Talk:Ethernet/Archive 4

Telecomm or not
There seems a slow edit war over the Telecommunications template. I can understand both sides to some extent. Certainly historically Ethernet was a local area network, a very different world than telephone company networks. But over the years the boundary has blurred. With Ethernet in the first mile, Metro Ethernet and Carrier Ethernet, the traditional "phone companies" are now in the Ethernet business, and Ethernet standards have accumulated features normally associated with telecom companies. So I would vote to keep it in. The article is in the telecom wikiproject, and in the template itself for that matter. Do not see the harm in it, since it comes up unexpanded anyway. It could use some culling, e.g. ARPANET and Vint Cerf etc. are questionable. Any chance of a consensus? W Nowicki (talk) 18:26, 26 August 2011 (UTC)
 * Not everything that has recently been tagged with the telecom template actually belongs in that category. One might as well tag electricity and bolts, since you need both to make a telecommunication system. Every last ham radio article has been tagged; really, tag just the main article "amateur radio" and leave the low-level noisy details of repeaters and contesting out of the "telecomm" ambit. Ethernet runs for 100 metres and is used inside a building.  If you can run it between cities, it's not Ethernet any more. Inside a building is not telecommunications. Templates should not be added to articles needlessly. "Wiki Project Physics", anyone? --Wtshymanski (talk) 19:17, 26 August 2011 (UTC)

Not sure if you are talking about category or project above? But your claim about Ethernet is very outdated. The single-mode fiber physical layers with longer wavelengths, such as, say 1000BASE‑LX can extend ten kilometers. For many years switches had SFP ports that can support both short and long reach PHYs. A packet coming out of the switch has no idea if it is going into a laser that makes it travel kilometers between cities, or a twisted pair cable that goes centimeters. With a media converter it can even go back and forth on the same link, short reach, long reach, short reach. Certainly packets have no clue about building boundaries. And as end-users are increasingly using wireless, fairly soon if not now it will be at least a plurality of telecomm companies who buy copper and fiber Ethernet products. W Nowicki (talk) 20:58, 27 August 2011 (UTC)


 * "Ethernet runs for 100 metres and is used inside a building."


 * 100BASE-TX and 1000BASE-T (IIRC 10BASE-T has similar limits but they wern't spelt out as explicitly) run for 100 meters and are usually used within a building (you can run them between buildings but it's generally considered inadvisable, especially if the buildings have independent electrical supplies) but they are only a tiny part of what Ethernet is.


 * "If you can run it between cities, it's not Ethernet any more."


 * Once CSMA/CD was replaced by switching there was no longer any fundamental limit on the length of an ethernet link other than that imposed by the physical layer in use. I don't see how fiber physical layers are any less "Ethernet" than twisted pair physical layers are (neither was the original physical layer and both fiber and twisted pair physical layers started to appear at about the same time). Plugwash (talk) 00:42, 29 August 2011 (UTC)


 * As per the categories, Ethernet is a telecommunications technology. I don't think it has been miscategorized. Telecommunications is a broad topic and the new template has been pasted widely. I'm not convinced that these navigation templates are particularly useful but, if we're going to have them, this article is a much better target that some of the other places it has recently landed. --Kvng (talk) 22:11, 29 August 2011 (UTC)
 * I wasn't aware that you could get rid of CSMA/CD and still call it "Ethernet". Oh well, in a world where we click on "Start" to shut down the computer, I shouldn't expect better. We'll be reduced to pointing and grunting within a decade, at this rate. --Wtshymanski (talk) 22:54, 29 August 2011 (UTC)

