CCD-world: Optical Link in CCD Controller

[Bruce Atwood]

The following was posted to CCD-world:

We have been using AMD (Advanced Micro Devices) TAXI chips
(AM7969/AM7968) for the parallel to serial conversion and HP fiber parts
(HFBR 2416/HFBR 1414) with ST connectors for over ten years.  We have a
120 Mbit/sec bi-directional link which is fully synchronous (only one
clock in the whole system).  The circuitry between the TAXI's and the
fiber parts is from the HP's Application Bulletin 78.

Yes we do have the "sequencer" at the computer end of the fiber link
(hence the need for high speed in both directions)  We don't need a
reset in the telescope mounted electronics because it just does what it
is told to do by the sequencer data coming up the fiber.

Bon chance.



DE KAT Jean DAPNIA wrote:
> 
> The following was posted to CCD-world:
> 
> Hello everybody,
> 
> as discussed in the ESO WorkShop, I find that will be great to present in
> this forum some solutions and not only questions. (Maybe this would be more
> efficient than some WorkShop roundtables ...)
> 
> Then I start with a common problem for all CCD controllers developpers i.e.
> the optical link to evacuate the CCDs data.
> 
> I have recensed in the last 2 years the available optical links in the
> market because I didn't want to develop a new one for MegaCam (for time to
> develop and maintenance reasons).
> 
> I needed a performant link (100MB/s), easy to implement in my custom CCD
> controller, without any high level software driver (as in FiberChannel,
> gigabitEthernet, ...) and directly pluggable into commercial VME boards.
> 
> I found only 2 links like that. One from Systran is called the Simplex Link,
> has a PMC form factor at both side and implement a FPDP->Fiber->FPDP
> protocol. (FPDP = simple parallel communication protocol).
> Unfortunately, this products seems virtual since it exists only on the web
> for many months ...
> 
> The other is the SLink that I finally adopted. We plan to plug it into a
> Motorola PPC board (VME board with PCI as internal bus).
> 
> Here is a description of SLink:
> 
> - SLink was developed at CERN (Particle Physics in Geneve). When we need
> some optical links for our controllers, they need thousand of links for some
> experiments...
> - SLink won't vanish in few years; it is now sold by private companies and
> the CERN experiments will need it for at least the next 10-20 years.
> - SLink is not a common standard. Only the mechanics, the connector and the
> protocol on this connector is defined. That allows users to benefit from
> future laser/fiber/coding... technologies by just unplug the old link and
> plug a new one.
> - SLink allows transfers on fiber at rates > 100MB/s in one direction. There
> is a second fiber but is used to secure dialog between emitter/receiver. It
> can however be used but at a lower transfer rate. That doesn't matter in
> most of our controllers where we have to download few commands to the
> controller and upload large CCD data (except for those who believe in remote
> CCD sequencing ...).
> - On the controller side, SLink is seen as a synchronous 32 bits Fifo. That
> is to say that the controller developer has just to implement somewhere the
> SLink connector, to reserve the space for a SLink daugther board (PMC form
> factor) and to interface its system to a 32 bits Fifo.
> - On the other side of the fiber, we have to plug a same SLink board on a
> adaptator. Adaptator for PMC, PCI, VME exist. A more compact board with a
> SLink board merged with an PMC adaptator exists too.
> - Various software drivers are already developped (except mine ... : we have
> had to develop our driver because it didn't exist yet (PMC, VxWorks).
> 
> To summarize, SLink is easy to implement, performant, and there is nothing
> to develop in VME crate.
> The only drawback I found, is that you can't use it to send a reliable reset
> to your controller. You need another channel just for that.
> Note too that, in most of cases, the SLink will not be the speed limit of
> the whole system. In my system for example, the PPC board is the 1st limit
> with 89MB/s on PCI bus.
> 
> I think that SLink could be a good solution for most of us.
> 
> For more info, you can see the WebSites:
> http://www.cern.ch/hsi/s-link/
> http://www.systran.com/fibrextreme.htm
> 
> But maybe someone has another solution ?
> 
> Another question:
> Is anybody happy with remote CCD sequencing (sequencer at the other end of
> fiber) ? What are the performances in term of sequencing speed,
> synchronisation, power dissipation ?
> 
> jdK
> 
> ******************************************
> jean de Kat - Electronics Engineer
> CEA-Saclay, Dapnia/Sei
> 91191 Gif-sur-Yvette, FRANCE
> Tel: 33 1 69 08 85 05   or  33 1 69 08 62 81
> Fax: 33 1 69 08 31 47
> Email: jean.de-kat at cea.fr
> *********************
> - -- CCD-world -- --
> CCD-world is fully moderated. Send posts to CCD-world at astro.ku.dk
> Standard replies will go to the list; address personal replies manually.
> For more information, please go to:  http://www.not.iac.es/CCD-world/

-- 
Bruce Atwood
Director
Imaging Sciences Laboratory
Astronomy Department
The Ohio State University
174 W. 18th Ave.
Columbus, Ohio  43210-1106
(614) 292 6279 (V)
(614) 292 2928 (F)
- -- CCD-world -- --
CCD-world is fully moderated. Send posts to CCD-world at astro.ku.dk
Standard replies will go to the list; address personal replies manually.
For more information, please go to:  http://www.not.iac.es/CCD-world/