Experimental Physics and
Industrial Control System

David Purcell <[email protected]> · Tue, 18 Mar 2003 15:34:04 -0500

All,
SNS is also developing quite a few PCI based systems, so we would be
interested in in sharing experiences.  There are about 290 IOCs planned for
accelerator instrumentation, with about 20 already deployed.  Ultimately,
270 will be PCI based and 20 will be VME based.  The balance of SNS
instruments (neutron instrument systems) will consist of about 150 PCI based
PCs.  All of these systems are based on rack mounted PCs consisting of both
32 bit (132MByte/sec) and higher performance (>500 MByte/sec) PCI
interfaces.

David Purcell
Spallation Neutron Source Project
701 Scarboro Road MS6473
Oak Ridge, TN  37830
Telephone:  (865) 241-9834

-----Original Message-----
From: Townsend Zwart [mailto:[email protected]]
Sent: Monday, March 17, 2003 6:06 PM
To: Billy R. Adams; [email protected]; [email protected]
Subject: Re: PC-104/Plus as IOC's / PC PCI bus IOC

Dear Bill,
	We have been looking at this same problem and have done a little bit of
work developing a Linux IOC using the PCI bus. Dylan Maher has driven
this effort. Look for him at the next EPICS conference.
	I agree with Bob that longevity may be an issue for PC based
instrumentation. He points out that you can still buy CAMAC hardware,
but how much CAMAC hardware are they using at the SNS? A sensible
requirement seems to be to have reliable instrumentation platform that
will outlive the facility (or application) or be prepared for a
significant upgrade at some future point. Having a plan that includes an
instrumentation upgrade at a ten - fifteen year interval seems like a
possibility. Unfortunately this is somewhat longer than the 5-10 year
time horizon of PC buses.
	Just as important as the bus longevity are the availibilty of
replacement boards. We are currently a little bit short on our VME pepII
bitbus cards which is a big inconvenience for us as they are controlling
a few thousand pwower supplies. Maybe the message should be to buy lots
of spares at the outset.
	We looked at using the PC-104 cards but by the time you put together an
enclosure that will keep the dust out and something to mount it on, you
have spent enough to buy a rack mount PC with ~7 PCI slots. Further the
PC-104 cards have a stacking topology that makes servicing them a little
bit awkward. What if the card in the middle of the stack fails?
	I gather that many people are also shying away from the non real-time
Linux IOC because of concern about lack of real time performance. I do
not understand this. It seems that anything that really requires
real-time should be pushed down into the hardware.
	So far we are quite happy with our two Linux IOC's. They are rack mount
for easy servicing and they have a wonderfully low cost and relatively
robust industrial packaging. This will allow us to expand to a much more
distributed system with relatively little impact. Anything that requires
real-time will continue to be supported in our VME/Motorola/VxWorks
platform. We are in the process of evaluating which PCI cards we would
like to use. The people at Comedi, www.comedi.org, have done a nice job
of assembling a collection of linux drivers for various PCI boards. One
thing that would be nice is to select an established vendor who is
supportive of the Linux platform. We would welcome any comments anyone
else has about their choice of PCI cards.

						Townsend

"Billy R. Adams" wrote:
>
> Bob,
>
>       Thanks for the info!
>       This is the type feedback that I need.  We were considering
> a hardware which used the PCI bus as its interface, but this is
> really not a requirement.
>       Basically, I am hoping to get experience like yours, so
> that I can reduce the number of my mistakes.
>
>       I forwarded your reply-to-me to tech-talk since this could
> help other people as well.
>
> Thanks again.
>
> -------- Original Message --------
> Subject: Re: PC-104/Plus  as IOC's
> Date: Mon, 17 Mar 2003 12:56:58 -0700
> From: "Bob Dalesio" <[email protected]>
> To: "Billy R. Adams" <[email protected]>
>
> What requirement are you trying to meet that makes you
> prefer the PCI bus? Is there a family of I/O in this
> format that you prefer?
> I worry about putting instrumentation on a computer bus.
> If you look at the history of the computer buses, they
> last 5 years and then you can no longer buy boards or
> backplanes.
> I prefer to put my instrumenation in instrumenation buses
> - like vme, vxi etc... You can still buy CAMAC modules.
> So, I wouldn't change the VXI backplanes out.
> If you have I/O boards in them that are not using the VXI
> addressing modes, just the VME addressing modes, you could
> use the VXI crate as a low noise VME crate and buy any
> given VME CPU to put in there.
> At SNS, we are using a Motorola 2100 series Power PC in a
> VXI crate along with an extender and we are seeing no
> problems.
> Bob
>
> On Mon, 17 Mar 2003 14:15:01 -0500
>    "Billy R. Adams" <[email protected]> wrote:
>  >
>  >       Is anyone using PC-104 or PC-104+ cards as IOC's?
>  >Do you recommend them?  Problems?  Advantages?  Sources?
>  >Any other info on them that you can offer?
>  >
>  >       We are considering using these cards to replace
>  >some
>  >KineticSystems Corporation Model VXI 195-21 IOC's as part
>  >of EPICS systems upgrades in the Research Facilities
>  >Branch
>  >at NASA-Langley Research Center.  We would like to use
>  >the
>  >PC-104+ to get the PCI bus, but perhaps the PC-104 could
>  >be
>  >used instead.
>  >
>  >Thanks,
>  >
>  >Billy Adams    <mailto:[email protected]>
>  >
>
> Bob Dalesio
> [email protected]
> 410 557 0297 Maryland
> 505 667 3414 Los Alamos
> 505 699 1632 Cell Phone
> 505 667 6087 Our Secretary (Lisa Marie)
>
> --
> Billy Adams    <mailto:[email protected]>

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System