Rick Merrill wrote:> Rene wrote: > >> "Rick Merrill" <RickMerrill@comTHROW.net> schrieb im Newsbeitrag >> news:JV8ic.10862$aQ6.875998@attbi_s51... >> >>> Tauno Voipio wrote: >>> ...> Admitted, the signal is RF, but the wavelength (even at the >>> >>>> third harmonic, 188 MHz) is well over a meter (little over a >>>> yard, for the foot-pound people), so a couple of millimeters is >>>> nothing. >>> >>> >>> That reminds me of a common error when people wire up their own >>> ethernet routers: they use as short a cable as reasonable, but they >>> should never use less than a couple of meters single length, right? >> >> >> >> As you may have seen in my thread in this newsgroup I am facing >> a problem with short cables and a specific PHY. So far I've never >> heard of a "minimum cable length". Can you supply some details. > > > In brief, an ethernet cable is an antenna: If it is too short the > reflected signals will create interference. The "minimum cable > length between workstations" (and between hubs) is "8 feet". > (That is for 10baseT) For 10Base2 the minimum is "20 inches". > > - http://support.microsoft.com/default.aspx?scid=kb;en-us;Q97550 > > It is also vital to note that while many configuations will work, there > are not "supposed" to be more than 4 hubs between ANY TWO > NIC. This topology restriction derives from the time delays inherent in > a hub. (Switches have different rules.) - RM >A properly terminated cable does not have reflections. Old ham radio knowledge is: If your system is fussy on the length of feed line, check for SWR (Standing wave ratio, a measure of proper termination. Please check the pairing on your cables. If your transmit signal is paired with the receive return line, it very often works on 10 Mbits/s, but gives all kinds of weird cable-length problems on 100 Mbits/s. Tauno Voipio (OH2UG, for well over 40 years) tauno voipio @ iki fi
Simple tests of an Ethernet port?
Started by ●April 21, 2004
Reply by ●April 23, 20042004-04-23
Reply by ●April 23, 20042004-04-23
Tauno Voipio wrote:> Rick Merrill wrote: > >> Rene wrote: >> >>> "Rick Merrill" <RickMerrill@comTHROW.net> schrieb im Newsbeitrag >>> news:JV8ic.10862$aQ6.875998@attbi_s51... >>> >>>> Tauno Voipio wrote: >>>> ...> Admitted, the signal is RF, but the wavelength (even at the >>>> >>>>> third harmonic, 188 MHz) is well over a meter (little over a >>>>> yard, for the foot-pound people), so a couple of millimeters is >>>>> nothing. >>>> >>>> >>>> >>>> That reminds me of a common error when people wire up their own >>>> ethernet routers: they use as short a cable as reasonable, but they >>>> should never use less than a couple of meters single length, right? >>> >>> >>> >>> >>> As you may have seen in my thread in this newsgroup I am facing >>> a problem with short cables and a specific PHY. So far I've never >>> heard of a "minimum cable length". Can you supply some details. >> >> >> >> In brief, an ethernet cable is an antenna: If it is too short the >> reflected signals will create interference. The "minimum cable >> length between workstations" (and between hubs) is "8 feet". >> (That is for 10baseT) For 10Base2 the minimum is "20 inches". >> >> - http://support.microsoft.com/default.aspx?scid=kb;en-us;Q97550 >> >> It is also vital to note that while many configuations will work, >> there are not "supposed" to be more than 4 hubs between ANY TWO >> NIC. This topology restriction derives from the time delays inherent >> in a hub. (Switches have different rules.) - RM >> > > A properly terminated cable does not have reflections. > > Old ham radio knowledge is: If your system is fussy on the length > of feed line, check for SWR (Standing wave ratio, a measure of > proper termination. > > Please check the pairing on your cables. If your transmit > signal is paired with the receive return line, it very often > works on 10 Mbits/s, but gives all kinds of weird cable-length > problems on 100 Mbits/s.Hi Tauno! All true, but many old NICs do not perfectly terminate nor match the NIC at the other end. Let's hope Rene solves her problem with our tips!-)
