Hacker News new | past | comments | ask | show | jobs | submit login
Why does gigabit internet via coax cable not offer symmetrical speed like fiber? (superuser.com)
354 points by personjerry 9 months ago | hide | past | favorite | 113 comments



I do quite a bit of DOCSIS consulting.

There are both technical and business reasons for the gap between upstream and downstream speeds

There are a few technical reasons. There is less upstream frequency available in the upstream. 42 MHz is the standard upstream/downstream split, 85 MHz is a midsplit, 200 MHz is a high split, and the downstream goes up to 1Ghz or even 1.2Ghz. The upstream modulation runs at 64 QAM (~27Mbs) and downstream modulation runs at 256 QAM (~38 Mbps). The upstream is also more sustainable to interference and impairments. So the technology is inherently imbalanced.

From a business prospective it is very difficult to cost justify. Moving the upstream/downstream split and/or upgrading to DOCSIS 3.1 is very expensive. The reality is that the typical customer use very little upstream bandwidth. It isn't uncommon for me to see a node with 24 or 32 downstream channels at 80% usage and 4 upstream channels are at 25% utilization or less. We see a downstream to upstream usage ratio of about 15:1. If you move the US/DS split that means you are taking away from available downstream frequency which means less DS data bandwidth or fewer video channels. It's hard to make a business case to invest in increasing upstream bandwidth when the data says people aren't going to use it.

Newer technology like DOCSIS 3.1, Docsis FDX, Remote PHY and modern plant designs like N+0 or N+1 (which means fewer customers per node) will bring higher upstream bandwidth, but it will be awhile before we see anything close to symmetrical speeds.


Just wanted to mention the info that the biggest or second biggest (depends on how you count, it's a duopoly) cable ISP in Serbia is finishing up the migration to DOCSIS 3.1, and will provide gigabit speeds to cable customers. However, the upload will still be 50 Mbps maximum. Currently the fastest they have is 300/15 Mbps, which is better than what they will offer with DOCSIS 3.1 -- 1000/50 Mbps.

Most of their passive and active equipment in areas which have been fully "digitalized" is 3.1 compliant, they just need to finish the whole city and then they will start marketing the offer. Modems will have to be changed as well as equipment on the customer's side which they haven't done yet, so it will take quite some time.

They "promised" it by the end of 2019. (in 2016.), seems it will be 2025. By that time the Chinese government and Telekom Srbija will have FTTH to every home.


This doesn't surprise me. Most providers that are deploying DOCSIS 3.1 today are only doing OFDM in the downstream and the upstream is still doing ATDMA (same as DOCSIS 3.0). I'm sure there are some but I don't know of anyone doing 3.1 in the upstream yet.

If they did upgrade all their actives then 3.1 upstream is likely in the works. DOCSIS has some pretty amazing diagnostics and troubleshooting - I can look at a modem and say "approximately 71 ft from the modem there is a break in the cable causing an RF impairment" or "These 14 modems in the same neighborhood show the same impairment - there is likely a fault at this specific segment of cable". DOCSIS 3.1 expands on that even more and as more 3.1 CPEs get deployed operators are able to leverage that additional data to get the plant ready for 3.1 upstreams. OFDM frequencies are also not backwards compatible with 3.0 CPEs but a 3.1 CPE can bond both types of channels, so as the ratio of 3.1 CPEs increases that allows operators to shift more frequencies from legacy modulation to newer 3.1 modulation.


> 300/15 Mbps, which is better than .. 1000/50 Mbps.

How is 300/15 better than 1000/50?


It's cheaper per megabit ($120 vs $20 per month) and a better DS:US ratio.

edit: the ratio is the same, I just haven't powered up my brain yet.


The ratio is 20/1 in both cases


Oh, I haven't even thought to check. Sorry.


>The reality is that the typical customer use very little upstream bandwidth.

I’m not trying to disagree with this especially since you are talking about coax, but the new mmwave WiFi installations at places like Ohio state stadium have shown the opposite. The vast majority of their traffic were photo and video uploads to social media. Consumers want upload; it’s just overshadowed by streaming.


I imagine usage patterns in the home and at an event are very different.


But he does make a decent case that we generally upload more since the explosion of social media. Not sure by how much since people stream SO much.


That's interesting, I wouldn't have considered that but it make sense. If I'm at a football game or concert I'm much less likely to be streaming content and much more likely to be sharing content than a user at home.

