Talk:Twisted pair

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Common mode[edit]

@Kvng: Why have you removed all mention of common mode and the link from the explanations section? It's an important concept here and a link would be useful. I agree that Common-mode rejection ratio was not the best link to point people to, Common-mode signal would be better. By the way, I think some merging needs to happen here, we also have Common-mode interference which makes two articles too many. SpinningSpark 09:13, 28 March 2018 (UTC)[reply]

Look carefully at the diff. I added a link to Common-mode signal as I removed the link to Common-mode rejection ratio. I'm sure you appreciate that the failing described in the second paragraph of the Explanation section is not the result of CMRR issues at the receiver but non-CM noise picked up by the transmission line.
Common-mode interference doesn't seem to be about differential signals so I don't think a merge with Common-mode signal is appropriate. ~Kvng (talk) 14:17, 28 March 2018 (UTC)[reply]
Ah, so you did. Sorry, I missed that, I had only seen the removal.
On the merging, what do you think of merging to a new title like common mode (electrical circuits)? The article common-mode interference is about differential signals (and interference thereto) except for the second 1037C def. I think that should be removed as it is a different topic, and by policy we have one topic per page. It is also a rather unusual use of the term. SpinningSpark 16:37, 28 March 2018 (UTC)[reply]
By my read the first part is about interference in single-ended signaling. Thes second part is the same noise affecting multiple otherwise unrelated signals. ~Kvng (talk) 23:32, 29 March 2018 (UTC)[reply]
The first definition reads "Interference that appears on both signal leads (signal and circuit return), or the terminals of a measuring circuit, and ground." "Appears on both signal leads (signal and circuit return)" can only mean common-mode interference on a pair of conductors. Similarly, "the terminals of a measuring circuit" is plural. They mean common-mode interference on both terminals. Perhaps you were misled by the final clause "and ground". The comma indicates that clause is applied to both parts. That is, the common-mode interfernce on both legs of the circuit is identical as measured to ground, not that single-ended ground return signalling is being considered. SpinningSpark 08:24, 30 March 2018 (UTC)[reply]
I think we've started splitting hairs here. In the end, there really is no such thing as single-ended signaling. All measurements are performed with respect to a reference so all signals are differential. But that's not how introductory sources generally treat things - I guess they don't want to risk blowing anyone's mind at the outset. Anyway, I'm reading that 1037C definition differently than you are and we'll have to leave it at that for now. ~Kvng (talk) 20:25, 30 March 2018 (UTC)[reply]
Differential signalling uses either a differential amplifier or transformer to convert to single ended (ground referenced) signal. Single ended means that the ground reference has low enough impedance to be a useful reference. In any case, remember that signals don't travel in wires, but in the electromagnetic field between wires. (That is most often used to explain split-pairs in ethernet cables, but is true in all signal cables.) In the early days of computers, transistors (or even vacuum tubes) were expensive, relative to wire. Systems used a lot of wire, to save on transistors. Now, we use transistors to save on wire. With a transformer, the differential signal can have a very large common-mode component. With a differential amplifier, not as large. (There are stories of running UTP ethernet with 220VAC common mode. Be careful where you put your hands.) Gah4 (talk) 21:19, 30 March 2018 (UTC)[reply]

Balanced line and differential signalling[edit]

@Constant314 and Omegatron: Haven't we had this discussion on another article? A balanced line does not imply differential signalling (although diff. sig. can be used on a balanced line). Typically, balanced lines are transformer floating so saying the signals are equal and opposite has no meaning as it is not possible to state the voltage to earth, or any other reference. It is even possible to tie one leg of the pair to ground (through an impedance) and still maintain the balance. SpinningSpark 23:29, 2 August 2019 (UTC)[reply]

