WWWombat wasn't discussing anything to do with target SNRMs. He was referring to the fact the OP is capped at 40mb with an SNRM of 10.3dB, so there's 4.3dB spare.
He's guessing the OP would get 8mb more if they upgraded package, and removed the 40mb cap.
This is true.
However, William's line happened to make use of "spare SNRM", though only 3dB, through the 5-4-3dB trial.
I'd expect lines to see roughly(*) similar speed improvements, no matter what method was employed to make use of an extra 3dB margin.
(*) My rule of thumb was originally calculated by examining the "bits/tone" graphs at various speeds, and estimating how many tones were in use at different speeds (explained below), but it has proven to be remarkably resilient so far.
As a 3dB improvement in SNRM allows 1 extra bit to be carried per symbol, on each tone, the improved speed comes from multiplying the count of usable tones by 1bit per symbol, by 4000 symbols per second.
A complete downstream set of tones is 2,748 tones. 1 bit on every tone amounts to 11Mbps. That's the best you can hope for.
The gains fall into 3 main bands because of the way spectrum is split:
- 10-11Mbps for lines using D1, D2 and D3
- 6-8Mbps for lines using D1 and D2
- 2-3Mbps for lines using D1
Beyond this, however, there are many variations in behaviour. Some lines get 40Mbps by putting a few bits/tone on a lot of tones - these are likely to gain more than the rule-of-thumb. Some lines get 40Mbps by using a lot of bits/tone on fewer tones, and are likely to gain less by this change. It will depend on the interaction between Hlog and QLN graphs.
Its rough - but that is the definition of a rule-of-thumb, isn't it?