Ah, one musn't show one's age. :-) Ethernet started phasing out CSMA/CD in the 1990s with switches and the full-duplex PHYs. The BASE-T chips still support it for backward compatibility, but the 10 Gigabit and above PHYs do not even try. Irony of course is it remains in the title of the 802.3 standards documents. We should make it clearer in the article. There is a great quote from Metcalfe (we need to get a cite for) on something like the 25th anniversary of Ethernet, when he was asked what computer networks would be like in another 25 years. He said he had no idea, except that they would still be called "Ethernet" (perhaps after the quote about Fortran which turned out to be less true). W Nowicki (talk) 18:30, 30 August 2011 (UTC)
 * You mean if I invite a bunch of buddies over for a LAN party using a nasty little hub I bought at Staples, we're not using CSMA/CD? Staples still sells hubs, don't they...last time I was there they were right between the 5 1/4 floppies and the carbon paper, right across from a display of Gestetners.  Oh dear. When did I get old? --Wtshymanski (talk) 19:32, 30 August 2011 (UTC)
 * "if I invite a bunch of buddies over for a LAN party using a nasty little hub I bought at Staples, we're not using CSMA/CD?" -- if you are using a hub (technically a multiport repeater) you are using CSMA/CD, if you are using a switch (technically a multiport bridge) then you most likely aren't using CSMA/CD. Note that it's not unheard of for switches to be mislabeled as hubs.
 * "Staples still sells hubs, don't they...last time I was there they were right between the 5 1/4 floppies and the carbon paper" -- i'm not in the habit of buying my network gear from staples but it's YEARS since i've seen a hub for sale at any place I normally buy computer stuff (there still seem to be new old stock hubs floating around the internet). Plugwash (talk) 01:00, 31 August 2011 (UTC)
 * So CSMA/CD still lives! I was very disappointed that my "wireless router" from Staples couldn't tell me which RJ 45 8P8C port was which. It's not really a "router", at least on the wired side. --Wtshymanski (talk) 16:03, 31 August 2011 (UTC)
 * A typical "wireless router" is logically (physical implementations may vary) a combination of 3-4 devices.
 * A router with two ports (LAN side and WAN side) and NAT capbility (which is enabled by default but can usually be turned off)
 * An ethernet switch
 * A wireless access point
 * Sometimes a DSL/cable/whatever modem
 * -- Plugwash (talk) 14:42, 9 November 2011 (UTC)

Encoding?
I was looking at Manchester coding and I understand this is how bits are transmitted by Ethernet but I don't see any discussion of encoding in this article. It seems to just assume you can put bits down a wire! Perhaps I am missing something? Billlion (talk) 14:07, 31 December 2012 (UTC)
 * Here is a (non-authoratitive) source? http://www.thenetworkencyclopedia.com/d2.asp?ref=1182 Billlion (talk) 14:10, 31 December 2012 (UTC)


 * Different coding schemes are used on different Ethernet physical layers. Specifically, Manchester coding is used for 10 Mbit copper variants such as 10BASE-5, 10BASE-2 and 10BASE-T. I have improved Manchester coding to reflect this. -—Kvng 16:40, 31 December 2012 (UTC)
 * Thanks, do we need to update articles on each of physical layers to describe the coding? Are several different types of coding used for the faster variants?Billlion (talk) 17:21, 31 December 2012 (UTC)
 * I'm sure there's room for improvement in this area. Different encodings are used at different speeds and for different mediums. For example: 100BASE-TX uses MLT-3. 1000BASE-T uses MLT-5. -—Kvng 18:27, 31 December 2012 (UTC)
 * Do you know a good source for this stuff? I am new to it. Billlion (talk) 02:31, 1 January 2013 (UTC)
 * The primary source is the IEEE 802.3 standard. -—Kvng 05:33, 6 January 2013 (UTC)

Ethernet over barbed wire
What about ethernet over barbed wire?

http://www.sigcon.com/Pubs/edn/SoGoodBarbedWire.htm — Preceding unsigned comment added by 74.42.105.10 (talk) 01:59, 18 February 2013 (UTC)

Etymology
Should this article include the etymology of the term "Ethernet"? I was thinking of adding the following subsection to the History section: Etymology

According to Metcalfe, Ethernet was named after the luminiferous ether, a postulated, ever-present "substance" widely believed to be the medium through which electromagnetic radiation propagated in the late 19th century:

