Kitz Forum
Chat => Tech Chat => Topic started by: burakkucat on June 09, 2012, 03:02:23 AM
-
Quite recently, Asbokid provided me with a link [1] to an interesting document. After reading it, I decided to examine the "FUN-JIN" branded plug-in microfilter that TalkTalk had given me, when I first subscribed to their service five years ago. That microfilter was only used for a short period of time, when it was replaced with a filtered faceplate fitted at the NTE5/A -- an NTE2005.
Below are four images that I managed to capture with my flat-bed scanner. Notice the severed green wire? That was exactly how it was received. It had clearly been done as an afterthought, in the factory, maybe?
So there are four wires, yellow, green, red & black, that originate from the plug. The green and red wires just connected the corresponding pins of the plug and the (telephone) socket together. In terms of IDC#s, of which we are all familiar, they are IDC3 and IDC4 -- the "ring" wire and the earth wire. The severed, green, is the "ring" wire and the red, the earth wire. The other two wires, yellow and black, are IDC2 and IDC5 (the B- & A-wires) respectively. The yellow and black wires are connected directly to the inner pair of the RJ11 socket. From there, they then continue to the PCB (the filter circuit) and then to corresponding pins at the telephone socket.
From the above, I deduce that there is just a low-pass filter (for frequencies from 0 up to about 3.5 kHz) between the line and the telephone. After a moments thought, I assume that every modem/router has a band-pass filter for ADSL/ADSL2/ADSL2+/VDSL2 frequencies, as appropriate, which block the low audio range and those radio frequencies above the top of the VDSL2 range.
The low-pass filter in the "FUN-JIN" device consists of three components. Two chokes (wound on a common core) and a capacitor. As of yet, I have not determined the exact circuit due to "gunk" on the PCB!
More to follow, once I've de-gunk'd the PCB. ;)
[1] http://www.rsars.org.uk/ELIBRARY/BBCONNECT.pdf
-
This is an interesting investigation. After reading the linked PDF I was reminded that ADSL Nation claim that their XF-1e filter and XTE-2005 faceplate have a sharper response because they are active filters, so it would also be interesting if someone with the appropriate gear could test their frequency response.
-
The current filters put the entire spectrum through to the router.
So the router's connection from the filter is direct connected to the incomming phone line and is totally unfiltered.
The document in the link details an experiment to pre-filter the incoming phone line. So that what gets though to the normal internal faceplate filter (or the ADSLnation/BT type dangling ones) is cut off above the ADSL1/2/2+ upper frequency limits.
Isn't this what the RF3 pre-filter is intended to do? - where its fitted and where there is a problem with interference from frequencies higher than the upper limits of V/ADSL on to the ADSL line - such as RFI (though goodness knows where the RF3 starts cutting off)
If so this would suggest that the filtering inside the router for these higher frequencies is not so hot - I suppose its all down to cost in the manufacture/design process.
-
I haven't had the time to "de-gunk" the PCB, so that little (in both senses of the word) mystery remains.
Having read both Eric's and c6em's posts, I guess we need to find someone who could provide a loan of both an ADSLNation XTE-2005 filtered faceplate and a BT80B-RF3. Somehow I think Walter will be that person. ;D
Next, we need to consider who would have the technical expertise and ability to perform the analysis. We need an electronics wizard. Eric would be an ideal -- if he still had access to suitable test equipment. So my thoughts turn to Asbokid. :-\
To be continued . . .
-
Hi, no I haven't got the kit, sorry. I've got a DSO but to perform proper tests on the DSL filters, it would also require a signal generator to sweep up from voiceband frequencies right through to 30MHz, while measuring the response on the scope.
All the same, that's a very interesting article from Stephen Wilson, G3VMW, formerly "engineering manager in charge of telecommunications for Yorkshire Electricity". It was generous of him to put it into the public domain. Has anyone encountered him on the internet?
Though it's not clear how a better filter would alleviate many of problems from amateur radio interference. Wilson talks about 160m band, 80m band, right down to the 10-metre band.
160m is around 1.8MHz to 2.0MHz. So that's at the top end of the ADSL2+ bandplan.
80m is around 3.5MHz to 4.0 MHz. That's going to interfere with VDSL2 signals in the DS1/US1 bands.
10m band goes right up to 30MHz. Again, VDSL2 (but only Profile 30) signals.
Ham radios have very narrow channel widths, so interference should only knock out a few DSL subcarriers. Filtering out voiceband frequencies *and* frequencies above 2.2MHz (the top end of ADSL2+) is all fine and dandy (for ADSL2+).
