Kitz Forum
Broadband Related => Broadband Technology => Topic started by: WWWombat on November 25, 2015, 01:12:16 PM
-
I've just been analysing a line with a strange problem over on the TBB forums.
It turns out that, hidden in the background, the line is "suffering" from a lot of FECs and (for 75% of the day) a lot of G.INP retransmission. I don't think this cause the original issue (a big jump in speed), but it is interesting all by itself.
This means you can see a whole variety of G.INP statistics in operation, via the MDWS website. I can see 5 days (* must remember to donate one day), but it looks like the problem only started about 5 days ago.
Analysis starts here: http://forums.thinkbroadband.com/fibre/t/4449595-re-overnight-15mbps-speed-increase.html
-
Was testing are new crimbo tree 3 weeks ago and my G.INP go's into full flow due to interference from its thyristor circuit but no extra CRC's or ES so thats the good news.
-
Oooh yes - the first time we'll have seen G.INP encounter the christmas lights phenomenon!
-
You can have a look now as the Xmas is up and running at 3pm 2 days earlier than expected I just do what I'm told ::) you can see straight away how G.INP is dealing with this extra internal interference.
As the DECT phone is close to the tree the interference is travelling into the phone cable and then into master socket because as soon as I remove telephone plug from SSFP the interference stops at 10pm :)
-
This sound the right time to get my modified BT80B RF3 out RJ11 in and RJ11 out, removed DECT phone cable and replaced with above plus a Micro Filter to telephone SSFP and XMas tree interference has been removed :fingers:
probably a twisted RJ11 to RJ11 cable would have had the same result but as I don't have a spare this will have to do.
-
Very interesting. I presume you have used the microfilter to just convert the RJ11/12 plug to a 431A/631A plug?
One of Walter's clients in Ewhurst has had a long standing problem with his VDSL2 circuit as a result of thyristor controlled water pumps (owned and operated by Thames Water) injecting noise into the shared D-side cable bundle. According to my notes, Walter now owns a couple of BT80 RF2s and is waiting for the "moment to be ripe" to fit one of those filters into the analogue telephone circuit, within the Thames Water equipment cabinet.
-
Very interesting. I presume you have used the microfilter to just convert the RJ11/12 plug to a 431A/631A plug?
That is correct I used to have one of those RJ11 to 431A converter plugs but can't find it must in the shed with the spiders thats why I used a Micro Filter.
It seems to be working as long as I keep the lights on a steady config but once you change the light config to say blinking or phased then the interfernce comes back yet not as bad since BT80B RF3 was fitted to DECT phone.
-
It seems to be working as long as I keep the lights on a steady config but once you change the light config to say blinking or phased then the interfernce comes back yet not as bad since BT80B RF3 was fitted to DECT phone.
A BT80 RF2 would be even better! ;)
Thank you for performing those experiments and sharing the results.
-
Wow - to both the impact of the lights, and the improvement through the RF3!
I note that the graph for "retransmission TX" peaks around 7,500 per minute, the "retransmission Corr" peaks at around 6,000 per minute instead, while "UnCorr" is essentially zero.
I guess that tells us that 6,000 blocks need retransmission per minute - 100 per second, or double the mains AC frequency; not a coincidence, I guess. Also that 1,500 of those blocks need re-transmitting twice before they get through correctly.
-
I note that the graph for "retransmission TX" peaks around 7,500 per minute, the "retransmission Corr" peaks at around 6,000 per minute instead, while "UnCorr" is essentially zero.
I guess that tells us that 6,000 blocks need retransmission per minute - 100 per second, or double the mains AC frequency; not a coincidence, I guess. Also that 1,500 of those blocks need re-transmitting twice before they get through correctly.
Ah, I see. Having read your analysis, it now makes sense! :)
-
A BT80 RF2 would be even better! ;)
Thank you for performing those experiments and sharing the results.
It was a surprise to see this interference but once you ask yourself how and why is this interference is getting into the master socket, there must be millions of DECT phone cables that are the flat type none twisted which act like an 1 or 2 meter receiver antenna.