Alternatively, it could be given its own section toward the end of the article if it interferes with the flow too much. ––67.180.24.57 (talk) 16:14, 4 March 2013 (UTC)


 * Nevermind. I don't feel it's significant enough to give it its own section, so I integrated the basic idea into the third sentence of the History section. ––67.180.24.57 (talk) 16:58, 4 March 2013 (UTC)

Collision-free?
I'm confused about this passage: "Performance on real networks is significantly better.[23] The 10BASE-T standard introduced a collision-free full duplex mode of operation that eliminated collisions. Modern Ethernets are entirely collision-free." First, my rusty memory from the computer networking course I took over a decade ago was that a core idea of Ethernet was collision detection with random delay before retry. I realize that in large networks, one would use switches not dumb hubs so any one cable can be connecting a computer to a switch, but my understanding was that hubs are still legit Ethernet and so collisions can occur in the standard. '''Is that wrong? Are hubs not allowed?''' Beyond that, the passage needs more explanation. In a switched network, if you have full-duplex communication, I can see a computer and a switch not colliding with one another since the "send" and "receive" conductor-pairs (as seen by either end) are distinct, but "Modern Ethernets are entirely collision-free" is still incorrect unless Ethernet really doesn't do collision-detection anymore. Did I totally forget how Ethernet is supposed to work? —Ben FrantzDale (talk) 12:14, 3 April 2013 (UTC)


 * There would probably be those who would argue that an Ethernet using hubs is not "modern". In that sense of "modern", no, hubs aren't allowed, just as big fat yellow cables with vampire taps aren't allowed.


 * ""Modern Ethernets are entirely collision-free" is still incorrect unless Ethernet really doesn't do collision-detection anymore". Well, it's true if the collision-detection hardware/firmware/software never detects any collision, as would be the case in a full-duplex switched Ethernet.  And, yes, a "modern" (switched, full-duplex) Ethernet isn't supposed to work the same way as a "traditional" (common-medium, half-duplex) Ethernet.


 * Perhaps "modern Ethernets are entirely collision-free" should be stated as "switched full-duplex Ethernets are entirely collision-free; most if not all present-day Ethernet networks are switched and full-duplex and thus collision-free", or something such as that, to make it clearer that, whilst collisions can, in principle, still happen, they don't happen in practice in most Ethernets. Guy Harris (talk) 18:02, 3 April 2013 (UTC)


 * Thanks. That makes sense. If I understand you correctly, I could plug three devices together through a hub and have them try to talk GigE and have packets collide and have them do the random fallback thing and it would still work. Yes? I'd go for something like "Presently, good network-design practice dictates the use of switches. Combined with the fact that individual cables are full-duplex, this means that collisions can never occur in a properly-designed network." —Ben FrantzDale (talk) 18:13, 3 April 2013 (UTC)


 * Actually, I guess if it is full duplex, then collisions shouldn't happen because a three-way-star topology over shared media doesn't make sense since as soon as one host sees a packet it didn't send on its transmit line, then it knows something has gone terribly wrong. —Ben FrantzDale (talk) 18:16, 3 April 2013 (UTC)


 * Gigabit Ethernet does define a half-duplex mode but AFAIK, no one has ever built a GBE hub. 10-gigabit Ethernet does not define a half duplex mode so no collisions are possible there. I support the proposed changes but want to be careful not to lead any readers to believe that they need to worry about collisions on today's wired Ethernet. -—Kvng 20:19, 3 April 2013 (UTC)
 * AIUI when it comes to medium access there are basically three "generations" of ethernet.
 * Early ethernet used a shared medium as the physical layer, collisions could happen on that shared medium, hosts had to deal with them and repeaters had to ensure that if a collision happened on one port the collision was repeated to the rest of the ports.
 * Later ethernet used physical layers that were naturally full duplex and point to point but those links were operated in a mode which simulated the affects of a shared medium to allow for the use of dumb repeaters (hubs) and to allow the new link technologies to be used with existing systems that expected a shared medium.
 * Modern ethernet uses full duplex point to point links, collisions simply cannot happen because each "channel" is only ever used by one transmitter. The CSMA/CD hardware is still there for backwards compatibility but it's disabled. If multiple devices are to be connected then switches are used which can buffer packets.
 * Of course the great thing about Ethernet is that these generations can be mixed freely on the same network. A port of a switch connected to a legacy end device or an old hub can happilly do CSMA/CD while the rest of the network runs in collision-free full duplex. Collisions can bappen with the "legacy" stuff but they are contained by the switch and don't impact the rest of the network. Plugwash (talk) 14:02, 7 May 2013 (UTC)

