Can cheaper lasers handle short distances? | Hacker News Reader

Can cheaper lasers handle short distances?

semiengineering.com

46 pointsPaulHoule15 comments

I am a big VCSEL fan. It makes tons of sense to have a laser that you can manufacture at chip scale. I think that VCSEL and SPAD technology is promising in both telecommunications and lidar. History tells us that the cheaper technology gets better faster than the expensive technology gets cheaper.

I wish the prototype human visible ones would reach production soon.

Imagine a display where each pixel has a red, green, and blue VCSEL. it would be even better than a microLED display!

Imagine laser pointers mountable to any flat surface!

The article appears to contain a mistake:

"The lasers typically operate in one of two bands, C or O. The former is centered at 1,550nm wavelength, while the latter is at 1,310 nm. The O band aligns with an optimal low-loss frequency through the fiber channel. The C-band range doesn’t have quite the same low loss, but it’s acceptable"

This is reversed from reality. The C-band (1550 nm) is where optical fibers have minimum losses, which is why this is the only band used for very long-haul links. The O-band (1310 nm) is where optical fibers have minimum dispersion and slightly higher losses.

Minimum dispersion means that data can be transmitted at high speed and at minimal cost (dispersion mixes adjacent symbols, so it limits the bit rate for a given cable length, unless it is compensated). For the 1550 nm band, the higher dispersion requires additional devices for dispersion compensation, but for long-haul links their cost does not matter in comparison with the savings provided by lower fiber losses.

The 1310-band can be preferable for short links because there is no need for dispersion compensation, but nowadays its main role is to provide a second frequency-separated band for bidirectional communication (upload/download) in residential Internet access through optical fibers.

Also, for long haul, you won't get by without dispersion compensation, so the lower noise from the higher signal levels getting into the amplifiers helps you.

There are limits to the amplifier output power due to how tightly single mode fibers concentrate the beam: if you'd send 1W through a 9μm core, you'd have about 1.5 MW/cm² power density. Even carefully packaged laser diodes break around 12~20 MW/cm² (according to Wikipedia; "catastrophic optical damage"), not to speak of open air patch connectors in a fiber hut.

Don't forget the fiber carries around 100 channels that share this power limit.

Sorry, waveguide? For light? I thought that’s what fibre was.

It is. But it’s not the only possible waveguide. The article discusses alternatives for a system that can transmit 5-10k signals in 32U. If you’re using fiber optic cables, that’s 5-10k pairs, in 32U. That doesn’t work at all. It’s light (ha, pun) on details for what the alternative waveguide is, though.

Just getting VCELs to 25G per lane would be nice, though, if you can match current prices. Upgrading to 100G-SR4 on the same cable plant would be nice.

> If you’re using fiber optic cables, that’s 5-10k pairs, in 32U.

not with 4 or 12 core MCF!

Fair! Would still be a nightmare to manage, though. I wish they had some details on the proposed alternative waveguide.

I think the alternative is shown in figure 2. SiN (silicon nitride) is different from optical fiber. Both versions shown in figure 2 lack detail, especially in the small boxes labeled GC, Mirror, and PD. Depending on the details, one might put micro optical assemblies between or in those boxes in the figure. In any case, SiN waveguides are small, so you can pack many lanes in a small space.

MCF has a bunch of challenges, eg no good pigtail connector, need for rotational alignment, inability to radix (MCF is great for point to point, less good if you want to fan out from one chip to multiple chips), etc

And then even after all that, it’s still 1-2 orders of magnitude lower density than waveguides

> Among the various issues the company says it has addressed are: ...Infant mortality

Huh?

In case you're not just stirring up controversy: thats the fairly standard (if unfortunate) term for failures within a short time of being placed into service.

Time scale varies with context. But, its why you have places that "burn in" new computer equipment for days/weeks.

If you think that's unfortunate, we used to call it "crib death".

Reply on news.ycombinator.com