Date   : Wed, 08 Aug 2007 23:48:17 +0100
From   : C.J.Thornley@... (Chris Thornley)
Subject: Emulating Econet hardware?

You probally have to use a PCI parrelle port as this does not have the ISA
to PCI bridge bottle neck.
To get interupt to work correctly with PCI devices you have to try and
convince your PCI card to use a low interupt IRQ7 or IRQ5 or alternative
(BIOS).

You can then enable the interupt register 
(Control Register 4) Base+2 Bit4

Which will trigger an interupt to respond to request on the acknowlede line
(Status Register 6) Base+1 bit6 at the PCI bus speed 66Mhz.

But the existing chip/module will sooner or later vanish in obsecurity. Only
to be bought at extonate prices on places like flea bay.

Chris


               />      Christopher J. Thornley is cjt@...
  (           //------------------------------------------------------,
 (*)OXOXOXOXO(*>=*=O=S=U=0=3=6=*=---------                             >
  (           \\------------------------------------------------------'
               \>       Home Page :-http://www.coolrose.fsnet.co.uk
 
-----Original Message-----
From: bbc-micro-bounces+c.j.thornley=coolrose.fsnet.co.uk@...
[mailto:bbc-micro-bounces+c.j.thornley=coolrose.fsnet.co.uk@...
uk] On Behalf Of Jules Richardson
Sent: 08 August 2007 09:28
To: bbc-micro@...
Subject: Re: [BBC-Micro] Emulating Econet hardware?

Jonathan Graham Harston wrote:
> You would need to find a +5v line.

One of the internal drive power connectors hung out the back of the case
would do the trick in absence of an external supply, I suppose. (I do wish
an aux power connector had been standard on PCs)

 > As Econet is a background
> interupt device, you would need to find a way of getting a signal from 
> the parallel port to generate an interupt.

Historically, one of the input lines on the PC's port was connected to one
of the interrupt lines (typically INT7 I think).

I believe the plan was for the attached printer (remember that original
ports were uni-directional and just intended for printer use) to signal the
CPU when it was ready for more data. What actually happened though was that
the PC's such a heap of junk, it proved far more efficient to just poll the
port's input lines than it did to go through the whole interrupt process.

> As has been pointed out, interupt latency is the limiting factor, but 
> most PCs nowadays run faster than the 2MHz BBC which managed to grab 
> incoming data adequately.

Although I think they *still* run the parallel port through the ISA bus
emulation layer, which means it's still stuck at ISA speeds, rather than
being a general purpose I/O port running at close to the CPU's speed.

> With a network clock running at 200KHz, up to 25,000 bytes fly past 
> every second. Each byte needs to be removed from the ADLC within 40us.
> The 2MHz 6502 has 80 clock cycles available in that 40us time-slice.
> A 200MHz 80x86 would have 8000 cycles to pull a byte from the ALDC 
> before it was trampled by the next byte.

See above - I don't think the PC ever moved past max ISA speed (which was
around 8MHz I think) regardless of CPU clock. Plus I think it takes a few
ISA clock cycles to actually transfer data over the bus...

cheers

Jules


_______________________________________________
bbc-micro mailing list
bbc-micro@...
http://lists.cloud9.co.uk/mailman/listinfo/bbc-micro