EWAN?
I noticed there's a type of internet connectivity known as Ethernet WAN? I'll quote what I read from here: "supports ADSL or Ethernet WAN connections (EWAN) that allow users to have the flexibility of different Internet connections among ADSL, cable or fiber modem using its interchangeable LAN/WAN port. This unique feature makes it easier when users need to change to fiber or cable services when necessary." Should we create an additional page for EWAN? I noticed there was like nothing on Wikipedia about it, or perhaps it's known as something else and EWAN should be a redirect. TinyEdit (talk) 23:28, 4 August 2013 (UTC)


 * From reading the User's Guide for that device, where it says "Enable EWAN function and you will be able to connect to Cable/FTTH/VDSL/ADSL device." at the bottom of page 6, I infer that the device is a combination of:


 * an ADSL modem, which can extract the ADSL signal from a telephone line plugged into the ADSL jack;
 * an Ethernet hub or switch with four ports (LAN1, LAN2, LAN3, and LAN4/EWAN);
 * a Wi-Fi access point;


 * and can:


 * receive packets from an Internet ADSL connection and transmit them on any of the Ethernet ports or the Wi-Fi network to a local host, so that the local host is receiving packets from a host on the Internet;
 * receive packets from a local host on of the Ethernet ports or the Wi-Fi network and transmit them over the Internet ADSL connection, so that the local host is sending packets to a host on the Internet;
 * receive packets from the LAN4/EWAN port, with an Internet access device of some sort (cable modem, fiber-to-the-home gateway, VDSL or ADSL modem) plugged into it, and transmit them on any of the other Ethernet ports or the Wi-Fi network to a local host, so that the local host is receiving packets from a host on the Internet;
 * receive packets from a local host on one of the other Ethernet ports or the Wi-Fi network and transmit them on the LAN4/EWAN port to the Internet access device, so that the local host is sending packets to a host on the Internet;
 * presumably receive packets from a local host on of the Ethernet ports or the Wi-Fi network and transmit them over any of the other Ethernet ports or the Wi-Fi network, so that it acts as an Ethernet hub/switch or a Wi-Fi access port, allowing local hosts to communicate with each other.


 * So EWAN is typically just regular Ethernet, used as a way of connecting Internet access devices to a local network, rather than a form of Ethernet used as a non-local-area network such as the "pure Ethernet" flavor of Metro Ethernet (to use the term in that article). I don't think the "pure Ethernet" flavor of Metro Ethernet is used as an access technology for home users; there are flavors of "Ethernet in the first mile" used for that, but they're different from the Ethernet physical layers used for LANs.


 * So, no, it's not a new form of Ethernet deserving of its own page; it's the use of Ethernet as a mechanism for connecting Internet access devices to a home network. Guy Harris (talk) 00:57, 5 August 2013 (UTC)


 * Sounds like a Residential gateway. ~KvnG 13:33, 7 August 2013 (UTC)


 * Right, the box that implements this kind of thing should be discussed in the Residential gateway article. The protocols that implement it are generally Point-to-point protocol over Ethernet (combined with Ethernet over ATM) or Point-to-Point Protocol over ATM for most of the DSL variants as I recall. (Aside, I once worked for a company that developed a true "Ethernet over telephone lines" protocol using DSL-like modulation bands, but it never caught on for Internet access, just within a home as a HomePNA protocol). Also many firewall boxes nowadays have Ethernet over twisted pair physical ports on both sides, and refer to one of the ports as the "WAN" port like above, which is the one that runs DHCP client vs. DHCP server etc. vs a single box with firewall and DSL modem combined. Ideally some of this could be described in the prose a bit better instead of just a litany of unexplained "see also" links. Although I do not think the EWAN acronym is that common. I was amused by what you get when you search for it! W Nowicki (talk) 15:55, 7 August 2013 (UTC)

