Print

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

[jack21]

Over the past 3 years, the biggest obstacle to my broadband experience has been the fluctuating/low-spiking SNRM on any of many routers I've tried. Usually fairly consistent within +/-1db, every so often, a large downspike or temporary drop of up to 4db wrecks my tuned SNRM (tweaked to 7 to get best sync rate), sometimes causes disconnections and has contributed to my high default SNRM (15).

My line is long, has an attenuation of 68.5db, but at brief good times, eg last January, enabled me to hold a sync of 2016 for 4 days at an SNRM of 6-7db. Normally, 1440 to 1728 kbps is my current target, at SNRM 9'ish, and I can usually keep a profile of 1250 (even 1500). At good times, my routers can hold on as low as 2.5 to 3db, but more often they don't like going below SNRM 5.....so I now allow for a 4db spike, and tune my routers to SNRM 9......a fairly successful procedure. Lowering the SNRM to 7 and below puts me into 'dangerous' territory.

On 2 occasions, I've identified the cause; once it was my neighbour's problematic 4-way power strip...its neon flicked unusually, and it made a buzzing noise. I replaced the strip for him and that cause ceased. The second time was when the same neighbour fitted a replacement PSU to his PC (I helped him). After the PSU was in place, whenever his PC was on, I lost 2db of SNRM. His house is 50m away, but we share the same telegraph pole for BT line; he moved house a couple of weeks ago, so that problem has also ceased. But there are other fairly regular drops, of predictable pattern,  which I can't identify, and there are mysterious very short-duration downspikes of 0.25 to 4db which don't have any pattern I can spot.

Because I believed the interference to be mains-borne (our vacuum cleaner also causes similar spikes), and trying off-the-shelf mains conditioners seemed to have no impact whatsoever, I'm now in the process of testing my theory by totally isolating my router from the mains supply, and powering it from a car battery via an inverter. The inverter connects to the battery, the router PSU plugs into the inverter output of 240v and thats it - no link from the router to the mains. I've determined that I can charge the battery without interrupting router operation, but of course charging gives a route to the mains, so I've yet to investigate whether thats a problem.

After 3 hours of operating, I'm impressed with the stability of the router (DG834 V4) SNRM - not even 1 downspike recorded, and thats unusual, and the variation is only +/- 0.1db. I don't yet know how long the (old and 16AH)  battery will power the inverter (its currently (!) drawing 1A at 12V, so might last for 12+ hours), but if the experiments conclude satisfactorily, I'll invest in a  55AH battery that matches the one in my car, and I could put the charger onto a pre-set timer during the quiet nighttime.  I'm waiting for the next predictable time for SNRM drop of (previously) 2db - tomorrow morning between 6.30 and 9.00 - to see if that is prevented.

Total cost so far: £16 for power inverter; battery was free from recycling heap at local garage.

You may wonder if the DG834 could be powered directly from the battery - its PSU says 12v at 1A, so maybe it could.....but the route I've chosen seemed the safest to me - it ought to give a more stable voltage to the router. And of course the direct battery route would not be immediately suitable for non-12v-DC routers, whereas this inverter method can be used for any mains-powered-PSU router. Why not use a UPS? - well most UPS are 'off line'; during normal operation the mains is fed to the output, and only when mains power is off does the battery drive the output. 'On line' UPS, which can provide galvanic isolation from mains  at all times, are very expensive indeed, but also offer a solution.

I'll update this posting once I get more results in.

[roseway]

Thanks for that report. It's interesting and entirely plausible. Like you, I don't like the idea of running the router directly from the battery because the voltage is a good bit over 12V when charged.

I suppose that a good alternative would be to use a UPS (and power the computer from it as well), but you would need to have the kind which operates permanently from the inverter while trickle charging the battery, not the cheaper kind which normally operates from the mains and only switches over to the inverter when there's a mains failure.

[sevenlayermuddle]

I'll certainly be interested in the outcome of this experiment.

To be honest I'd not expect much improvement, since the router is very effectively isolated from the mains by the 'power brick' in the 13A plug.  The better ones, such as Netgears, tend to have old-fashioned magnetic transformers inside, which is why they're so darned heavy and run so warm, but by doing so they can very effectively isolate the 12V output from the mains supply, and also avoid the interference pulses that can be generated by lighter and cooler switch-mode power-bricks.

