bill c. - Krista J.
>>I've read about SDMT is some of Krista Jacobsen papers.... From my understanding today... a DMT flavor will not make it into the standard on the upstream path. I'm taking the most negative view here, and as I learn more, hopefully it will turn positive for AMTX on this front. <<
You might say I've keep an eye on Krista's work.
I met her at SuperComm, she was incognito. No badge was pre-made for her so she wore a modified one of DR. Cioffi's. It said JOAN CIOFFI.
(Hi KJ, for a bunch of investors, we do dig deep :)
The battle has been going on so long, I've almost lost interest in keeping up, but unfortunatly I'm insatiable......
It was fun to watch, but the process is long and drawn out.
Here's some fun from the archives back in 95
Give-em hell Krista.......
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
THIS INFO DEALS WITH HFC, NOT POTS
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
September 12-14, 1995 Interim meeting
Tuesday, September 12, 1995
(Snip)
Upstream Ingress Noise Model
Walter Chen (of Bellcore) presented work on an upstream noise model for HFC (IEEE 802.14 -95/ 110).
Some of it was based on work by CableLabs and AT&T. He believes the wideband interferers on the CableLabs spectral plots are actually clusters of shortwave radios. A spectral plot with filters the drops proves that much ingress occurs in homes and drops. A slide using AT&T data showed very uniform per-subscriber noise contribution. A question was asked about the experimental methodology here. Chen said the results were from traditional cable plants, but admitted they weren't all the same. Chen converted measured ingress noise power to field strength and felt that the numbers matched expected field strengths due to shortwave and radio background freely well. Citta stated a belief that while statistically, ingress noise power will increase monotonically with number of subscribers, that the variance in the contribution of each subscriber is high. Chen thought that contradicted the AT&T data. Citta said, in essence, "trust me". Chen suggested that grounding problems at the drop could be an ingress mechanism Citta disagreed with his loss figures. There was an inconclusive discussion. Chen concluded with a description of a small cable plant they're building to measure ingress, and a short description of the ingress model mentioned earlier.
2.7 Performance Simulation for DMT proposal
Chen continued with a second presentation (IEEE 802.14-95/ 111) on a performance evaluation they'd made of Amati's DMT system. He thought the results showed good potential for DMT. He first recapped the parameters of the system. He showed an interesting plot of mean error vs carrier frequency deviation (between tones). Over a 0-10 ppm CF deviation, it showed an error increase between 10-16 dB. Timing deviation was much less harmful, at least as long as it didn't exceed the cyclic prefix length. He showed an ingress noise model slide, then one about performance with such noise. The performance slide assumed the good carriers used 2048-QAM. This caused general astonishment. In conclusion, Chen recommended a MAC scheme that could accommodate either single- or multi-carrier. He suggested the best way to manage the sensitivity to CF deviation would be to broadcast a master clock that all stations would lock to. He claimed lllal DMT could get 5-6 times the spectral performance of QPSK. Ungerboeck disputed this claim.
He pointed out that 6x QPSK is 12 bits/Hz. He referred to the disputes involved in standardizing V.34, saying there were extensive single-carrier versus multi-carrier debates, and that the standard ended up single-carrier. He claimed that from a spectral efficiency point of view, single-corner and multi-carrier are ultimately equivalents Krista Jacobsen objected that this was true "only under an optimal single-carrier [noise?] model". General disbelief was expressed that you can get 12 bits/Hz in real systems."""""
JW@KSC |
