Slashdot Mirror
Cringley On Bandwidth-Expanding Modulation Technology
jtappan writes: "Robert X Cringely has an article describing a new modulation technology that will allegedly allow cable modems to run 10 times as fast, and which will eventually allow existing cable networks to carry 500 HDTV channels."
Wow, so the cable companies can now run 10 or more times the amount of homes on the same network segment yay! (and you know thats EXACTY what they will do too) :( (obviously this number will be lower if they are offering internet connections through the same wire, but you get the idea)
BTW the way many cable companies digital TV works, is that when you change channels your tv sends a signal up the wire saying that you want to see such and such channel. So they start to stream channel X down your network segment. So having the ability to run 500 HDTV channels would only mean that they can now run 500 TVs on the same segment YAY
This is similar to what modems do. AFAIK, they still don't run any faster than 3750 baud (Hz),
but they can encode up to 15 bits per wave to get 56kbit/sec. If the line isn't so quiet, they cannot distinguish all 15 bits, so the modems have to negotiate a constellation with fewer bits.
My question is how this will work with an ethernet-like collison detection system that AFAIK cable modems use. The jam signals could get ugly, and I'm not sure you can carry as my info on broadband as baseband systems. Or how cable decoders will cope.
while i enjoy cringley columns, his mangling of the bottom layers of the OSI model made me cringe (pun intended).
encoding systems are physical (layer 1) technologies, not 2nd layer like he claims. he further states that ethernet and token ring are layer 3 technologies, which is blatently false - they are both data link technologies.
maybe i'm just being nitpicky....
I'd rather see advances in backbone speed than last mile speed, thank you. Cable modems are already capped at a fraction of their potential because of insufficient capacity at the ISP side. Give the ISP a couple of gigabit connections, open up the cablemodems to 10mbits, and I'll be perfectly happy, for a couple of months anyway...
Nobody said you'd be constantly streaming 10Gbps all the time and saving it to disk. To me it's more about how quickly a page downloads, not how much stuff I can download overall. How much time do you spend reading a page vs. downloading it? Take this comments page for example, I would easily spend 5 minutes reading everything. As it is, the page only takes 5 seconds to download, but, if that could be decreased to near instantaneous I'd love it.
An entire web page and all its related files (even graphic/sound/flash heavy pages) could easily fit in most modern PC's RAM. Stream it all direct to RAM and pop it up on the page? Why save it to disk at all? For your cache? You wouldn't need a cache if you connection were that snappy. And just think, we could actually stream streaming video instead of spooling streaming video... No disk involved.
I could see ISPs moving away from limiting your instantaneous banwidth (i.e. capping you at 1.5Mb/sec) and moving towards capping your average bandwidth (i.e. 5Gb/hr). I mean, so what if I choose to eat up my hourly bandwidth allocation (say, by downloading several linux distros simulataneously) in 0.5 seconds instead of an hour? (Technical issues of me saving off that much data that fast, aside.) The overall useage from the ISP is the same. OK, so maybe it takes me 2 seconds instead because there are 4 people queued up ahead of me with big downloads. It would still be very snappy in comparison to today's setups.
Wavelets are an alternative to Fourier Transformation of time domain data to obtain a functional decomposition of the waveform for analyis or processing. They are particularly useful with choppy or spikey signals.
It's a very fundamental mathematical tool for any kind of signal processing application. As such it has a wide range of applications. It came into wide use perhaps 15 years ago; perhaps you were out of school by then. I am sure that every EE undergraduate is getting exposure to wavelets these days.
Here is a link to resources on Wavelets:
http://www.mathsoft.com/wavelets.html
The wavelet modulation is sound, and I don't doubt that rates that high could be achieved. As the author notes, the advantage that wavelets have is their resilence to noise, in particular, interference from the signals already in the channel. However, what he doesn't note is that the interfernce is from the existing signals on the line, signals that aren't immune to interfernce, including interfernce from the wavelets. I don't think that will make the existing users very happy.
You are not being too nitpicky. Cringely is an idiot^Wjournalist, not an electrical engineer.
:-)
I, too, was cringing when I read the article. He JUST DOESN'T GET IT. Layer 1 is what defines token ring and ethernet, not layer 3 (network addressing). Even if this rainmaker technology wasn't a scam, layer 1 is where you define both the physical medium and the signal modulation that works best with the medium. Changing TV cable modulation would cause tons of knock on effects, with cross channel interference, harmonics, parasitics, and probably Nyquist reflections cancelling out other channels.
And I know far too much about QAM, as it is used in modems. QAM has existed for decades. It isn't used on cable systems because there is no way to keep the signal clean enough to recover a tight constellation on grungy, up in the air exposed to the elements cable systems. Shannon's limits on recovering signals from noise get slowly pushed back from time to time, but his model is still sound. Its not going to be replaced by wavelets or whatever the scam buzzword of the week is.
As for costing US$10, HA! The cable companies would have to replace their entire HFC plant, and every repeater, splitter and signal booster to work with signals that filled each 6MHz channel with wall-to-wall noise. Most of the cable companies offering internet have just placed a little piggyback backchannel filter around each of their repeaters to get a single channel back to the HFC headend. They haven't replaced all the repeaters or much of anything, and they still grumble about the cost.
Nope. rXc deserves to be kicked around for this shameful piece of drivel. And slashdot is just the place to do it
the AC
Hemos is like...sci-fi fans;he thinks technology is cool, but he hasn't bothered to understand the science it's based on
Meanwhile, CableLabs just rolled out DOCSIS 2.0 with new upstream PHY (two different modulations that MUST be implemented, because CableLabs couldn't deside which one to use!), so the roadmap is pretty much known for the next 10 years.
Mr. Z is correct -- moderators, promote the note I'm replying to! Cringely is wrong about layers. Layer 1 does all the bit stuff, including modulation and even ATM cells. (Layer 2 is about user-sized frames and error detection.)
But that's not what matters. Shannon matters. You can't defeat Shannon, and Cringely admits it. So let's see... Shannon basically says that the limit of bps is proportionate to the product of bandwidth times the log2 of the signal to noise ratio. So if you have an infinite SNR, you can have infinite bandwidth. But getting 33 Mbps (around the top end of DOCSIS cable modems) requires good SNR. My cable modem right now has 36 dB SNR and is running QAM64; DOCSIS adapts speed to line quality.
So even if wavelets were better than QAM (and I can't say, because Cringely doesn't tell enough to know if this is real or a scam), there's just not that much more you can do in 36 dB! (Shannon limit of 6 MHz at 36 dB is around 6M*12=72 Mbps.)
1024QAM does give you 25% improvement over 256QAM - after all, it packs 10 bits in space where 8 bits are now. The wavelet may do some additional magic with sidebands, but if you use plain old 256QAM on 18 MHZ channel, you will get about 120 Mbit/s. 40% improvement is good, but is it good enough to convince cable companies to change standards?