ADSL upload faster than download

In connection with the BDUK rural broadband plans, I've been soliciting speed measurements from people who live more than 3 Km from our village exchange, as the lines go. Most of them look sensible, but I've had three where the upload speed is higher than the download:-
Download 0.133 Upload 0.357 Mbps, Distance 6.6 Km
Download 0.220 Upload 0.350 Mbps, Distance 3.5 Km
Download 0.125 Upload 0.345 Mbps, Distance 4.1 Km

My exchange does not have WBC or ADSL2 so this is plain ADSL. The bandwidth that ADSL technology allocates to download is about 11 times that for upload, so I'd expect the download was always faster than upload. I suppose a strong RF noise source in the house might preferentially affect download?

Has anyone had experience of ADSL upload being faster than download, and what are the possible causes of it?

Edited by deleted (Wed 29-Aug-12 10:36:37)

Even though more tones are allocated to download, they are the higher frequency tones. It is these that suffer worst from long lines.

If any of them have appropriate routers, try using a program like DMT to have a look at the frequency spectrum in use. http://dmt.mhilfe.de/ The webpage is in German, but if you find the right version, the code is mainly in English and easyish to understand.

In reply to a post by pvfpvf:

Has anyone had experience of ADSL upload being faster than download, and what are the possible causes of it?

In my experience, that is a sign of there being other faults, be it poorly wired or there being a fault on the line.

I would expect to find an audible line fault.

You need to look also at attenuation e.g. if the 3.5km line is actually a 50dB attenuation line then something is wrong, very likely the telephone wiring inside the property, or they just have a bad choice of ISP.

If you can get the stats from them, then type them in at http://www.coolwebhome.co.uk/calc/calculator.php and if its a good line you'll get three sets of green traffic lights.

If things are slow, then fixing their speeds, e.g. remove ring wire or fit an i-plate. Remove unused or substandard extension wiring, can improve speeds with perhaps 30 minutes of fiddling, rather than wait 12 to 36 months for BDUK to deliver.

Thanks. Yes it makes sense to me that higher frequencies suffer greater attenuation.

Had a look at DMT and was not aware that some ADSL modems give so much info, but it seems not all.

I'm dealing mainly with non-technical users so there's a limit as to shat I can push them to do.

Thanks. Yes, I did send out an explanation of the ring wire/I-plate scenario, but not sure who will take it up. Users are mainly non-technical.

Unfortunately, I forgot to ask them to tell me their ISP, so don't know that.

Probably have to visit them to see the attenuation/SNR. Surprised you say house wiring could affect the attenuation: I can see it might pick up noise and hence affect SNR, but attenuation surely is mainly affected by impedance of the long line from exchange to house?

The house wiring will not affect the attenuation (or should not to a significant effect). but it does allow you to estimate the range of speeds you would expect from a line.

e.g. if the attenuation is 61dB and they are getting 1.8 Meg from speedtests, then probably not much more if any to be squeezed from the line.

You already have several useful comments, but a few other points.

The distances you refer to - are they straight line, by road, or actual line length?

Secondly, I have seen quite a few lines where the upload is faster than download - always on long lines where the higher frequencies just cannot be used.

The customer side equipment and wiring can affect the attenuation - for example if there are too many master sockets, or too many phones, or cheap phones ... they can adversely load the line, changing teh termination impedance which could result on several dB of additional attenuation.

When taking line stats, do it with just the modem connected to the test socket behind the removable front of the master - that disconnects everything else. Also try it wit the modem located in its normal place and see what the differences are in Attenuation, SNR and Sync ...

Thanks. My distances are actual BT line lengths (I wandered around and traced all the lines. There are only a few places where the route of buried cables is conjectural.)

Yes, I can see that if equipment is actually connected to the extension wiring then it could indeed affect the attenuation, and thanks for pointing that out.

I'll bear in mind the other advice should anyone invite me into their home to investigate!

If someone on a 3.5km line is really only getting under 0.5 Meg, then something is seriously wrong with the line length or something else.

