Last weekend we had two visits from two AT&T techs, and another today. The first visit was from a premises tech who concluded that our internal wiring was sub-standard, so he and I climbed around and drilled holes to run a length of CAT5e cable from the
DMARC (aka MPOE) box to a new isolated jack. He put a line filter into the DMARC itself, which makes the install very clean. Afterwards he found that our feed line had a couple of
bridge taps which would need to be removed, so he submitted a service request ticket for that.
Observation: The premises tech was a younger guy, and in the usual banter of running wire he told me he'd been in the US Marines where he did communications and technology work. During the clean-up phase of the wire install the tech managed to staple into the CAT5e itself. It happens, and it's not usually a big deal since you have four pairs to work with - you just pick another pair. I suggested that we "ohm out" the wires since we still had easy access to the ends, and that way we could avoid a debug session later. He didn't have an ohmmeter with him, which I thought was odd. Checking the wire pairs is fairly easy - you strip the insulation from the ends, twist one end of each pair together, and then measure resistance of each pair. If any are obviously higher than the others, that's your problem pair and you avoid it. Turns out he had no experience with things like this! His military technology education had been along the lines of "If this box goes bad, remove it and install a new box." When I went through US Coast Guard electronics school in the early 80's we were taught basic electronics, Ohm's Law, circuit-level troubleshooting, and even vacuum tube circuits before moving on to systems-level work. Clearly a major shift in military training has occurred since then, and not for the better.
On Sunday the line tech showed up. He removed two bridge taps from our line, and also cleared out a rodent nest from the pole-top splice boot. He declared the line clean, although the signal level seemed a bit marginal (about 10 - 11 dBm). As the day went on I monitored the modem's diagnostic page and noted that we were getting a LOT of forward error correction (FEC) errors. FEC errors are considered correctable errors (unlike CRC errors which are uncorrectable) but they still cause lower performance and indicate something is wrong. Tuesday night the network went offline - the logged FEC count in 48 hours had reached almost 100,000 and the connection speed was below 1 Mbps. I tested using the modem's diagnostics and saw IP errors, IPv6 errors, and a lot of DNS failures. Back on the phone with AT&T...
By now I've been on the phone with AT&T enough (almost every day) that they've given me a special Tier 2 support number and a passcode for it. This gets me straight into their domestic tech support line, which is nice. Another call to Tier 2, more tests from their end, another tech visit scheduled. Today's tech agreed that the FEC errors indicated a problem, and was about to call in an order to have my DSL line card swapped at the central office. I asked him if he had seen a lot of problems with the NVG510 modem. (He thought I'd already had a modem swap, but I hadn't.) I asked him if he considered the NVG589 more stable. The NVG589 is only for VDSL installs, but he said "Hold on." and came back with a new 5168NV modem. From what I've read, these are the "go forward" modems which AT&T will standardize on since they support everything from ADSL up through VDSL2. The 5168NV (
datasheet) also offers 802.11n 2x2 MIMO, 400 mW Wi-Fi power, and has a dual-core processor which speeds up recovery from retrains and allows faster adaptation to spectral interference. One side benefit of the 5168NV is that the downlink receiver has much better performance than the NVG510 - I'm seeing +18 dBm on the downlink versus the previous 10 - 11 dBm. So far the performance has been very good, Netflix picture quality is a lot better, websites are more responsive, etc.