But any ham radio transmissions on a frequency that falls within the DSL bandplan (e.g. a 1.8MHz signal from the 160-metre band), by definition, cannot be filtered out on the subscriber line. To do so would filter out the affected subcarriers from the DSL signal itself.
As the rollout of VDSL2 slowly becomes more universal, the problem of RFI from radio ham kit can only become more evident.
Mr Wilson is very fortunate.. To put up with internet disconnections every time he keys up, he must be blessed with very tolerant neighbours!
It's surprising that no independent tests have been published on DSL filters and faceplates. It's surely not a difficult set of tests to perform. And as a framework for the tests, there is already a BT 'SIN' (346) specifying how DSL filters should perform.
cheers, a
-
You could always try asking ADSLnation for the data on their own filters
I'm sure there must have been some certification done on them to comply with a BT SINE standard before being approved for connection into the UK telephone system. It seems likely, given their function, that a full spectrum analysis would have been done.
They may claim its proprietary information - but it's not like you are asking for part numbers of the components and the data could be obtained with the right equipment anyway.
-
Front of the ADSLNation PCB
(https://forum.kitz.co.uk/proxy.php?request=http%3A%2F%2Fi1053.photobucket.com%2Falbums%2Fs468%2Fblackeagle24%2F20120610_211915.jpg&hash=44c34e917bdcebfa8538e9b670e05291157f8ee3)
And the rear
(https://forum.kitz.co.uk/proxy.php?request=http%3A%2F%2Fi1053.photobucket.com%2Falbums%2Fs468%2Fblackeagle24%2F20120610_211935.jpg&hash=91d1abbb3da8bbe2fd5f0f84459e266ead745cb9)
-
:hmm: Hmm . . . I assume you are now busy drawing out the schematic diagram for us? :P
-
According to the ADSLNation site, it contains:
1 Spark Gap
4 Ferrite cores
5 Diodes
5 Capacitors
7 Resistors
2 Transistors
(But that's probably obvious :) )
-
Interesting stuff. I've always been a smidgen sceptical of 'wonder-filters', but I'm open-minded too.
I could've sworn there used to be a web page somewhere (maybe adslnation's own site) that published schematics for various filters, including the adslnation 'active' filters. But I'm darned if I can find it.
Anybody else remember seeing such a thing, or shall we put it down to a 7LM hallucination?
-
I have always been under the impression that a face plate filter, or another filter if connected effectively as the face plate filter (with all other telephone wires on the phone side of the filter) is easy to test.
Basically disconnect all of the telephone side so you just have the adsl/vdsl socket straight to the router (i.e. nothing to filter) and note the statistics (they should be the same as on a direct master test socket connection) or better monitor continuously with routerstats or dmt. Then add back the telephone side to the filter while keeping connected. If nothing changes then the filter is good enough. With the sync fixed, if the SNR drops a better filter may help. Infrequent noise sources are as always harder to test. A Pressac face plate passed this test for me with no discernible drop even with a phone call. An Adslnation filter was nearly as good but totally failed after a while! Other filters I had gave a drops of between 0.2 and 0.6 db in SNR on my wiring.
As an easier test which I have not tried---Possibly not messing with the basic wiring but monitoring the SNR and connecting to the telephone side with a long length of un-twisted or separated pair would simulate adding a bad home wiring situation. Equally some phones are good test noise sources and when a call is made or answered will impact the SNR with a poor filter. A phone test is good confirmation that filter is good but if the phone test fails the line may be faulty.
-
I could've sworn there used to be a web page somewhere (maybe adslnation's own site) that published schematics for various filters, including the adslnation 'active' filters. But I'm darned if I can find it.
There are a couple of schematics at the bottom of this ADSLNation page (http://www.adslnation.com/support/filters.php) but that's all I've seen.
-
So its basically a tuned circuit amplifier (really a switch in this instance) with some negative feedback then. ie one of the first transistor circuits every electronics engineer learns.
I can't see that doing much in terms of sharpening the response curve - it will simply remove some of the rolloff compared to a passive filter. Its not going to improve anything other than a faulty dect baseunit/early sky box and even then you'd probably do much better putting two passive filters in-line.
I can see an additional and quite unnecessary extra failure mode though - ie the transistors. The soldering on those pictures is appalling as well :)
-
One other point about the transistorised circuits interests me, and that is where do they draw power from?
I was under the vague impression that it would not be legal to deliberately draw current from the line's DC voltage when the phones were quiescent, or for any other purpose than powering the phones when a handset is taken off-hook?
