NSFCAM IC Computer Upgrade.
May 21, 1998
Tony Denault

The goal of this upgrade was to replace the NSFCAM IC computer's operating
system from LynxOS ver 2.2 to Linux. LynxOS, a 1994 vintage OS, only
supports obsolete PC hardware (disks, ethernet cards, isa bus) and does not
support newer PC hardware standards (EIDE, PCI). Also the development tool
provided were very basic and limited. This limits me from making any
improvements in the software.

Rather that upgrading LynxOS, I decided to port the IC applications and
device driver to Linux. The current Linux is superior to the current LynxOS
in terms of supported hardware, available software and development tools.
It's only draw back is that the kernel is not a design for real-time
application, but rather follows the traditional Unix timesharing OS
standards. 

A new PC was purchased and the application ported. The change over was
performed on May 18, during a week NSFCAM was not scheduled to be used for
observing.

Basically, we removed the nsfpc from MIM, transferred two DSP boards, their
array electronic interface boards, and the PC38 to the new computer.
Installed the new computer on MIM and started testing the system. Two major
problems were encountered:

1. We were unable to image the array. 

This was a major source of frustration, as DSP/Interface board imaged in
the old computer but not in the new one. This problem was very difficult to
diagnose and eventually involved Lars, George, and Peter  O. to pin point
the culprit. It seems the clock DSP has trouble initializing its array
interface board in the new computer. I added a redundant call to initialize
the interface board into the clock DSP program. This seems to works. We
(Lars & I) don't have confidences in this board and feel it is a potential
source of problems. 

2. The GPS timer board didn't fit in the slots. 

Although I did a board count and we only needed to transfer 6 ISA card ( 2
DSP, 2 interface boards, PC38, GPS), I couldn't fit these boards into the
new computer's 9 isa slot due to the fact many of the boards are wide 
and interfere with an adjacent slot. I am currently running an network 
time protocol software to help the PC keep accurate time.

Summary:

Overall the change over was successful. The newly ported IC program is
version 2.0. I tried to make this very similar to the LynxOS versions, as
this transition shouldn't change any staff or observing procedures.
Hopefully it's 'transparent to the user'. I did make one change in the 
DSP device driver to hopefully improve the lost interrupt problem.

The XUI application has not changed. 

Two new issue still needs to be addressed:

	1. The problem with the clock DSP & its interface board.
	2. The GPS board.