Sorry to continue sounding negative, but I'd not expect an inverter to be the ideal supply, as they are switch based, and quite likely to cause interference of their own.   In that case the interference wouldn't be occasional spikes, it would be constant in nature and probably reduce some of the bit loadings at selective frequencies.  It may be interesting to compare bit loading graphs as well, with the router powered from mains and from inverter.

Mains-borne interference (e.g. from a neighbour's house) can also find it's way into the router via coupling of the telephone/router cables, it doesn't need to be a direct electrical connection (tho' I grant you, calling it 'mains borne' is in that case misleading).

It's an interesting experiment anyway, and I'll be interested to hear the results.
 

One final work of caution, just in case it needs to be said... always be careful when trying experiments like this to avoid reconnecting more than a few times in an hour as that can upset BT's DLM, leading to a punitively high target margin.  Better still, keep it down to no more than a few reconnects per day.

-7LM

PS I see Eric's responded while I typed.  Glad to see he's more upbeat about it, as it makes me feel less guilty for being downbeat   

[waltergmw]

I believe a battery can be an excellent large capacitor to smooth out spikes.
However some care is necessary as a car battery is specifically designed to accommodate very large currents when e.g. the self-starter is operated.
As has been said voltages can be quite significantly in excess of the nominal 12 Volts.

I echo the comment that a UPS specifically designed for this type of application is likely to be a much safer option and will include batteries with a more suitable duty cycle.

In summary a cheap and cheerful battery solution is unlikely to provide the optimum solution and could prove quite costly if a mistake is made.

As an aside I remember a tale of a submariner who dropped his spanner across the main battery busbar and was a little surprised when it melted !

Kind regards,
Walter

[jack21]

Yes re the battery-direct possibility....I too am cautious about that approach - thats why I chose the inverter route. It converts the battery's 12v dc to 240v ac, then I plug the normal PSU into the inverter.

Further update:

Over the last 15 months I've bought/tested/resold over 60 modem/routers, in a quest to find the best for my line/conditions,  and the regular, predictable  SNRM dips and random spikes have been evident on all of them, tho the switching-type PSUs do seem to be affected most. So I'm doubtful about the PSUs being problematic.

So far, with 22 hours operating under my belt, I'm astonished at the very 'clean' RouterStats display - so far, none of the 'odd' downspikes I normally get . There has been one 0.5 db drop, which lasted for 90 minutes then recovered to as-was, but none of the larger drops I used to see. For the rest of the time, its +/- 0.1db

But this battery-inverter-psu-router setup has not cured the 2 morning evening SNRM dips that I see every day. Each morning, between 6.30 and 9 ish, the SNRM dips by around 2db, slowly increase by 0.3db, then around 25 minutes later returns to its previous 'normal' level.....this is usually repeated in the early evening.
This setup has maybe reduced the size of the dip by 0.5db, but not eliminated it.....its arising cause is nothing I can identify.

So at the moment, the setup has apparently eliminated the odd, unpredictable, short-duration downspikes, and the SNRM graph is remarkably consistent at =/-0.1db (apart from the 2 daily, predictable drops). The twice-daily drops I see must be either via the phone line or direct to the router via radiation, and I doubt there's much I can do......could try shielding the router, I suppose.

My final test will be to replace the DG834 with my trusty, speedy Billion 7300 (firmware-fixed SNRM of 6) and push the setup  for 24 hours, if possible. Normally, syncing the Billion in a morning, at high sync rate for here, would cause loss of connection during evening/night due to the short downspikes, so I'll see how that goes.

[stevie]

One solution to charging the battery would be using a solar panel constantly trickle charging the battery. Something like that linked to below or similar device.

http://www.maplin.co.uk/Module.aspx?ModuleNo=223250&source=1

I notice it states Peak output:   3.42A @ 17.5V, I`d think this would be ample to maintain a battery if it had light for perhaps 8 hrs/day. I`d be tempted to use a 12 volt regulator on the output  to the inverter (just to ensure it doesn`t get to high for the inverter - relatively easy to do/build).

If your findings are that use of battery & inverter work then maybe a deep cycle battery would be better option?

Interesting idea & project.

[HPsauce]

I'm a tad sceptical TBH.