Otherwise, we'd have 'LED night lights' and all manner of naughties plugged into phone sockets. If that would be legal, then perhaps 'Dragons Den summons' :D
-
My experience of using ADSL Nation faceplate :
Attenuation
Up Down
D-Link filter [1] 11.5 22
ADSL Nation faceplate 11.4 22
Though it also tidied up my bodged wiring and resolved the DSL noise on voice phone problem I've discussed in another thread.
So it did make a slight difference, but not much.
[1] D-Link filter as supplied with D-Link DSL-2780 modem from Talk-Talk. Exterior build quality looks good, I've never opened it up though.
Ian
-
I can see an additional and quite unnecessary extra failure mode though - ie the transistors.
That was also my initial thought when I first read about the ADSL Nation filtered faceplate, some years ago.
The soldering on those pictures is appalling as well :)
I'm glad you've said that, so now I know that my eyesight is still reasonable! :-\
-
I can't see the point of an active filter. The 3dB points on either active or passive filtering are going to be broadly similar and the guard band for ADSL is enough that filter rolloff is unlikely to be an issue. If you wanted to have a nice sharp response then simply add more poles - a four-pole passive filter is cheap enough to make and will do the job fine.
You only need an active component for notch filters operating below a few hundred MHz. The ADSLNation faceplate looks like overkill to me.
-
With you obvious knowledge on the subject, would you be willing to sketch out a few example schematic diagrams, please? ???
-
I'll knock something up later.
Having thought about this I don't believe the ADSLNation filter is actually "active" - ie its not a tuned circuit amp/switch.
I think the purpose of the two transistors is to switch a secondary filter in when the phone is in use, then switch it out when the phone is not in use.
The reasoning behind that would be that the default filter would filter everything out above say 50Hz (ring signal is 25Hz in the UK isn't it?), then when the phone is picked up the secondary filter switches in and allows frequencies up to 4kHz to pass through so voice works.
Quite an elegant solution as it effectively prevents the phone portion of the subscriber wiring passing any noise above 50Hz to the ADSL modem until the phone is actually in use.
On the page Eric linked, the second filter picture shows two reed relays in the top right hand corner. Assuming I'm right in my assumption, those relays will do the same job as the transistors - ie switch filters in and out. I think I'd rather have the relay solution than the transistor solution, less prone to failure.
-
I think the purpose of the two transistors is to switch a secondary filter in when the phone is in use, then switch it out when the phone is not in use.
That would address my earlier question, about the legality of drawing power (by/for the transistors) when the phone is not off-hook.
But I would have concerns about this approach, or the equivalent relay solution. If it actually conferred a benefit (which I accept, it may do in some cases) then surely that benefit would evaporate as soon as a phone were lifted off hook?
-
Not really.
You can design a Chebyshev passive filter with a nice sharp cutoff response for on-hook and then use a Butterworth filter for off-hook. The cutoff response of a Butterworth filter isn't as good as a Chebyshev but you don't end up with ripple on the passband, which isn't something you want being superimposed on the voice signal. That way you get the best of both worlds.
A quick scribble suggests that a 5 pole Butterworth lowpass filter with a cutoff frequency of 5kHz would have an insertion loss of 70dB @ 30kHz. The equivalent Chebyshev filter would have an insertion loss of near enough 100dB @ 30 kHz but it'd have 1dB ripple on the bandpass frequencies (0-5kHz).
http://en.wikipedia.org/wiki/Chebyshev_filter - look at the section entitled Comparison with other linear filters, specifically the Butterworth and Chebyshev type 1 graphs. That might help if you don't understand what I'm on about, which is quite likely as I'm no teacher :)
For any BT guys - is the termination impedance for a line 600 ohms? That figure rings bells so to speak :D
-
@Rizla,
Now I'm confused, I thought the suggestion was that the effective cut-off of the filter could be radically shifted, from a few kHz to a few tens of Hz, when the phone is on-hook vs off-hook?
I must admit I'm way out of my depth in the level of detail you cited, and further detail would probably drown me. :D
But can you at least confirm whether I'd misunderstood that point about shifting the cut-off?
-
No you're quite right in your thinking.
You can (with sufficiently large chokes) have a Chebyshev filter designed to cutoff at 100Hz which will have an insertion loss of >90dB @ 500Hz. That means virtually no noise will pass between the voice port and the DSL.
I think I've worked out WHY you need to have a switch - its CLI. You have to be able to pass CLI info to the phone and that's in the voice band.
So when a ring signal is received you switch filters so that CLI can be passed to the phone. So for example you could switch from your 100Hz filter to a 5kHz filter when the ring signal is received.
