The example you describe would certainly lead you to believe there was a problem between the TX and CMTS, but that definitely doesn't have to be the case. Upstream SNR is more like MER and is reported from the CMTS; therefore, anything that degrades the upstream carrier from the modem to the CMTS will degrade upstream MER, noise/ingress is just one of the possible problems.
Some of the other things to consider:
RF drive levels into the TX must be correct for the system upstream loading; channel loading, NPR curve and dynamic range must be considered. You could have the situation you described above and have horrible MER back at the head end.
The type of upstream platform makes a huge difference. As in this thread; BDR II perform only fairly (as compared to other systems) even when set up for optimum performance. BDR I platforms with 10 bit RX will have trouble in a 64 QAM system...etc.
Many older modems don't play well in the newer DOCSIS 2.0/3.0 environment and will report bad MER.
Modems at the end of longer cascades can show MER problems due to group delay.
Any upstream mismatch affecting the carrier will degrade MER. Remember, if you use sweep to balance the return; you are not sweeping through the DOCSIS carriers, therefore may be missing problems. The bandwidth not being swept is getting very large...other tools must be used. If the DOCSIS carrier is running into path problems, the upstream MER will be affected.
The reported number is based on MER taken from good packets only (Bad packets, the ones possibly lost to noise or problems, aren't included!) and there must be a certain amount of packets over a short time for the number to be close to the real number. Once the MER gets over 30 or under 22, it is more of an estimate.
That's all I can think of for now; I hope this helps.
