I'm having problems keeping up here, with trying to sort out my power problems (and some sort of access rights problem I'm going through with HG612 Modem Stats), but I thought I'd add something here...
Who is going to interpret/explain all those extra graphs to a new member which is having a line issue ?
At the moment a rtx_tx per min above 90 will give me 1 errored second on DSLstats.
& what about rtx_c & rtx_uc?
I THINK it will be the rtx_uc that causes one or more CRCs, which trigger the ES (but I'm not sure).
It is pretty tricky to interpret overall. Having seen the effect of a dicky power supply on DLM, and as DLM intervenes more, I can say that some things happen backwards now: The FEC/interleaving settings seem to reduce while the INP setting increases.
However, I think the retransmission counters work like this:
rtx_tx gets incremented, at the sending side, for at each retransmission. If something gets retransmitted more than once, the counter gets incremented multiple times.
rtx_c gets incremented at the receiving side when a retransmission is received correctly.
rtx_uc gets incremented at the receiving side when a transmission is determined to have never been received correctly, i.e. after a timeout with perhaps multiple retransmissions having been attempted but failed.
Therefore rtx_tx >= (rtx_c + rtx_uc).
On my statistics output (Billion 8800NL, Broadcom), I think the rtx_tx value in the first column aligns with the rtx_c and rtx_uc values from the second column, and vice-versa. The columns are probably best labelled as the near-end and far-end, rather than downstream and upstream.
Except ... I think rtx_uc can get incremented a lot in the event of a loss of service - perhaps because of DLM triggering a resync. I have one such counter set to a large value (72,000) which is an order of magnitude more than the other counters, and it hasn't increased in days. I suspect it got set that high when DLM de-intervened, but the power issues mean I haven't tracked the changes well enough to know.
I don't think any of these rtx_* counters impact on the CRC or ES counters.