This article is about twisted pair and not about balanced line. There is no requirement to use differential signaling or equal source impedances for the two wires making up the twisted pair. I reverted Omegatron’s edit not because it was wrong, but because, in my opinion, what came before was marginally more useful to an uninformed reader. A mention of a common type of signaling means more than a discussion of source impedances. But, there is room for both a discussion of impedances and signaling. There is plenty more that can be said. The old TTL differential drivers had one impedance when driven high and a different impedance when driven low and the high and low levels were both positive with respect to ground. The telephone folks use a clever arrangement of transformers to use both the differential mode and the common mode to create a “phantom” channel. Constant314 (talk) 00:25, 3 August 2019 (UTC)[reply]
My point is that saying In a balanced line, the two wires carry equal and opposite signals... is definitely wrong. This could be reworded to something like Twisted pair can be used as a balanced line carrying equal and opposite signals... which is both true and avoids introducing other concepts. I am fully aware of phantom circuits, I wrote most of that page, but I fail to see the relevance of that, or differential TTL to this discussion, other than to emphasise the wrongness of your preferred version of the article. SpinningSpark 00:51, 3 August 2019 (UTC)[reply]
No-one is going to make a twisted-pair with different wire sizes for the two wires of the pair. It obviously works best to be both current and voltage balanced, but that might not always be needed. How well you maintain balance depends on how much EMI (both directions) you can stand. Near field the voltage (capacitive coupling) is more important, far field current (inductive coupling). As well as I know, if you drive a twisted (or untwisted) pair with an unbalanced voltage, the current (at the important frequencies) will be close to balanced. (Probably depends on appropriate bypass capacitors on the drivers.) Also, ferrite rings are commonly put on cables to reduce common-mode currents (they won't see voltage) and so reduce EMI. Otherwise, you don't need to say the voltage to earth down to DC, if it is balanced at the important (high) AC frequencies. But I think the important part about twisted-pair is that it can be balanced and have minimal RFI. Gigabit ethernet through UTP cable is pretty amazing. Gah4 (talk) 01:05, 3 August 2019 (UTC)[reply]
Who said anything about the two wires having different gauges? Balance of the circuit depends on the impedance inserted into the circuit by the source and load as well as the impedance of the line. It does not depend on the symmetry of the applied signal one bit. SpinningSpark 01:21, 3 August 2019 (UTC)[reply]
My preferred version is to say something about the signals and not just about the impedances. Lines without signals are not very useful. I have no objection with saying that a twisted pair can be used as a balanced line. I have no objection to making it less wrong. The TTL driver is a counter example to the line having the same source impedance on each wire of the pair. Constant314 (talk) 02:07, 3 August 2019 (UTC)[reply]
A fairly common case is flat ribbon cable with odd wires grounded. (There are twisted pair ribbon cables, but they are more expensive, and so relatively rare.) If you drive one wire with the rest ground, the signal will go down that wire, with an appropriate impedance, and most of the return current going through the two adjacent ground wires. The far end is the terminated in the appropriate impedance to ground, and so minimizes the reflection. If there is no reflection, then source impedance mismatch doesn't cause much problems. If you drive a coax cable from TTL with the shield grounded, current will go down the outside of the center conductor, and return current on the inside of the shield. If you drive a UTP cable the same way, current will go out on one conductor, and mostly return on the grounded conductor. If the common mode impedance is relatively high, the common mode current will be relatively low. Close enough for common usage. So, the current is close to balanced, even when the voltage is not. Gah4 (talk) 07:35, 3 August 2019 (UTC)[reply]
All those examples are not true balanced. Ribbon cable obviously isn't. It is not relevant that most of the return current goes down the grounded wire (and that's ignoring the elephant in the room that the return current is split between multiple wires). The issue is that the induced common mode current will be radically different between a hard grounded wire and one with kΩ of impedance. Likewise, differential TTL is not balanced so is not a counterexample. Differential signal != balanced line. I don't know why we are even discussing these cases. Constant314, you say you want to write something about signals, but remember, you are writing it in an article about a line. The properties of the line should be the focus, not the signals it might carry. SpinningSpark 09:57, 3 August 2019 (UTC)[reply]
I don't have any other comments. Do what you want. Constant314 (talk) 17:49, 3 August 2019 (UTC)[reply]

"I reverted Omegatron’s edit not because it was wrong, but because, in my opinion, what came before was marginally more useful to an uninformed reader. "

@Constant314: But "what came before" was wrong, and you reverted to it. — Omegatron (talk) 13:45, 3 August 2019 (UTC)[reply]

I've rewritten this without mentioning either impedance or differential signalling. SpinningSpark 17:55, 4 August 2019 (UTC)[reply]

As well as I know, twisted pair was first used for telephone wiring, not so long after RFI sources were created near phone lines, and with balanced signalling. Balanced differential signalling is a common, but not the only, use for twisted-pair cable. Gah4 (talk) 04:56, 5 August 2019 (UTC)[reply]
Balanced lines were used in the telegraph era in some places to overcome interference from power cables, particularly from trams. However, these were not twisted pairs which, as you say, was invented specifically for telephone lines. SpinningSpark 15:23, 26 November 2019 (UTC)[reply]

ENGVAR[edit]

I've changed this article back to British English. The very first version of the article was clearly BritEng by the use of "cancelling" rather than "canceling". The article was changed to American English with this edit with the rationale "spellchecker". Per WP:ENGVAR there is no standard variety of English on Wikipedia and per MOS:RETAIN the original established variety should be retained unless there is consensus to do otherwise. SpinningSpark 17:53, 4 August 2019 (UTC)[reply]

Today, most of the millions of kilometres of twisted pairs in the world are outdoor landlines, owned by telephone companies, used for voice service, and only handled or even seen by telephone workers.[edit]