Why? Because back in 2000, BT specced ADSL as giving 2Meg at 3.5km line length, and plenty of lines now show it running at 7 to 8Meg now that BT has relexed the various limits.

Also as its an IPStream exchange, rather than WBC, best to get people to check what their IP Profile is, via www.speedtester.bt.com

I hope to get measurements from their neighbours in due course, then it'll be easier to see if it's something peculiar to them.

Are the speed-versus-linelength figures in your faq here still correct?

I have one person who is 3.5 Km from the exchange who measured download speed at 12.8 Mbps. My current view is that that is impossible (with ADSL 1) and there must be something wrong with the measurement (e.g. a faulty realtime clock on his pc).

Well as ADSL1 tops out at 8Meg, then yes impossible.
With ADSL2+ given the errors involved in estimating distance it is possible.

Why do you think it is the real time clock in the PC? More likely to be a user error, or a false reading from a speedtest, some do over estimate or interactions with AV suites cause issues.

As for the FAQ, yes they still stand, but would recommend abandoning distance use, and relying on attenuation where it is available.

In reply to a post by pvfpvf:

I have one person who is 3.5 Km from the exchange who measured download speed at 12.8 Mbps. My current view is that that is impossible (with ADSL 1) and there must be something wrong with the measurement (e.g. a faulty realtime clock on his pc).

Inaccurate measurements of download speeds can sometimes be due to anti-virus programs - e.g. Kaspersky IS 2011 could inflate download speeds by a very wide margin.

Thanks. But how does an antivirus program inflate the download speed? It cannot actually increase the rate at which data is received, so I imagine the only other thing it could do is give a false reading of elapsed time, which sounds implausible to me.

All the speed test readings I'm soliciting are from broadbandspeedchecker.co.uk - sorry about that, but I just thought the way to run it and its presentation was simpler for the lay person than the alternatives, including tbb's estimable one. I think it unlikely he misread the displayed result, so was trying to think of any technical effect that could give a totally wrong reading like 12.8 Mbps (in my view more than twice what he can attain).

There is no ADSL2 on my exchange.

A possibility that's just occurred to me is one of those software accelerators. I presume those work by redirecting the browser to a server, which gets the wanted data quickly, compresses it, sends it to the PC in compressed form, and then expands it in the PC, before giving it to the browser. That would give falsely fast readings, unless the file that is downloaded by the speed test cannot be compressed. That suggests the speed testers should use test data that cannot be compressed: e.g. a JPG photo that has been Winzipped.

Can't get attenuation data from lay users: too complicated.

Speed tests done with some AV products will show wildly different and erroneous results due to the scanning of network traffic and because of buffering of traffic. This only affects the numbers shown and is not related to the actual download speed

e.g. http://forum.kaspersky.com/index.php?showtopic=227894

Edited by 4M2 (Thu 30-Aug-12 13:02:03)

I can confirm that Kaspersky used to cause some speed tests to give absurd results, such as 100Mbps on an ADSL line.

This only applied to some testers. My memory is that speedtest.net and dslzone suffered, but that thinkbroadband and some others did not. Someone said it was flash based testers that failed with Kerspersky; I can't verify that.

The bt speed tester was usually OK - can not use it now because I'm LLU'ed however I think the thinkbroadband tester and iPlayer diagnostic results are about right when using KIS IS...

As others have suggested then sounds like that tester may have been affected by something like the AV software on their machine.

What happens is that the AV software delivers data in a bursty fashion, and unless the speedtest coder is aware of this, then the wrong results can be displayed.

Our Flash based tester is as simple as that one to use, and can post result to twitter for people.
http://www.thinkbroadband.com/speedtest/broadband.html
Broadband Speed Test Result - 6.27 Mbps (down); 0.84 Mbps (up) (via @thinkbroadband)

Thanks. Do you think your flash speed test would avoid being fooled by anti-virus burst delivery? If so, I'll ask him to try it.