said by BadMagpie:It looks like the 30a bandplan has an upstream channel going from 22 MHz all the way to 30 MHz. How much far-end crosstalk would there be if this entire band were used, with little of the other upstream bands, given that its frequencies are beyond anything 17a uses? (Feel free to feed me technical documents and explanations, I have a background in engineering)
30A wouldn't help at all, it has the same problems as 17A (which has an additional upstream band that 8B doesn't). 30A's U3 band goes from 23-30mhz, which sounds great, but because the frequencies are so high they get no bitloading past about 150m (30A is only 100mbps symmetric at up to around 130m). At which point the upstream performance deteriorates to 'slightly better' than 17A.
30A uses a wider tone spacing (8.625khz) and twice the symbol rate (8 ksymbol/second), this coupled with the wider bandwidth causes the line card density to be reduced by 50% (additional FFT/iFFT logic is required on the line card DSP's). Thus reducing a 192 port DSLAM to 96 port. On the CPE side, the BCM6368 chipset supports 30A, but it requires an additional AFE and a different line driver to support single-line 30A mode. Hopefully, it's clear why 30A is almost never deployed in practice =).
If you want to learn how to remove far-end crosstalk, read some of John Cioffi's papers on vectoring:
»
www.stanford.edu/group/c ··· pub.htmlThe ITU-T has standardized on G993.5 as a method to use linear precoding to remove far-end crosstalk on VDSL2.