Data transfer over coaxial cable

Factual Questions
1

As data transfer rates increase, we see perennial upgrades to ethernet, HDMI, USB etc. And ISPs are obviously upgrading what’s out on the street. But isn’t most of the coaxial cable infastructure inside our walls decades old, from an era when data rates were a tiny fraction of what’s now possible? Yet I’ve never seen that cited as a constraint on data transfer. Comcast now offer gigabit internet, which they explicitly say can travel over that same old coaxial cable, albeit only in the download direction.

I don’t believe that we explicitly had the foresight to install infrastructure inside our walls that was future-proof decades ahead as data transfer rates have increased 3 orders of magnitude. So what’s going on here? Am I just mistaken, and the cable guy does run some higher spec of coaxial cable through your wall when he upgrades you?

2

The main cable in the street needs to be able to carry all the data to everyone on your street simultaneously. The cable inside your walls only has to carry yours.

3

Well obviously, but how is that responsive to my question? HDMI, ethernet & USB cables, which also only carry my data, have been upgraded many times to new specs to handle much higher data speeds. I’m asking why the same thing hasn’t been required of the coaxial cable in my walls (or if I’m mistaken and it has been upgraded).

4

In spite of what the manufacturers of expensive, gold-plated cables would like you to think, the technology of the cable from the street to your computer hasn’t improved much in the past fifty years. All the routers between the street and your ISP? There’s been some improvement there.

5

All you need in the walls is good old RG-6QS. This has been the default standard for cable TV for probably 40+ years. It’s a slightly thicker brother of RG-59, but the QS denotes quad shielding - four layers of foil and braid to keep the wanted signals in and interference out.

(Cable geeks - please don’t roast me - this is meant to be very simplified!)
As to how coax can carry more data, USB is essentially pulsed DC - most of the developments in its cabling is based on keeping the pulse shapes intact so the receiver can read them properly at distances measured in single-digit meters. Technologies like VDSL operate on wiring originally designed to carry telephone audio signals and are meant to be run over a couple thousand feet. Coax, however, is meant to carry either a single “channel” of video or hundreds of TV channels at frequencies ranging from close to DC on up to multiple gigahertz for miles, so there’s a huge frequency spectrum for its equipment to operate in.

6

My understanding is that most cables handle a single baseband signal. Coaxial cables, such as the RG6 cable used for tv can handle broadband signals. Broadband uses modulation to send multiple signals or channels over the wire at the same time. DOCSIS, the standard for internet over cable typically uses a lot of channels for the downstream and fewer channels for upstream. This is why cable internet is typically much faster for downloading than uploading.

7

Old-fashioned analog video used a LOT of “bandwidth” and that’s what coax was designed for.

Actually, there have been upgrades to coaxial cable over the years. The current version is RG-6, which can carry up to 100 mbps on a signal of 750mhz. Even the older, cheaper types (RG-5 and RG-59) have plenty of data bandwidth.

8

HDMI carries uncompressed video signals. A 4K 60 Hz monitor requires 3840 x 2160 pixels * 3 bytes (1 for each color) * 60 = 12 gigabits per second. And HDMI 2.1 (which I think is just becoming available) can do 48 gigabits per second.

9

The simple answer is just that it’s just amazing how much bandwidth there is in a coaxial cable through clever modulation and broadband multiplexing techniques. It’s essentially the same technologies that have advanced speeds in other areas – consider that in the early days of 10 mbps Ethernet there was initially a thick coax specification for the physical layer, and later a thin version, with considerable doubts as to whether Ethernet could be run over twisted-pair (essentially, ordinary telephone wiring), yet today we have gigabit Ethernet running over twisted-pair copper per 1000Base‑T. Likewise look at the speeds of USB3.

As someone already noted, analog video took up a lot more bandwidth than modern digital channels, though even in the old days there was room for hundreds of analog channels, which is an astounding amount of bandwidth. But that’s one reason that cable companies were quick to switch over to digital-only, which freed up bandwidth for other services, most importantly high-speed internet, but also telephone services and other things like security monitoring. AIUI the biggest limitations have not been in the home cabling, but in the typical “last mile” infrastructure, so that in order to offer faster speeds cable companies have had to upgrade their local infrastructures. I was an early adopter and at the time the only available speed was 3 mbps down, which I thought was fantastic, coming off dialup as I was at the time! Today 100 mbps+ is routine, and 1 Gbps is available in most places. DOCSIS standardization has also made it far more reliable, and at least around here, the cable company is far better now at maintaining area segmentation to prevent overloading the local segment.

10

Almost anything can carry almost anything if the distance is short enough

My telephone wires, which used to give me 2400 max, then 9600 max, mostly because they were connected through a telephone exchange, are now (sometime in the next 6 months) going to give me 96M – 40000 times what theory and practice told me was the practical maximum – partly because I’m going to connecting to a digital subscriber line access multiplexer less than 150 yards away.

Part of that is an increase in power levels. With the short lines and the few people connected right here, they can crank the power up a bit and it won’t cause problems for other people.