On an unrelated note - I've been pretty disappointed with the WiFi at The Shoe. The first game I tried to use it I couldn't event join the network, the second time I could join but it wasn't usable. It was early in the season though so they could have still be working out the bugs - I can't even imagine how much goes into supporting a system like that. Trying to get 100k devices on a WiFi network is impressive.


This kind of comment is why I read HN


You should take a look at Australia, where we try to shove hundreds of customers onto a single node (Up to 900 per node, up to 250 per segment).

Edit: Although it wouldn't be fair to leave out that it's a EuroDOCSIS 3.0 setup with 16/4 downstream/upstream, with very limited 3.1 being used on a small scale.


Dave at EEVblog entered the chat ;)

IIRC, Australia has limited broadband options, they're expensive and data is expensive relative to other countries. Also, isn't data transiting outside of AU more expensive than intra-national data?

In the US, there's a 1 GbE variant of Comcast "cable modem" service called Xfinity that often has a 1000 GiB "data plan" (varying by state, some states have unlimited data) that imposes a charge per GiB beyond that. Where I use it, the network is all optical with copper from poles to customer. During power outages, it only lasts about a day... and we now have 3-6 day "public safety" outages in the summer.


Isn't telstra just riddled with stories of corruption over time?

https://www.channelnews.com.au/key-telstra-5g-partner-admits...


Telstra is the equivalent of Comcast here. They’re monipolies in large parts of the country so they can get away with a lot


And NBN's wholesale charges are comparable to what a retail ISP in a competitive market (e.g. Stockholm) charges for symmetric 10Gb/s FTTP.

https://www.itnews.com.au/news/nbn-co-puts-future-residentia...


You could run a lot of websites on that.


Yikes! I mostly work with small providers that don't have the same resources as the big guys so I'm dealing with networks that are Node + 7 or more actives (vs the N+0 or N+1 the big guys do now) and anything about about 200 modems is when we start planning for node splits.


Also prior to the NBN (wholesaler) taking over the network, "take rate" (number of homes actually connected to the cable) was ~30%. Changing a network designed around that to 80% is quite a task


It also doesn't make sense to turn your millions of residential cable customers into DDoS cannons. Best to leave them as DDoS pea-shooters.

Though on second thought, it might help out your ratios.


If they all get together and increase upstream then they can start selling do’s protection, like the phone companies did for spam calls.


> So the technology is inherently imbalanced.

I won't call it inherently, or even "technical reason(s)".

Nothing stopped the standard to split the frequency bands for up and downstream in a more way, or even swap the frequencies for each. They're simply configured/designed this way is because of the business reason.


Typo: "more even way"


Of course cable companies don’t mind turning the internet into a consumer/producer model as well.


Why isn't there any dynamic allocation of the up-/downstream channels?


From the linked article:

". I should also note that since the cable operators' infrastructure was built around sending the same TV signals to all houses, they basically just run one cable to your neighborhood, and then use passive splitters and dumb amplifiers to split that one cable to all the houses in your neighborhood. So you don't have a dedicated coax cable directly from your house to their headend equipment, so they can't change the frequency split on a house-by-house basis."


Because of the equipment out in the plant.

There are passive devices out in the field that split the upstream and downstream (such as directional couplers). Passive devices are favored over active devices whenever possible because they are much less likely to fail, are less expensive, and require less maintenance.


>The upstream is also more sustainable to interference and impairments.

s/sustainable/susceptible/ ?


Awesome post, thanks for that.


Couldn't they use something like code division multiplexing?


While I stewed after the dotcom crash during early 2000s, I had a short 2-year gig working as a headend tech for a small cable company.

It’s not about demand, it’s about climbing poles and rolling bucket-truck crews to replace filters and other expensive equipment on a massive scale.

The top comment in the link there is spot-on. https://superuser.com/a/1519918

I was involved with the “digital switchover”, moving the traditional analog channels over to the digital QAM channels, potentially freeing those now duplicated analog channels for upload channels. Maybe more upload is technically possible but the work on the poles is going to be expensive.

[edit] oops, pasted duplicate.


Priorities, right? I wonder what their business would be like if they had the foresight to not make the split so lopsided.


Probably less lucrative than the government blessed monopoly on data and media distribution they developed that required virtually no investment in anything other than lobbying.


Capex as a percent of sales for the dominant cable companies is generally 15 to 20 percent. Why do you think there was no investment?


I still don't understand why DSL connections have the same issue


They don't, it's a question of DSL profiles. Commonly used consumer DSL profiles have a 8:1 to ~2:1 ratio because that seems to match well with consumer usage. Business DSL generally is symmetric.

