• Hello, Guest! Welcome back, and be sure to check out this post for more info about the recent service interruption and migration.

Building a new LocalTalk (and LToUDP) to Ethernet gateway

cheesestraws

Well-known member
Following on from the 'What should a hobbyist-focussed AppleTalk router do' thread in the Networking subforum, some new prototype boards arrived in the post a few days ago. I haven't brought all the peripherals up yet, but here's a thread to whet people's appetites.

RenderedImage.jpg

This board contains all the hardware needed for a new LocalTalk to Ethernet gateway. It's about the same size as an AsantéTalk.

The microcontroller, in the bottom left, is an ESP32. The prototype board has a version with extra RAM: I'm unsure if this'll be necessary in production units. I managed to order the wrong one, and got one with an onboard antenna, which is obviously pretty useless over all that ground plane. It was meant to be the one with a u.FL socket on it, so that if I decide WiFi is a good idea, an offboard antenna can be used.

The main reason for using the ESP32 (aside from liking the tooling), however, is that it has an Ethernet MAC on board. Above the ESP32 you can see a small chip, which is a 10/100BaseT Ethernet PHY, specifically a Microchip LAN8720. This has really nice documentation adapted to the meanest understanding, that even a software idiot like me can more or less follow. A modern Ethernet interface will give a big advantage over "period" gateways already, simply by being able to do things like MDI-X, autonegotiation and 100BaseT good enough for picky crappy home switches.

To the right of that is the LocalTalk interface, which is only partially populated at the moment. This is essentially the same LocalTalk hardware that's in AirTalk, with a nice pretty filtering network (which is identical to the filtering networks in a period Mac). Below that are a pair of RS485 drivers—unlike certain popular projects, I know what a line driver is for and am not attempting to do this off a GPIO pin of a microcontroller ;-). It's interesting—I found a handful of RS485 drivers that are kind of enhanced clones of older TI chips, offering a wider supply voltage range and some other nice improvements that the parent chip doesn't have. The datasheets are all in Chinese, but the tables are the important thing (and I'm starting to spot phrases like 'unit load' in them now, this is a bad way to learn languages, please send help).

To the right of that is a USB UART. This is intended for use as a firmware flashing interface, but it could also be used as an application serial port, and perhaps support for ARA over serial or something could be implemented. Who knows.

Obviously, this is nowhere near a production product yet. I don't even know if the Ethernet interface works properly yet, let alone how to make it into a nice physical object. But people seemed pretty interested, so here we go...
 

Johnnya101

Well-known member
:OOOOO

That was fast! Can't wait! These will be an absolute hit. No more Asantetalk issues, no more hard to find ethernet cards, etc.
 

mactjaap

Well-known member
Holy cow! Impressive! Really cool. Will follow this thread with much interest. For me as hardcore MacIP fan this is really… IT!
 

cheesestraws

Well-known member
Will you be selling this or will you provide the files to make the PCB?

The nearest thing I have to a plan at the moment is to sell fully made ones but also release the files under some kind of permissive license so people can build their own / build some for other people. As with AirTalk, I want to make the prototyping costs back but I'm not in this for profit particularly.

But I want to get the thing working before I commit myself too much about how other people can get one: expectations management is important.
 

Scott Squires

Well-known member
I take it that you have never programmed a microcontroller before. Because that statement is simultaneously hilarious and frightening. 😄
 

Scott Squires

Well-known member
But to be fair, tashtari, you're a true PIC wizard. Us mere mortals would require 8 times as much memory. So, like 2 KB.
 

NJRoadfan

Well-known member
The Shiva Fastpath has 256k to 512k RAM standard depending on the model, but I'm assuming that K-STAR was all written in 68k ASM. They did manage to cram in MacIP and a complete implementation of EtherTalk Phase 1/2 along with DecNET and RTMP/ZIP.
 

ErinRF

Active member
Exciting! I’m eager to see where this will go!

I remember talking on the IRC channel about off the shelf transformers for direct LocalTalk/phonenet access, I should finish my tinkering with that for a complete solution.

EDIT: Disregard that, it helps to read the entire post before opening ones mouth.

Regarding the wifi module, perhaps a dual design that could accommodate both the internal and external antenna module? Even if just for flexibility in sourcing the parts.
 
Last edited:

cheesestraws

Well-known member
The 512k RAM on the ESP32 is going to make for some pretty tight programming.

That's why that's got the module with the extra 4meg on it ;-). I don't have much confidence in my ability to write an AppleTalk routing stack that'll run in 512k. We can't all be @tashtari...

The Shiva Fastpath has 256k to 512k RAM standard depending on the model, but I'm assuming that K-STAR was all written in 68k ASM. They did manage to cram in MacIP and a complete implementation of EtherTalk Phase 1/2 along with DecNET and RTMP/ZIP.

They were also much, much better at writing embedded code than I am :)
 

tashtari

Well-known member
off the shelf transformers for direct LocalTalk/phonenet access
This is definitely worth doing, possibly for a future rev of this, or AirTalk, or any number of different projects involving TashTalk, where the LT or PhoneNet transceiver and connectors are right on the board. The transformer specs are on page 19 of this PDF, where they make it sound like these are easily obtained... perhaps they were in the 80s, heh.
We can't all be @tashtari...
Well, to be fair, TashTalk only had to translate, it wasn't even necessary to keep the whole frame in memory at once... routing is a much more complicated animal.
 

ErinRF

Active member
This is definitely worth doing, possibly for a future rev of this, or AirTalk, or any number of different projects involving TashTalk, where the LT or PhoneNet transceiver and connectors are right on the board. The transformer specs are on page 19 of this PDF, where they make it sound like these are easily obtained... perhaps they were in the 80s, heh.

Yea I did take a look at that and did a dive into parts suppliers, found a few units that were close, but none that really matched or gave me warm fuzzies. Its going to boil down to doing some direct measurements on a phonenet transformer which is far different from what I can tell, or buying a few close contenders and trying them out in circuit. It shouldn’t actually be too hard to build up a transformer to their spec, plenty of pot core kits out there with similar magnetics, but that’s not all that easy to produce in mass without some gear and a lot of time.

It’s on my list of things to play with.
 

superjer2000

Well-known member
But I want to get the thing working before I commit myself too much about how other people can get one: expectations management is important.
I hear you on this and I am looking forward to seeing the results of your completed throughout tests by Monday or so!
 
Top