Kitz Forum
Broadband Related => Broadband Technology => Topic started by: Weaver on March 29, 2017, 11:35:53 AM
-
Is my upstream any good?
I get 537 / 429 / 440 kbps sync on my three lines. The upstream attn is 42.4 / 42.4 / 42.1 dB. The lines are 8 km long according to BT, if I have read the engineer’s notes correctly - would be good to check this actually. (It is 7.3km by road anyway, and the line does definitely (visibly) follow the road - as discussed earlier.)
The downstream sync is currently 2.78 / 2.83 / 2.87 Mbps with attn 65dB.
Is that ratio - 0.19 / 0.15 / 0.15 - anything like that which you get?
And is there any way to improve upstream?
It's a shame that I can't reduce the upstream SNRM to 3dB, as I have successfully done for the downstream. But there seems to be no option to do that - presumably a 21CN limitation. Makes no sense though.
I noticed today what a pain it is uploading big photos, even with the lines combined, not brilliantly, to give a measured total 0.88 Mbps. Some speed testers give a combined upstream of only 0.7 Mbps.
-
I have an upstream attenuation of 7.4dB on ADSL2+ which gives me a sync speed of 1243Kbps. I would be interested to see the math behind signal attenuation to be honest. It looks as if it follows some sort of exponential curve, but that is a discussion for another thread.
-
I've just asked my ZyXEL VMG1312-B10D to disclose the current values --
[Duo2 tmp]$ ssh admin@ap
admin@ap's password:
BusyBox v1.20.1 (2016-10-18 15:41:48 CST) built-in shell (ash)
Enter 'help' for a list of built-in commands.
$ xdslctl info --state
xdslctl: ADSL driver and PHY status
Status: Showtime
Last Retrain Reason: 8000
Last initialization procedure status: 0
Max: Upstream rate = 980 Kbps, Downstream rate = 5832 Kbps
Bearer: 0, Upstream rate = 963 Kbps, Downstream rate = 5120 Kbps
$ Connection to ap closed.
[Duo2 tmp]$
Using the "Bearer" synchronisation speed (and not the predicted "Max" synchronisation speed) for the circuit I find a ratio of 963 / 5120 = 0.188
Now a throughput speed test, noting that the time is currently 1812 hours --
[Duo2 tmp]$ speedtest_cli --help
Usage: speedtest_cli [options]
Command line interface for testing internet bandwidth using speedtest.net.
--------------------------------------------------------------------------
https://github.com/sivel/speedtest-cli
Options:
-h, --help show this help message and exit
--bytes Display values in bytes instead of bits. Does not affect
the image generated by --share
--share Generate and provide a URL to the speedtest.net share
results image
--simple Suppress verbose output, only show basic information
--list Display a list of speedtest.net servers sorted by
distance
--server=SERVER Specify a server ID to test against
--mini=MINI URL of the Speedtest Mini server
--source=SOURCE Source IP address to bind to
--timeout=TIMEOUT HTTP timeout in seconds. Default 10
--secure Use HTTPS instead of HTTP when communicating with
speedtest.net operated servers
--version Show the version number and exit
[Duo2 tmp]$ speedtest_cli --version
0.3.4
[Duo2 tmp]$ speedtest_cli --simple
Ping: 61.411 ms
Download: 4.33 Mbit/s
Upload: 0.79 Mbit/s
[Duo2 tmp]$ speedtest_cli --simple --secure
Ping: 60.637 ms
Download: 4.34 Mbit/s
Upload: 0.80 Mbit/s
[Duo2 tmp]$
Using the latter throughput values I find a ratio of 0.80 / 4.34 = 0.184
So let's say the ratio is between 0.18 - 0.19 for my circuit, operating in G.992.3 mode.
-
I don't think there's much in the ratio of upstream line rate : downstream line rate. The frequency range usable for the downstream varies with line length, but the upstream frequencies are all lower than the downstream, so the usable upstream frequency range will normally be all of it, only the upstream SNR will be different for different line lengths.
-
Ejs, I don't follow you. I'm being a bit thick. :-) We are saying that the ratio isn't constant aren't we? I'm interested to know how the ratio varies with length. Also whether those upstream sync rates are any good for that upstream attenuation and that length.
The attenuation upstream will be different for different line lengths and the bit loading surely will be different? Even though clearly you will use all of the upstream bins. Assuming your upstream isn't capped then upstream speed has to vary, agreed?
My upstream rate isn't (effectively) capped, as I pay extra for IPstream Max Premium and I am nowhere near fast enough to hit the 800kbps limit or whatever it is - mind you I am 21CN now so I don't know if the rules are different. Anyway I do indeed beat the old 448k limit (I hope I remember that figure correctly)
Btw Could anyone tell me what the story is for 21CN and Premium / non-Premium?
-
I'm saying it seems pretty pointless to compare the upstream speed with the downstream and calculate the ratio. Although there is the EP 2 747 401 A1 patent about fault detection by comparing the upstream attenuation with the downstream.
There is no such thing as Premium on 21CN, you do not have to pay any extra for uncapped upstream.
-
@ejs Would you say that is so because there are better methods of deciding whether or not the upstream is any good compared with what you would expect for a typical line with such physical characteristics?
Yes, I suddenly thought that I had a vague memory that BT had dropped the old 448k cap for some reason.
Off-topic: The only reason that I am willing to pay the huge extra charge per month (it is a huge lot) is because I get traffic prioritisation - discussed in several earlier threads - despite having been warned by AA that the details of how it works in BT-land are unspecified and the true benefits are unknown. I also get traffic prioritisation inside AA's network as well. I think it is an excellent idea because if done correctly it should allow some customers to save a lot of money and preserve customers' choices. [See earlier threads]
-
I am also trying to work out why the upstream sync on one of my lines is so much better than the others - 25% faster in fact.
