This is for PON technology. 1 fibre can be split 32-ways to feed, you guessed it, 32 customers. 50g over a fibre that is split 32-ways with a minimum of 15db loss is impressive.
I guarantee those 100gbps circuits are a single fibre all the way from the provider to the customer. And they are expensive, very expensive.
Kazumara@discuss.tchncs.de 2 weeks ago
I wonder if they use semiconductor optical amplifiers in the receivers, or if they can make do with avalanche photodiodes.
The 100G stuff I’m looking at has 18.5 dB budget with APDs, that seems rough considering you want a few kilometers of fiber, a few splices and a few connectors (probably LC/APC) as well.
Nfamwap@lemmy.world 2 weeks ago
I work on PON and XGPON. Officially we work on a -25dB maximum, but I’ve seen circuits stable at around 30dB.
It’s surprising how many bad splices you can ignore before it gets problematic.
-18.5dB is going to limit you to either a really good fibre path, or a really short one. Unless you have options with long-range SFPs? The constant progress keeps my job interesting at least.
Kazumara@discuss.tchncs.de 2 weeks ago
I’m working on long range stuff so I’m not so familiar with PON specifically. Maybe I made some bad assumptions. Stable at -30 dBm receive sounds really impressive.
The one I was talking about is this, with 18.5 dB total budget, that is, min +4.5 dBm transmit, and min -14 dBm receive. This one is built with an APD.
In my kind of application, without splitter, this will get you about 30-40 km. We’ve got one of a slightly older type with 18 dB budget running between Fribourg and Bern for example.
I realize that PON stuff is quite different with the time slitting and I think wavelenght splitting too, at least in XGS-PON, but I was thinking the pure laser and diode physics would need to be the same.
For -25 dBm minimum the most similar of the ones we currently have would be this one which manages -26.9 dBm and is one of the ones with a SOA built in, or for the 10G stuff this one, which manages min -23 dBm but with only an APD and no SOA.
I’m thinking their 50G stuff must be closer to 100G than 10G transceiver design. So I wonder if they manage to make it without SOA.