1000 MBit/s downstream, 35 MBit/s upstream is an 30:1 ratio, far more extreme than any DSL profile.


DSL, at least ADSL commonly found in homes, is designed that way as a good way to split the frequency available on the line to match what people need. It also shares a good portion of the frequency with POTS telephone service. There are SDSL (symmetric DSL) standards which provide fully symmetric bandwidth on the circuit. Not sure how popular it is these days but when I supported it at an ISP about 10 years ago, it was a cheap alternative to T1-class circuits.


They don't technically. I remember there being a provider here in Germany where you could switch your connections profile around. (afaik something like 6:1, 1:1, 1:6) This was way back when speeds really were an issue, so freelance web developers etc loved them. Haven't seen anything like that for a while though, at least in the home offerings. (but I also would expect demand to be quite low for it)


DSL is dedicated all the way back to the central office, so it doesn't have the same problem as cable has. That said, what others posted is correct, home users download far more than they upload, so it makes sense from a marketing perspective to give far more downstream bandwidth than upstream. Home users tend to appreciate having a high download speed much more than having equal, but much lower upload and download speeds.

Other interesting info: ADSL is dreadfully slow by today's standards, but VDSL can actually hit some decent speeds. The big problem with both is attenuation of copper wire is fairly significant over the distance the lines generally travel, much more so with VDSL, due to it's higher frequency range. That's why companies like AT&T, with their U-Verse brand, bring fiber to the neighborhood, to lop of a significant distance of copper. They can increase the distance they service by making use of two lines paired together. AT&T takes a portion of legacy ADSL service now and puts VoIP over it, with the VDSL service having more than half it's capacity generally devoted to IPTV bandwidth. One single good condition line could generally carry 118 Mbps, but they would split that up for TV, leaving only 45 Mbps. Now they are mostly pushing FTTH/FTTP though, leaving no option for POTS/copper. That is all good though, because most the copper run is such garbage that it absolutely sucks being a technician. You would be gigged for the line runs getting water in them and making service drop out, even if everything tested great when you installed, or even if you turned a ticket in the line techs might kick it back and say everything is fine.


I'd also like to mention something that was tried in Serbia (but ultimately failed, though in a good way):

Our national telecom operator, Telekom Srbija, had a monopoly on internet access in most areas due to the fact they own all the phone lines. In cooperation with Huawei and the Chinese and Serbian governments, they made a plan to kill POTS service per line and use ALL of the bandwidth for VDSL, as well as building DSLAMs on street level, to provide very high speeds (100/10 Mbps was the max they achieved before stopping) with short copper lengths. Phones would work over VoIP, so all new modems were VDSL compliant with IAD (integrated access device) "certification" (a SIP client and 2 POTS ports).

In the end the Chinese lost interest and it was all forgotten and covered with ash. The new idea, which is why I say it failed in a good way, is to skip the street DSLAM buildout and just do FTTH. And they did, though not with symmetric gigabit speeds, because reasons (bla bla customer demand, costs, etc..) The highest speed is 1000/400 Mbps for $120/mo. (which is super expensive compared to what was promised by the Chinese and also compared to general purchasing power in the country).


ive got 1000/1000 unmetered in the US for ~$110/month


Well.... yes, you do, and some people in the US have 4/0.5 Mbps for the same price. It really depends.

For example, in Niš, which is the most developed city when considering internet infrastructure and the number of ISPs, you could have 600/400 Mbps for $10/mo with ~300 TV channels on unlimited* TVs. You'd even get a [VoIP] landline.

* all the TVs in your home, basically up to 5 for 99% of users

Then Telekom Srbija got the Chinese money and bought 10 largest ISPs other than them and SBB (the other side of the duopoly), and now there's 200/20 Mbps on FTTH for $30/mo on up to 3 TVs with 260 channels and a landline that has call routing problems. Users which sign new contracts are downgraded and pay more, those that don't sign new contracts suffer with TV and Internet dying, and are told to "sign or sue us".

Ironically, I chose the second option, and won (https://news.ycombinator.com/item?id=21733553) but the thing I "won" is free contract termination and being repaid fees that were charged but are not mentioned or required under the TOS and Contract at the date of signing (given I have not signed anything else giving permission or agreeing to anything new, and they didn't officially force new contracts on people so that people can't cancel to run away)...