The article says: Today, most of the millions of kilometres of twisted pairs in the world are outdoor landlines, owned by telephone companies, used for voice service, and only handled or even seen by telephone workers. Withe large amount of networking cable being installed, is that still true? Do you count by pairs, so a Cat 5 Ethernet cable counts as four? Gah4 (talk) 22:26, 25 November 2019 (UTC)[reply]

I'm not even sure what point that statement is trying to make. Is it saying that most twisted pair is buried rather than on poles? Hasn't that always been the case? Most cable for any purpose is only "seen" by the technicians installing it. Those that are visible generally don't have the cores visible to the user. I'm inclined to remove it as a vague piece of hand-waving of little benefit to the article. SpinningSpark 15:32, 26 November 2019 (UTC)[reply]
@Zojj: who wrote it. SpinningSpark 15:37, 26 November 2019 (UTC)[reply]
I suppose I believe that it includes both buried and on poles. In the past, the other uses for twisted pair would have been relatively small, until UTP networking came along. Our house has an intercom system installed by the previous owner, using its own special four pair UTP cable. UTP cable part of a company's own PBX system is not owned by the telephone companies. Funny, I now remember living in a college dorm years ago, and being notified by the phone company that I now owned my inside wiring. (Presumably the school and not me.) There is a lot of UTP cable owned by the phone companies, but also a lot owned by companies for their own networking (and phone) systems. Gah4 (talk) 20:49, 26 November 2019 (UTC)[reply]
By the way, is there an easy way to find out who added some statement? I have a few times done it going back through the change history, but often that isn't an easy way. Seems like something that computers should be good at. Gah4 (talk) 20:49, 26 November 2019 (UTC)[reply]
It looks like this change was made by Zojj, but mostly changes the wording of statements already there. I didn't search it farther back. Gah4 (talk) 21:07, 26 November 2019 (UTC)[reply]
There is the Wikiblame tool, which I used to use quite a lot but nowadays I more often manually use the binary search algorithm. On a very heavily edited article you might need to set the limit on the number of entries displayed in the history log to a very large number (like, in the thousands) by manually editing the url. Not so good if you have a slow internet connection, but it works well enough for me. You are right, this goes back further than Zojj's edit. It was User:Jim.henderson in this edit. SpinningSpark 22:55, 26 November 2019 (UTC)[reply]
A lot of UTP network cable has been installed since 2006, so it might have changed since then. Actually, I have no idea of the scale for either, but it was a better bet in 2006 than today. Reminds me of the condo that I used to live in, where we had lots of problems with phones. At some point, I was on the condo board and found where the wires came though. The main panel used springs to hold the wires, as in older than type 66. I finally got the phone company to replace it, and hopefully send the old one to a museum. I was remembering that with the mention of using 25 pair cable indoors, while I believe much is used outdoors, and specifically to the panel in our condo. (Though I didn't count them.) Are you against removing it without asking? I suppose I can ask Jim.henderson. Gah4 (talk) 00:49, 27 November 2019 (UTC)[reply]
So, it seems I've been caught, after all these years. I am still confident in my hand waving, since telephone wires are long and computer wires are short. Many places, probably a majority, that have a wired telephone also have a computer, and a similar proportion of those with a computer also still have a wired phone. The telephone wire pair generally goes to a panel, to connect each pair to a pair in a cable which extends to an exchange a km or more away. My own local loop is only half a km, but that's unusual. What's also unusual is for someone not in the business to see the individual wire pairs, anywhere from dozens to about 10,000, coming out of the cable and running through a fanning strip to their termination.
In comparison, everybody sees the twisted pair Ethernet cable from a computer at home. It runs usually only a few meters to the router, which connects it either to the telephone wire or to the CATV drop. Thus, if you could gently extract the telephone wire pair from its underground or overhead cable, it would usually be between a few hundred and a few thousand times as long as the computer cable. So, the hand waves confidently enough even if wired phones have become a minority, and even if the computer cable were to have a dozen wires.
Big business offices tend to have less of a contrast, as PBXes are not rare and are sometimes colocated with the Ethernet wire hub, but those cases of approximate equality remain a minority. I spent a couple hours today with a banker in a nearby big bank office that I happen to know has its desk phones directly wired to the same exchange, about 1.5 km away. In any case, I don't expect a reliable source to publish something that is so obvious to anyone with experience with both kinds of wiring, and that ties the hands that would wish to appeal to a higher court to contest a deletion. Jim.henderson (talk) 02:49, 27 November 2019 (UTC)[reply]
Yes, I didn't try so hard to actually figure out how much wire is where. Since business and other non-residential installations often run two four-pair cables to each location, it adds up faster. And business with PBX need fewer lines going to the central exchange. As well as I know, all wire internal to a house/school/business is now not owned by the phone company, though it might have been in years past. (That is for US, I don't know at all how other countries do it.) For business use, it is much easier to install more than might be needed in the first place, instead of installing a little at a time. But I don't have a very good guess for the actual value for either. Gah4 (talk) 03:03, 27 November 2019 (UTC)[reply]
