|
Recently following this MLG I read a rather interesting discussion in which people complained about their bandwidth cap being exceeded. People understandably expressed frustration about their caps, and seemed to blame it more on the cable companies being greedy then any real need. As someone who designs telecommunications infrastructure equipment for their job I was a little shocked to see the lack of understanding about the state of the infrastructure, and why it's so hard and expensive to increase capacity.
How your cable modem works:
Your cable modem connects to the internet through a system called "hybrid fiber cable". First the data bits from your router or PC are modulated into something called 64-QAM(this is the most common, you may have 128-QAM or 32-QAM). This is the same type of modulation system used by many cell phones, and a number of other modern telecommunications systems. Your modem is assigned a frequency and a bandwidth window by a piece of equipment back at your local regional center called a cable modem termination system. 64-QAM is a clever way of encoding data: a group of 6 bits is encoded into a sin wave by storing the information in the amplitude and phase of the wave, not just the amplitude like AM or the phase like FM. This allows you to transfer 6 bits per symbol versus just one for AM, but because the signals for different combination of bits are very similar you are more susceptible to noise, so you need a higher signal to noise ratio. Each person on a physical channel of a cable modem termination system(within the same neighborhood) has a different frequency, and it is all combined onto a common line. The system tries to be smart about allocating your bandwidth in order to make sure everyone can get internet, but isn't that smart about it: its pretty hard to finely control such a thing.
*I have been corrected by a more knowledgeable poster about my understanding of QAM symbol rates, for anyone who cares: *
On October 18 2011 11:43 Myrmidon wrote:
Uh, there is NOT one modulation symbol per cycle of the carrier (sine wave). The modulation symbol rate is usually much lower than the carrier frequency.
Also, you can get more than 1 bit per modulation symbol when doing just amplitude modulation or frequency modulation. Normal square 64-QAM is just 64-QASK (8-ASK on in-phase component and 8-ASK on quadrature component). Though in practice, non-square QAM constellations are often used.
QAM = quadrature amplitude modulation
QASK = quadrature amplitude-shift keying
Here is a very simple example to anyone following, for sending 3 bits per symbol via 8-ASK:
Send 000: amplitude -7
Send 001: amplitude -5
Send 011: amplitude -3
Send 101: amplitude -1
Send 110: amplitude 1
Send 111: amplitude 3
Send 101: amplitude 5
Send 100: amplitude 7
If you are the receiver and detect an amplitude of 2.1, assume that 111 was sent (most likely it was 111 since that is the closest value on the list). If 111 was sent and you detect an amplitude of 1.8 and assume 110 was sent, then the last bit was in error.
Each time you want to send a new set of 3 bits, send a different amplitude value corresponding to the above mapping.
Because you need a lot of power(which generally leads to a high signal to noise ratio), you need to be very close to the transmitter, or you will lose a lot of power to attenuation over the cables. In order to solve this problem cost effectively, a system called hybrid fiber cable was devised. In this system, at a point somewhere in your neighborhood, the QAM signals carried over coax cable are modulated into fiber. This is not the same as connecting to fiber optic internet: it is an analog QAM signal, but it is modulated up to fiber optic frequencies to be able to travel through fiber a long distance with low attenuation. The thing in your neighborhood is not a router, it is a simple non-regenerative modulator: it just multiplies the signal coming in by a light wave to carry it along the fiber optic cable.
When your signal gets to the cable company it is demodulated back to regular QAM over coax. This is because we don't really have the technology yet to directly process analog signals carried over fiber. It is then routed to the cable modem termination system. The cable modem termination system is basically a router, that takes in QAM modulated cable at one end, and feeds out to the internet on the other end.
Why it is so hard to upgrade:
We are basically maxed out in capacity. Typical cable infrastructure has a bandwidth between 5MHz-1GHz, but that is shared across any number of homes. In order to increase capacity one of two things need to be done:
-Increased signal to noise ratio: this can be done by either reducing the noise, ala more expensive modems and recievers, or more power. If you can increase the signal to noise ratio, you can increase the number of bits per symbol, but you have to upgrade the equipment both in the home and in the CMTS. An example of this is DOCSIS3.0 that can support up to 256-QAM. Because it is darn expensive to increase the SNR by any other way, this basically boils down to having more neighborhood sites closer to homes to boost the power(you cant just increase the transmit power, or you will start blowing out components). For 128-QAM you typically need to receive a little less then 1 milliwatt of RF power, which is quite a lot when talking about telecommunications systems. As a point of reference your cell phone receives somewhere around 10^-9 to 10^-12 watts.
