
dabbede at gmail.com
Mon Feb 24 22:35:47 CET 2025
Dear Gavin,
thanks for your precious help! I had no idea that all the slaves must obey
the same ESC register as Beckhoff's ones... I was assuming that the ESI
file was needed to specify these sorts of things.
I was also assuming that there was a "lower level" (hardware) memory area
with addresses not necessarily aligned with the "higher level" (logical)
registers...
I guess it makes more sense, now, but I'm still in doubt if I miss
something in my "vision" of every piece in the puzzle...
Anyway, so now I downloaded the register manual for Beckhoff ET1100, and I
can confirm that register 0x09A4 is used to set the offset - not the cycle
- between Sync0 and Sync1.
(I was not doubting your explanation, I was just checking if I could find
the same info there, for future reference).
Regarding ecrt_master_reference_clock_time(), you correctly spot the error:
I was calling it after ecrt_master_sync_slave_clocks(). I'm still confused
why it should make any difference where I call the query, but I've learned
the lesson and I'm not gonna move it.
I think I'm almost done, but unfortunately I still think that something
fishy is hitting my configuration... As I previously wrote, I cannot select
the reference clock using ecrt_master_select_reference_clock(). No matter
my choice, the CLI command "ethercat master" always shows that the ref
clock is slave 0 (the coupler EK1100). In theory I would not mind that, but
what bothers me is that if I issue a sync check with
ecrt_master_sync_monitor_queue() and ecrt_master_sync_monitor_process(), I
always find "big" numbers, well above 1000 and often 10000 or more (at
application start, the number is even bigger, so for sure some
synchronization has been performed, just not as good as hoped). At the same
time, if I call "ethercat reg_read -p1 -tsm32 0x92c" from the CLI, I
receive small numbers, in the order of ~10 for slave 1 and 2, while I get
big numbers only for slave 0...
If EK1100 supports DC, it should be able to get below 1000 without any
problem... Is my slave defective? Is there anything else I can change to
force the synchronization to a slave in position > 0?
Moreover: the command "ethercat reg_read -pX -tsm32 0x92c" seems a reliable
way to ensure that DC is properly configured... is there a way to get the
same output within the (userspace) Application Interface?
Thanks again, best regards,
Davide
On Wed, Feb 19, 2025 at 1:35 AM Gavin Lambert <gavin.lambert at tomra.com>
wrote:
> The behaviour of Sync1 is common to all masters and slaves; it’s a quirk
> of how Beckhoff specified the EtherCAT ESC. If you have access to the
> hardware datasheets, it’s documented in the behaviour for the DC
> configuration registers; unfortunately, I’m not allowed to reproduce my
> copy of it here. The Etherlab master simply writes the values given
> directly to the corresponding registers during slave configuration.
> Unfortunately, Beckhoff’s own slave documentation (which should specify
> this sort of thing) only focuses on configuration via the TwinCAT UI and
> ignores other masters. Still, if you dig into the XML files or inspect the
> configuration packets it’s possible to extract the actual configuration
> that it sends, which follows those rules.
>
>
>
> Regarding ecrt_master_reference_clock_time, it will return EIO if the time
> synchronisation datagram has not circulated yet. Ensure that you call it
> before calling ecrt_master_sync_slave_clocks and after receiving
> datagrams. It will likely still error out the first time it’s called.
>
>
>
> As for the “DC reference time” in the “ethercat master” output, it looks
> like that is only updated (based on ecrt_master_application_time) when the
> master is activated; it’s used as a reference to calculate various DC delay
> offsets, not as a representation of the current time. The “Application
> Time” is the current time (still based on ecrt_master_application_time).
>
>
>
> *Gavin Lambert*
> Software Engineer
>
>
>
>
> [image: tomra facebook] <https://www.facebook.com/TOMRA.Food/> [image:
> tomra linkedin] <https://www.linkedin.com/company/tomra-food/> [image:
> tomra twitter] <https://twitter.com/TOMRAFood> [image: tomra youtube]
> <https://www.youtube.com/playlist?list=PLDD3B1A7BAE919EC6> [image: tomra
> instragram] <https://www.instagram.com/tomrafood/>
> *TOMRA Food (ANZ) Limited* | 4 Henderson Place | PO Box 13 516 |
> Onehunga 1061 | New Zealand
>
> Phone: +64 96 34 00 88 | https://www.tomra.com/food
> The information contained in this communication and any attachment is
> confidential and may be legally privileged. It should only be read by the
> person(s) to whom it is addressed. If you have received this communication
> in error, please notify the sender and delete the communication.
