With some basic SERDES knowledge I wonder what could be done in 2022, now that we have cheap FPGAs. Ronja is limited in that it piggybacks on 10BASE-T Ethernet, which in turn makes in simpler. It wouldn't be easy to analogously piggyback on 100BASE-T due to PAM-3 transmission. An FPGA would allow for a bespoke PHY/MAC to be implemented in lieu of Ethernet.
There are many techniques for pushing bits into a wire faster, like CTLE, DFE, pre-emphasis, to name a few. Probably the best starting point would be Behzad Razavi's "Design of Integrated Circuits For Optical Communications".
For starters, the LED driver could use current switching with some idle current (pre-bias) in "0" state to achieve more bandwidth.
As I see it, RONJA's nature as a hack of ethernet was one of its benefits, it could be integrated with ordinary ethernet gear.
I think if one was to try to develop a followup, basing it on one of the fiber variants of ethernet seems natural enough considering the idea is more or less fiber without the fiber.
In theory up to 25G ethernet could be attempted with something like this, anything bigger than that uses multiple lanes or multilevel encoding, but obviously in the real world the practical limit for something that could be built at a DIY level is maybe gigabit realm.
Yep. Could see a low cost FPGA with a being used for a new, more advanced ome with better linecoding, FEC and DSP on the receive side.
And all that cheaper than the original RONJA was back in the day.
As someone who recently started looking into free space optics for communication, seeing this suddenly pop up here is a bit surprising! This is the project I pointed at when my friends asked if you could do fast communications with visible light.
The project has cataloged several installations (http://ronja.twibright.com/installations.php) but I suspect there's many more that haven't been registered. I'm curious if any of these systems are still in place.
We built and used several ones. Much better than 802.11b wifi links in 2.4 GHz, but finicky, pain to align, horrible mechanical design, requires super-stable mast, prone to bad weather, and also stopped to work during dawn/sunset, if the link was aligned with sun :-).
We mostly abandoned them when 802.11a wifi links in 5 Ghz came. Later, we bought some 100Mbps professional FSOs (free space optics) that were much better than Ronja, but still more problematic than radios.
Can’t imagine a use case. I’m sure they exist. But I’ll bet I could count them on fingers and toes. And anything truly practical would require active stabilization of the aiming.
Small non-profit wireless ISP. In early 2000s, when only affordable alternative was 2.4 GHz wifi, which was already heavily congested and have low throughtput and 100ms latencies under load due to collisions and retransmissions, the 10 Mbps full-duplex optical link with low latencies and immunity to interference (outside of sun :-) ) is pretty good deal.
My father once brought this device home. I saw computers using it pinging, but having a look at the hardware was awesome in its own way (I was around 6, 7?). Very sophisticated and at the same time simple.
It makes me happy and sentimental to see this post show up on hackernews. I was a contributor to the Ronja project in my teens. Karel (`clock`) taught me so much about open source and truly believed in pushing open technology forward. I was 14 at the time and he just finished university and taught me a ton mostly late at nights on the swiss linux user group's IRC channels.
A lot of the replies point out that cheap PCBs were not readily accessible. Clock and I were both living in Switzerland and prototyping PCBs especially in Switzerland was very costly. This kept us away from developing PCBs for a while.
We did want to build a 100mbit version and that was significantly easier using SMD components and a multilayer PCB so we started working on that. When we had the first prototype I actually took a train to Prague when I was 14 years old and went to a PCB fab to pick up 50 or so along with all the parts. This saved us a few thousand dollars and we tried to recoup the cost by selling kits. Fun times :)
The philosophy of the project was also to build it only with open source projects (something that wasn't feasible for the PCB design - we used Eagle which at least was free to use). The dead bug style worked very well and was extremely simple but of course not aesthetically pleasing when compared to what you can order online today. But it also meant that you could build it yourself without the need to order a PCB.
Dead bug style circuit assembly has been popular in the ham community for a long time. It’s definitely a legacy thing but still quite common for hobbyist projects.
It’s quite suitable for specific RF situations provided you have done the right characterization work.
1) First and foremost, the board or box the components are soldered to forms a very good ground plane
2) The parts are organized so that the circuits fit the enclosure
3) The part leads are used to provide relatively rigid support for the parts
4) Where they can't, it looks like the builder used epoxy
You could argue that the builder should have cleaned off the solder flux and you'd probably be wrong, because you probably don't know what a pain it is to really clean off solder flux to the standards of high reliability circuits.
Folks that are used to breadboards and perfboard construction have a hard time appreciating the care that goes into good dead-bug assembly. That is probably a fairly high performance circuit and is probably much more rugged than the typical perfboard sewing card.
Initial RONJA prototypes and versions ware built before the time of cheap chinese pcb fabs.
They would have likely had to pay between 150 and 500eur per pcb round.
We live in such nice times these days for electronics design that even hobbyists can afford multiple prototype rounds with assembly.
Some do, but it's stupid. There's no reason to do it that way other than laziness. And even that's the wrong word as that's much more difficult than using a board.
You don't get a ground plane with perfboard, unless you submit yourself to the pain of working with vectorboard, and perfboard construction is inherently more fragile than dead bug.
If you get good at dead bug construction, you can make changes and do experiments as easily as with a solderless breadboard.
There are many techniques for pushing bits into a wire faster, like CTLE, DFE, pre-emphasis, to name a few. Probably the best starting point would be Behzad Razavi's "Design of Integrated Circuits For Optical Communications".
For starters, the LED driver could use current switching with some idle current (pre-bias) in "0" state to achieve more bandwidth.