Slashdot Mirror
Technical Analysis Of VMSK
Phil Karn writes: "Regarding the Slashdot article on VMSK that appeared August 22, 2000, I have
written a detailed technical analysis that shows it to be
snake oil." I'm convinced.
← Back to Stories (view on slashdot.org)
Mr. Karn was just doing his job for Qualcomm, and doing a service for the rest of us.
I suspect it's much more than that. For those not familiar with the name Phil Karn, he was one of the original tcp/ip implementation guys way back in the day. He also has significant interest in wireless data transmission. He was the author of a networking/packet radio package called ka9q, one of the first packages that supported virtually every networking protocol and supported packet radio -> internet. I believe his reasons for investigating the VMSK proposal were largely linked to personal interests. The fact that he is employed by qualcomm only gives him even more credit.
BTW: a Google search on "Phil Karn" will yield virtually infinite interesting results. I highly recommend doing it.
** Martin
and will continue to be "experimental" for as long as people buy into this snake oil stuff.
Walker's first patent for this "technology" has been filed in 1986 and he has probably been working on it for a while before that.
I am familiar with this type of personality - the more refusals they get the more they believe they are some kind of Galileo that one day will be vindicated. The problem is that for every Galileo that is eventually proven to be right about the Earth not being the center of the universe there are a hundred crackpots that believe all kinds of bullshit with the same amout of self conviction.
----
Stop worrying about the risks of nuclear power and start worrying about the risks of not using nuclear power.
I can send 32 bits per cycle! But I have to be able to put 4294967296 different discrete voltage levels and/or timing shifts, or some combination thereof, into each cycle, WITHOUT NOISE MANGLING IT, to accomplish that.
If the noise is a mere 1 microvolt, I'd still have to make each step be at least 2 microvolts to get over it, and then 0xFFFFFFFF would be over 8 thousand volts!
Much of the engineering that takes place now is finding ways to get around the noise, or to reduce the noise, or to minimize the impact of noise. Spread spectrum, for example, statistically changes the effect of the noise in ways that allows retransmissions to have a chance to make it through the next time.
Much of the science is already done and it is very mathematical and statistical. What remains to be done is more and more engineering to work within the relatively well known confines.
now we need to go OSS in diesel cars
We haven't found ways around physical laws. We have found physical laws that we hadn't previously known of. How things were made of molecules, how molecules were made of atoms, how atoms were made of particles, how particles were made of ...well, quarks, probabilities, wave functions...
"At hyper frequencies (1.2ghz and higher) I could see some kind of a spread spectrum system working, but anything lower is plain impossible."
There's nothing magical about high frequencies (well, except maybe popping corn). Spread spectrum can be done at any frequency, but does not allow any faster data transfer than if you had a similar transmitter locked on an unchanging frequency.
"You have a specific capacity at X frequency, Example 108 MHZ - you have a maximum of 108Mbps if you were able to encode on every cycle of the carrier (impossible without generating nasty things)"
Well, maybe you have to define "nasty things", such as "bandwidth". You'll have to encode your signal somehow. A phone line has a bandwidth of about 3KHz, but we can send more than 3Kbps. Part of that is due to sending several frequencies, but most of it is due to using one signal change to indicate more than one bit signal. For example, using two tones to indicate four bits, loudness of the two tones to indicate more bits... You can see a summary of the technology, although real modems are more complex.
The hard part is knowing at what point engineering ends and fiction begins. Sometimes it's easier when someone points out the fiction to you, such as when a picture of a board for sale shows a chip is longer than the socket it is plugged into.
Wake up. Local identification has been dead for a long time.
Have you driven cross-country in the last decade?
Every single radio station sounds exactly like its counterpart in the next market over. For example, pick any two "Modern Rock" stations at random. Chances are they're using the exact same playlist. They pay big money to a consultant for the privilege of being told what to play. Ditto for "Urban Contemporary", double ditto for "Classic Rock."
I agree with you that radio sucks, but look for the culprit elsewhere (hint: deregulation).
ps: The "Alternative" playlist on Music Choice is actually pretty good, with no commercials!
pps: Don't even get me started about local news. "Your kids may be about to EXPLODE!!!!!!! More at 11."
--
whuppy enjoys smelling like diesel fuel
But seriously -- local television gives you a bad case of tunnel-vision. I do feel that it's important to learn about local events, issues, etc. but local television stations don't provide much outside of talk shows and sitcoms. If you want to learn anything, you have to flip to CNN, CSPAN, TLC, Discovery or VH1 (behind the music, of course.)
Local television pisses me off, mmmkay.
--
Wooden armaments to battle your imaginary foes!
Instead of using shorter and shorter symbols, as you propose, the key in the past has been to use longer, more easily identified symbols, and encode more bits into them.
QAM is an often- used multiple-bit-per-symbol system. Each symbol encodes two bits.
Unfortunately, as you add bits to the symbol, it degrades noise immunity. I think that this would be the big breakthrough for comm systems, and the natural law that would have to be gotten "around."
Jeff
Jeff
Here's a totally off base suggestion. What about downloading radio streams to broadband connections in homes, then using wireless ethernet to communicate with special streaming radios?
- I don't care if they globalize against free speech. All my best free thoughts are done in my head.
I don't watch local TV because it's shite.
Headline stories about kittens being left in dumpsters while the real news gets mentioned in a 30 second world round-up. My roommate and I resorted to watching a French language channel just to get some detailed coverage of the protests in Yugoslavia.
Even the cable networks don't cover anything in any kind of depth. If you want decent news you have to get it from overseas. It's sad but I guess that's what the American public likes.
but what I want to know, is why bother writing a detailed paper debunking something that is still experimental?
The guy who wrote the original paper is from Qualcomm- they have a strong interest in being at the forefront of technology. When a new technology like this comes up with such a great stated promise, if it has a remote possibility of being useful, it is in their best interest to investigate it. Otherwise, this situation will happen, which I have been on the receiving end of: Suit comes in for a project review and asks why we haven't been spending his limited money researching this latest and greatest technique he read about in a trade magazine. This causes a shake up all the way down to the people that actually do the work, we investigate, and find that our previous cursory investigation that showed that idea to be worthless was entirely accurate. Total time wasted- about a man month.
Mr. Karn was just doing his job for Qualcomm, and doing a service for the rest of us.
In the mean time, AmWay has announced that it will release plans for it's next generation DSL technology that will increase existing DSL speeds by 150%.
--
Wooden armaments to battle your imaginary foes!
The Shannon-Hartley equation (usually referred to as Shannon's limit) is this:
capacity=bandwidth*log(base2)*(1+Signal/Noise)
The 33Kbit/second limit for a traditional analog/analog phone line comes from this- Signal/Noise is about 256:1 (8 bit sampling), 4KHz bandwidth-- capacity of a 4KHz phone line is about 32000 bits/second.
108MHz leading to 108MBit/sec would only be from simple on off keying- which makes no use of the signal/noise ratio. If you had about 48dB of S/N on 108MHz that leads to a capacity of 860 Mbit/second.