Hi.
>
> On 15 Apr 98 at 21:49, Herbert Graf wrote:
>
> > >   I don't see where 6EA comes into it.  Maybe you're talking about
> > > something else.
> >
> >         I believe what the person was talking about is correct. You see, some
> > cards do not properly decode the full address, they only decode certain
> > address ranges present in the original XT. Because of this, some cards
> > might respond to more than one address, it is a common problem on older
> > cards designed under budget. TTYL
>
>   Oh, okay, I see.  Javier's suggestion was based on the assumption
> that the conflict at I/O address 2E8 was an artifact of the original
> PC design decision to use only the lowest 10 bits of the address.
>   But that's *not* what's happening in this case.  COM4, by default,
> wants addresses 2E8-2EF and 8514/A calls for addresses 2E0-2E8,
> and neither provides high bits that would avoid the conflict at 2E8.
>
> David G
>

        Sure? I've posted that the only clear mention of the problem
I've read referred to a port such as 22E8H.
        Indeed, an address range 2E0-2E8 is strange, since it occupies
9 addresses; if it were the case, perhaps it would occupy 16 (2E0-2EF).
All this comes from the use of classical TTL decoders, such as 74LS138,
not seen in today's cards. But one never knows with such intelligent
designers that for saving $0.1/card have introduced so many
incompatibilities. (Probably, cards have been designed by someone in
the financial departments).
        BTW, although COM4 has 2E0-2EF assigned, only 2E0-2EE are
real registers in the classical 8250. I don't know if port 2EF has
ever been assigned some use, but it would be occupied with a
classical decoding scheme (which would provide "air", but active levels).

************************************
Javier Vizcaino. Ability Electronics. [log in to unmask]
  Starting point:        (-1)^(-1) = -1
  Applying logarithms: (-1)*ln(-1) = ln(-1)
  Since ln(-1) <> 0, dividing:  -1 = 1     (ln(-1) is complex, but exists)