[Sephiroth]

A primer (I hope).

It's all suitably complex. You have a two way circuit from your perspective.

Downstream power is needed to send data TO you from the server (CMTS). Upstream power is needed to send data to the cable modem FROM the CMTS.

The CMTS controls everything. It sends out its data at a fixed power value of 50 dBmv (I won't explain the realtive power values here), calculated in theory to arrive at your modem with 0 dBmv (which represents a nominal power of, I think, 75 milliwatts - BBings will jump on this if I'm wrong). The cable modem is designed to work with lower or higher power levels because the path to the modem is not ideal; it may be nearer to or further from the street box, for example. IMO (and others have their own view) a range -3 dBmv to +6 dBmv is unlikely to cause problems. Too much power brings too much noise.

The cable modem estimates the signal to noise ratio (SNR) based on what it has demodulated with/without errors. This is presented in the modem stats. For reasons too complex to explain in this primer, the data density used in your downstream (256QAM!) needs the SNR to be higher than 30 dB at all costs; > 33 dB to handle effects at the margin and 35 dB to be on the safe side. The CMTS can only react to a poor SNR by lowering the modulation to 64QAM which has a lower low SNR threshold.

In the other direction, the CMTS estimates the SNR on the route from the cable modem based on initial ranging requests. It then commands the cable modem as to the power level needed to overcome the estimated noise impairment. The CMTS does this very frequently so as to maintain its upstream circuit and only when it fails does the modem event log report something. The cable modem maxes out at 58 or 61 dBmv depending on model. If you're over 55 dBmv upstream, there's a struggle to get data to the CMTS.

The potential causes of upstream impairments are many - from loosely screwed in coax in the home or on the other side of the wall, through RF noise ingress due to corrosion (dipole effect), badd connexions in the street box, failing laser, noise reflection - it's amazing it works at all because so much of the circuit is copper and modulators/demodulators, combiners and so on.

I might have misunderstood what he's fitted by way of an attenuator. He would have fitted a forward path attenuator to lower downstream power output - possibly because you're close to the streetbox and possibly connected to the wrong attenuation block in the street box.

After all that spiel above, perhaps you could pst your modem stats so we can attempt to judge what's going on, plus the event log REMOVING ANY CM-MAC ADDRESS INFORMATION.

HTH.