I needed to know the answer to this question for an upcoming Marketwatch column. So I asked superstar engineer Milo Medin what he thought since he was the chief architect for the @Home cable modem rollout and a cable modem pioneer. The barebones answer is about 100 Mbps down. But the real answer is more complicated and interesting enough to post on the blog.

Question: What is the fastest practical Internet access speed that can be delivered over the cable Tv/Cable modem network?

Answer: Well, doing something straight forward would be taking your best plant, 860 Mhz, which has 135 6 mhz channels in the forward direction and digitizing all of it with a 256 QAM modulation scheme. That would yield 135*42 Mbps = 5.670 Gbps if you digitized all the forward on a modern cable plant.

Now, that is a shared network, but you can reduce the sharing quite a bit without getting too expensive by using DWDM wavelengths over fiber. A couple test markets in the US went down to 50 homes passed per optical node so they could eliminate line amps (all the homes were driven by the fiber node). If you then took that entire digitized payload and fed 50 homes by that, each home could be running more than 100 Mbps downstream continuously. So there is a ton of bandwidth available there.

Now, there are some issues:

1) You aren’t going to be able to remove all analog from the plant. Basic cable will almost always be analog, and while that’s inefficient, it’s not likely to change. Still, that’s maybe 15 channels, so there is still a good chunk left.

2) The upstream will be a problem. You have to have enough upstream data capacity to at least service TCP acks for the downstream data, at least if you are using TCP for transport. Given you only have 5-42 in the reverse, as oppose to 50-860 in the forward, that’s a larger limiter. You can subdivide that down so fewer and fewer users are sharing that, but to the extent that is also down in the forward, it doesn’t give you much leverage. On top of the fact it’s only 37 Mhz, it’s a noisy environment, so only about half that is useful. So call that 18 Mhz. Now, that also cannot be coded at the same rate, because of the noise, so you can get maybe a max of 16 QAM in the reverse. That means about a max of 57 Mbps in the reverse, or about 1/100th that of the downstream. That limits you, as you need probably a minimum of 1/10th of the download speed in the reverse to effectively use the downstream.

3) Cable modems typically don’t tune across a large swath of channels like that. That is, each 6 Mhz channel does about 42 Mbps of data rate in the downstream (256QAM). To be able to go faster than that, an individual model must be able to stripe data across multiple channels, and that means being able to tune multiple channels and decode them. That affects cost of the device.

Milo

Now you know.



  1. QWERTY says:

    Fastest speed? If you carry it, about 20 MPH. In a car, more like 60 or 70 MPH (100 km/h). Airplanes will buy a bit more speed, but if you want the cable modem to go really fast, put it on a rocket!

  2. ECA says:

    and in the END, thats why they settle for the Cheapest format, they can use, that has a few options.

  3. xrayspex says:

    In the words of Andy Tannenbaum (as best I can remember)

    “Never underestimate the bandwidth of a station wagon full of 9 track tapes going down the interstate at 70mph.”

    You’re looking at 3,000,000 ms ping times, though.

    Milo’s experiment has a latency issue, too. Not that I wouldn’t give it a shot.

  4. James Hill says:

    Great write-up. Hopefully this means an increase in speed in the near future.

  5. JoaoPT says:

    TCP involves a lot of overhead. UDP is better for media, because it doesn’t need acknoledgement…and arguably such a high bandwidth is suited for TV or audio.
    But for critical data TCP is crucial, since it has mechanisms of acknoledgement and retry of missed or corrupt packets.
    Anyway if such high bandwidth is available, I guess there can be programs using UDP and some way of requesting lost packets, working on top of standards without breaking them, to send data.

  6. James: LOL!!!

  7. rctaylor says:

    This is way off topic, but the Google ads that popped up on the site was for 2 Roman Catholic sites, a religious statuary store, a motorcycle tour of Alaska, and a church of Satan. That should cover all of your readership.

  8. James Hill says:

    I’m glad you got the joke.

    While this is interesting, we’ll never see these speeds over coax. The only potential mover in the bandwidth game is the idea that Google is going to blanket the nation with free ‘net access, which sounds about as real as Santa and the Tooth Fairy.

    Hopefully fiber to the curb movements will continue across the country, as one currently is here in Utah.

  9. andrewj says:

    We’ll see those speeds after they run another cable to every home, just for the upload. Cable was invented for one way data transmission, only.

  10. Mike Voice says:

    8 Hopefully fiber to the curb movements will continue across the country, as one currently is here in Utah.

    Ditto.

    My town’s monthly newsletter says we’re on Verizon’s list for 2006-2007.

    Not holding my breath, but ready to jump… 🙂

  11. techno-bla-speak-ka-bibble. The language of the valley.

    Makes sense to me!

    =100

  12. Rich says:

    Why are we still screwing with using existing wiring (eg POTS dial-up, DSL, cable tv) for Internet access? I know it would be expensive, but couldn’t we run dedicated coaxial cables? Or are we reserving that tactic for a dedicated fiber-optic scheme? Imagine the bandwidth THAT could accomodate.

  13. ECA says:

    As one person pointed out to me…

    Its SOP, that when something becomes popular and sells very well….Prices go up.
    Can you see all the corps buying out all the Coax, and the prices go UP 4 times, the price…STUPID aint it.


0

Bad Behavior has blocked 5631 access attempts in the last 7 days.