>
> *From:* dabbede at gmail.com <dabbede at gmail.com>
> *Sent:* Tuesday, 18 February 2025 10:32 am
> *To:* Gavin Lambert <gavin.lambert at tomra.com>
> *Cc:* etherlab-users at etherlab.org
> *Subject:* Re: [Etherlab-users] DC and oversampling with EL2262
>
>
>
> Dear Gavin,
>
> thank you for the help. The Sync1 specification was completely obscure to
> me. If it is something "general" (not slave specific) I think it should be
> better described in the documentation.
>
>
>
> Regarding my example, with EL3702 I think I'm starting to approach a
> working solution: the slave is OP and oversampled data appears every cycle.
> What I still don't completely understand is why if I run in the terminal
> the command "ethercat master -v" I see that the DC reference time does not
> increment.
>
> To better understand this, I've tried to check the return values in my
> code after calling "ecrt_master_sync_reference_clock_to()" or
> "ecrt_master_sync_reference_clock()" (used exclusively one or the other)
> and they both are 0.
>
>
>
> Then, I tried to read the reference clock using the provided function
> "ecrt_master_reference_clock_time()" and, in this case, I get the return -5
> (-EIO) and the print message "Failed to get reference clock time:
> Input/output error"
>
> In another console, if I check the DC residual error by continuously watch
> "ethercat reg_read -p1 -tsm32 0x92c", I see that the number starts big and
> gradually reduces to ~<100.
>
>
>
> This information seems to contradict each other: on the one end the
> application seems to work, and the time error is minimal, on the other it
> seems that the reference clock is not properly set...
>
> I don't know who to trust :-)
>
>
>
> Have you any idea why ecrt_master_reference_clock_time() could fail? Is it
> an error in the slave configuration or in the master?
>
>
>
> Thanks again, and best regards
>
>
>
> Davide
>
>
>
>
>
>
>
>
>
> On Sun, Feb 16, 2025 at 11:36 PM Gavin Lambert <gavin.lambert at tomra.com>
> wrote:
> >
> > The EK1100 is DC-capable (it has a DC clock and can be assigned as
> reference clock) but does not actually use DC-application functionality, so
> you don’t call ecrt_slave_config_dc on it (thus it doesn’t need an
> AssignActivate word).
> >
> >
> >
> > Note that if you have Sync0 = 1000000 and Sync1 = 1000000 then you are
> not 1x sampling, you are 2x sampling (Sync0 pulses at 1000000ns and Sync1
> pulses at 2000000ns, so your application cycle interval should be
> 2000000ns). Sync1 is an offset; it works differently than Sync0. Blame
> Beckhoff.
> >
> >
> >
> > If you actually want 1x sampling at an application cycle rate of
> 1000000ns then you need to specify Sync0 = 1000000 and Sync1 = 0. For 4x
> sampling at 1000000ns then Sync0 = 250000 and Sync1 = 750000. You may also
> need a non-zero Sync0 shift value; for that you’ll need to check the
> documentation of your slave.
> >
> >
> >
> > Gavin Lambert
> >
> > Software Engineer
> >
> >
> >
> >
> >
> > TOMRA Food (ANZ) Limited | 4 Henderson Place | PO Box 13 516 | Onehunga
> 1061 | New Zealand
> >
> > Phone: +64 96 34 00 88 | https://www.tomra.com/food
> > The information contained in this communication and any attachment is
> confidential and may be legally privileged. It should only be read by the
> person(s) to whom it is addressed. If you have received this communication
> in error, please notify the sender and delete the communication.
