Or maybe the question should be why do we have cables that are not crossover.
My first encounter with this concept came when connecting serial ports together. Turned out there were two kinds - data set, and data terminal.
Not content with that, when UTP cables came out, we had a similar situation.
Maybe I've missed something, but it's always seemed to me that a logical approach would be to define some pins as input and some as output, and for cables to connect input pins to output pins, thus obviating the need for two different ways of wiring up connecting cables.
Did I miss something? Is there a reason this situation persists?
Consider the case when you want to connect two of the same type together. For example, a normal PC is considered a data terminal. If you want to connect two together, the transmit pins on one must connect to the receive pins on the other. Hence the crossover cable, often called a "null modem" for the serial cable case.
The same reasoning applies to CAT5 cables, but in this case there are a total of four pairs, two transmit and two receive. Both sets are not always used, but the ones in use must be "crossed over" to connect two of the same type equipment. Many routers can sense the type of cable required and will automatically do the crossover internally if required.
If the transmit and receive pins on a data set (typically a modem) were transposed in the original design, then the same cables could have been used in all cases.
I agree, it does make a great deal of sense to set up the interfaces in a symmetrical fashion whenever possible. I've grown fond of the Yost method of doing serial-port hookup... there's no differentiation between DTE (e.g. terminal or PC) and DCE (e.g. modem), and you can hook either to the other.
10BaseT could also have been designed with this sort of symmetry in mind, I suppose, if we'd started out using the sort of highly-intelligent auto-adaptive interfaces that we use today on e.g.
100BaseT.
However... these sorts of symmetrical cabling systems don't tend to arise when there's a big functional asymmetry between the sorts of devices at either end. In the case of serial ports, there are some functions provided by the DCE (e.g. carrier detect, ring indicate, and master-clock timing in the case of synchronous ports) that the DTE is physically or philosophically unable to provide... the DTE is a consumer of these, not a provider. The Yost serial standard deals with this issue by simply avoiding it - it doesn't carry those signals at all and thus they aren't available to the DTE.
In the case of Ethernet, a similar asymmetry existed when the technology was developed... hubs had functions that clients did not, and tended to incorporate more-expensive electronics. It's only more recently that the "interface intelligence" in a client (e.g. PC) and a switch have become more comparable and symmetric.
Also (minor issue) it's a trifle easier to extend a straight-through cable with another... you just use a second straight-through cable and a one-to-one butt-heads splicer. You can't do this with two crossover cables, as you'll end up crossing everything over twice and creating the equivalent of a straight-through cable... your splicer needs to include a *third* crossover!
--
Dave Platt AE6EO
Friends of Jade Warrior home page: http://www.radagast.org/jade-warrior
I do _not_ wish to receive unsolicited commercial email, and I will
boycott any company which has the gall to send me such ads!
Same reason. MAKING an RJ-11 cable is easier if the crimped on connector is wired the same way ALL the time. Especially during field service scenarios. The error rate in cables is much higher if specific cross-overs have to be made from end to end. Also, one would need to observe the one completed end to reference what would be needed on the other end. Doing them all the same practically guarantees success. Requiring cross-over practically guarantees a much higher prime pass yield in manufacturing circles, and a higher failure rate in field installations as well.
The problem is that when these systems of interface were designed, there was nearly no automated assembly.
Hand assembly means failures, unless error conditions are reduced to a minimum. Wiring both ends identically means that less errors were made in manufacture of said interface devices and systems. Making the switch at the hardware itself was easy, and 100% repeatable.
Prime pass yield was a huge consideration in labor intensive hand operation production procedures, and still is. That is why most interface cables are pin-for-pin. Particulalry those that have the same or very similar connectors on each end.
The reasn this situation persists is that it is written into the international standards, and a huge installed base of hardware out in the field conforms to those standards.
I used to know about this when I worked for ITT-Creed in the U.K back in 1979-1982, in a group that used to send people to the CCIT standards committee meetings.
The concept dates back to the Telex and Teleprinter networks. The ASR
33 Teletype printer was orginally a data set (IIRR - 1982 is the last time I was seriously involved) produced in huge numbers of the AT&T network, and its use as a computer terminal was never more than a minor spin-off.
Western Electric ALSO hand assembled practically everything back then, and they knew about error rates in cable assemblies, and they knew how best to reduce them by making remembering the wiring procedure an easy thing to do.
OK, I can accept the field installation error issue, though the existence of two different standard ways of wiring the plugs seems at least partly to defeat the goal of reducing errors.
I'd have thought that in a mass-production environment it would be as simple as having one person do one end and another person do the other end.
this would presumably require the DTE to have outputs for carrier and ring although thise concepts apply only to DCE.
the standard was developed before the invention of the smartmodem(tm) which could indicate those two conditions as serial data.
Early modems ware loosely speaking analogue filters coupled to serial line drivers, if you were lucky you could pulse dial by toggling the DTR line with the correct cadence...
Just means those pins would be tied to ground. This would have required more than the 9 pins on modern serial ports, but would have been easily done with the original 25 pin standard.
Did something similar with a telephone in my teens, by pulsing the handset rest.
Shame no TV suspense movie ever used it - devious villain leaves a phone with its dial detached so that he can call his imprisoned victims - but one captive knows better and calls the police despite the absence of a dial.
For modern ethernet, you don't need cross-over cables. The devices figure out the pair sorting. All the NICs, routers, and switches I've come across in the past 3 to 5 years have auto-sorting of the pairs.
On Wed, 11 Mar 2009 16:24:45 +1100, Sylvia Else put finger to keyboard and composed:
An even more logical approach is used by USB OTG ("On The Go"). A device can be either a host (power provider) or a peripheral (power consumer) depending on the status of a fifth pin. After power-up, both devices can negotiate to swap functions. The USB data interface is bidirectional.
See
formatting link
- Franc Zabkar
--
Please remove one \'i\' from my address when replying by email.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.