Re: Opinion on timing system for synchrotron beamlines

[Márcio Paduan Donadio <marcio.donadio@lnls.br>]

    Sorry, I didn't send this answer to tech-talk, only to David.

-------- Mensagem encaminhada --------

Assunto:	Re: Opinion on timing system for synchrotron beamlines
Data:	Fri, 11 Nov 2016 21:50:19 +0000
De:	Márcio Paduan Donadio <marcio.donadio@lnls.br>
Para:	David Vine <dvine@lbl.gov>
CC:	Márcio Paduan Donadio <marcio.donadio@lnls.br>

        Hello, David.

    Personally I see no problem to send you our document, but I need to check this with the head. But there are some issues:
    - It is not ready yet, so, it is incomplete and parts are only sketched.
    - It is in Portuguese.
    - As we already use EPICS in beamlines since 2010, this is not the focus of the document. The focus is to describe requirements in terms of frequency to acquire sensor values, encoder values, and frames. Also to describe minimum jitter, response time to trigger events, amount of generated data (not considering detectors data) and data transfer rate. The document describes the type of encoders and sensors that need to be read, types of trigger signals and other inputs and outputs. It also proposes a test case that needs to be accomplished as we can define if we are satisfied with the solution. But EPICS in fact is barely cited.

    Do you think it could be useful even considering those issues?

    Best regards,

-- 
Márcio Paduan Donadio
Engenheiro de Computação - Software de Operação das Linhas de Luz
Laboratório Nacional de Luz Síncrotron – (LNLS)
Centro Nacional de Pesquisa em Energia e Materiais (CNPEM)
+ 55 (19) 3512-3520 - marcio.donadio@lnls.br
www.lnls.cnpem.br

Em 10-11-2016 14:53, David Vine escreveu:

Hi Marcio,

I am contributing to the conceptual design for the Advanced Light Source upgrade project at Berkeley National Lab. One item we are considering is adopting EPICS for the beamline control system. I see you mention you developed a requirements document for the beamlines. We are going to make a similar document and I would be very interested to learn from yours. Is this a public document, and if so could you send me a copy?

Best regards,
David Vine

On Thu, Nov 10, 2016 at 8:47 AM, Márcio Paduan Donadio <marcio.donadio@lnls.br> wrote:

        Hello, Bob.

    We are in the project phase of Sirius' beamlines control system. We did the requirements gathering with the researchers and no one cited the synchronization with the accelerator timing system, with one exception: the pump and probe experiments. So it seemed that it was a special case.

    But some time ago I was evaluating a Zebra Box and asked a friend that works with beamlines in the USA to comment about his impressions on this system. Among some things he told me that Zebra Box lacks from direct connection with the EVG and that is an important feature because he needed to get the global timestamp from it.

    So, I thought that, maybe, we don't know yet that we will need to synchronize events during some other experiments like quick exafs, fly scan mapping, and tomography. That's why I'm talking to you guys, that have more experience on beamline's timing system.

    Thank you very much,

-- 
Márcio Paduan Donadio
Engenheiro de Computação - Software de Operação das Linhas de Luz
Laboratório Nacional de Luz Síncrotron – (LNLS)
Centro Nacional de Pesquisa em Energia e Materiais (CNPEM)
+ 55 (19) 3512-3520 - marcio.donadio@lnls.br
www.lnls.cnpem.br

Em 09-11-2016 17:52, bob dalesio escreveu:

First a disclaimer: The actual EPICS developers always have a better answer about the technology as is. 

That said, I will take a shot at it:

Hi Marcio,

This is really a question that starts at the beam line.

What do they want? Do they care about the energy? at what precision?

There are independent timing systems implemented in FPGAs with different specifications. If you cannot buy hardware, working with this firmware allows you to build your own.

EPICS time stamps can be derived from different timing sources. If you use PTP, as I understand it, you have to purchase a master as well (is that right?). So, it is important to know what your beam lines want and your cost constraints.

EPICS itself in version 3 has time to the nsec. The MRF timing hardware time stamp resolution is 8 nsec. The jitter is < 10 psec. It can easily be extended to the beam lines. Slow signals, like beam energy, can be sent over the timing system along with time stamp information that could be put into the metadata for the experiment. LCLS II timing has sub nsec time derived at the EVR from the EVG (both homemade at SLAC). PVAccess has 64 bits for subsecond instead of 32 bits. It also has a user defined extension for pulse ID or other user defined identifier.

- Bob

On Wed, Nov 9, 2016 at 2:26 PM, Márcio Paduan Donadio <marcio.donadio@lnls.br> wrote:

     I'd like to know your experience regarding integration of the
accelerator timing system with the beamlines:
     - Which accelerator events are useful for the beamlines' control
system.
     - Do the beamlines use a global and very precise timestamp sent
from the accelerator timing system? What is the time precision?
     - Do you think that a PTP system is sufficient for timestamping
experiments in beamlines?
     - What is the EPICS role regarding such a timing system?

     Thank you,

--
Márcio Paduan Donadio
Engenheiro de Computação - Software de Operação das Linhas de Luz
Laboratório Nacional de Luz Síncrotron – (LNLS)
Centro Nacional de Pesquisa em Energia e Materiais (CNPEM)
+ 55 (19) 3512-3520 - marcio.donadio@lnls.br
www.lnls.cnpem.br