Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[sheddyian]

Okay, so in my ever-obsessive zeal for squeezing that extra bit of bandwidth out of my ADSL2+ connection, I wondered if isolating my modem from the mains might eliminate any mains-borne interference and give me faster speeds.  The PSU is a switch-mode one, and generates a lot of noise on an AM radio held nearby.  This noise is radiated along the power lead as well.  My logic goes that this can't be doing any good, but good things might happen if I could eliminate it.

The modem is a TalkTalk branded D-Link DSL-2780, with an unusual sized DC power connector. I got one as near to the size as I could, although it was thinner and the centre hole was a tight fit.  But it made a good connection.  I checked the polarity of the PSU : centre +, then wired my new plug onto a 12v cordless drill battery that'd I'd had charging.

Under no load, the cordless drill battery gave 13.8v according to my meter, but as the PSU measured 15v (very slowly dropping) when tested unloaded, I wasn't too concerned by this.

To establish a baseline, I power cycled the modem twice whilst powered by the PSU, and noted the speeds, attenuation, S/N margin etc each time.  I then tried a few combinations of tests, and recorded the results.


Synch Speed   Attenuation       S/N Margin          Test performed
UP   DOWN     UP    DOWN        UP    DOWN   

1020 21070    11.5   22        9.5      6            Usual PSU, 2 wired Lan connections
1020 20964    11.5   22        9.2      6            Usual PSU, 2 wired Lan connections
1020 21066    11.5   22        9.2      6            12v battery, 2 wired Lan connections
1020 20756    11.5   22        9.7      6            12v battery, 1 wired Lan connection
1020 20921    11.5   22        9.4      6            12v battery, 0 wired Lan connections
1020 21114    11.5   22        9.3      6            12v battery, 0 wired Lan connections
1020 21078    11.5   22        9.4      6            Usual PSU, 2 wired Lan connections
1020 21078    11.5   22        9.5      6            After 2 1/2 hours running on usual PSU


Note : All measurements were taken within 50 seconds of the modem synching, except the last one.  "0 Lan connections" means I disconnected all cables except the 12v battery and the ADSL during power up and synch. Once the DSL and Internet lights lit on the modem, I reconnected the Lan to log in and get the results.
I tried this because I wondered if noise might enter through the wired Lan connections.

The wireless was active throughout, as I could not find a way of having it switched off at startup (it can be disabled in software once the modem is running).

So, unless some whizz statistician can spot something in there, I would say that it made no difference at all running the modem on a battery instead of the noisy 12v switch mode PSU it was supplied with.

I had intended to try it with an old PC PSU supplying the 12v, but that seemed rather redundant after these readings, and I was getting twitchy that the exchange would drop my synch speed because my modem appeared so unstable with all my power cycling!  I therefore stopped at that point.

Ian

[les-70]

  Thanks for the interesting post.  I tried adding a few ferrite filters to the output lead of a Netgear PSU that did not have any ferrite.   It seems to reduce my CRC error background but not the odd heavier CRC burst. It is not an easy thing to judge.  If your battery lasts long enough and your errors have any sort of steady rate, you may want to see if you can see any difference in errors.

[burakkucat]

The modem is a TalkTalk branded D-Link DSL-2780 . . .

Would I be right in presuming you therefore have a TalkTalk broadband service? If yes, I would suggest you request that the DLM be turned off for your line. It was decidedly beneficial to me, by having the advantage that the MSAN totally ignores however many resyncs that may occur. 

[sevenlayermuddle]

Thanks for that experiment, Sheddyian.

I've always been suspicious of routers with switch-mode power bricks as, like yourself, I'd noted the whopping interference on AM radio.  It is one reason I've stuck with Netgear kit which uses big, heavy, old fashioned, but quiet, PSU bricks. But the evidence you've provided would suggest that my suspicions were perhaps ill-founded.