Evolution
The "Evolution" section, second paragraph, states: "(...) An EtherType field in each frame is used by the operating system on the receiving station to select the appropriate protocol module (...)". This is correct only for the Ethernet_II variant, as 802.3 defines the 3rd field on the header to be the size of the frame. Zekkerj (talk) 23:33, 13 November 2013 (UTC)


 * No, actually, current versions of IEEE Std 802.3 allow both; from IEEE Std 802.3, 2000 Edition:


 * 3.2.6 Length/Type field
 * This two-octet field takes one of two meanings, depending on its numeric value. For numerical evaluation, the first octet is the most significant octet of this field.
 * a) If the value of this field is less than or equal to the value of maxValidFrame (as specified in 4.2.7.1), then the Length/Type field indicates the number of MAC client data octets contained in the subsequent data field of the frame (Length interpretation).
 * b) If the value of this field is greater than or equal to 1536 decimal (equal to 0600 hexadecimal), then the Length/Type field indicates the nature of the MAC client protocol (Type interpretation). The Length and Type interpretations of this field are mutually exclusive.
 * When used as a Type field, it is the responsibility of the MAC client to ensure that the MAC client operates properly when the MAC sublayer pads the supplied data, as discussed in 3.2.7.


 * This dates back to IEEE Std 802.3x-1997. Guy Harris (talk) 23:53, 13 November 2013 (UTC)

Speed evolution
This article needs a chart showing the evolution of speed along the years. — Preceding unsigned comment added by 186.59.87.44 (talk) 19:49, 25 May 2015 (UTC)


 * I can offer a secondary source for this information in a paper I wrote. The paper is also available at . Obviously, I shouldn't be the one to add this to the article. ~Kvng (talk) 16:35, 30 May 2015 (UTC)

Opening illustration


Hello, ! As already noted in, showing an Ethernet cable and saying that's the physical layer makes very little to no sense. If we'd take that route, then the RJ45 plugs and sockets would also be the physical layer, which would also make no sense. Hope you agree. &mdash; Dsimic (talk &#124; contribs) 21:41, 22 December 2015 (UTC)


 * Hello ! I am not sure what the confusion is. The ethernet physical layer is the wire. Why is this so hard to understand? RJ45 is a modular connector. Please explain to me how this is confusing you so I can fix this. Thank you, --Wyn.junior (talk) 00:28, 23 December 2015 (UTC)


 * I would like input from additional people if they think this picture is as good as I think it is.--Wyn.junior (talk) 00:40, 23 December 2015 (UTC)


 * The physical layer is far more than the cable - most of the interesting stuff is how the signal is put on the medium. Some physical layers don't go over twisted pair, and twisted pair can carry a number of different physical layers and can also carry stuff that has nothing to do with Ethernet. Guy Harris (talk) 00:53, 23 December 2015 (UTC)


 * Please see page 9 in this document, for example. In a few words, the physical layer of Ethernet deals with the low-level electronic way in which the signals are transmitted over the cabling, and the cabling doesn't equal the physical layer. &mdash; Dsimic (talk &#124; contribs) 00:54, 23 December 2015 (UTC)


 * Are you saying the physical layer is composed of things besides the Ethernet cable? Like the physical layer could be part of what is inside of the Ethernet computer port?--Wyn.junior (talk) 22:11, 29 December 2015 (UTC)


 * The physical layer is primarily about how the signals are transmitted over the cabling. &mdash; Dsimic (talk &#124; contribs) 22:25, 29 December 2015 (UTC)


