Hacker News Reader

Comment by CharlesW | Hacker News Reader

> Next, each of these channels was split into 223 data chunks that existed in individual slices of the electromagnetic spectrum.

As someone who's naive about fiber-optic communications, I'm very surprised that this was not already happening. Is it that the idea is new, or that it's just recently become possible/practical?

EDIT: It looks like the idea itself goes back to at least 1980: https://ieeexplore.ieee.org/document/4102348

View on HN

As others have said it's not the concept of DWDM that's new here. That has been around for a while. What is interesting here is that they have managed to integrate a DWDM transceiver onto a single silicon-photonic chip. As someone who has worked in similar silicon-photonic research, the challenge is to create the required filters, modulators, and demodulators that are stable enough over environmental conditions and manufacturing process variation for commercial production. Most silicon photonic structures are quite sensitive to temperature in particular.

I worked with glass fibre multiplexers that used multiple colours of might back in 2000. It's called DWDM. I moved on to different things so I don't know if it's still used, I assume so because it had no real drawbacks. But it was definitely a thing used on many expensive long distance fibres.

On the short ones it wasn't used because the equipment was more expensive than just laying more fibre.

Whatever happened with polarized light? I recall at one point them deciding it needed either a different chemical formula for the fiber or tighter tolerances on the surface? Either way the problem was that yes they could shove more bandwidth down the pipe but you needed to run new fiber to get it. I didn’t hear anything after that.

If you have to run more fibre, there is no point. Because the main cost of running fibre is the work, not the material. And it's so thin you can run as many as you want.

So, if you're going to run new fibre anyway, just run 10x as much as you need. In fact this is pretty standard procedure with most telcos.

Not sure about the polarisation by the way, when I was working with this it was not yet a thing but it's decades ago.

Managing all those frequencies on a single solid state chip is the novel part here, based on my (admittedly dated) understanding.

Reply on news.ycombinator.com