Fiber can also have asymmetrical speeds (see GPON, NGPON2 etc.). The issue is not the media per se but the concept of shared media. Take any media, wireless, coax, fiber, even copper (ADSL) you have X amount of spectrum to work with so you have to decide how allocate said BW. Since most users demand is higher for downstream than upstream (typically you are downloading a lot more than uploading) it makes sense to allocate more to downstream. Of course it is possible to have full duplex symmetrical speeds. Say you want to allocate 2 fibers, on for upstream and one for downstream, then you can have symmetrical speeds. But typically due to costs, you want to have one fiber shared between a large number of users (e.g. GPON or NGPON, aka FiOS) or one coax shared amongst many users, or some set amount of RF spectrum, in any case you need to make the same type of trade off choice that the cable provider makes or that the DSL provider makes or wireless providers. In the case of wireless the reason that uplink is so much weaker is due to the limited power of the user equipment, also smaller antenna etc. refer to Shannon's law, uplink will be weaker. But generally speaking, it makes sense to have asymmetrical speeds unless you have the luxury of having the same spectrum and power available for both directions.


Yeah I got FttH gigabit down 50 meg up for $65 here in upstate NY. It's the best in the area and it's a small company.


Directional couplers are a thing. Thomas Edison patented multiplexing techniques for telegraph wires that didn't even need them.


I think the fiber industry prefers wavelength division multiplexing over simple directional couplers. You can certainly have multiple 10Gbps customers on a single strand of fiber with passive equipment, and many ISPs will happily sell you such a circuit.

(Gpon, mentioned above, is mostly time division multiplexing. I have worked for 2 ISPs that use gpon, and we've never sold anyone an asymmetric plan, except for the "free plan" that Google Fiber had. Gpon is perfectly happy to be symmetric, but it does have a limited bandwidth shared between multiple subscribers.)

The biggest problem I've found is that the Internet isn't really ready for customers that can download at 10Gbps. When I set up a 10Gbps connection for myself, I had a lot of trouble finding anything on the Internet that would send me data that quickly. Even my own servers were AWS instance types that only supported 5Gbps burst. So that is probably the reason why you aren't seeing consumer ISPs selling you 10Gbps circuits... even the servers don't have 10Gbps! A new wave of upgrades are needed for this to become viable; 100Gbps networking equipment is still pretty expensive.


The use case for faster connections isn't usually so that you can download faster from a single source. It's so that you, the spouse, three kids and the dog can each watch their own 4k stream while downloading software updates for six laptops, six phones, your fridge, your microwave and two dozen smart lightbulbs.


> The biggest problem I've found is that the Internet isn't really ready for customers that can download at 10Gbps.

TCP is dumb and if it can saturate your connection it will increase your latency. I feel like being unable to saturate a connection is a good thing.

There are other measures you can take to avoid saturation, but simply having an oversized pipe is hard to beat.


This may be relevant to your interests: https://apenwarr.ca/log/?m=201808


But to really get best results, you need that management on both sides of a chokepoint, and you still want every other link in the system to be oversized.

Fast links are a simple solution to a lot.


I don't buy the 'no demand for it' line. It has never been offered by the big players. They don't want consumers uploading large amounts of data. I suspect this was originally done due to piracy concerns. But with so many homes having Nest products (and the like) that constantly upload, we need higher uplink speeds.


There was a company I like claim something like "only 5% of our users <do a thing that we dont support>" as argument for not supporting the thing.

To me, I heard "A full 5% of our users want this badly enough to hack around and get this for themselves"

This would be like Comcast saying "Less than 1% of our customers are using 10 gigabit connections, so clearly this isnt a thing people want"... You dont even offer that speed so of course the user base is low (assuming even top businesses can get those speeds). It's your job to increase the user base.

Yes, I want symmetrical broadband. But I would rather have 200/20, than 10/10. Don't mistaken that for "lack of demand" but rather "settling for what the area monopoly provides"


ISPs do not care even a little bit about piracy. The only reason piracy is an issue for them is because it hogs bandwidth, not the other way around.


Well, that's only true for ISPs who are not also content producers/owners.

Many ISPs who are also cable companies sell the content that people pirate, so they definitely do care.

In Canada, Rogers and Bell (two major quad-play telcos) are at the forefront of pushing for legislation to crack down harder on piracy by eroding privacy rules that protect consumers' Internet usage.


Fair point, but I think that was more true maybe 5-8 years ago. Since streaming services have taken off, piracy has gone way down, and providers see subscription-based streaming services as their existential threat. If any ISPs haven’t realized this, they’re just behind the curve. Bell clearly has, hence Crave TV.

Just think, who do you know outside of some geeks that bother with torrents anymore? Even most geeks who know how to pirate video and music don’t bother. They just give their $5 to $15 to netflix/hulu/amazon/spotify and don’t bother with the hassle.