Reply by ●April 23, 20042004-04-23
"Rick Merrill" <RickMerrill@comTHROW.net> schrieb im Newsbeitrag news:2Vdic.9422$cF6.415245@attbi_s04...> Rene wrote: > > "Rick Merrill" <RickMerrill@comTHROW.net> schrieb im Newsbeitrag > > news:JV8ic.10862$aQ6.875998@attbi_s51... > > > >>Tauno Voipio wrote: > >>...> Admitted, the signal is RF, but the wavelength (even at the > >> > >>>third harmonic, 188 MHz) is well over a meter (little over a > >>>yard, for the foot-pound people), so a couple of millimeters is > >>>nothing. > >> > >>That reminds me of a common error when people wire up their own > >>ethernet routers: they use as short a cable as reasonable, but they > >>should never use less than a couple of meters single length, right? > > > > > > As you may have seen in my thread in this newsgroup I am facing > > a problem with short cables and a specific PHY. So far I've never > > heard of a "minimum cable length". Can you supply some details. > > In brief, an ethernet cable is an antenna: If it is too short the > reflected signals will create interference. The "minimum cable > length between workstations" (and between hubs) is "8 feet". > (That is for 10baseT) For 10Base2 the minimum is "20 inches". > > - http://support.microsoft.com/default.aspx?scid=kb;en-us;Q97550 > > It is also vital to note that while many configuations will work, > there are not "supposed" to be more than 4 hubs between ANY TWO > NIC. This topology restriction derives from the time delays > inherent in a hub. (Switches have different rules.) - RMRick, meanwhile I found some notes stating that 802.3 requires a minimum length of approx. 2.5meters. It seems to be as you mentioned. Most of the times it works, even if the standard tells otherwise. Are you sure about the reflection. If the impedance of the network cable is correct and matches the "endpoint" impedance there should not be relevant reflections. Is my assumption valid in real life scenarious ? - Rene
Reply by ●April 23, 20042004-04-23
"Rick Merrill" <RickMerrill@comTHROW.net> schrieb im Newsbeitrag news:zeeic.9570$cF6.417251@attbi_s04...> Tauno Voipio wrote: > > > Rick Merrill wrote: > > > >> Rene wrote: > >> > >>> "Rick Merrill" <RickMerrill@comTHROW.net> schrieb im Newsbeitrag > >>> news:JV8ic.10862$aQ6.875998@attbi_s51... > >>> > >>>> Tauno Voipio wrote: > >>>> ...> Admitted, the signal is RF, but the wavelength (even at the > >>>> > >>>>> third harmonic, 188 MHz) is well over a meter (little over a > >>>>> yard, for the foot-pound people), so a couple of millimeters is > >>>>> nothing. > >>>> > >>>> > >>>> > >>>> That reminds me of a common error when people wire up their own > >>>> ethernet routers: they use as short a cable as reasonable, but they > >>>> should never use less than a couple of meters single length, right? > >>> > >>> > >>> > >>> > >>> As you may have seen in my thread in this newsgroup I am facing > >>> a problem with short cables and a specific PHY. So far I've never > >>> heard of a "minimum cable length". Can you supply some details. > >> > >> > >> > >> In brief, an ethernet cable is an antenna: If it is too short the > >> reflected signals will create interference. The "minimum cable > >> length between workstations" (and between hubs) is "8 feet". > >> (That is for 10baseT) For 10Base2 the minimum is "20 inches". > >> > >> - http://support.microsoft.com/default.aspx?scid=kb;en-us;Q97550 > >> > >> It is also vital to note that while many configuations will work, > >> there are not "supposed" to be more than 4 hubs between ANY TWO > >> NIC. This topology restriction derives from the time delays inherent > >> in a hub. (Switches have different rules.) - RM > >> > > > > A properly terminated cable does not have reflections. > > > > Old ham radio knowledge is: If your system is fussy on the length > > of feed line, check for SWR (Standing wave ratio, a measure of > > proper termination. > > > > Please check the pairing on your cables. If your transmit > > signal is paired with the receive return line, it very often > > works on 10 Mbits/s, but gives all kinds of weird cable-length > > problems on 100 Mbits/s. > > Hi Tauno! All true, but many old NICs do not perfectly terminate nor > match the NIC at the other end. Let's hope Rene solves her problem > with our tips!