how does the DSLAM control upstream power ? I can see how it might send data to the modem but the modem does the transmitting and the DSLAM merely listens to it.

There is still a maximum power per tone, which is what I was driving at, as fixed by the power mask. If the lower speed means less tones, rather than less bits/tone, then there'll be less power. Power cutback is seldom seen in the wild for "optimisation" but a regulatory mask is A Different Thing (tm).

Conversely if the power mask is "in the way" there'll be less SNR per tone and hence less bits.

The ANFP requires upstream power masking per line based on D side length, just like it requires downstream masking based on E side length.

well the modem is the device doing the transmitting, but most likely the parameters are sent to the modem during the sync process.

To make this even more interesting, my 36mbit upload attainable bitloading with the highest TX power at 6.6 is also not using all the tones for the upstream, since it had so much excess snrm the tones between 1000 and 1250 werent used. So it was using more power across less tones.

Edited by Chrysalis (Fri 15-Feb-13 22:45:27)

I have seen different strategies used by different kit where they easily meet the required speed like in your upstream case. One modem used a small number of bits per bin across the whole range and another piled lots of bits into the low bins until it met the speed spec.

The first would allow power cutback to be used across the range, the second used less power due to less tones.

With a limit on total power and on power per tone there are many permutations I guess, without throwing "Green DSL" into the pot and getting 80% of achievable speed for 50% of the power or whatever. I don't think we're seeing power cutback used other than on lines at maximum sync, are we ?

no idea.

Crosstalk can also cause crc errors even without an impact on speed, so it may be there is lines in any given area (maybe my area) that are having stability issues so then power is cutback to reduce crosstalk to manage the errors, maybe not I dont know.

To be fair I havent read documentation anywhere that states CP's use powermasking to manage downstream crosstalk (only upstream) and on downstream to protect adsl services, I just assumed if its used on the upstream it logical its used on the downstream as well. Then I made further assumptions based on what some end users have reported and what I am seeing on my own line. But in the VDSL spec's power can be cutback on individual tones if desired.

An end user eg. is more likely to notice a 15mbit speed drop off to 5mbit than someone on 80mbit drop off to 70mbit. So a CP making choices on who to hinder for the sake of fairness and all that maybe chooses to help the 15mbit user over the 80mbit user. Indeed I havent reported this as a fault, but if I had a 15mbit sync dropoff to 5mbit I would almost certianly be crying about it to tech support on a regular basis.

In terms of crc errors. on the 5.8 power my upstream had about 50% of the downstream crc error rate. On 6.1 its now about 20% with the extra margin, on 6.6 it was extremely low.

To me it looks fairly clearcut my upstream power is been fiddled with, but far less clearcut on the downstream. The downstream could easily be just down to huge crosstalk.