The real geeks are setting up media servers that grab all content automatically and on-demand using Usenet and torrents.


How is a streaming service a threat to a company whose product is letting you use the service?


When you pay your cable company $X/month to add a premium channel, they get to keep some of that money. When you pay the premium provider directly, your cable company gets nothing. (You were going to have an Internet connection either way, so they're not getting any additional money from you.)


Click the article. The guy asking on Superuser says that fiber services like Century Link have symmetric upload speeds.


Pretty much all fiber services around Seattle offer a symmetric gig (CenturyLink, WaveG, Google Fiber/WebPass).

The only caveat I’ve heard of is that in some older buildings, WaveG offers only non-symmetric 1Gb down/100Mb up, but i am yet to encounter that myself. I didn’t look into this myself, but i have a feeling that one might actually be just a DOCSIS instead of the real fiber.


Most fiber services use GPON which also reserves less bandwidth for uploads.


> I don't buy the 'no demand for it' line.

There is measurably no demand even if you look at fiber. Most fiber deployments are GPON which is asymmetrical as well. They even make symmetrical deployments which however means you can’t (in theory) use much of your upload. However since almost nobody uses their upload nobody notices.

In Austria even fiber uses a very low upload (500/70, 1000/100 are common) because cable also doesn’t have more and there is little customer demand to increase upload.


There is no demand because of the limited upload bandwidth and ISPs only providing limited IPv4 addresses. When customers are forced use NAT with very little upstream bandwidth, they are de facto a lower-class network host that cannot use many types of network software.

I personally stopped working on a few P2P communication and personal server projects in the late 90s. Letting people communicate and share files with their friends directly with their own local private server could have been a way to fight back against the FAANGs centralizing the internet. Unfortunately, the limitations from the ISPs made this impossible for most people.


Seems more likely they want price discrimination for their business offerings. I don’t know how well that would explain things in competitive market areas though, but in monopoly areas it would make sense.


The business offerings for my office are basically the same as the consumer offerings but 4x as expensive. Getting 75/15 at work cost as much as 1000/35 did at home.


I always assumed this was part of it. ISPs want you to have to pay more if you're going to be hosting a server (and possibly making money). also explains why port 80 is usually blocked.


You cannot block port 80 like you imagined. The web would not function normally. HTTP uses port 80 both for data and control.


It's extremely common for ISPs to block inbound packets with a destination port of 80, 443, 25, etc. Of course they don't block them outgoing (with the exception of port 25 commonly). He didn't "imagine" anything and it's quite out of form on HN to claim something like that.


Port 25 is SMTP, it needs to be allowed in outgoing direction. Otherwise you will be locked to proprietary web mail.


There's a dedicated SMTP submission port. Try dialing e.g. gmail's MX record on 25. It might/should probably block.


If you mean port 465, it is the port for SMTPS, not a dedicated submission port.

I received timeout, not block, when I queried gmail on port 25. If it were blocked, I would get a TCP reset.


Port 587.

Not nearly all consumer isps block port 25, but a lot do.


I always thought that this was the reason... so they could charge more to businesses because they tend to upload more then download (not sure if it is still true today).


Possibly, they shot themselves in the foot by creating demand for pushing everything "into the cloud"? If the telco demands monopoly prices for hooking up your servers on premise, chances are you move your servers to a datacenter where there is actual competition, or use services of someone who does.


Well, it was originally for cost concerns, when you have a limited number of channels. eyeball Network tend to want down speed.


Besides this, coax is 'half duplex' until you apply Frequency division. Hence every bit of upload costs you bandwidth on the download side.

At the same time, fiber is full duplex so this is much less of an issue.

The difference is that applying a voltage at the transmission side in coax will be noticed by the receiver, (cause electricity goes both ways) whereas shining a light into fiber doesn't get to the detector (because light only goes one way). At least, this is my limited understanding. Would love if anyone more knowledgeable could correct me.


Both cable and FTTH(single-strand fiber to the home, such as Fios) are full-duplex via frequency division.

Cable operates in RF bands, and frequencies are denoted in MHZ and sometimes GHZ(included in the linked answer)

Fiber operates in optical bands and frequencies are denoted in nanometers(wavelength) -- downstream data, downstream video, and upstream data are all operated on different wavelengths on the same piece of glass, and they are received at both the transmitting end and the receiving end.