Switch mode bricks usually have the advantage of running cooler compared to magnetic transformers (hence costing less to run, and longer life expectancy) so I may be more open-minded in future.

[sheddyian]

I tried adding a few ferrite filters to the output lead of a Netgear PSU that did not have any ferrite.   It seems to reduce my CRC error background but not the odd heavier CRC burst. It is not an easy thing to judge.

I've wound the PSU output cable through several ferrites I recovered from an old PC PSU.  At the time I thought they had made small improvements to my synch speed, but looking back on the figures I measured, I think it was more down to general fluctuation than improvement.

However, I haven't done much to study the number of errors I get.  Maybe that's the next project!

Ian

[sheddyian]

Thanks for that experiment, Sheddyian.

I've always been suspicious of routers with switch-mode power bricks as, like yourself, I'd noted the whopping interference on AM radio.  It is one reason I've stuck with Netgear kit which uses big, heavy, old fashioned, but quiet, PSU bricks. But the evidence you've provided would suggest that my suspicions were perhaps ill-founded.

Switch mode bricks usually have the advantage of running cooler compared to magnetic transformers (hence costing less to run, and longer life expectancy) so I may be more open-minded in future.

I was of a similar opinion, which is why I was keen to try isolating it from the mains and noisy PSUs. 

Of course, all my experiment shows is that it made no perceivable difference for me.  A different modem might not have such good supply filtering (perhaps), or might be more susceptible to noise, or the PSU noise is overwhelming etc etc.

Whilst being slightly annoyed that I hadn't defeated (imaginary) noise and gained extra bandwidth, I was also left impressed that the filtering etc is such that the detectable AM interference from the PSU doesn't affect the modem in any way that I could measure.

Ian

[sheddyian]

Would I be right in presuming you therefore have a TalkTalk broadband service? If yes, I would suggest you request that the DLM be turned off for your line. It was decidedly beneficial to me, by having the advantage that the MSAN totally ignores however many resyncs that may occur. 

I may well do that. May I ask what benefits it gave you?  Is it just that you have the ability to reboot your modem 100+ times a day without the exchange being bothered, or is there some other improvement to be had as well?

My line seems pretty good, I think I've only ever seen 1 resynch that wasn't down to me, excepting the early days when the DLM resynched several times in the first few weeks to lower the noise margin until I ended up where I am now - 6db, and pleasantly surprised

Ian

[burakkucat]

The various benefits, in no particular order, are:

(1) Ability to resync with the exchange MSAN (I'm reliably told it is an MSAN, as my line is fully LLU'd) as often / frequently as I like.
(2) Retain total control of the link parameters (profile) via a proxy (the TalkTalk Members' Forum OCEs) and not have some software process (with a non-disclosed algorithm) constantly meddle with it.
(3) I am using the 6 dB target SNRM profile, with interleaving. If there should happen to be a severe noise event during the hours of darkness (when my DS SNRM could already be depressed by 3 dB), the SNRM may reach zero. If that occurs, it depends for how long the situation prevails before the 2Wire 2700HGV decides to drop the PPPoA session, re-sync with the MSAN and re-establish the PPPoA session. With DLM active on the line, it could take the previous event as justification to increase the target SNRM. (Not wanted, thank you!)

[asbokid]

Very interesting study, Ian.  Thank you for posting your findings 

Is the DSL-2780 a Broadcom chipset, or it is a MediaTek (was RaLink was Trend) like the DSL-2680 with its TC3162/3086 chipset?    Sorry, just realised we had this conversation a month ago! [1]

Maybe the QLN ("Quiet Line Noise") dataset could shed some light on the problem, and whether it is due to RFI.

This QLN graph below shows probable interference from a switched-mode PSU, circled in green.

Most of the noise in the ADSL2+ spectrum is from AM broadcast radio. But the noise in the green circle is outside the LW/MW commercial broadcast bands.