Apparently (according to SIN346) +/- 1dB ripple is permissable on the voice signal so Chebyshev would be the filter of choice due to the nice sharp rolloff.
Do note that this is all speculation as to how the XF-1e and FDX100 filters work, although I'm buggered if I can work out what else relays would do in a passive filter :)
Edit - are there any chunky looking capacitors on the ADSLNation filter and if so what are the markings? There will be one 1.8µF cap for the ring circuit, just curious about the rest.
-
Just had a look at the filters Sky send out.
Its a 4-pole filter, two chokes in series on both the A&B lines, couple of caps across A/B after each choke (so its a series filter rather than shunt filter, as you'd expect), ring cap across pins 4 & 5, couple of 82 ohm resistors after the second chokes for impedance matching purposes.
Datasheet is here : http://www.lea-networks.co.uk/upload/pdf/65_131583446053.pdf although it looks like they just copied the requirements of SIN346.
Looks perfectly adequate to me and if you stick a couple of them in series that will sharpen up the rolloff nicely.
I think the trannys/relays are overkill and just add another failure mode.
-
are there any chunky looking capacitors on the ADSLNation filter and if so what are the markings? There will be one 1.8µF cap for the ring circuit, just curious about the rest.
Yep, 1.8µF electrolytic & a 1µF electrolytic (250v, 85°) next to the bridge reccy. Q2 is a 2N2222 and Q1 is an MPS2222A, so basically same transistor.
There are no values on the caps below the coils, but the mfg numbers are F393J (on the left) and F562J to the right of it. No markings on the coils at all apart from a coloured dot !!
-
. . . (ring signal is 25Hz in the UK isn't it?) . . .
I'm not aware of any changes, so I'll answer with a tentative "Yes" and wait for Black Sheep's definitive answer. ;)
For any BT guys - is the termination impedance for a line 600 ohms? That figure rings bells so to speak :D
"Yes", says the b*cat, emphatically. I don't think we need the wooly-back's assistance for that query. ;D
-
The 1µF cap will just be there for some rough smoothing on the bridge rectifier.
Looking at the board and trying to work out what's happening I think that there is only one filter and I've been overcomplicating things :)
It rather looks like the phone is electrically disconnected from A/B in the NTE while the phone is on-hook and no ring signal is being received.
If the phone is taken off the hook then Q2 is switched on, which in turn switches Q1 on and connects the phone to A/B via the filter.
If a ring signal is received then Q1 is switched on, which in turn switches Q2 on and connects the phone to A/B via the filter.
So basically its one filter with a switch which only connects the phone when its off-hook or a ring signal is received. At all other times you can effectively consider the phone - and all associated wiring - as unplugged.
-
So basically its one filter with a switch which only connects the phone when its off-hook or a ring signal is received. At all other times you can effectively consider the phone - and all associated wiring - as unplugged.
Which would bring us back, would it not, to my earlier criticism... that the benefits evaporate with ( possibly substantial) DSL degradation whenever the phone goes off hook, or ring tone received?
-
Yes, once the phone is in use it will be dependent on the filter design which (going by the number of chokes) is a 4 pole series filter. As the chokes have no value on them we can safely assume they're custom wound for this particular application. That should make for a nice sharp filter.
I think its a rather elegant design though. I don't think the idea of disconnecting the voiceband cabling when not in use would have occurred to me.
The main failure mode would be lightning causing Q2 to go open collector, hence the spark gap just in case. Looks like a "one-shot" job which is fine as its secondary to the NTE spark gap. They've used bipolar trannies too so no worries about static damage, only likely lifespan reduction would be excess heat when soldering (which is pretty bad on that board).
I'm quite impressed with that filter - I can see why its recommended for people with "dodgy" internal wiring.
-
I use an ADSL Nation filtered faceplate, having previously had problems with ADSL noise on the voice telephone, despite using correct filter. I assume there's something odd with my wiring, though it all works ok.
Anyhow, with the ADSL Nation filtered faceplate I get the same S/N margins and line attenuation with the extensons wired through the filter on the faceplate as I do if I plug my modem direct into the test socket with everything else disconnected. The synch speed was about 250K lower with the filtered faceplate at the time I tested, but I reckon that's easily within margin of error. (I've later had it a bit higher than the "test socket only" synch speeds)
So : I certainly like it. Didn't study the soldering quality on my one, so can't comment - but maybe I'll open it up soon and take a few pictures.
