Hacker News Reader

Comment by Uberphallus | Hacker News Reader

The laws of physics didn't change, it's great but suffers most of the problems of IR communication, namely air diffraction/absorption (severely hurting range) and mostly line of sight communication. Through advanced modulation schemes one can reach gigabit speeds but it's such an extremely short range[1] it's no surprise there seems to be no commercial product.

Terahertz signal processing, though, has a much brighter (hah) future ahead, in medical and security imaging[2].

[1] http://www.embednet.com/Giga-IR_General.pdf

[2] https://codeburst.io/terahertz-thz-gap-tech-for-speedy-commu...

View on HN

I'm aware of at least one research project[1] which is currently prototyping Terahertz communications in the range of several hundred Gbps over distances of several meters.

[1]https://www.ee.ucl.ac.uk/research/photonics/eu-2018beyond-5g...

What about interconnects inside a computer?

I’m just speculating. Theoretically if you have vacuum sealed CPU, it could work. This brings other problems. How would you remove heat released by the chips in vacuum if there is no airflow.

That does not make sense.

The distances are so short that you would not have much attenuation from air at normal pressure / humidity.

Also: (not that it matters), but you can remove heat from chips in vacuum through IR radiation and through heatpipes connected to the guts of the chips in the normal fashion (through a heatsink connected to the top of the device).

https://en.wikipedia.org/wiki/Nanoscale_vacuum-channel_trans...

Comes down to another issue discussed in the article: extremely high power and current density, which require another greal deal of engineering to cool down and make sure things don't vaporize, melt or simply break down.

Physics is still a problem there as well. 1 THz = 10^12 Hz. The speed of light is ~3E10 m/s so at 1 THz light can only move ~0.0003 m/cyc = 300 um/cyc. Electric current moves slightly slower than the speed of light through copper and silicon but its close enough to approximate here. That means in a CPU running at 1 THz it would take multiple clock cycles for a signal to move from one side of the CPU to the other. You also have to take into account the switching speed of transistors and the distance from the CPU to memory. All together, you would just end up wasting a lot of clock cycles achieving nothing and driving the current required to run a CPU at those speeds would generate a lot of heat.

Reply on news.ycombinator.com