They are typically separated and filtered by optical prisms to ensure the correct wavelength hits the correct receiver(most receivers are wide-band and will receive a wide range of optical wavelengths)

The medium is typically shared at a neighborhood level, in that a single optical strand may be split 16 or more times to serve multiple houses. Bandwidth is shared on that strand by the houses served by it, typically via TDM(time division multiplexing)

End users don't typically notice this because the line rates are higher than the maximum bandwidth package sold. IE downstream rate 2.5Gb/s, upstream 1.5Gb/s, max package available is 1Gb/s.

Edit: removed incorrect statement regarding gepon/csma


Full duplex DOCSIS 3.1 has theoretical maximums of 10Gbps down and up. It relies on precise timing and echo cancellation to allow the cable modem and CMTS to transmit simultaneously on the same frequency bands. Kind of like how gigabit Ethernet allows both ends to transmit simultaneously on the same four pairs.


The question is how practical that mode of operation is on real network, or in other words whether it actually makes sense to run DOCSIS on network that can reliably support that mode (think various xDSL technologies that can push >1Gbps over <100m of what essentially is Cat5 cable)

DOCSIS is somewhat inherently asymmetric due to nature of the CaTV network. Even if the upstream and downstream channels would have same bandwidth, the upstream direction tends to have significantly worse SNR. Historically this was caused by splitters along the way, but also the cable modem is relatively cheap device installed on customer premises and thus has inherently lower transmit power and worse SNR than CMTS. This is the main reason why CMTS tend(ed) to have more upstream receivers than downstream transmitters.


This is the generally correct answer for most time divided (tdm) services. I worked on dsl back in the 90s and we experimented with pretty much all the flavors (SDSL, ADSL,etc) and while some of the plant that was built to recent modern standards could handle a maxed-out capacity, a lot of the plant (50-80+ years old) simply wasn’t suitable for it. Telcos had a choice - give everyone the lowest common denominator, deploy multiple solutions to different neighborhoods or overbuild the plant with new plant. Verizon largely chose to do the later and you got fios as a result. Other telcos did overbuilds as demand materialized. Regardless of the technical later you still had the commercial offering to consider. One way to preserve “business” line economics was to create an asymmetric residential offering. For the most part it also reflected the underlying technical capabilities. WDM fiber I don’t see happening as a residential offering anytime soon. With current technologies 200G per wavelength is capable but the card costs alone for that level of service implies many thousands in monthly recurring to recover them. 10g cards obviously are lower but gpon is much better suited as a residential architecture both economically and technically.


How come fiber can't use the same frequency for uplink and downlink? Coherent and focused laser light only travels in one direction, so I'd expect the transmission not to interfere with the receiver.


Duplex fiber where a separate strand for TX/RX can use the same frequency.

Simplex fiber, that is point to point, may in fact be able to use the same frequency, but it's not done in practice. Possibly due to the difficulty in directing the outgoing and incoming signals to different pieces of hardware(receiver and transmitter)

GPON networks(typically used by Fios, etc...) have multiple passive splitters on the strand. For example, a common PON splitter would have 1 strand in from the OLT(optical line terminal, essentially the head-end), 32 strands out. The light received from the OLT is split x32 and sent to each downstream port. The devices on the downstream ports(ONU/ONT, Optical Network Terminal) receive this signal and also transmit it on the same strand. That transmitted signal must be coupled back into the single upstream port, from every attached ONT. I imagine there is significant difficulty getting the light to traverse the splitter and exit only via the upstream port. So a different wavelength is used so that it doesn't matter if the light is received by the other units, and also it can be more easily filtered (by wavelength instead of direction).

My knowledge of optics is not sufficient to fully explain why it's easier to filter by wavelength instead of direction.


> My knowledge of optics is not sufficient to fully explain why it's easier to filter by wavelength instead of direction.

Reciprocity. For most optical devices, if x% of light coming in port A comes out port B, then x% of light coming in port B comes out port A. Non-reciprocal materials exist, but they’re unusual and may be rather inefficient and complicated to work with. Wavelength splitters are reciprocal and are relatively easy to construct. In fact they’re cheap enough that people use them for art. (The usual splitter is a piece of dichroic glass.)

The Wikipedia article is so-so:

https://en.m.wikipedia.org/wiki/Reciprocity_(electromagnetis...


The reason why it is easier to filter by wavelength is that almost any optical path reflects significant amount of light back, primarily from various discontinuities along the fiber (connectors, splices, and also the actual receiver on the other end) but also from imperfections in the fiber itself. On copper interconnects there are similar issues, but for reasonably controlled cabling (eg. 1000-base-T) they can be minimized enough, that what remains is either insignificant or can be compensated for by DSP magic.