 * So the biggest part of the physical layer is inside the chip or chips attached to the Ethernet computer port (the PHY chip), not inside the cable or the receptacle into which it plugs. I.e., it's not visible in the image below. Guy Harris (talk) 23:03, 29 December 2015 (UTC)



How about this? Minus the physical layer.--Wyn.junior (talk) 01:20, 23 December 2015 (UTC)


 * Well, the pictured connector isn't modular either. It's the permanently overmolded type. &mdash; Dsimic (talk &#124; contribs) 03:09, 23 December 2015 (UTC)


 * I think the pictured connector is a modular connector. If it isn't then please reference what it is.Wyn.junior (talk)


 * You're right, it's an overmolded modular connector, which makes it permanently attached. However, I find the picture currently in the article to be better, because it shows a real-life environment. &mdash; Dsimic (talk &#124; contribs) 03:50, 23 December 2015 (UTC)


 * As a note, it isn't that the non-overmolded modular connector isn't permanently attached. &mdash; Dsimic (talk &#124; contribs) 03:52, 23 December 2015 (UTC)


 * This type of connector is named "modular", no matter how it's attached to the cable. Actually, nearly all cables are crimped, but receptacles are usually soldered. --Zac67 (talk) 09:15, 23 December 2015 (UTC)


 * ... except for the receptacles on patch panels, which are also crimped. :) &mdash; Dsimic (talk &#124; contribs) 01:15, 24 December 2015 (UTC)


 * I also think the current illustration is better. One other problem with the one here. There's no visual context for the "port". This could be a patch panel, wallplate or port on a piece of network equipment. ~Kvng (talk) 02:00, 24 December 2015 (UTC)
 * You are correct.--Wyn.junior (talk) 17:20, 25 December 2015 (UTC)

Galvanic isolation -- requirement and design issues
I recently learned that standard copper/RJ45 Ethernet is electrically isolated using a network isolator&mdash;a small isolation transformer right next to every RJ45 jack (sometimes built into the jack itself). However, I've seen very little mention of this on the Wikipedia Ethernet pages.

First, this page seems to be in error, claiming that an advantage of fiber is isolation: "Fiber optic variants ... offer ... electrical isolation". They offer more isolation, sure, but isolation is built into the standard.

Second, this is an interesting design decision that solves problems that non-EEs wouldn't think of, like ground loops.

Third, the implication of this design decision is that (if I remember my E&M) the signaling protocol had to be designed so that transitions (not constant voltage or current) conveys information. As such, this would mean the the electrical design directly affects the data encoding, which is interesting.

On that note, I'm having trouble finding information on how data is encoded as it goes over Ethernet; I'm sure it's somewhere but it could do with better linking to this article. Thanks. —Ben FrantzDale (talk) 12:13, 3 April 2013 (UTC)


 * Aah: Ethernet links to Ethernet physical layer which links to Manchester code, which opens:
 * "In telecommunication and data storage, Manchester coding (also known as Phase Encoding, or PE) is a line code in which the encoding of each data bit has at least one transition and occupies the same time. It therefore has no DC component, and is self-clocking, which means that it may be inductively or capacitively coupled, and that a clock signal can be recovered from the encoded data."
 * Surprisingly, despite mentioning coupling, that page doesn't explicitly mention the isolation benefits. I just fixed that. —Ben FrantzDale (talk) 12:27, 3 April 2013 (UTC)


 * Different flavors of Ethernet use different signaling. 10BASE-T uses Manchester coding, Fast Ethernet uses MLT-3 and so on. You should find all that linked from the Ethernet physical layer page. -—Kvng 20:24, 3 April 2013 (UTC)


 * Seems to me that the use of a network isolator, although acceptable, isn't a required practice. So, if you're not advised to use it, probably you'll not use. In the other hand, optical fibers just don't conduct electricity, so they are totally immune to any electrical hazard, with no little additional boxes required. Zekkerj (talk) 23:27, 13 November 2013 (UTC)


