Date   : Thu, 15 Jul 1993 08:52:44 +1200
From   : David Andrew Sainty <David.Sainty@...>
Subject: Re: Tube connections (was RE: colour Graphics) 

Date: Wed, 14 Jul 93 10:17:23 +0100 (BST)
From: "I Stephenson" <ian@...>

>>Hmmm, but does the tube ULA mask out the illegal address?
> 
> possibly possibly not, but your best bet is to mask out the extra addresses
in  
> your board, then attach the 2P to that. This will probably be necessary
anyway,  
> as the tube is sensitive about what loads it can drive. If you add a buffer  
> then you an exted the length of the cable, and I would sugguest passing
through  
> NTUBE only for addresses in the range E0->EF, keeping F0->FF for yourself.

Sounds sensible.... I'm sort of keen for software to be able to auto-detect
the extra hardware though, and if the tube misbehaves when writing the
"auto-detect" addresses, that'll cause problems...

>>But what about the other direction? Are the tube registers polled on the IO
>>processor? I always assumed that they were tested on IRQ, in fact, it
>>seems strange that Tube IRQ is an output, as the IRQ line is used for all
>>the hardware that the 2nd P shouldn't care about. Are you sure about that??
> 
> There are no references to IRQ's or NMI's on the circuit diagram that I have.  
> Though its possible that they've been missed out (unlikely), it makes sense
to  
> have NO interupt connections accross the systems.
> 
> Any interupts on the 2P should be generated by hardware on that machine -  
> otherwise what is it going to service! No doubt certain writes to the tube
chip  
> DO generate interupts, but these are strictly internal.

Another assumption I always made was that there was no special tube hardware
on the BBC side of the tube, so any interrupts generated by Beebward
communications would be generated on the 2P on-board ULA, and sent across
the tube connector....

Is this right or wrong?

> The Host can operate via polling, as its not actually DOing anything else.
The  
> timing isn't critical (for most comms), as the tube chip ios designed to
allow  
> the systems to run asyncchronously. Using interupts would lock the two
machines  
> together, and drastically reduce performance

Well, the connection can still be buffered, but an interrupt to say that
the buffer was indeed pending processing would be infinitely useful....

Dave.