-Increased number of physical channels: in order to do this, you must increase the number of physical cable channels running to the regional center. The regional center must purchase more cable modem termination systems(which run around 3 million dollars a piece), and must serve less homes on a single modulator, and run appropriate fiber to the modulators.
Both options boil down to having the fiber modulators closer to the home, and both options are extremely expensive to implement. The ISPs have found that the majority of people are not willing to pay more for more speed, and with the huge expense it then becomes hardly worth it.
Conclusion:
When looking at the modern cable modem infrastructure, you will notice that it is basically one giant kludge. This is because it incrementally evolved to minimize cost while increasing speed. If anyone was to design a new system it would look nothing like the current one, but to install an entirely new system is incredibly expensive(hence why most people don't have FIOS access, and it is not much cheaper then cable).
 
|
Amazing read, thanks for the insight
It's something I haven't actually thought about very much, though it's a bit technical at parts
5/5
|
Very interesting. Is there anything you can say for bandwidth throttling?
|
ISPs are greedy. They offer internet speeds that they "can't afford" on their current infrastructure, then they charge the clients up the ass for a fake bandwidth cap issue.
Why is it I'm being charged as much for a GB during off hours as I am during prime time, which in Canada, ISPs stated they might throttle your internet speed during prime time if not enough bandwidth is available, which is understandable... But then they go ahead and claim its a precious resource that should be used sparingly, because everyone should have their fair share.
Bandwidth is only a big deal during prime hours, where most people use it, the rest of the time, it's available freely to just about anyone that uses it.
Also, if ISPs charge you for bandwidth caps, shouldn't they at least use it on upgrading their infrastructure, rather than increasing your speed?
|
This allows you to transfer 64 bits per symbol versus just one for AM,
6 bits, not 64. It's 64 different symbols, each representing 6 bits of data. 2^6 = 64
btw, where do you work?
|
People are upset (and justifiably) over usage caps, commonly and erroneously referred to as bandwidth caps, not literal digital bandwidth and the limitations thereof
|
On October 18 2011 11:15 Seth_ wrote:6 bits, not 64. It's 64 different symbols, each representing 6 bits of data. 2^6 = 64 btw, where do you work?
Yeah you are right  I will change it, I do RF stuff not your silly base-band stuff. I work for Aeroflex we ODM a lot of the components that other companies then mark up and sell.
|
I'm curious as to why you tell us why bandwidth caps are a regular thing that occurs to us, even though we pay more than other countries that have faster internet than us.
I will be insanely biased here, but whatever. South Korea pays around $30 a month for broadband (going off last years numbers at 28.80, but i rounded up a dollar just in case) vs our $45 a month on broadband. They also get faster internet than we do, and I haven't heard much about a bandwidth cap from them (although I did not do a search for this specifically, I never hear about it in the news).
Why do you think that we have to pay more, for a worse system that isn't going to support the need of internet in the future. You say caps are needed, but how are we to eliminate them if the future is going to have us getting internet to more people, thus making this a big problem.
|
On October 18 2011 11:21 floor exercise wrote:
People are upset (and justifiably) over usage caps, commonly and erroneously referred to as bandwidth caps, not literal digital bandwidth and the limitations thereof
Well they are one and the same. The majority of people don't ever get near their capacity, but want their internet to be snappy fast when they do use it. If you can increase bandwidth capacity you have no need for usage caps.
|
very quality post There are a lot of technological upgrades that are coming from places like 3M that is making a lot of your points better but yeah its still a huge problem. Great insight loved the read.
|
On October 18 2011 11:23 Alventenie wrote:
I'm curious as to why you tell us why bandwidth caps are a regular thing that occurs to us, even though we pay more than other countries that have faster internet than us.
I will be insanely biased here, but whatever. South Korea pays around $30 a month for broadband (going off last years numbers at 28.80, but i rounded up a dollar just in case) vs our $45 a month on broadband. They also get faster internet than we do, and I haven't heard much about a bandwidth cap from them (although I did not do a search for this specifically, I never hear about it in the news).