-)In about two weeks I will have more devices to test (with a newer version of the PHY). I will repeat my tests with these and keep you informed of my findings. Yours - Rene PS. Rene is not a female, that would be Renee. Sorry to disappoint you :-)))
Reply by ●April 23, 20042004-04-23
Rene wrote:> "Rick Merrill" <RickMerrill@comTHROW.net> schrieb im Newsbeitrag >... snip ...>> >> Hi Tauno! All true, but many old NICs do not perfectly >> terminate nor match the NIC at the other end. Let's hope Rene >> solves her problem with our tips!-) > > In about two weeks I will have more devices to test (with a newer > version of the PHY). I will repeat my tests with these and keep > you informed of my findings. > > PS. Rene is not a female, that would be Renee. Sorry to disappoint > you :-)))Bad move! You just cut off assistance from half of the testosterone crazed daydreamers here :-) -- Chuck F (cbfalconer@yahoo.com) (cbfalconer@worldnet.att.net) Available for consulting/temporary embedded and systems. <http://cbfalconer.home.att.net> USE worldnet address!
Reply by ●April 24, 20042004-04-24
CBFalconer <cbfalconer@yahoo.com> wrote:> Rene wrote: >> "Rick Merrill" <RickMerrill@comTHROW.net> schrieb im Newsbeitrag >> > ... snip ... >>> >>> Hi Tauno! All true, but many old NICs do not perfectly >>> terminate nor match the NIC at the other end. Let's hope Rene >>> solves her problem with our tips!-) >> >> In about two weeks I will have more devices to test (with a newer >> version of the PHY). I will repeat my tests with these and keep >> you informed of my findings. >> >> PS. Rene is not a female, that would be Renee. Sorry to disappoint >> you :-))) > > Bad move! You just cut off assistance from half of the > testosterone crazed daydreamers here :-) >And thats why I use gender neutral pronouns unless I know who I'm talking about. Although I would suspect most of those who frequent this newsgroups are male. :P -- Wing Wong. Webpage: http://wing.ucc.asn.au
Reply by ●April 24, 20042004-04-24
Rene wrote:> "Rick Merrill" <RickMerrill@comTHROW.net> schrieb im Newsbeitrag > news:2Vdic.9422$cF6.415245@attbi_s04... > >>Rene wrote: >> >>>"Rick Merrill" <RickMerrill@comTHROW.net> schrieb im Newsbeitrag >>>news:JV8ic.10862$aQ6.875998@attbi_s51... >>> >>> >>>>Tauno Voipio wrote: >>>>...> Admitted, the signal is RF, but the wavelength (even at the >>>> >>>> >>>>>third harmonic, 188 MHz) is well over a meter (little over a >>>>>yard, for the foot-pound people), so a couple of millimeters is >>>>>nothing. >>>> >>>>That reminds me of a common error when people wire up their own >>>>ethernet routers: they use as short a cable as reasonable, but they >>>>should never use less than a couple of meters single length, right? >>> >>> >>>As you may have seen in my thread in this newsgroup I am facing >>>a problem with short cables and a specific PHY. So far I've never >>>heard of a "minimum cable length". Can you supply some details. >> >>In brief, an ethernet cable is an antenna: If it is too short the >>reflected signals will create interference. The "minimum cable >>length between workstations" (and between hubs) is "8 feet". >>(That is for 10baseT) For 10Base2 the minimum is "20 inches". >> >>- http://support.microsoft.com/default.aspx?scid=kb;en-us;Q97550 >> >>It is also vital to note that while many configuations will work, >>there are not "supposed" to be more than 4 hubs between ANY TWO >>NIC. This topology restriction derives from the time delays >>inherent in a hub. (Switches have different rules.) - RM > > > Rick, > > meanwhile I found some notes stating that 802.3 requires a minimum > length of approx. 