POET is building 2 types of 400G optical engines LR4 and FR4 both at the more advanced 100G Lambda standard(4 lanes of 100G)

The 400G-LR4 transceiver does not use CWDM filtering (no mux/demux) and as such it is a lower cost transceiver but requires 8 fibers (4 pairs at 100G send and receive) So LR4 achieves cost savings only for  shorter distances. There are other factors however that complicates the costing. LR4 can use 100G ports as each of the four pairs of fiber use a 100G signal but the actual reach is also limited compared to the FR4 (CWDM).    

The other 400G optical engine POET is building is the 400G-FR4 which uses CWDM. Four 100G lanes at separate wavelengths placed on one pair of fibers (2 fibers instead of 8 fibers required for the LR4 standard).  In addition to the cost savings associated with less fiber requirements for FR4 POET also adds cost disruption to 400G FR4 by replacing the mux/de-mux TFF –Thin Film Filter (the glass box Suresh spoke of) with POETs optical interposer waveguide filtering to achieve integrated CWDM capability.

46:07 (Suresh): This one I’m particularly proud of. We started the year thinking this was going to be our high-risk item. And we’ve made some progress here. It’s the industry’s first, I believe, FR4 400G compatible filter. On the left is what a traditional filter looks like [slide 23]. This is what is used: a gold box. You can see the gold box there. And in the box there is something called TFF-based demux: that’s a glass block that’s sitting inside the gold box. It’s a fantastic filter. You can see the response of that particular filter at the bottom: it’s nice and flat, nice and box-shaped. There are two concepts here that are important to understand. One is bandwidth, which is how wide the box is. So let’s go back to my analogy of people getting out of a car and going back to a hotel room. Well now let’s assume those guys are now coming back from Oktoberfest. They need a wider door to be able to enter their room. If you give them a really narrow door they’re likely not going to make it. The same thing is true here: the lasers move. So you need a wide window to allow that particular wavelength of light to go through it. It’s called bandwidth. When we started the year we were there: the box that says 2018. You can see how narrow it was. And what we needed to do was take that signal and make it look as flat and wide as the signal on the left. Well people say that looks so good why can’t you use it? We can’t use it because we can’t integrate it at wafer scale. It has to be done one at a time, “? glass block?”, it has to be hermetically sealed, it’s expensive. What we want to do is create a wafer-scale version of that signal in a low cost manner and that’s what that data share shows in 2019, which is a flat wide response that meets the requirements for bandwidth and crosstalk, which is what happens at different wavelengths—you don’t want them to be confused with each other. 

48:19 So I hope you can visually see there has been tremendous progress. Even if you don’t understand it, you can say Hey, that is narrow and pointy, this one is wide and flat. Ok? You want wide and flat and that’s what we’ve been able to do. Or you want wide and flat in filters, not for ourselves. That is important. This was hard. FR4 filters are, of all the filter technology, the most difficult combination of bandwidth and crosstalk of all. The fact that we can do this, we can do other filters now because we now have the knowledge capability to understand how to do it.