 * Zekkerj, Kvng, etc. -- I find the network isolator article confusing. My understanding is that *every* device where I can plug in an Ethernet cable has a few transformers inside that device, often called "the magnetics". Often the magnetics look a lot like an integrated circuit, soldered to the same PCB that the socket is soldered to. Other times, the magnetics are built into the Ethernet socket.
 * Those transformers are already built in, "no little additional boxes required", and they *are* required practice for every Ethernet standards-compliant device.
 * I think these are the transformers Ben FrantzDale was talking about.
 * Those built-in magnetics (the built-in transformers) seem to fit the definition of "network isolator" currently given in the lede of the network isolator article.
 * Are those built-in magnetics one kind of network isolator, and the "additional boxes" another kind of network isolator?
 * Or should the definition in the network isolator article be changed to clarify that the built-in transformers are technically not a kind of "network isolator"?
 * Either way, I think the Ethernet over twisted pair -- and perhaps the Ethernet physical layer article and this Ethernet article -- should say a few more words about those built-in transformers -- and perhaps also the "additional boxes" described in the network isolator article. --DavidCary (talk) 20:43, 21 May 2014 (UTC)

The Ethernet over twisted pair variants are required to provide up to 5000 volts of isolation. This is for safety and to prevent ground loops. 5000 volts is not enough to protect against lightening strikes so there are supplementary surge protectors you can use to (try to) address that. I agree that Network isolator is confusing. The information in that article is not verifyable so it shouldn't necessarily be trusted. ~KvnG 14:07, 22 May 2014 (UTC)


 * I thought the usual UTP ethernet transformers are required to hold off only 500 volts. The transformers are so small that 5000 volts can probably go around them. It is recommended that one not use them for outside links where lightning is possible, such as from one house to another. A 500 volt isolation will protect against a connection to the power line, though. Not only do the transformers provide isolation, but ensure that common mode signals are ignored. UTP ethernet is believed to work with a 240 VAC common mode voltage on the cable. (I believe someone even tested this.) Gah4 (talk) 17:44, 30 May 2015 (UTC)


 * It looks like I misremembered the isolation requirement. The requirement is 1500 v RMS or a bit under 2500 v peak. See IEEE 802.3 section 12.10.1, for instance. Do you know where your 500 v thought comes from? ~Kvng (talk) 21:02, 30 May 2015 (UTC)


 * Each PHY has it's own isolaton clause but they're all identical as far as I can see (even for fiber ;-). IEEE 802.3-2012 14.3.1.1 says:
 * 1500 V rms at 50 Hz to 60 Hz for 60 s, applied as specified in subclause 5.2.2 of IEC 60950-1:2001.
 * 2250 V dc for 60 s, applied as specified in subclause 5.2.2 of IEC 60950-1:2001.
 * A sequence of ten 2400 V impulses of alternating polarity, applied at intervals of not less than 1 s. The shape of the impulses shall be 1.2/50 μs (1.2 μs virtual front time, 50 μs virtual time of half value), as defined in IEC 60950-1:2001 Annex N.
 * --Zac67 (talk) 09:55, 31 May 2015 (UTC)


 * I think I once looked up the data sheet for the ones in a NIC that I had. But it was long enough ago that I am not sure now which one it was.  It might have been for a 10base2 port.  Also, it was long enough ago that 802.3 wasn't freely available, so I couldn't look there. Gah4 (talk) 17:13, 1 June 2015 (UTC)

Also, there is a transformer on each end of a TP link, so device to device it is twice the transformer value. Coaxial ethernet has isolation inside the transceiver. The signals might go through a transformer on the host end, but the power supply doesn't. Some parts of the transceiver run on one side, some on the other. On a 10base2 card, you can see the isolation on the PC board, as the ground plane doesn't go through. There is a DC to DC converter to power the isolated side. You are supposed to ground the shield of the coax at exactly one point. It mostly doesn't matter if it stays inside the building, but if you go outside, you should be sure to ground one point. Gah4 (talk) 07:33, 3 July 2016 (UTC)