> >
> > From: dabbede at gmail.com <dabbede at gmail.com>
> > Sent: Saturday, 15 February 2025 12:29 am
> > To: Gavin Lambert <gavin.lambert at tomra.com>
> > Cc: etherlab-users at etherlab.org
> > Subject: Re: [Etherlab-users] DC and oversampling with EL2262
> >
> >
> >
> > Dear Gavin,
> >
> > thank you for your insight and sorry if I didn't reply earlier... your
> message has been classified as spam and I missed it...
> >
> >
> >
> > Yes, I know that usually the reference clock is assigned to the first
> DC-capable slave: in my simple setupthis slave is the bus coupler EK1100,
> but other EtherCAT masters refuse to assign the reference to a coupler, and
> generally use the first "normal" slave. This is the case for TwinCAT and
> Koenig Pa., for example. Moreover, I don't have an "AssignActivate" word
> for EK1100... Anyway, I've tried both with and without the specification of
> "ecrt_master_select_reference_clock" and the result does not change: the
> reference clock is always Slave 0, and the DC reference time does not
> advance.
> >
> >
> >
> > Since I was having issues on sending EL2262 to "OP", later I've tried to
> use a different slave with similar characteristics (DC capable and
> oversampling), which is EL3702. With that, I could go "OP" without problem,
> but all the other issues related to the reference clock staying assigned to
> Slave 0 (despite ecrt_master_select_reference_clock ) and DC reference time
> which does not advance remained the same.
> >
> >
> >
> > Of course I strictly followed the example dc_user (that was my starting
> point), and in fact my cyclic task has this form:
> > clock_gettime(CLOCK_TO_USE, &time);
> > ecrt_master_sync_reference_clock_to(master, TIMESPEC2NS(time));
> > ecrt_master_sync_slave_clocks(master);
> >
> >
> >
> > I don't know if I'm allowed to attach files to this mailing list. I'm
> going to try with a minimal example and, if it does not work, I'll copy in
> the next message the content of the file.
> >
> > I don't know if anyone of you owns a EL3702 to try replicate the error,
> but this is what I can see with the attached minimal example:
> >
> > - the file compiles properly, and does run "as expected" without
> printing any "if-exception"
> >
> > - the output is
> >
> > Activating master...
> > Using priority 97.
> > Starting cyclic function.
> > 2 slave(s).
> > AL states: 0x02.
> > Link is up.
> > AL states: 0x08.
> > Domain1: WC 1.
> > Domain1: State 2.
> >
> >
> >
> >
> >
> > - the command ethercat config -v gives
> >
> > Alias: 0
> > Position: 0
> > Vendor Id: 0x00000002
> > Product code: 0x044c2c52
> > Attached slave: 0 (OP)
> > Watchdog divider: (Default)
> > Watchdog intervals: (Default)
> > SDO configuration:
> > None.
> > IDN configuration:
> > None.
> > Feature flags:
> > None.
> >
> > Alias: 0
> > Position: 1
> > Vendor Id: 0x00000002
> > Product code: 0x0e763052
> > Attached slave: 1 (OP)
> > Watchdog divider: (Default)
> > Watchdog intervals: (Default)
> > SM0, Dir: Input, Watchdog: Default
> > PDO 0x1b00
> > PDO entry 0x6800:01, 16 bit
> > PDO 0x1a00
> > PDO entry 0x6000:01, 16 bit
> > SM1, Dir: Input, Watchdog: Default
> > PDO 0x1b01
> > PDO entry 0x6800:02, 16 bit
> > PDO 0x1a80
> > PDO entry 0x6000:02, 16 bit
> > SM2, Dir: Input, Watchdog: Default
> > PDO 0x1b10
> > PDO entry 0x1d09:98, 32 bit
> > SDO configuration:
> > None.
> > IDN configuration:
> > None.
> > Feature flags:
> > None.