Of course, I can't comment on how it compares to a standard BT issue filtered faceplate, or similar - I've not tried those. Someone should do some independent tests :)
Ian
-
All fine and good, nut I can foresee the following problem....
Subscriber is out, gets an incoming call. Router is connected with (say) 6dB SNRM. Internal wiring is in poor state.
1) The internal wiring is connected, the SNR degrades, and the router drops the connection.
2) Caller gets no answer, and hangs up, router meanwhile is starting to reconnect.
3) Internal wiring is disconnected again
4) Router completes reconnection, achieving same speed and same (fragile) SNRM as before.
...All of above may seem harmless enough. But suppose it is repeated ten times inside an hour, ie somebody is keen to call you... that is the threshold for which BT's DLM reacts to slap an extra two ' stability steps', i.e. 6dB increase in target SNRM.
One assault by DLM is bad enough, but if same thing happens just one more time then the subscriber would find himself in DLM's notorious 15dB prison, from which it is very difficult to escape. :no:
Admittedly, the likelihood of a second occurrence would be reduced as the higher target margin would provide some immunity, but it is not inconceivable.
-
Well in that instance you'd already have been DLM'd down anyway had you used another filter. I appreciate what you're saying about the BT system but that's a system problem and not a technical problem. DLM is there to save OR/ISPs money, not to keep users happy :)
Your scenario may however complicate diagnosis of fault by the ISP - eg "it only happens when the phone rings" - which then may land the EU with a charge from OR (via the ISP) when dodgy internal wiring is found.
Its still an elegant design.
-
Well in that instance you'd already have been DLM'd down anyway had you used another filter.
With the greatest of respect Rizla, I don't think I quite agree on that.
With a conventional filter, the noise from internal wiring is present all the time (or at least when the router initially connects) and it is catered for in the negotiated connection speed. Admittedly that speed may be reduced (compared with transistorised filter) but, even during calls, the connection would be sustained at error rates that ought to be acceptable to DLM.
In contrast, the transistorised filter would allow the router to connect at an artificially reduced noise level (and hence negotiate a faster speed), but then to suffer a sudden outbreak of reality every time the phone is used. If that causes the error rates to soar, or worse, the connection to drop, then DLM may well intervene.
-
What I meant was you'd have a far lower "normal" sync rate than without the filter in question.
The fact BT's DLM throws a wobbly IS a system problem, it really doesn't detract from the design of the filter. Its a technically good design which pre-dates DLM. DLM is the problem not the filter :)
-
Is there any possibility that an expert in such matters can now sketch out the schematic diagram, please? :-\
-
If you mean the ADSLNation one then it'd be easiest for Blackeagle - there's some tracks around the chokes/caps I can't see on the front of the board, also around the connector. I had to assume its a 4-pole filter due to number of chokes/caps but it could well be slightly different.
-
B*eagle is a little busy at present, but will draw a schematic as soon as time allows !!
-
Thank you, B*Eagle. :)
-
...that's a very interesting article from Stephen Wilson, G3VMW, formerly "engineering manager in charge of telecommunications for Yorkshire Electricity". It was generous of him to put it into the public domain.
Stephen Wilson's article is missing from its original site, but it can be found here now:
http://qso365.co.uk/wp-content/uploads/2012/02/ADSL-filter.pdf
cheers, a
-
Very interesting article :)
-
There's a hefty book chapter here on the purpose, function, and design of xdsl splitters:
It is excerpted from Golden et al's 2007 book Implementation and Application of DSL Technology:
http://docs.google.com/file/d/0B6wW18mYskvBN2hiU1VFR2FHNGs/edit?usp=sharing
cheers,a
-
Thank you. That chapter has now been downloaded, ready for a 'quiet moment'. :)
-
That's really interesting. I too will await a 'quiet moment'. :)
-
Heh, group delay on a telephone pair at voice frequencies. Seem to remember that was part of a module in electrical principles, along with working out Zo from first principles.
Blast from the past :)
-
A belated comment on rizla's point on active filters.
I think the point is the cutoff is sharper than a plain RC filter, NOT than is possible with an LC filter. At low frequency the trade off between cost and bulk of an LC filter and (for a Sallen and Kay Active filter, independence of source and load impedance ) is appreciable but my slight experience and understanding of theory is that an active filter can't do anything a passive one can but it may be more convenient.
A supplementary hypothetical question - AIUI xDSL is channelised partly to overcome line group delay variation over the frequency. In a case of interference from Droitwich on 198 kHz, would it be worth using a notch filter and sacrificing one or two channels to prevent a modem going into overload and loosing the lot, or would the extra delay variation make the cure worse than the disease ?