exxos's Atari, Amiga & retro forum

But how does it differentiate between accesses to the RTL8019 and to the USB chip? At least the circuit to generate /IOR and /IOW must be different. I'd look at that...

The access is done the same way using the same 74ls245 and also using a 74ls374. It uses one of the address bit to select between the RTL8019AS and the USB controller.
I can post more screenshot of the schematics later today.

rpineau wrote: ↑Mon Apr 30, 2018 2:57 pm The access is done the same way using the same 74ls245 and also using a 74ls374. It uses one of the address bit to select between the RTL8019AS and the USB controller.
I can post more screenshot of the schematics later today.

The address decoding to select between Ethernet and USB controller adds further delay. Especially because LS is not a particularly fast logic family.

I agree. May be they should be replace with F or ABT versions.

another screenshot that shows more of the selection logic.
It would be easier if redone in Eagle for all of us so that we can potentially improve the design.
RoRo

I've been giving this whole approach of cartridge based ethernet some thought and I'm not convinced that if someone were to do something today they should possibly go this route at all.

Surely now it's "cheap" and easy to get Arduino boards/chips a better approach would be to put some intelligence onto the interface to run the ethernet chip and then communicate in a more simple method with the Atari, possibly through the bi-directional parallel port which also provides the possibility of using interrupts to do data reads. The Strobe/Busy lines could be used to set the data direction and data being passed to the board would be pairs of metadata+data byte pairs sent serially.

I wonder how fast this sort of interface could run.

It's all moot as I don't have the skills to do any of it. Just a hypothetical pipe dream.

If I were to build a network card for the ST I would use the (will use ) something like the Microchip LAN 9218 ( http://www.microchip.com/wwwproducts/en/LAN9218 ) on the CPU bus.

I'm afraid that anything which doesn't use one of the standard ports and requires modification of the machine wouldn't interest me in the slightest. Changing the TOS version or replacing the floppy chip with an Ajax one on machine types which had those fitted at the factory is as far as I will go as far as modifications.

So, it's either kludge writing to the cartridge port without interrupts or try to use the parallel port in an inventive way with some off-load. Of course, on the TT and MegaSTE there's VME but I don't know how complex the interfacing for that would be.

There is (was) a version using the DMA port (ethernea) if I remember well, which would be faster than the parallel port.

Based on the ASCI schematic above and the NetUSBee schematic, may be we can combine them to build something like this :

If someone what the Eagle file to double check what I did I'll be more than happy to provide them.
Then we can build a proto and see how it goes.
The ASCI /IRQ doesn't seem to be used so not sure how fast that thing would be without proper interrupt.
Rodolphe