Stuck Profile - and Interleaving question

[#1]

I'll admit to that Mand !

[#2]

Me too. 

[#3]

I agree I've been on the lucky side when it comes to broadband, from both being able to get it in the first place, to the speed it is capable of.
And just its presence has made a difference to the choices you make over work & life. If I moved now, checking the broadband capability would be about the primary check now.
It was rather harder sorting my parents out a year ago though, when it would change daily (at dusk) from easily sustaining a 5MB link, to not being able to hold a link up over a 30 minute span.
The question is though... what turn out to be good figures & bad ones?
I've seen something suggesting that
- things are OK if CRC < 0.1% of SF
- getting iffy at 0.5%
- definite problems when > 1%
Or that
- A level of 5-10 ES in 15 minutes (900s) is manageable - again around the 0.5 - 1% level.
Are there any rules of thumb you use on these error counts, as opposed to the noise margin (of which there is a lot of info out there)?
  Mike

[#4]

Just manipulating a few numbers, I think Errored Seconds represent complete seconds when no data gets through due to problems (errors). As you stated 1% of 900 seconds is 9 seconds, so that level could be assumed to be equivalent to the CRC 'definite problems' limit (1% of SF) that you mentioned.
So possibly a rule of thumb is that 1 ES per 15 minutes is tolerable, 5 ES per 15 minutes is 'iffy', 10 ES per 15 minutes is 'dodgy'. Personally I'd feel more comfortable with a number less than 1 per 15 minutes. Rather an intangible question though.

[#5]

Here's an update on my experience of turning interleaving off, so far, and some observations.
Remember that, according to the attenuation level of 30dB, my line is *just* capable of sustaining the 8MB level.
1. Turning Interleaving off has also turned off the "reed solomon" forward error-correction.
It is the loss of the RS that turns the sync rate up from 7616 to 8128, as you regain the extra error-correction bits in the bandwidth.
Loss of RS has gained approx 7% raw bandwidth (ignoring the effect of the profile change, if a change ever happens).
2. The reed-solomon stats posted earlier in the threads suggest that the RS would correct around 97% of the errors on the line, leaving 3% to become CRC errors.
Those CRC errors would end up at the rate of 1 fault per hour.
This suggests there were perhaps 30 faults per hour on the line, of which 29 were corrected.
3. Without reed-solomon, the line is now seeing more CRC errors.
The stats (graphed as attachments) suggest a rate of 7 faults per 15 minute period, or 28 faults per hour.
Those faults need retrying, and will reduce the raw bandwidth slightly. The ADSL frame rate seems to be 60 per second, so if the CRC applies to a full frame, then 28 faults per hour is around is around 0.01%.
4. I have no perception of that level of CRC faults - everything seems to be working perfectly well.
5. The reported value for "Errored Seconds" usually changes identically to the "CRC" counter.
That suggests that 1 single  CRC error in a second-interval is enough to make that second flag as "errored", rather than it being "all frames in the second".
I haven't yet triggered anything that flags as a "severely errored second".
6. I have triggered values of the "UAS" (I assume it means "unavailable seconds"), but only when I power-cycle the modem to force a resync.
This kicks the UAS value by somewhere in the region of 10-20 seconds.
I hadn't quite realised that the interleaving change would have the knock-on impact on the RS change, and the consequent impact to the CRC rate.
The jump from 1 CRC per hour to 7 CRC per 15 minutes seems to be manageable though...

[#6]

An Update:
2 days ago, I had a go at upgrading the firmware on the router. This didn't go as smoothly as desired, so I will have had a few disconnections on the line.
I disconnected it from the BT line for the most part, but there will still have been 5 or 6 disconnections over a 4 hour period.
This had a positive effect... the BT profile jumped up to 7150 sometime during Monday, and the Plusnet profile has jumped up to match it sometime today.
I think that's the first time that the profile at either end has gone up automatically...
Thanks for the help on this guys!
  Mike