2.5meters. It seems to be as you mentioned. Most of > the times it works, even if the standard tells otherwise. > > Are you sure about the reflection. If the impedance of the network > cable is correct and matches the "endpoint" impedance there should > not be relevant reflections. Is my assumption valid in real life > scenarious ? > > - ReneI have found discussing the theoretical vs the practical is not always conclusive via Usenet :-0 Now that you have identified 802.3 as your standard, and have reassured us that all your NIC are up to date in impedance and protocol, make sure your network adheres to all the standards, including total length from any point to any point, and total hubs from any point to any point are not exceeded. - RM p.s. My spell checker does not recognize "Rene" :-(
Reply by ●April 26, 20042004-04-26
Hi Tauno,> The impedance shown on the twisted pair should be 100 or 150 ohms, > depending on the cable type: STP or UTP. In practice, small > reflections do not harm, so as a compromise, you could start > with 120 ohms.Thanks for this info, it is exactly what I wanted. Unfortunately, it seems to have deepened the mystery for me quite a bit. I connected the suspect device to a hub via standard UTP patch cable, and created some activity. I see ~0.3V of signal. The open-circuit impedance across the RJ45 pins seems to be about 40 ohms. WTF? I think - without much concrete evidence - that the person who designed this circuit originally cut-n-pasted an app note for the MAC, then later had some trouble finding the right PHY, and substituted a different part that happens to have the same PCB footprint. So I really don't know what I can trust in this design.
Reply by ●April 26, 20042004-04-26
Lewin A.R.W. Edwards wrote:> Hi Tauno, > > >>The impedance shown on the twisted pair should be 100 or 150 ohms, >>depending on the cable type: STP or UTP. In practice, small >>reflections do not harm, so as a compromise, you could start >>with 120 ohms. > > > Thanks for this info, it is exactly what I wanted. Unfortunately, it > seems to have deepened the mystery for me quite a bit. I connected the > suspect device to a hub via standard UTP patch cable, and created some > activity. I see ~0.3V of signal. The open-circuit impedance across the > RJ45 pins seems to be about 40 ohms. WTF? >The RF impedance? The signal is on the tiny side, which points to an unsuitable transformation ratio, wrong filter components (10BASE-T magnetics on 100BASE-TX) or improper termination resisitors. Actually the slow filter should not distort the link pulses (which you probably are seeing).> I think - without much concrete evidence - that the person who > designed this circuit originally cut-n-pasted an app note for the MAC, > then later had some trouble finding the right PHY, and substituted a > different part that happens to have the same PCB footprint. So I > really don't know what I can trust in this design.Can you get the data sheets of the PHY and magnetics? Any termination / matching resistors between PHY and magnetics? Tauno Voipio tauno voipio @ iki fi
Reply by ●April 26, 20042004-04-26
Hi Tauno,> The RF impedance?Well, no, DC :)> Can you get the data sheets of the PHY and magnetics?Actually I am on crack, it's been a stressful week. The PHY is integrated in the MAC. I meant to say the magnetics were possibly chosen randomly. The part on the actual board is an XFMRS XFATM2Y-CLYG1-4MS, 1:1. Of course, the datasheet for this cannot be downloaded from XFMRS' web site, it 404s. The webmaster of this site deserves whipping. The CLYG1-4MS suffix probably means some special characteristics, but without the datasheet I can only guess.> Any termination / matching resistors between PHY and magnetics?Yes, but what is in the schematic does not match the board. Argh. I think I should abandon this project.
