Original Post with the start of the conversation: https://www.dslreports.com/forum/r30623834-Just-got-Gigablast-Here-s-the-story-with-pics~start=42
said by uid://1931403 :So it is a sales pitch when all the vendors, CableLabs and other people doing research says that remote PHY gives several benefits? That digital fiber improves snr, which leads to higher modulation and it gives you higher bandwidth?
Back to my provider. They have contracted Huawei to upgrade/rebuild the network. I must believe they are doing it for a reason. Why not fiber all the way, you might ask... They say it will cost them 8 times as much.
Sure, wouldn't be the first time tons of papers have come out, standards have been made, and people have attempted brainwashing to push something that has little to no advantage. There is about 20 new products a month I have some vendor calling and trying to sell me on, that literally do nothing except for do it different "because so-and-so expert says do this" or "such-and-such big vendor is trying it, it must be good". Ive even had these "cmts nodes" pitched at me as well with bold claims like "you can run hundred of customers off this!". Bad news hotshot, that would be a step backwards for me, because I don't even run nodes with that density, not because of bandwidth, but because of noise on the copper plant after that much cascading. Didn't have the heart, or time, to tell him his product wouldn't be a silver bullet to that problem in the real world either, but would actually be worse if he expected me to start combining nodes to get those "hundreds of customers" on his doo-dad.
I think your missing the big picture I've been saying post after post. What improvement? I literally see no discernible difference between a plant with, and without a node involved, and I touch this stuff every day. I work with real life numbers, not with lab cooked white papers, and my numbers say its not there. A plant with no node is infinitely better curbing optical noise than anything every market vendor is "saying" their crap can do, because guess what? There isn't any optics. And I'm not seeing enough of a difference between optical and non optic sections to curb any future modulation profile.
If your provider is seeing "8x as much" of a cost to run fiber back to the headend v.s this, then that is also another reason I won't trust their judgement. They weren't smart enough to build out the fiber right the first time. There is so little cost difference in fiber count v.s. construction, that your stupid for putting up anything smaller than a 48 branch count or 96 trunk count anywhere you build, unless its literally a customer drop.
Regardless of "bandwidth" its still at least 1 fiber per active, if you going to build this like any competent company would. If we are taking advantage of bandwidth and "less fiber" by chaining actives, it is being done wrong. Nothing ruins your day when node 1 dies, and nodes 2 - 48 die as well because of it. May as well have stayed an HFC plant, same problems, but at least there is 50 years of research, hardening, and testing behind the copper side of that system to back up it's stability.
And the big picture is you still have a copper plant to maintain, which is really where your SNR issues arise. That fancy "cmts node" isn't going to do a damn thing for that portion of the system.
Just for giggles next time I fire a new node up I'll hang a modem off it before I cut the plant into the node and screenshot the "optical snr" of the forward, as that would be the only thing determining SNR at that point. It will only be confirming what I saw on the last node we did this on, however, which was a forward SNR of roughly 45dB (Which btw is right at the forward SNR I get when hooking a single modem up to dedicated forward / reverse ports on the CMTS directly, with the correct padding in place to hit the target numbers). Last I looked at the DOCSIS 3.1 specs, that was 7dB over what was needed for 4096QAM.
If it becomes a problem 10 years from now when we are actually liable to even see that level of modulation in mainstream smaller systems, then I'll tackle it. Odds are it will be pointless, because 10G-PON will become cheaper to retrofit in than upgrading all the actives with "cmts nodes", or any form of amplifier for that matter.
::EDIT:: Even better, I have a screenshot of last week's node turn up, that I've attached. This is literally worst case scenario type stuff. This is three miles of plant that has never seen reverse signal. This is a modem hanging off the end of that three miles, with the reverse amps setup "guessing", as the technician hand't yet gone back and balanced the node from where we did the cut in and turned on the reverse laser. This is running on a 4x2 bonded Docsis 3 modem, 256Qam downstream, QPSK upstream (again, QPSK because there is no noise floor from the reverse not being setup, likely there is tons of over and under driving going on between cascades). Either way, the upstream isn't important because that isn't what we are talking about. We are talking about that clearly visible 45dB of forward SNR, through 3 miles of unbalanced plant. I don't expect that noise floor to drop as modems are added to existing subs, because that is not how forward works. There are plenty of video only subs that could be causing ingress already if there was an ingress issue. News subs would be specc'd as they are added, also eliminating ingress. This is what a functional forward plant looks like.
↧