All I have BC is two BT80B RF3's one is that modifed cable and the other is unused.
-
It was a surprise to see this interference but once you ask yourself how and why is this interference is getting into the master socket, there must be millions of DECT phone cables that are the flat type none twisted which act like an 1 or 2 meter receiver antenna.
Indeed. There is a distinct lack of ready availability of UTP cable with a 431A/631A plug at one end and an RJ11/12 plug at t'other.
-
Just seen a clip on ITV to warn christmas tree lights may slow your broadband and said there is an app for smart phones to see if there is interference, what is this app called ?
-
Just seen a clip on ITV to warn christmas tree lights may slow your broadband and said there is an app for smart phones to see if there is interference, what is this app called ?
Ah, I believe that is the subject of a thread (http://forum.kitz.co.uk/index.php/topic,16535.0.html), Ofcom WiFi Checker, which was started by 7LM earlier today. :)
-
I prefer Wifi Sweetspots by Assia
-
Ah, I believe that is the subject of a thread (http://forum.kitz.co.uk/index.php/topic,16535.0.html), Ofcom WiFi Checker, which was started by 7LM earlier today. :)
Oh I see it's a wifi thing already have an app on phone to test signal strength and channel overlaps.
-
This ofcom app tests ping and packet loss to/from router
-
This ofcom app tests ping and packet loss to/from router
May be ofcom should purchase DSLstats for £4 million to get to the truth of what really happens to Broadband customers lines it is as close to a JDSU as we will get ;)
-
This ofcom app tests ping and packet loss to/from router
May be ofcom should purchase DSLstats for £4 million to get to the truth of what really happens to Broadband customers lines it is as close to a JDSU as we will get ;)
They'd need to be be able to monitor the whole country remotely though. Um... Maybe invent a universal router that has a built in uploader to MDWS.
:doh: where did I store that Tandem NonStop I used to interface with.... :cool:
-
They'd need to be be able to monitor the whole country remotely though. Um... Maybe invent a universal router that has a built in uploader to MDWS.
I wonder if there's enough room in a Huawei HG612 to implement such an uploader?
-
I'm just following up this post to add another graph to the data I reported in the first post.
Where "tenbyboy2" had G.INP retransmission happening in full flow, with some blocks being retransmitted more than once, there is obviously a knock-on impact to latency for *some* packets.
That impacts can be seen in a TBB BQM - where the yellow "max latency" bar is thicker by perhaps 5ms from 10pm through to 3pm. I guess that tells us the maximum additional latency that G.INP introduces.
-
WWWombat I would have thought my line should get a latency below 20ms with G.INP but it does seem to stuck at 28ms and no amount of gateway jumping resolves this, I can remember being served by maidenhead 2 years ago.
But only seem to have two gateways dundalk or douglas these have higher latencys while maidenhead was close to 15ms what has changed on this line ?
-
I don't know enough about your line or history to recognise this. Can you post some details?
-
I don't know enough about your line or history to recognise this. Can you post some details?
Not much to post just whats in MDWS this a BT wholesale only exchange and FTTC 40/10 is fed by a larger exchange 2.5 miles from local telephony exchange, the first year with FTTC the latency was like 12-15ms and a year later is was 25ms then it hit a high of 45ms and then when G.INP was switched on it now sits at 28ms.
It is funny just passed my FTTC Cabinet number 1 to-day it has sticker on it saying fibre available it has been fore more than 3 years wonder why they do that :-\
-
I'm just following up this post to add another graph to the data I reported in the first post.
Where "tenbyboy2" had G.INP retransmission happening in full flow, with some blocks being retransmitted more than once, there is obviously a knock-on impact to latency for *some* packets.
That impacts can be seen in a TBB BQM - where the yellow "max latency" bar is thicker by perhaps 5ms from 10pm through to 3pm. I guess that tells us the maximum additional latency that G.INP introduces.
nice thanks for the data, that is still way better than either packet loss or permanent increased base latency via traditional interleaving
Sent from my SM-G901F using Tapatalk