Why do you think that we have to pay more, for a worse system that isn't going to support the need of internet in the future. You say caps are needed, but how are we to eliminate them if the future is going to have us getting internet to more people, thus making this a big problem.
The big difference is that, as far as I know, they are not working off of a hybrid fiber cable infrastructure. I believe they are basically pure fiber(correct me if I am wrong, I know very little about what they have in place). Its a lot cheaper to upgrade pure fiber then to upgrade hybrid-fiber cable. We already had the cable infrastructure in place so it was cheaper to go to hybrid fiber cable then to pure fiber, and most people don't want to pay a lot more for more speed or usage capacity.
|
Wow, what an informative post. Could tell you know more than 99.8% of normal people about this.
Gonna look into it more. Slightly confuzzled atm.
|
Uh, there is NOT one modulation symbol per cycle of the carrier (sine wave). The modulation symbol rate is usually much lower than the carrier frequency.
Also, you can get more than 1 bit per modulation symbol when doing just amplitude modulation or frequency modulation. Normal square 64-QAM is just 64-QASK (8-ASK on in-phase component and 8-ASK on quadrature component). Though in practice, non-square QAM constellations are often used.
QAM = quadrature amplitude modulation
QASK = quadrature amplitude-shift keying
Here is a very simple example to anyone following, for sending 3 bits per symbol via 8-ASK:
- Send 000: amplitude -7
- Send 001: amplitude -5
- Send 011: amplitude -3
- Send 101: amplitude -1
- Send 110: amplitude 1
- Send 111: amplitude 3
- Send 101: amplitude 5
- Send 100: amplitude 7
If you are the receiver and detect an amplitude of 2.1, assume that 111 was sent (most likely it was 111 since that is the closest value on the list). If 111 was sent and you detect an amplitude of 1.8 and assume 110 was sent, then the last bit was in error.
Each time you want to send a new set of 3 bits, send a different amplitude value corresponding to the above mapping.
|
On October 18 2011 11:23 InvalidID wrote:Show nested quote +On October 18 2011 11:21 floor exercise wrote:
People are upset (and justifiably) over usage caps, commonly and erroneously referred to as bandwidth caps, not literal digital bandwidth and the limitations thereof Well they are one and the same. The majority of people don't ever get near their capacity, but want their internet to be snappy fast when they do use it. If you can increase bandwidth capacity you have no need for usage caps.
Not really, it's just the convenient excuse used to fleece customers. The only possible time for bandwidth capacity to reach peak is in peak usage hours. Usage based billing does not in any way curtail that, it's clearly an exploitative 'solution' to what may or may not be a problem at the end of the day.
How are they one in the same? I don't deny the existence of bandwidth limitations or even network congestion in certain hours, but how does usage based billing in any way effectively tackle the supposed issue of everyone coming home and using the internet at the same time?
Very greedy people decided to arbitrarily assign a cost to transmitting data and charge us this fee regardless of when or how we use this data, because there might be a problem in certain areas between the hours of 8 to 11 due to the technical limitations of our current infrastructure.
I appreciate your post for what it is but it completely misses the mark as to why Canadians are upset
|
On October 18 2011 11:43 Myrmidon wrote:Uh, there is NOT one modulation symbol per cycle of the carrier (sine wave). The modulation symbol rate is usually much lower than the carrier frequency. Also, you can get more than 1 bit per modulation symbol when doing just amplitude modulation or frequency modulation. Normal square 64-QAM is just 64-QASK (8-ASK on in-phase component and 8-ASK on quadrature component). Though in practice, non-square QAM constellations are often used. QAM = quadrature amplitude modulation QASK = quadrature amplitude-shift keying Here is a very simple example to anyone following, for sending 3 bits per symbol via 8-ASK: - Send 000: amplitude -7
- Send 001: amplitude -5
- Send 011: amplitude -3
- Send 101: amplitude -1
- Send 110: amplitude 1
- Send 111: amplitude 3
- Send 101: amplitude 5
- Send 100: amplitude 7
If you are the receiver and detect an amplitude of 2.1, assume that 111 was sent (most likely it was 111 since that is the closest value on the list). If 111 was sent and you detect an amplitude of 1.8 and assume 110 was sent, then the last bit was in error. Each time you want to send a new set of 3 bits, send a different amplitude value corresponding to the above mapping.