I'm not sure that you would find any product on the market which shines the same frequency of light in both directions on a single fiber. All of the bi-directional on a single fiber products that I know of send a different frequency of light in each direction.

Therefore it would seem strange to me to claim that a single strand of fiber can do full duplex, as frequency "B" of light could instead be used to send data in the same direction as frequency "A". So really you are just choosing to allocate some part of the overall capacity of the single fiber to return traffic.


You actually can separate out the forward and reverse waves on a coax, but it requires good impedance match on both ends, and some transformer based magic-Ts. So you can do full duplex on a single coax, sharing the same bandwidth. It’s just not practical in long runs since you’ll only get 20 dB isolation between directions.

A common use is carrying Ethernet over coax when retrofitting analog CCTV cameras to IP cameras, and you don’t want to run new cable.


Fiber is very often frequency divided to give full duplex on a single strand to customers. Or the concept is taken to the extreme e.g. GPONs.


GPON is time divided, not frequency divided. You have one wavelength for Tx and one for Rx and everyone on the PON has to split it.

It is certainly POSSIBLE to do WDM on a PON, but you need separate matched pairs of transceivers for each connection which kind of defeats the purpose (you now need a mux/demux at the end which adds more attenuation to the line, as well as more expensive electronics as you went from needing one GPON port to N Gbe ports). As such, wavelength division is typically used over point-to-point links to get more discrete links over a single fiber (or pair).


GPON is both time divided and frequency divided.

There are typically 3 wavelengths.(frequency division). Two for data(downstream and upstream) and one for video (downstream).

The downstream channels(both video and data) are broadcast and not time divided. The streams are filtered by the ONT. The upstream data channel is time divided to permit multiple devices transmitting without collision.


I'm sure you could do full duplex coax if you really wanted to. But it's not worth the effort here.


That information is quite outdated. It's more than possible to get symmetric speed on (EURO)DOCSIS 3.1. The main reason it does not happen is because there is no consumer demand for it.


The reply was actually pretty spot on. It’s not a demand issue. It’s true that there is not overwhelming demand for higher upload, but that’s pretty far down the list of what prevents it from being offered.

DOCSIS 3.1 allows it on paper, yes. But in the real world, bandwidth is a huge constraint, especially if you’re also serving digital cable. No cable ISP in North America today on D3.1 has a network that can support it even if they wanted to.

For example, Comcast has 4 upstream 6mhz sc-qam channels (which is basically using none of the advantages offered by 3.1) and 32 downstream, and their video service is 720p across the board and overly compressed. They are constantly trying to free up data channels and migrate to newer codecs, and eventually go to IPTV, but that requires changing hardware both in their network in the customer’s home, or the customer’s TV goes black and their internet goes out and you get to hear their rage when you tell them they have to spend a few hundred dollars on new cable boxes/modems so that Comcast can build out their network. It’s a massive infrastructure upgrade to use DOCSIS 3.1 to its full potential that is further complicated by a need to support all of the legacy hardware during a transition.

Add to the mix that upstream channels are far more susceptible to unwanted ingress, most homes are full of amateur RG59 coax cable, and it becomes a support/troubleshooting nightmare as well. When you get into higher upstream speeds, these deficiencies become a much bigger problem.

I have observed speed tests on an uncapped docsis 3.1 modem on the comcast network getting full bandwidth on all channels, and it was still way below 100mbps upstream (around 80mbps).


in Germany Vodafone deployed DOCSIS 3.1 partially and while still not symmetric i get 100Mbit/s down and 50Mbit/s up more or less consistently with DOCSIS 3.0 where i live unless something goes wrong (it does sometimes interfere with LTE here). DOCSIS 3.1 is also available with lots of [0]channels using the newer 1024QAM. they offer up to 1Gbit downstream and 50Mbit upstream for about 70€ a month. i also made a [1]screenshot of the channel usage and stats page of the modems web interface.

[0] https://helpdesk.vodafonekabelforum.de/sendb/belegung-11.htm...

[1] https://gist.github.com/mercora/ef66f5bcd95b060ac25269b17cc6...


It's not required to have DOCSIS 3.1 to have 100Mbit/s symmetry, 3.0 can provide that. Vodafone in Spain has a mixture of HFC footprint (the widest), its own small FTTH footprint, and Telefonica (Spain's DTAG) wholesale FTTH. Prices and speeds for the three of them are the same, but if you are on an FTTH zone you get symmetric upload speeds, while on HFC zones they provide 100Mbit/s upload with all their plans (min download is also 100MBit/s).