> > DC configuration:
> > AssignActivate: 0x0730
> > Cycle [ns] Shift [ns]
> > SYNC0 1000000 0
> > SYNC1 1000000 0
> >
> >
> >
> >
> >
> > - the command ethercat master -v gives
> >
> > Master0
> > Phase: Operation
> > Active: yes
> > Slaves: 2
> > Ethernet devices:
> > Main: 00:12:cd:07:a3:4a (attached)
> > Link: UP
> > Tx frames: 2102849
> > Tx bytes: 139702270
> > Rx frames: 2102848
> > Rx bytes: 139702206
> > Tx errors: 0
> > Tx frame rate [1/s]: 1000 1000 995
> > Tx rate [KByte/s]: 64.2 64.2 63.9
> > Rx frame rate [1/s]: 1000 1000 995
> > Rx rate [KByte/s]: 64.2 64.2 63.9
> > Common:
> > Tx frames: 2102849
> > Tx bytes: 139702270
> > Rx frames: 2102848
> > Rx bytes: 139702206
> > Lost frames: 0
> > Tx frame rate [1/s]: 1000 1000 995
> > Tx rate [KByte/s]: 64.2 64.2 63.9
> > Rx frame rate [1/s]: 1000 1000 995
> > Rx rate [KByte/s]: 64.2 64.2 63.9
> > Loss rate [1/s]: 0 0 0
> > Frame loss [%]: 0.0 0.0 0.0
> > Distributed clocks:
> > Reference clock: Slave 0
> > DC reference time: 6591555116393
> > Application time: 6842454116393
> > 2000-01-01 01:54:02.454116393
> >
> >
> >
> > Notice that DC reference time and application time are different and
> that, while the latter updates at each call, the first does not.
> >
> >
> >
> > The original code was more complicated, to account for the oversampling
> also in the PDOs, but even this minimal one with oversampling = 1 still
> does not work, so I don't understand if I'm missing something or if these
> types of slaves are not supported...
> >
> >
> >
> > Thank you all for your patience and for any guidance.
> >
> >
> >
> > Regards,
> >
> >
> >
> > Davide
> >
> >
> >
> >
> >
> > On Tue, Feb 11, 2025 at 4:15 AM Gavin Lambert <gavin.lambert at tomra.com>
> wrote:
> >
> > Typically, the reference clock should be the first DC-capable device on
> the network (often an infrastructure device like a coupler, if present – as
> in your case), because the DC time only flows downstream from that. It’s
> unusual for it to be your DC-consuming device directly unless you don’t
> have any others. Thus, it’s not usually necessary to call
> ecrt_master_select_reference_clock unless you have special requirements
> (such as DC synching across multiple networks).
> >
> >
> >
> > Since you’re using ecrt_master_sync_reference_clock_to and you’re saying
> that the DC time doesn’t appear to be advancing, ensure that your code
> actually does advance the time in your realtime loop – that’s your
> responsibility when you use that method rather than the alternatives.
> >
> >
> >
> > There can be many reasons why a device refuses to transition to SAFEOP
> (or does but then faults back to PREOP). You’ll need to review the syslog
> to see what error codes it’s reporting, if any. (But it could just be due
> to the DC time not advancing.)
> >
> >
> >
> > Try comparing your code against the existing examples such as dc_user,
> and pay close attention to your slave’s documented requirements for
> ecrt_slave_config_dc – in particular note that the SYNC1 time is specified
> in an unusual way.
> >
> >
> >
> > Gavin Lambert
> >
> > Software Engineer
> >
> >
> >
> >
> >
> > TOMRA Food (ANZ) Limited | 4 Henderson Place | PO Box 13 516 | Onehunga
> 1061 | New Zealand
> >
> > Phone: +64 96 34 00 88 | https://www.tomra.com/food
> > The information contained in this communication and any attachment is
> confidential and may be legally privileged. It should only be read by the
> person(s) to whom it is addressed. If you have received this communication
> in error, please notify the sender and delete the communication.
> >
> > From: Etherlab-users <etherlab-users-bounces at etherlab.org> On Behalf Of
> dabbede at gmail.com
> > Sent: Monday, 10 February 2025 9:54 pm
> > To: etherlab-users at etherlab.org
> > Subject: [Etherlab-users] DC and oversampling with EL2262
> >
> >
> >
> > Dear Etherlab community,
> >
> > I have been using the EtherCAT master for a few days. I think I
> understand more or less how to configure the PDOs and domains for one or
> more slaves, but I'm having trouble with the distributed clock (DC).