I would have thought that a good (i.e. not generating RFI) power supply at a suitable distance from the router, possibly with ferrites on the DC lead to suppress any "leakage" would be equally effective.

I happen to know people who both work on designing such equipment AND are involved in the RFI testing (a very interesting topic in itself) and quite frankly, as long as the phone line, ADSL lead and router are kept away from close contact with AC mains (and AC powered devices) there should be minimal interference.

Of course that doesn't stop other sources, especially faulty ones, generating problems from anywhere along the route of the phone line.

[roseway]

I'm really interested in this experiment, and it may have relevance to my own situation. I think it's an undeniable truth that powering the router from the cleanest possible supply should be considered best practice, so the only questions are how much benefit (if any) it provides and what's the best way to provide a clean supply. Obviously the car battery is a bit cumbersome and won't be a solution for many people, but if it proves the point that some mains-borne interference can be eliminated this way, then it's a valuable addition to the sum of knowledge.

Keep up the good work!

[jack21]

Thanks Roseway, and others, for the comments;  it is an experiment and I'm prepared for success or failure....or somewhere in between!

Already I've found that it doesn't eliminate certain large interference/SNRM drops, but it does seem to have a tidying effect on the overall SNRM, and see latest results following...

Today, already having a default SNRM of 15 due to a village-wide phoneline plague a few weeks ago, so no low default margin to protect, I've been pushing the DG834 SNRM downwards, in stages, to regions I've never previously been able, and my sync rate is the highest I've ever found achievable on this line in the whole 44 months I've had broadband - a whopping 2368 ! (at SNRM 4db - it surely can't last!!). And yet for the past few weeks, using SNRM-tweaking,  I could not reach and keep 1728 at 7db!! without spikes and connection drops.  No downspikes seen as yet ........remaining hopeful!

CRC rates are high - 23/min, whereas the norm has been between 0.3 and 5 per minute.

[jack21]

Update:

Having run for 3+ hours at the highest sync rate I've ever seen at home, via the battery-inverter method, with the straightest, least variation SNRM (+/- 0.1db) and with, very unusually,  no spikes whatsoever, I thought I ought to backtrack a little and recheck the previous mains-based setup.......I did not expect to get any higher than 1728kbps.

Oh, my....switching back to mains and retweaking the SNRM by the same amount (65445) now gave me exactly the same SNRM and sync rate (2304) as the battery/inverter method, and it has remained steady ever since - no down spikes, slightly wider SNRM variation +/- 0.2db and higher CRC/ES rate.....something I've never seen before today.

This has astonished me - my setup has never run so well, and just 2 days ago I couldn't hang on to 1728kbps !! Is it possible that somehow I've affected the exchange/line/router behaviour by my experiments, I wonder. I do notice that at some stage today, the downstream interleaving value has risen from 8 to 16......I'm not sure why, how, and at what point it did so, and whether it contributes to the better overall state of play.

[HPsauce]

Hmmmm. It's the weekend.  8)

Interference from commercial activity, especially of an industrial nature, largely missing.......

(my scepticism is maybe a bit more than a "tad" now)

I'd keep monitoring over days, maybe weeks, to see if a pattern emerges.

[sevenlayermuddle]

I'm afraid I have to say I'm as unconvinced as ever 

I argue that if any significant mains-borne reaches the router via the DC power inlet, then the router PSU 'brick' is either broken or badly designed.    Mains-borne interference is radiated as EMI from the mains wiring and then induced into the router & phone cables, and I'd have expected that induction effect to be massively greater than any interference vie the DC power inlet.

Moreover, having a 240V inverter anywhere in the same house as DSL, let alone connected to the router, just has to be a recipe for more interference, as they contain pretty vicious SMPSUs.   Even my kitchen halogen downlighters' PSU costs me quite a bit of speed.    Have you tried an AM radio near the inverter, to see how much noise it is kicking up?

More scientifically, it would be interesting to see some RS graph plots of comparitive bit-loadings, with the router fed from inverter, and with the router fed from mains (and inverter switched off)?

