[evla-sw-discuss] Re: MIB Study Group-Search For the MIB Processor

[Wayne Koski]

Barry Clark wrote:

>What we have been talking about for a MIB is a pretty powerful machine,
>and it is not clear to me that any modules will need to have a second
>microprocessor in them.  The only one I can think of that might is a 
>redesigned ACU.  So I would downgrade the importance of dual ported RAM.

I don't think that it is clear to the hardware people that the MIB is
to be closely integrated with the module.  We are use to a history of
Computer to Interface Device to Module.  Processors in the hardware
such as the FRM, VLBA BBC's, etc. are used to close hardware loops,
and not be concerned with high-level communication protocols while
doing so.  Isolation of functions/tasks game.

I understand this point, but further communications with the system
engineer will be necessary in order to cover this topic.

>I think a watchdog timer is absolutely critical.  If we don't have one,
>we must build in the capability to remotely cycle power to any MIB.

I tend to always see watchdogs as standard equipment nowadays, so
it doesn't trigger my critical need flag as it should.

I intend to have power supervision in any design I might do
so that power perturbations will also provide the MIB with a clean
reset.  I'd like to have a software reset function in the MIB as well
as a possible hardware MIB reset.  Also a commandable User Device
reset should exist. I know people think I'm reset crazy, but I'd
rather design it in and not use it, instead of needing it and not
have it.
 
>I don't know what a SPI interface is, nor a JTAG.

SPI = Serial Port Interface.  A synchronous serial port that can
multidrop.  Consists of a Serial Out, Serial In, Clock, and individual
enable lines.  A multitude of uses such as digital pots, flash, A/D,
D/A, etc.  Rather nice to be able to put an A/D device really close to
the measurment point, and then read the value digitally instead of
having long analog runs that could be perturbed.

JTAG = Joint Test Action Group (IEEE 1149.1)  Also known as
Boundary Scan.  A method by which you can test, develop, etc. devices
on a PCB. Came about to solve test access problems for Ball Grid Array
devices.  You have the same problem with surface mount device, so
JTAG is being built in most modern devices. I'd expect this to be used
in the correlator development and test, as well as NRAO produced
needs.  Another system engineer topic.

>One timer is a necessity.  Two are occasionally useful.  Three are overkill.

Well, I won't turn down a device if it has more then two timers.

How do you see these timer(s) being used? I'm quite concerned about
timing issues and the MIB. 

>We haven't really discussed the philosophy of loading the particular 
>software of the MIB.  Possibilities include:
>    1.  Software for all possible MIBs lives in the on-board flash; MIB
>        decides what sort of MIB it is, starts appropriate tasks.
>    2.  Software lives in on-board flash, OS has routine to easily reload
>        flash if the M/C system detects that it is for the wrong device
>        or the version ID is out of date.
>    3.  On boot, MIB realizes what sort of device it is in and requests
>        M/C system to supply it with the requisite software.
>All have their advantages and disadvantages.  It is quite likely that some
>packages will support one better than the other, so we ought to make up 
>our minds which one we want.

I'd rather explore a bit to see what nabs us processor wise, then
resolve this issue. But, I know that this is important and will be
looking at devices with this in mind.

>But in my mind the driving necessity is ease and compatibility of software
>development.  I suspect that we shall spend as much money on the software
>as on the design and construction put together.

The MIB is where the software and hardware meet.  My concern is that
if we focus too much on software and not hardware, we end up with
something that looks good to the Computing group, but the hardware
group has to do the equivalent of standing an elephant on its nose in
order to get things to work. :^}  And I don't want the software people
to have to do the same.  So I desire to have a compromise such that it
works out for all as best as it can.

-wayne-