DMT 46  +32dBm/Hz       198kHz  -- BBC Radio 4LW, Droitwich (500kW on 198kHz)
DMT 161 +25dBm/Hz       694kHz  -- BBC Radio 5 Live, Droitwich (150kW on 693kHz)
DMT 244 +10dBm/Hz       1052kHz -- Talk Sport, Droitwich (500kW on 1053kHz)
DMT 282 +10dBm/Hz       1216kHz -- Virgin AM, Droitwich (105kW on 1215kHz)
(As an aside, the noise on 854kHz is/was from a pirate AM station!)

In the area circled in green, note the periodicity of the spikes - they are spaced every 50kHz (roughly). The channel spacing of xDSL is 4.3kHz, so the frequencies of the noise spikes are not exact.

That regular 50kHz periodicity suggests a noisy PSU ingressing the DSL line somewhere between the modem and the DSLAM.



cheers, a

[1] http://forum.kitz.co.uk/index.php?topic=11142.15

[sheddyian]

Very interesting study, Ian.  Thank you for posting your findings 

Maybe the QLN ("Quiet Line Noise") dataset could shed some light on the problem, and whether it is due to RFI.

It's not really a problem, I was merely curious as to whether the modem was being affected by it's own PSU, which generates a fair bit of noise on a nearby AM radio.  Looking at my findings, I'd say it wasn't being affected.  I don't get any strong interference near the modem location on an AM radio when the modem is powered off.

How might one go about making a QLN graph, or getting the data in the first place?  This might be interesting in regard to my experiment of grounding the cat5 phone cabling that I tried a month or so back (and which seemingly made no difference to synch or SNR).  Again, I wasn't trying to fix a fault, merely to see "what would happen if..?" 

Ian

[asbokid]

How might one go about making a QLN graph, or getting the data in the first place?  This might be interesting in regard to my experiment of grounding the cat5 phone cabling that I tried a month or so back (and which seemingly made no difference to synch or SNR).  Again, I wasn't trying to fix a fault, merely to see "what would happen if..?" 

Most all Broadcom-based devices have a tool (found in the firmware under various guises: xdslcmd, adslcmd, adsl) to obtain QLN tone data.  It normally provides only static data - data that was obtained during the initialisation of the link.  So QLN is no good for identifying sporadic noise, not unless a resync is forced manually while the noise is present, or else the noise gets so bad it causes a resync.

cheers, a

[sheddyian]

Most all Broadcom-based devices have a tool (found in the firmware under various guises: xdslcmd, adslcmd, adsl) to obtain QLN tone data.  It normally provides only static data - data that was obtained during the initialisation of the link.  So QLN is no good for identifying sporadic noise, not unless a resync is forced manually while the noise is present, or else the noise gets so bad it causes a resync.

Ah, so we're back to the sort of stats I was trying, and failing, to get out of my DSL-2780. 

However, I have been loaned a Belkin F6D4630-4 V2 (it says here) wireless modem router.  Any idea if that might have a helpful chipset?

I've not yet connected it up or even taken it out of the box !

Ian

[asbokid]

I have been loaned a Belkin F6D4630-4 V2 (it says here) wireless modem router.  Any idea if that might have a helpful chipset?

No idea about it, sorry. Although from looking briefly at a firmware image for the device on the Belkin website [1], it doesn't look like it has a Broadcom chipset.

cheers, a

[1] http://www.belkin.com/uk/support/search/?lid=enu&search=F6D4630

[sheddyian]

No idea about it, sorry. Although from looking briefly at a firmware image for the device on the Belkin website [1], it doesn't look like it has a Broadcom chipset.

Gah!

I shall continue to acquire loaned or donated cast-offs until I get one with Broadcom chipset

Other than being unable to get more detailed line quality, bitloading etc stats out of my modem, the DSL-2780 has done everything I've wanted, and been very reliable - can't fault it really.  Especially as it's a bog-standard ISP-supplied unit.  Can't really justify buying a different modem ... yet. 

Ian