BUT I'd still be willing to be proven wrong.  Believe it or not, it's happened before 

- 7LM

[waltergmw]

Gentlefolk,

I agree with 7LM. The ADSL modem is not really a modem as it used to be known just modulating and demodulating a signal to produce a stream of digital bits.
It is a highly sophisticated special type of computer performing some seriously complex algorithms as Kitz explains in her dissertations on the subject.
The reason for all the complex algorithms is to maximise the number of digital bits that can be transmitted on the maximum number of frequency tones over kilometers of aluminium and copper twisted pairs which are subject to many types of radio frequency interference transmitted either through the air or in crosstalk between different twisted pairs. On a few occasions with crossed pairs the modem is further exercised by not have a balanced pair signal at all.

In these circumstances the power supply should be designed to provide a stabilised power source for the computing engine. Certainly the 2Wire modem is also provided with a Faraday cage to help to reduce local noise pickup and I expect other modems have been designed to do so too. Remember many have an actual wireless transmitter within close proximity of the computer.

Kind regards,
Walter

[jack21]

Update:

Remained on mains, lovely and steady - no spikes, and hi sync - until 20:40 when SNRM zero'd, CRC spiked by 1500, and connection dropped. Connected again, SNRM began to edge downwards - steadily and smoothly - until 22:00, when SNRM zero'd, CRC spiked by 2200 and connection dropped. It returned, but with an unsteady SNRM which fluctuated rapidly by +/-1.5db.  Wallander on TV ended, and I then saw what was happening and switched to battery power - SNRM continued to rapidly fluctuate, but by a reduced +/-0.5db.......so I let it run for 30 mins and then switched back to mains, whereupon the fluctuations returned to around  +/-1db.

I left it on mains overnight, and by morning the SNRM settled to +/-0.2db but with downspikes of  up to 2db roughly 10/hour. Switched back to battery, immediately saw 4 * downspikes of up to 1.5db within the first 4 minutes, and then no further downspikes for the next hour or so, when a 2db downspike happened. I switched back to mains, and the more frequent spiking (5 in 20 mins) of up to 2.5db was again present.

Its not very conclusive, is it? The battery-inverter-router method does seem to have a minimising effect on the SNRM-fluctuating levels, and may even cut out some of the downspikes, especially the smaller ones. But I can't say that the performance difference is such that I'm prompted to make a permanent change towards this method.

I haven't studied the evening performance drop for over 6 months - PC is usually turned off by 6pm - but its still clearly present, and very significant......its what is preventing me achieving a higher continuous sync rate and I/P profile. So this evening I'm going to switch to battery early on and remain on battery for the worst of the period; see what emerges.

Additionally, I've got a Billion 7202 12v DC router that I'm prepared to try the battery-direct approach with....later this week.

A further experiment, in due course, is to try shutting down the router a few hours before dusk and switching on again between 6 and 7. It looks as if I can push the SNRM to very low levels during the daytime, and I wonder if the BT I/P profile adjusting mechanism will eventually recognise the higher sync rate......its worth a try!

[jack21]

Hello 7LM and Walter,

Thanks for the comments; all enlightening.

I don't keep the RG bitloading graphs, but for my own extensive router-testing trials, I do keep a summary of them - the summary provides a total of the number of bits in the tones 35-144 (until yesterday I never reach more than tone 132), the number of occupied tones, the number of tone-gaps, and the average bit-per-tone for the tones 35-104 (104 over time proves to be the normal higest tone where more than 2 bits are present).  In my experiments, prior to pushing the sync rate, the inverter method and mains method produced almost identical results (the graphs looked the same, too).

My current maintainable values (at default SNRM 15 but with adslctl config --snr 1) are around  (mains) 447  85/1  5.9  (at 05:00) and  482  91/1  6.1  (at 09:15) ......and for (battery)  457  89/2  5.8 (at 05:50) and   497  89/3  6.4  (at 07:00),  but of course with the variability of DSL, conditions vary quickly.

whereas last Jan the norm for maintainable service was  557  82/8  7.5   at SNRM 8 -  Yesterday, I achieved the highest results ever -  680  103/1  8.2   but that was not maintainable after dusk fell.

The 2700HGV being one of my top routers, and an especial favourite once I have had a low default SNRM, it does produce marginally better results to the normal DG834 (Broadcom) variants and Billion 7300 if performing at a given SNRM - enough to make it my first choice when I've had a default SNRM of 6. But it can't produce the same results as the tweakable ones once the default SNRM is 9 or above.

I can't say that the inverter produced any noticeable interference, either when online or just running on standby, which shows up on routerstats - I've no means of checking otherwise.