Part of it is some electronics that was simply inconceivable back in the day. When they’ve got 4 of us locally talking to the same DSLAM, they can synchronize all 4 discussions so that I’m not shouting when somebody else is whispering (it’s digital encoding not voice, but I like the analogy). The transmitters at each end listen to each other and discuss line conditions, so they can compensate for echo and distortion. If there’s a sudden “click”, we can go back and fix up the bits that were lost (there aren’t as many sudden clicks as there used to be, but back in the day of 2400, that was a really limitation on bandwidth)

11

I remember when I had 56k, being told that “the copper can’t physically carry any more information”. Years later (when DSL showed up), I realized that what they meant was 'we haven’t figured out, yet, how to make the copper carry more information".

12

Ultimately Shannon tells you how much data you can get down a length of cable. What is often missed is that the information rate is governed by two things, not one. The bandwidth of the channel and its signal to noise.

In the real world with lengths of cable, the two metrics are somewhat intertwined, as there is no single figure of merit for either, rather the signal loss depends upon frequency, and the signal to noise is related in-part to the signal loss, and there is no hard cut-off, rather a frequency dependant loss.

Noise comes in the form of intrinsic noise in the system (thermal etc) and external interference.

Managing these, is the cornerstone of how we get serious data rates.

Coax versus twisted pair isn’t a simple answer, both have advantages and disadvantages. Coax has the advantage that it is self shielding to some extent. That means it tends to have better signal to noise than twisted pair. A perfect twisted pair can reject common mode noise, but this is limited by real world constraints in manufacturing tolerances and proximity to sources of interference. It is also limited by the ability of the receiver electronics to reject common mode noise.

In long cable runs you run into problems with impedance mismatches. These can come about due to jointing the cables, or from mechanical distortion of the cable. A mismatch causes reflections of the signal, and this ends up looking like even more interference - so the signal to noise drops - and so does your information rate. The massive hike in data rates over twisted pair cables - whether Ethernet or DSL cabling is mainly down to insane levels of signal processing that dynamically characterises the cable run, identifies the impedance glitches present, and creates an appropriate convolution for a DSP engine to use to ameliorate the cable imperfections. This is close to science fiction for those of us that grew up screwing 10-Base-2 vampire taps onto coax as thick as your thumb.

ADSL won big by dynamically characterising the interference it is seeing (and because it shares a multicore cable with all your neighbour’s signals, it sees a lot of interference.) By looking for frequency slots with less interference, and ignoring slots with more, it can pick and choose the total bandwidth, and optimise itself to only use the good bits, and not be hamstrung by the noisy bits. This trick is not restricted to ADSL. (The reason ADSL is Asymetric is because when all the cables come together at the exchange - in the DSLAM - there is a huge amount of noise from all the signals being sent out to all the customers. So signals coming into the DSLAM from customers, already attenuated because they have come the full length of the cable, arrive to be presented with maximum interference - so the signal to noise for data going upstream is bad, whilst data ravelling downstream is received by the customer in comparative quiet at the home, and much better signal to noise is found, allowing much better data rates.)

The point? Coax versus twisted pair isn’t a direct determinant of what can be delivered. Assuming it isn’t badly degraded due to its age, coax, in general, will have both significantly greater bandwidth and better signal to noise, and if you had a private connection you could see insane data rates. The fact that it has to be shared is where you see a drop in real life use. But given most people are not slurping constant high data rates, and most use is highly peaky, you can get a lot of happy customers on a shared resource. High resolution movie on demand streaming is the elephant in the room here.

Here in Oz there was a bit of a scandal when the coax infrastructure of one of our telcos was purchased with the intend of being used for providing high speed internet reticulation, but proved o be in such poor shape that the entire investment was written off, and roll out of services in quite substantial fraction of the country delayed by over a year.

Most people I know who have HFC (Hybrid fibre coax, where coax is used to provide the last mile reticulation) are pretty happy. I have fibre that terminates in a pit outside my house and the last few metres comes in over the old phone pair. I get a real 80M/s down and 40Mb/s up. But those on HFC can get comparable speeds.

13

Let’s hope that she shows up soon, then.

14

Claude Shannon is a personal hero of mine, so forgive me if I didn’t find the joke quite as amusing.

Note that the bandwidth theorem Shannon developed is in the case of additive, white Gaussian noise, which is a very friendly assumption to make. Real world noise tends not to be white, nor is it always totally uncorrelated with signal. Or rather, it’s rarely white and rarely uncorrelated with the signal.

But it still provides a pretty good rough approximation of how much information can be transmitted at whatever noise level is present. And it also gives a very good idea of the relative importance of the width of the passband and SNR to the bandwidth.

15

Wasn’t a joke–I thought it was an autocorrect fail for some original word that I couldn’t discern, and there was noting in the context that made me think otherwise. I didn’t know Claude Shannon from Adam.

16

And of course, the reason why you can get more data along coaxial cable is because you send all the 0s down the outside cylindrical shell, and turn the 1s on end to fit down the narrow wire in the middle.