Design details Was:Re: [evla-sw-discuss] RTOSes ?

[Bill Sahr]

John,

  I should be doing other things right now, but I saw this posting
and could not resist taking at least a few minutes for a reply.
Your questions are good ones, and I will do my best to give some
quick answers which can be elaborated at a later date.

Bill

John Ford wrote:
> 
> Kevin Ryan writes:
>  > Thank you for your comments John.  I don't believe we've met; my
>  > name is Kevin Ryan and I'm a programmer in the ASG group.
> 
> Hi.
> 
>  >
>  > At the risk of straying away from the multiple RTOS subject, I
>  > would like to explain some of where we are with the MIB.  Keep in
>  > mind that nothing is set in stone and lively debate is ongoing.
> 
> Thanks for the info, it was very enlightening.  I've gone and read the
> rest of the archives, but much of what you wrote was not in there
> anyway.
> 
> <snip>
>  >
>  > So, no matter how we do the second phase, our MIBs will have to
>  > be at least complex enough to support the frontend and backend
>  > drivers - with the Ethernet frontend being the most demanding.
> 
> Indeed.  I would think that the communications protocol will take most
> of the cycles of the MIB if you use a real TCP/IP stack.
> 
> I have so many more questions that I will go to the Web documents and
> read them for answers, but here are a few things I first thought of:
> 
> Why Ethernet?

In the beginning.  (How's that for a snappy opening ?  However,
I think it's been used before.)  We did a survey of fieldbusses.
That was several years ago.  We found nothing we really
liked.  CAN was considered most strongly, but as people in
the Electronics Division began to work with it for ALMA,
they began to request that we look elsewhere.  Most
of them found what they felt to be inefficiences concerning
payload size & overhead to be objectionable.

We feel there are enormous benefits to ethernet, most of them
issues of pragmatics.  It won't go away anytime
soon, there will be upgrade paths for many years to come,
the hardware is widely available and benefits from volume 
pricing, there are many people with the requiste knowledge, 
givving us a large pool to hire from, there are many middleware
and "software infrastructure" type packages which use 
ethernet as the underlying wire protocol (not particularly
relevant to the MIBs, but important for other components
of the system), and, if I thought longer & harder, I'm
sure I could come up with more reasons.  Interestingly,
at the earlier stages of the decision making, it was the
hardware designer who worked on the CAN based AMBSI-II
board for ALMA that spoke against CAN and in favor of 
ethernet.

> 
> How are you planning to connect up all 50-odd of these MIBs in each
> antenna into a network without RFI and without breaking the bank?

We are worried about RFI, and the use of fiber within the antenna
has been mandated, cheifly from concern for RFI.  It will be
expensive, but the money is in the budget.
> 
> Isn't deterministic performance needed for the MIB->control computer link?