GNet vs Siemens vs Thomson - DownStream SNR Margin & Rx CRC
Finally...
I wondered about this one for a long time. Something told me my GNet could compare with my SpeedStream in terms of noise but it's DownStream SNR Margin was not expected to beat those of my Siemens SS4200 & of my Thomson ST546 altogether. The mix of optimum SNR Margin and of Noise performance may result from lower Attenuation levels - which could relate with the input stage design. Should it be true, the Attenuation levels observed while evaluating different DSL MoDems may happen to be an important parameter for persons trying to decide which product to buy! There are other criterias to consider than noise performance alone but it seems possible Attenuation level variations between products might be more significative than i suspected.
I had one large capture gathered this winter which remained to be parsed... It's an interesting sample for me because i've noticed a "bias" effect after my record was separated into two sets: one which has nicely aligned time-stamps and another where those hour marks reveal some "jitter". See what happened when i decided to superpose the curves instead of averaging them as usual!... In my opinion, it's partly due to the rejection of marginal "sparkle" events: it defines/flattens a curve when those aren't allowed to contaminate it. As a consequence, this suggests i'll need to pay more attention to the duration of my future tests (if i decide to go on) in order to make sure that each curve undergoes a same amount of filtering, hence removing the same amount of bias...