> > In particular, my setup is currently as follows:
> > - master
> > - Beckhoff EK1100
> > - Beckhoff EL2262
> >
> > The latter is a digital output with oversampling capabilities. It can
> output a sample every Sync0, which in turn can be set to trigger faster
> than the application cycle. I used this slave successfully with other
> EtherCAT masters, so I know it is working properly.
> >
> > In my code I've set the PDOs using "ecrt_slave_config_pdos" and assigned
> them to the domain1 using "ecrt_slave_config_reg_pdo_entry".
> >
> > Then I've set this slave to be the reference clock using
> > ecrt_master_select_reference_clock (master, sc); // Set this slave
> to be reference clock
> > and finally I've set the DC settings (and Sync0/1) using
> "ecrt_slave_config_dc" (by the way, the AssignActivate data is 0x0730 for
> this slave).
> >
> > Then, for every application cycle I call, in order:
> > - ecrt_master_application_time
> > - ecrt_master_receive
> > - ecrt_domain_process
> > - check_domain1_state
> > - ecrt_master_sync_reference_clock_to
> > - ecrt_master_sync_slave_clocks
> > - ecrt_domain_queue
> > - ecrt_master_send
> >
> > The problems I see are:
> > - the outputs are constantly OFF, even if I've set an alternating
> pattern of ON and OFF,
> > - the slave stops at PREOP,
> > - if I inspect the master from the cli using "ethercat master -v", I see
> that the reference clock is set to Slave 0, instead of Slave 1. Moreover,
> the DC reference time is set correctly every time I restart the
> application, but it does not update during execution. Conversely, the
> application time increase as expected
> > - during " check_domain1_state", I see that the wc_state==EC_WC_ZERO (no
> process data exchanged)
> >
> >
> >
> > I am willing to share my code if it is required.
> > Thank you for your help and, especially, thank you for providing such
> powerful software as open source.
> >
> > Best regards,
> >
> >
> >
> > Davide
> >
> > --
> > Etherlab-users mailing list
> > Etherlab-users at etherlab.org
> > https://lists.etherlab.org/mailman/listinfo/etherlab-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.etherlab.org/pipermail/etherlab-users/attachments/20250224/de97f0ba/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: TOMRA_CMYK_final_size_times_two_cd761a01-1d1f-446e-9316-8012271820b6.png
Type: image/png
Size: 6455 bytes
Desc: not available
URL: <https://lists.etherlab.org/pipermail/etherlab-users/attachments/20250224/de97f0ba/attachment-0001.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: TF-FB-icon_b77c57e4-4990-4f9d-b3a2-8e6ab45df7f2.jpg
Type: image/jpeg
Size: 10123 bytes
Desc: not available
URL: <https://lists.etherlab.org/pipermail/etherlab-users/attachments/20250224/de97f0ba/attachment-0005.jpg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: TF-LinkedIn-icon_d54c4829-dcb9-450c-9187-34b26e85ebaa.jpg
Type: image/jpeg
Size: 10214 bytes
Desc: not available
URL: <https://lists.etherlab.org/pipermail/etherlab-users/attachments/20250224/de97f0ba/attachment-0006.jpg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: icons-social-media-twitter_small_2_4bae5ad2-4add-4314-a352-5b317f784956.jpg
Type: image/jpeg
Size: 2256 bytes
Desc: not available
URL: <https://lists.etherlab.org/pipermail/etherlab-users/attachments/20250224/de97f0ba/attachment-0007.jpg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: TF-Youtube-icon_8b2c830c-70d9-48da-a4db-db9191d346ba.jpg
Type: image/jpeg
Size: 10218 bytes
Desc: not available
URL: <https://lists.etherlab.org/pipermail/etherlab-users/attachments/20250224/de97f0ba/attachment-0008.jpg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: TOMRAinstagram_45b30c55-490a-4f32-8fd3-998c152e3494.jpg
Type: image/jpeg
Size: 10303 bytes
Desc: not available
URL: <https://lists.etherlab.org/pipermail/etherlab-users/attachments/20250224/de97f0ba/attachment-0009.jpg>
More information about the Etherlab-users
mailing list