Representing Intel is Richard Jones, Senior Principal Engineer. He leads the laser development team at Intel’s Silicon Photonics Product Division.

https://on24static.akamaized.net/event/26/71/51/1/rt/1/documents/resourceList1602546460149/richardjones1602546458264.pdf

One of the key attributes that you need to connect your laser to silicon photonics die is to look at coupling loss. This is one of the challenges around the remote lasers. For integrated lasers at Intel we measure about .5dB coupling loss between the laser and the silicon photonic waveguide. For remote lasers we are estimating coupling loss  to an optical fiber to be about 2dB with a similar coupling loss between the fiber and the silicon photonic chip.  You can tweak these numbers if you can improve the coupling there but you are about 4dB coupling loss for remote lasers which needs to be accounted for and minimized where possible.

My comment: They are estimating losses associated with fiber couplings so they really don’t know and what is that telling you.

So why would you actually do this? It has been mentioned before but thermal environment. So switches are getting hotter as the bandwidth increases independent of your choice of optics whether it is pluggable or co-packaged. The addition of Co-Packaged Optics drives different thermal decisions particularly around the thermal cross talk between the electrical and photonic chips.

Our internal assessment at Intel is that both fan cooled and liquid cooled are possible for next generation switches and that is applicable to ether CPO with integrated lasers or remote lasers.

The performance of lasers varies over temperature.

Slide on the right – key points discussed

The lower temperature of the remote laser. The graph on the right shows measured laser bias required to emit different output power for different temperatures.  The grey bar shows estimated temperatures for remote and integrated lasers. So as you can see remote lasers can work at lower temperatures but the lower temperature does not compensate for the additional 4dB of loss as demonstrated by the circles on the graph.

So it is really key for remote lasers to reduce that coupling loss. Although operating the lasers at lower temperature does not compensate for the additional 4dB of loss as demonstrated by the circles on the graph. So it is really key for remote lasers to reduce that coupling loss. Although operating the lasers at lower temperature does allow them to emit higher output powers.

For integrated lasers the real question is how hot the environment they are working under is. Pushing the laser at temperatures above 100’C will be challenging.

So looking at some of the laser reliability here I am assuming that the mature manufacturing process has been used weeding out premature failure with burn-in. So the focus will be on the random failures that occur during the operating life known as failure in time or FIT rate.

Refer to slide 8 and particularly to slide 9.

Remote laser

Field serviceability allows relaxation of acceptable failure rate.

Integrated laser

If the laser fails then the entire switch package needs to be replaced unless redundancy has been added.