I am aware well of the terminology, and the basic concept of QAM and you are probably right, as far as the number of symbols per period, I said before I do RF not baseband. Everything I work with is well past modulation, I was just repeating what I remembered from classes as far as the modulation goes. I will update it to be accurate.
|
@op: are you kidding me? Then how come the rest of the fuckin developed world has way better and cheaper internet than us? It's a fucking joke. We've been monopolized for far too long and instead of increasing speeds overtime, they smack on more and more restrictions. I know that distance is an issue since our countries (canada and states) are huge, but companies could easily focus their efforts on where traffic is highest (i.e. major cities). No need to set up networks over the praries and shit. Start somewhere and expand.
|
10387 Posts
On October 18 2011 11:23 Alventenie wrote:
I'm curious as to why you tell us why bandwidth caps are a regular thing that occurs to us, even though we pay more than other countries that have faster internet than us.
I will be insanely biased here, but whatever. South Korea pays around $30 a month for broadband (going off last years numbers at 28.80, but i rounded up a dollar just in case) vs our $45 a month on broadband. They also get faster internet than we do, and I haven't heard much about a bandwidth cap from them (although I did not do a search for this specifically, I never hear about it in the news).
Why do you think that we have to pay more, for a worse system that isn't going to support the need of internet in the future. You say caps are needed, but how are we to eliminate them if the future is going to have us getting internet to more people, thus making this a big problem.
Korean government subsidizes a lot of the costs though, which is why Korea's internet is so good and relatively cheap. Until the US government decides to start investing heavily into internet, US will always be more expensive and slower
also, Korea is much smaller than US .. so for US to invest into internet infrastructure would be incredibly expensive .. (but better use than that damned military budget ..)
|
On October 18 2011 11:48 InvalidID wrote:Show nested quote +On October 18 2011 11:43 Myrmidon wrote:Uh, there is NOT one modulation symbol per cycle of the carrier (sine wave). The modulation symbol rate is usually much lower than the carrier frequency. Also, you can get more than 1 bit per modulation symbol when doing just amplitude modulation or frequency modulation. Normal square 64-QAM is just 64-QASK (8-ASK on in-phase component and 8-ASK on quadrature component). Though in practice, non-square QAM constellations are often used. QAM = quadrature amplitude modulation QASK = quadrature amplitude-shift keying Here is a very simple example to anyone following, for sending 3 bits per symbol via 8-ASK: - Send 000: amplitude -7
- Send 001: amplitude -5
- Send 011: amplitude -3
- Send 101: amplitude -1
- Send 110: amplitude 1
- Send 111: amplitude 3
- Send 101: amplitude 5
- Send 100: amplitude 7
If you are the receiver and detect an amplitude of 2.1, assume that 111 was sent (most likely it was 111 since that is the closest value on the list). If 111 was sent and you detect an amplitude of 1.8 and assume 110 was sent, then the last bit was in error. Each time you want to send a new set of 3 bits, send a different amplitude value corresponding to the above mapping. I am aware well of the terminology, and the basic concept of QAM and you are probably right, as far as the number of symbols per period, I said before I do RF not baseband. Everything I work with is well past modulation, I was just repeating what I remembered from classes as far as the modulation goes. I will update it to be accurate.
I'm much more familiar with wireless communications systems, though I focus less on the RF hardware, filtering, DSP, etc. side of things.
Anyway, I thought cable telecommunications systems did not use baseband, for better propagation characteristics? Also, if there's some kind of frequency-division multiple access as you're describing (and there is), that implies modulating information onto different carrier frequencies, not baseband.
|
You described what I deal with at my job on a daily basis. I'm guessing we do somewhat of the same work. Well put I'd say as a general brush-over, even without referencing dB, dBm, dBc, Eb/n0, etc.
It's amazing how long technology has been around before it's even brought mainstream, only to then be seen as that "hot, new technology product".
|
Really funky system, thanks for the information. Gotta admit though I would trade my internet for that. I'm paying 60 a month plus have to have a phone line on with my ISP and only get 3 down and .5 up. I live in a small town Texas but still, DSL basically costing 85 dollars a month for such shit is ridiculous. Hope one day all those ISP's out there take a big fucking hit that jostles there comfy little industry, cheapskate bastards. -_-
|
|
|
|
|
|
EPT