Even otherwise technically sophisticated people often fail to appreciate how massive is communications infrastructure, especially in urban areas where it's all, well, infra, that is underground. For a decade or two it has been shrinking slowly, in places that were already economically developed in the 20th century, but still massive. Usually as the system approaches the periphery, spare capacity increases as a percentage, since usage is less predictable. Thus, in about 1970, the standard for residential apartment riser cables in my inner city area increased to three pairs per housing unit, but the subscriber cable had usually less than two, and spare switch terminations were usually 10% or so. Nowadays yes, the small parts of the loop have a different owner, but no need for this article to mention it. Almost half a century ago I was working 80 hours a week, installing and inspecting a new system intended to replace PBXes, which was pretty much successful but several big companies hung onto theirs and even updated them, partly to be different from nearby competitors. They had a ratio of approximately 5 to 10 phones for each DID or DDCO (both underground) twisted pair. Jim.henderson (talk) 05:11, 27 November 2019 (UTC)[reply]
Yes no reason to mention it, but if you want to calculate wire owned by the phone company and wire not owned by them, then the distinction is needed. Gah4 (talk) 05:32, 27 November 2019 (UTC)[reply]
I don't really dispute the accuracy of what Jim says, just the need to say it. It was the "only handled or even seen by telephone workers" bit that caught my eye. It is just so unnecessary to say that buried cables are only seen by the people who put them in or dug them up. BTW, in my part of the world, I have seen so many telephone street cabinets with missing doors (through vandalism or technician incompetence I don't know) that I'm pretty sure that everyone has "seen" the multicoloured twisted pair cores. I agree with Jim, though, that few people have seen the massive distribution frames in exchanges. These were even larger in the days when solder tag blocks were used, and probably some still in service.
Stating that twisted pair runs to "millions of kilometers" is more encyclopaedic, although it would be better if we had some actual numbers, at least for one country. I note this has been changed from Jim's original billions of loop-feet, presumably by someone who does not understand the difference between miles of line and miles of cable. This number is bound to be going down with the rise of fibre for internet connection and the move to mobile devices. Many people in my country no longer have a landline phone at all.
On ownership, in the UK the cables are owned by Openreach, set up as a separate division of BT so that BT's competitors could be (supposedly) treated equally. Wiring in private homes is owned by the resident. It used to be owned by BT but was changed in (I think) the 1980s around the time that BT was privatised. I agree with Jim that this makes little difference to the amount of line owned by the telephone companies. SpinningSpark 10:29, 27 November 2019 (UTC)[reply]
Sorry about the section name. When I ask a question about a quoted statement, I often put it in the title. They aren't usually so long. Gah4 (talk) 11:33, 27 November 2019 (UTC)[reply]
From landline there are 1.263 billion main telephone lines in the world (2003, and about the same later), and maybe 1km each. I think I was mostly wondering about the US, but obviously should consider the whole world. Yes many only have cell phones, but in many cases the wires will still be there. Well, at least where I am, the wires from each house go to a nearby junction box, and then to more such boxes before they get to the central office. That makes the number of pairs a little more efficient. But there is also a lot of wire in networking, in four pair cables, and often many to each room. When I did a school networking project some years ago, we put six jacks, two places, in each classroom, so 12 four pair cables. That was set by the school district. (To be used both for phones and computers.) Anyway, there is a lot of network cable out there. Gah4 (talk) 11:33, 27 November 2019 (UTC)[reply]
Yes, in large building infrastructure, twisted pair has still got some life, but I thought we were talking about landlines, and in that role twisted pair is alive but definitely dead man walking. My ISP now offers a fibre connection delivered through a wireless router – nothing comes into the house at all. I suspect that in the next decade or so businesses are going to move away from wiring infrastructure as well. SpinningSpark 12:26, 27 November 2019 (UTC)[reply]
Remember also the reason IT in enterprises went to twisted pair in the first place. Technically, it really didn't have a lot to recommend it. Coaxial ethernet was the previous standard method. This was technically better in every way and easier to install to boot, the only drawback being stupid users breaking the loop by disconnecting their PC. The spec of twisted pair cable had to be massively tightened before large bandwidths were possible with this method. Enterprises reason for going to twisted pair was that their networks could by piggybacked on to the existing telephone cable infrastructure, thus saving big bucks on installation costs. This original rationale has long since become moot since old telephone cable is unsuitable for modern networks. Their own cable has to be laid anyway and wireless solutions are now available. SpinningSpark 12:38, 27 November 2019 (UTC)[reply]