Spectranet - ethernet for your Spectrum

[Gulliver]

I finally took the plunge and got a Spectranet ethernet adapter for my Sinclair Spectrums (I have a 48K "rubber devil" and a 128K toastrack these days)

I have to say, I am impressed. The unit worked right out of the box, connecting to the manufacturer's server which is pretty useful itself. Though I found Mark Round's site even better. And there's even Platoterm - a Plato terminal with access to much of the old Plato material from the 70s and 80s. I enjoyed playing Plato at battleships for the first time in 41 years.

The humble Speccy is online. I never once thought I'd hook up a Spectrum to the internet! There's even an irc chat client and a telnet client! But most of the pages designed for the spectrum are a bit of a cross between the old BBS and modern web pages. It's rather like surfing the web on your Spectrum.

[ortek_service]

I got one of those Spectranet assembled-PCB from the original Designer, Dylan Smith, (back in 2013 at the Silicon Dreams Computer-Show Weekend, held at Snibston Discovery Museum in Coalville, near Leicester) - Who I believe also gave a talk about it, (alongside Chris Smith's ZX Spectrum ULA Rev Eng talk & selling of the book Chris wrote describing the Spectrum ULA & the Rev-Eng process he used)

But I've never got round to trying it, so there might be some firmware updates since then. And I see you can now get cased versions: https://www.bytedelight.com/?page_id=3515
So I wonder if it's possible to buy a case, separately. Or can download a file to 3D-print one.

It's surprising that it's claimed to be about the only Spectrum Ethernet interface. I remember someone I know showing me the Spectrum PC ISA interface he'd designed & built on veroboard, that allowed him to connect an old PC Network card to it.But not sure how far he got, with creating software to support this, and I can't see anything about it on his website: https://www.mike-stirling.com/
But he was involved in the early days of reverse-engineering how parts of the ULA worked, to create an FPGA-based emulation of it.

[paulsherwin]

I've no personal experience of this, but I'd imagine implementing a TCP/IP protocol stack on a Spectrum would be more of a challenge than providing a raw packet level interface. An impressive achievement though.

[ortek_service]

Originally Posted by paulsherwin

I've no personal experience of this, but I'd imagine implementing a TCP/IP protocol stack on a Spectrum would be more of a challenge than providing a raw packet level interface. An impressive achievement though.

I think on this particular one, that much of this is handled by the interface board, being as well as the WIZnet iEthernet WS100 Ethernet 'PHY' interface, it also has these IC's:
Xilinx XC9572XL-TGC100 100pin TQFP FPGA +
AM29F010B-90 1Mbit FLASH and an
IDT71024S12YG 1Mbit 12ns SRAM

So may be a processor core running in the FPGA, with rather more computing power than the Spectrum's Z80, that this is off-loaded to and doesn't present much overhead to the Spectrum's Z80.

[cmjones01]

Yes, those Wiznet chips are really handy. They have the Ethernet interface and TCP/IP stack all built in, so the host microcomputer only has to say "open a socket to this address please" and it will. They make implementing internet applications on an 8-bit micro quite straightforward.

All the smarts on the Spectranet must be in the Wiznet chip. The XC9572 is a very old CPLD which doesn't contain much logic - just enough to make a few counters and gates, so it's probably used as glue logic, or maybe to create an SPI interface to talk to the Wiznet chip (though the W5100 has a parallel bus interface), as well as doing some bank switching to allow the RAM chip to be used (more memory comes in really handy when doing network stuff).

Chris

[Michael Haardt]

Originally Posted by paulsherwin

I've no personal experience of this, but I'd imagine implementing a TCP/IP protocol stack on a Spectrum would be more of a challenge than providing a raw packet level interface. An impressive achievement though.

The question is how minimal you want to go. uip and lwip have very different memory requirements, but both could work. See

https://github.com/adamdunkels/uip/b...uip-refman.pdf

for uip. You had to link the stack into the application, which means if the application ends, the system would not respond to a ping any more, much like NCSA telnet on DOS back then. uip needs an interrupt and a timer to handle timeouts.

Michael