The fastest line, cwcc@a.1 = BBEU20700042 is the one that was first installed. I think that the second one installed (cwcc@a.3 = BBEU20700055) was a second pair in the same cable - sound plausible?
[Btw the 'line' identifiers that are used by aa and which have this kind of syntax cwcc@a.n do have index numbers that follow chronological order, in my case at least. They aren't contiguous though, unfortunately because @a.2 is a non-DSL 'line', actually a 3G data SIM account.]
-
A year ago typical speeds for me might have been 4002k downstream, 804k upstream. Currently they are 3643k downstream, 804k upstream. The downstream is not as good as it once was, the upstream is still the same, I'd say the change in the upstream : downstream ratio is meaningless.
I admit that Weaver's upstream speeds and attenuations don't seem a typical match for the downstream speeds and attenuations, but it's not clear if that is due to the downstream performance being unusually high or the upstream performance being unusually low.
-
I absolutely agree that the ratio of upstream to downstream taken alone is meaningless but I would be interested to know how the ratio varies with length to see if upstream remains decent because the range of frequencies of downstream become really limited whereas with upstream all that can happen is that the bits per bin loading get reduced (well, that happens with downstream too clearly) but no buns are abandoned, and the noise spectrum at the low end vs higher regions may not simply be proportional to length for various reasons and may not be the same function of length either, plus the fact that we are interested not in (absolute) noise but in SNR, and the S behaviour is different.
One simple question is : is the ratio roughly constant with respect to line length? (Or attenuation.) I don't even know that.
I think that my downstream is unusually high, when compared to some others. But my upstream seems pretty good too, so I don't know who wins out in contributing to the value of the ratio. I have seen some very rubbish upstream values locally amongst G.992.1 users, never mind the old 448k cap.
-
I would have thought the ratio of upstream speed to downstream speed tends towards 1 for the extremely long lines. For shorter lines, the upstream speed only increases gradually compared to the downstream, eventually reaching about 0.05 for the maximum 1.2Mb upstream and 24Mb downstream.
-
What kind of numbers do we get for other DSL users here (and attenuation and line lengths?)
Does anyone know if any rough curves have been published?
-
The patent (https://www.google.com/patents/EP2747401A1?cl=un) I mentioned does have a graph plotting attenuation ratio (downstream / upstream) against upstream attenuation.
-
. . . but no buns are abandoned, . . .
Time to eat, methinks. :yum:
-
Thanks to ejs for that. I wonder how I can map lengths et al to attenuations and attenuations to sync speeds.
For the latter I would unfortunately also need to bring in noise, which presumably has two parts, a contribution from the ends and a contribution that is per-unit-length, and then each breaks down into crosstalk plus other kinds of environmental noise. So it would be a bit of a nightmare.
Someone ought to have done a big fat spreadsheet, with physical parameters plus a model of noise and crosstalk sources.
-
I dont think there will be any direct correlation. Firstly theres such things as PSD masks and Power cut back. There's also several different PSD masks in use depending on line length. Its the reason why if you get Annex_M those very same downstream tones converted to upstream most likely wont get the same bitload.
Finally crosstalk. Crosstalk can and does also affect ADSL/2/2+
I saw my BE* line go from a full 24Mbps down to more like 20Mbps by the time the MSAN was filled with O2 users. It only really affected my downstream. You cant predict how much crosstalk will affect an individual line.
-
Kitz makes an excellent point, I forgot about power limitations. I have probably had it.
It is still interesting to know how you compare against similar lines though so that you can detect if you have a problem. Here my question I suppose comes down to "What might it mean if someone has relatively good or bad upstream sync compared to another user’s similar line, but their downstream doesn't compare with the other’s in the same way?"
-
I found some good stuff here
https://www.ericsson.com/ag/res/thecompany/docs/publications/ericsson_review/2006/vdsl2.pdf
which I noticed because of a post by asbokid.
-
I have put a request in to my isp to ask if the up-stream can be put to 3dB target to see if it will improve my up-load.
-
Tickmike what kind of line is it?
I wish I had the option to change my 21CN ADSL2 lines to 3dB upstream as well but there seems to be no such choice visible in the UI on AA'S clueless server for 21CN.
-
Tickmike what kind of line is it?
21CN ADSL2+ line 4.3kl long, 51dB Line attenuation, Target now = 6dB and 6dB, Not had a reply from the isp yet.
-
Tickmike, I fear that it isn't technically possible, I hope to be wrong. What do others think? Do please let us know how you get on. I would love to do likewise
-
Are you on 'Annex I' The voice frequency ranges are used for additional 256kbps upstream, I can not do that as we want a POT's line, we need the phone.
http://www.kitz.co.uk/adsl/linestats_explanation.htm
-
Just really wish I were on Annex I but I don't think anyone does it in the UK. Deutsche Telekom offers Annex J in Germany though. But Annex I would be superb. For me those frequencies are just being completely wasted.
-
I did not know that :o
-
No not used in the UK. Different countries tend to stick with certain Annex's. Start mixing them up on the same MSAN and you get high risk of crosstalk.
Openreach have strict limitations on what can and cant be used on their network.
The target SNR profiles are up to the owner of the MSAN. BE* used to allow both 3dB for up and down. Cant recall what TT use.
If you are on adsl1/2/2+ can you not tweak your own downstream target margin?
-
It's supposed to be worth an extra 256kbps. I presume that means 'in increased sync rate', and you have to take protocol overheads out of that for ATM and above. So it's not peanuts.
-
My reply from my isp a bit late "I cannot set the upstream any lower than 6dB, unfortunately. I don't have control over the PSD" .