They argue they will be able to have symmetry on 1Gbit on HFC when they complete their DOCSIS 3.1 rollout, but that has been going on for years with no news. Ironically their own FTTH footprint is pretty small so they end up, in most cases, providing better upload speeds when reselling Telefonica infrastructure than with their own.


Thats approximately what is offered in the US as well. The reason that 100/50 seems close to symmetric is because the 100 is low. The fact that it’s still only 50 up with the 1 gig downstream suggests that they have the same constraints as cable ISPs in NA.


> there is no consumer demand for it.

The most demanding thing more than, probably, 0.01% of customers do is livestreaming. If your line can actually sustain 10mbps (vs advertised/claimed), you're good. The only other real scenario is someone learning that they have to back their data up, for the first time, which is really a once in a lifetime usecase for fast upload.

What would likely improve the experience of 99%+ users is latency, where fiber has consistently proven to be king (for me at least).


There is and the government had provided resources to provide the supply, but the market is a duopoly.


And Comcast has the seventh largest lobbying budget of any individual company or organization in the United States.

ISPs should be regulated like public utilities.


[flagged]


Please don't post insinuations of astroturfing to HN unless you have some sort of objective evidence to point to. This is one of the most common and most poisonous internet tropes around, so we have an explicit rule against it in https://news.ycombinator.com/newsguidelines.html. More explanation at https://hn.algolia.com/?query=by:dang%20astroturf&sort=byDat... if interested.

Edit: looks like you've done it at least twice before. That's too much. Please read https://news.ycombinator.com/newsguidelines.html and stick to the rules when posting here.


> Cable TV, before DOCSIS cable modem internet service became a thing, didn't need much upstream bandwidth. All the upstream was needed for was to authenticate the cable descrambler set-top boxes, authorize occasional pay-per-view transactions, and maybe do a few minor "interactive TV" things, like letting you check your cable bill from your descrambler box.

Cable TV didn't have any upstream signaling once upon a time; it was a purely a system of repeating and replicating an analog signal in the downstream direction.

Early scrambling methods for "pay TV" channels could be undone using circuits that were published in electronics magazines.


Mostly true.

But upstream signalling predates cable modems by a certain period; originally the upstream was intended for things such as city 'public access' stations to feed back to the headend (to then be transmitted back out to customers). It's existence was certainly handy for the use of cable modems.


> Mostly true.

Where lies the falsehood?


Because simply US (upstream) frequencies are limited to 42 MHz or 85 MHz. They start at 5 MHz.

While in most cases downstream frequencies start from 54 MHz to 1 GHz


1 gbit non-fiber is all but marketing. For starters they have a small bandwidth caps, which keeps you from doing anything interesting. Another thing like the article mentions is that it’s a meager upload ability.

I have symmetrical gigabit fiber at my house. When a new Ubuntu torrent drops I’m able to seed a TB relatively quickly. Bandwidth caps would make it so much less useful and interesting.


I don't think we need symmetrical, but could we get 4:1 ratio instead of 20:1 please? PIN fiber that Verizon FiOS uses is 4:1 at the physical layer, or it was with BPON and GPON standards. I did not follow. I imagine they can sell symmetrical since most people don't use much upstream. Oversubscription is standard and works great in almost all cases.


I think it's a conspiracy/scam, so they can milk every dollar out of every company that wants to peer with them.

The only way to peer for free is if your traffic profile is 50% send and 50% receive. Cable companies offer these asymmetric pipes to their customers to make this balanced traffic profile impossible to achieve, so everyone ends up paying the cable companies big bucks for transit on their network.


Because ISPs are cheapskates and don't want to upgrade their networks and hardware.

There are better exceptions, like Altice who are in the process of replacing their coaxial network footprint with fiber optics. Interestingly, according to them it will actually cause them to save money in the long run due to lower maintenance costs for fiber optics.

I guess other coaxial ISPs are too greedy for short term profits to look ahead.


I can about triple my internet bill to double my uplaod speeds and still be about 50% off what I need for things I run at home. Media streaming from my own server is possible for 1-2 people at 720p/mbps and I want that pushed to 1080p minimum. A business class connection might get me there.


My ISP will not sell me anything better than 1G/100M on fiber. Maybe if I looked into business plans maybe.


A fiber network Hotwire offers max 1 Gbps download and 50 Mbps upload in South Florida. Oh and it’s a 2-year contract if you want those speeds.


Not mentioned, they don’t want to support residential customers serving web sites (or anything else)


Is bandwidth speed?




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: