VDSL Performance & Tuning

Forums › New Zealand Broadband › VDSL Performance & Tuning

1 | ... | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | 75 | 76 | ... | 285

1598 posts

Uber Geek
+1 received by user: 252

Trusted

#802451 19-Apr-2013 18:49

Thanks for the update redstorm! This gives us a better idea of the upper limit of CRCPM allowed on DLM-1. Did it stay at round 12 before the resync this morning? (post the 'statistics' page from your Fritz if you're not sure).

redstorm

56 posts

Master Geek

#802479 19-Apr-2013 19:22

SamF: Thanks for the update redstorm! This gives us a better idea of the upper limit of CRCPM allowed on DLM-1. Did it stay at round 12 before the resync this morning? (post the 'statistics' page from your Fritz if you're not sure).

The bar on 3am = 187

SamF

1598 posts

Uber Geek
+1 received by user: 252

Trusted

#802484 19-Apr-2013 19:31

Thanks mate.

That's a bit odd how it drops off so much without you making any changes.

SamF

1598 posts

Uber Geek
+1 received by user: 252

Trusted

#804323 23-Apr-2013 14:25

Ok, so I've been doing a bit of testing and research lately, trying to discover the upper limit of CRCPM errors while on DLM-1 (I'll post the results later). However, last night, something odd happened. I'm currently on -6db with a sync rate of 56200 and a subsequent SNRM of 5. CRCPM errors were ~9 averaged over the last 3 days, but then started rising rapidly later in the day to round 14 (over a single hour). Then later in the evening the SNRM dropped 3db to 2 (no resync / connect speed change) and CRCPM went DOWN to about 1/5th of what they were prior to the SNRM drop!? Real world throughput (Speedtest) is unchanged.

Subsequently the SNRM is still changing by itself and the errors are still sitting around 4 CRCPM! What kind of DLM trickery is this!? The only comparable thing I can think of is the 3db SNRM drop that some stevehodge experienced just prior to being swapped onto a 17a profile, but my attenuation (9) is way too high for that. Also, stevehodge's error rate went UP after the SNRM drop.

Has anyone seen anything like this before? It's making no sense to me! Granted I am pushing the boundaries of 'normal', but still, a very odd outcome!

redstorm

56 posts

Master Geek

#804331 23-Apr-2013 14:46

SamF: Thanks mate.

That's a bit odd how it drops off so much without you making any changes.

I suspect it occurs when all the mum and dads turn off their dsl for the night reducing the crosstalk on the bundled lines from the cabinet.

SamF

1598 posts

Uber Geek
+1 received by user: 252

Trusted

#804333 23-Apr-2013 14:49

Hmm, I dunno man, that's a MASSIVE drop!

Lego

Shop now for Lego sets and other gifts (affiliate link).

sidefx

3775 posts

Uber Geek
+1 received by user: 1295

Trusted

#804341 23-Apr-2013 15:10

And who actually turns off their dsl these days? :)

Weird. Maybe something that the fritzbox doesn't show has changed?

Or maybe it's going to bump you to 17a... *fingers crossed* ;-)

"I was born not knowing and have had only a little time to change that here and there." | Octopus Energy | Sharesies
- Richard Feynman

richms

29114 posts

Uber Geek
+1 received by user: 10229

Trusted

Lifetime subscriber

#804623 23-Apr-2013 23:13

Doesnt need to be turned off. Adsl 2 backs off when not needed so on and idle is going to interfere less

Richard rich.ms

SamF

1598 posts

Uber Geek
+1 received by user: 252

Trusted

#804625 23-Apr-2013 23:26

As I mentioned earlier, recently I have been exploring the boundaries of DLM-1 and the number of CRCPM errors that is tolerated on this profile. My basic conclusion is that I got up to a connection average of 9.15 CRCPM and am still on DLM-1 after 4 days. Based on this and information from other posters, I believe that the CRCPM limit for DLM-1 is highly likely to be 10.

During my experiments I collected some info which you guys might find useful / interesting. Obviously this is only for my line and other lines may have different results, but there are a number of trends which I believe would be applicable to any VDSL line. Also, the fact that, in order to produce this many errors on my line, I had to push it significantly beyond the 50mbit 'limit' means that this may have yielded different results to what would otherwise be seen on a non-tweaked line. However, I think the principal is similar enough and the results do seem to be supported by other people's real world experiences. Here is the raw data I collected:

Date   SNRM Tweak   SNRM Actual   Sync   Avg CRCPM   Max CRCPM   Speedtest
7 M          0.0             11            49.5      0.09      0.13          46.00
11 N      -4.5      7      54.5      1.85                2.88         51.00
14 M         -5.0            6             54.8          2.00            3.20          51.40
16 N       -5.5         5-6         55.5    5.95            7.95         52.05
19 N       -6.0        5         56.2    8.75         20.41      52.75

Notes:
Date - M = Morning, N = Night
Avg CRCPM = Average CRCPM over the entire connect session
Max CRCPM = Maximum CRCPM in a single hour

To make this information easier to visualise / understand, I have made it into a graph and made some extrapolation estimates to fill in the blanks. There are a number of factors which could change the results in practice, but there does seem to be a general pattern to it:

I think that the difference now vs last time I was on DLM-1 was that I was very careful not to initiate too many resyncs. I limited myself to 1-2 resyncs per 24hrs and this seems to have made a big difference.

Another interesting thing I noticed over the testing period was that CRCs tended to be low for the first few hours and then climb up to 3-4x the initial rate and stay there:

This kind of pattern was fairly consistent throughout my testing.

So there you have it, I reckon the magic number is 10 :)

My other mystery with the odd SNRM and error drops is ongoing. The SNRM is back to 'normal' now and the errors are still down significantly (as compared to before the drop):

I'm going to leave it as is for now and see what happens!

SamF

1598 posts

Uber Geek
+1 received by user: 252

Trusted

#805098 24-Apr-2013 19:17

Well my connection was still being erratic today so I've dropped back to -4.5db SNRM.

Damager

2125 posts

Uber Geek
+1 received by user: 37

#806165 27-Apr-2013 00:03

Hi Guys

My current stats. Has my profile updated?:

Mode:VDSL2Traffic Type:PTMStatus:UpLink Power State:L0 DownstreamUpstreamLine Coding(Trellis):OnOnSNR Margin (0.1 dB):122168Attenuation (0.1 dB):00Output Power (0.1 dBm):135-64Attainable Rate (Kbps):4059213184 Path 0 Path 1 DownstreamUpstreamDownstreamUpstreamRate (Kbps):403231033000 B (# of bytes in Mux Data Frame):23923800M (# of Mux Data Frames in an RS codeword):1100T (# of Mux Data Frames in an OH sub-frame):64600R (# of redundancy bytes in the RS codeword):01600S (# of data symbols over which the RS code word spans):0.18940.73490.00000.0000L (# of bits transmitted in each data symbol):10136277600D (interleaver depth):1100I (interleaver block size in bytes):24025500N (RS codeword size):24025500Delay (msec):0000INP (DMT symbol):0.000.000.000.00

- Telstra HTC Touch Pro2 - Energy ROM WM6.5.5 20 Oct/Cyanogen Mod Froyo 2.2 - R.I.P
- AT&T Galaxy S Captivate 16GB on XT (now with brother)
- Samsung Galaxy S2 on XT- Runs ICS 4.0.3 Resurrection Remix 9.2
- Business Hours - Work In The Electricity Industry, After Hours - DJ/Turntablist - Will Scratch Vinyl For Free'
- What's next??? S3?