pat@hprnd.HP.COM (Pat Thaler) (04/07/90)
> From: duncan@siesoft.co.uk (Duncan McEwan) > > Hello, > > (Description of a Thicknet segment with cascaded fanout boxes.) > Could someone tell me whether this is legal and/or how well it should > work (the cascading is always only one level deep -- ie we always cascade > from a directly connected box) If by legal you mean does some standard (802.3 perhapse) allow it, there is no standard for fanout boxes. There specifications are determined by the manufacturer and vary from one to the next so it is not possible to say "x" works with fanout boxs. They are "extensions" to the standard that exist by striving to look as transparent (or as much like a bit of AUI cable) as possible. If you designed a box which degraded a signal passing through it no more than say 10 m of AUI cable, you could presumably use it in combination with 40 m of AUI cable and present a signal to that looked no worse than one which had traveled through 50 m of AUI cable. So as a manufacturer, you would feel comfortable telling customers it would work with AUI compatible 802.3 DTEs and MAUs (transceivers) when used in place of 10 m of AUI cable. What situations fanout boxes work in depends generally on how well they accomplish the goal of looking like a piece of cable. Since they generally buffer the signals, they do preserve the signal level. The hard part is preserving signal timing. Traveling through the AUI cable is allowed to add up to 1.5 ns of jitter to a signal, so to look like a bit of AUI cable, a fanout box could only add a small fraction of 1.5 ns. That is pretty hard to achieve. Failing to acheive it means you are relying on the jitter contributions of other components on the segment to not add up to their worst case value. (When I tested a couple of fanout boxes, I saw between 1 and 3 ns jitter. You could lower jitter by retiming in the fanout box making it more like an 802.3 repeater, but that carries other implications and I don't know of any fanout box that does that.) The only basis you have to judge whether it works or not is your (and other people's) experience with the product, whether the vendor says it works, and whether you believe the vendor's claims. Whether it works can also be affected by various variables in your set-up, primarily involving how much you are pushing the limits on other specs. For instance: AUI cable lengths, coax cable length, jitter performance of your MAUs, how well the decoders in your DTEs work in the presence of jitter. > > Now, if the above *is* ok, could someone comment on the following. > > We are about to install an ethernet in a new building -- this time we > are going to use thin-ethernet with a 4 port thin-ethernet repeater > which also happens to have a single female AUI connector for connecting > to a tranceiver (to allow connections to other ethernets, ie thick, or > I guess even fibre or 10baseT?). What I would like to know is, is it > legal to connect a fanout directly to the AUI port -- ie is it the > same as cascading the fan-out boxes I described above? As above, legal has no meaning for fanout boxes if by legal you mean does a standard allow it. The only meaning it could have is do the vendor's involved (including the repeater vendor and the fanout vendor) support that usage or does it work. > > One person I talked to muttered something about the repeater not doing > SQE and the fanout expecting it, so it wouldn't work, but they didn't > sound very convincing! He was also the same person that told me that > cascading fanouts is perfectly OK! I could understand having problems > I tried to connect to the fanout's female connector (ie the place where > the drop cable from the ethernet to fanout would normally go) using > some wierd female-female AUI cable, but if I used a regular male-female > cable and connected to one of the fanout's male ports that should be > the same as cascading fanouts shouldn't it? A MAU connected to a repeater must not send SQE test signals to the repeater. Fanout boxes generally send an SQE test signal. (They generate their own SQE test signal to send to the DTE if they when there is no MAU attached.) If you can not turn off the SQE test signal, the fanout will not be suitable for use with a repeater. Assuming the fanout supports repeaters, it would be more like using a single fanout than like cascading since the repeater retimes the signal and outputs a signal which meets the same specs as a DTE signal. I'm not sure how you want to use the fanout. If you have some DTEs near the repeater which you want to connect to the fanout along with the repeater that may be a problem since you may find your choice is to either have SQE test off for all ports or on for all ports of the fanout. > The obvious solution if the above it not legal is to have a minimal > ethernet with two transceivers, one connected to the fanout and the other > to the repeater's AUI port. But that leads to more questions. Such > as, what is the shortest length ethernet I could use? Also, I recall > someone once telling me that it was not legal to have a transceiver > plugged directly on to the end of an ethernet segment -- rather that > there was some minium distance between the last transceiver and the > terminator. Can anyone comment on that? There is no spec on minimum distance between a station and a terminator. There is a spec of a mimimum distance of 0.5 m between MAUs on thin coax (2.5 m on thick). So the shortest thin coax segment you could have which obeys the 10BASE2 specs for MAU spacing is 0.5 m (assuming two MAUs; I have run self-tests and such with a two terminators on a T connected to a MAU). > > Thanks in advance for any help. I would prefer mail replies -- I will > summarize -- partly to avoid clogging this group, but also because our > newsfeed is currently running about 5 days behind, and I need this info > quickly... > > Duncan > Internet: duncan@siesoft.co.uk Bang: ...!uunet!mcsun!ukc!siesoft!duncan Any opinions stated above are mine and not necessarily the views of IEEE or the 802.3 Working Group. Pat Thaler pat@hprnd.hp.com