markw@remote.halcyon.wa.com (Mark Ward) (08/02/90)
Hm. So far, I've come up with a nifty transceiver chip from national. Using that chip and a couple opto isolators, I should be in business. 'Course the local high priced chip place wants $30 for the chip, but I'll find it cheaper somewhere else. Why would it be especially hard to tie into thick ethernet? I've got just about any coax fitting I need, and I get a few a week for a really good price (free), but they are all used. I also have enough rg8, rg213, and rg214 to run around the inside of my house a few dozen times... -mark
rpw3@rigden.wpd.sgi.com (Rob Warnock) (08/04/90)
In article <Hgo9m1w162w@remote.halcyon.wa.com> markw@remote.halcyon.wa.com (Mark Ward) writes: +--------------- | Hm. So far, I've come up with a nifty transceiver chip from national. | Using that chip and a couple opto isolators, I should be in business. +--------------- *Don't* try to use opto-isolators. They won't be able to meet the *tight* Ethernet specification on duty-cycle. Use transformers, instead. You can get them fairly cheap from "Mini-Circuits, Inc.", I believe. The duty-cycle has to be 50% +/- 0.5 nanosecond, in order to avoid: - Problems with other station's PLL clock recovery circuits. - Problems with other transceiver's input circuits. - Driving a false D.C. level on the cable, and thus either causing collision-detect when there's no collision, or in the opposite case, causing some transceiver to *miss* detecting a collision when there *is* one. +--------------- | 'Course the local high priced chip place wants $30 for the chip, but | I'll find it cheaper somewhere else. Why would it be especially hard | to tie into thick ethernet? I've got just about any coax fitting I need... +--------------- There are tight specs on the impedance of connectors used on the thick cable -- type "N" UHF connectors are specified -- and there are tight limits on the amount of wire you may "stub" off the 50-ohm controlled-impedance cable -- no more than 3 centimeters, and no more than 2 picofarads of stray capacitance. If you violate these specs, your system *may* work, but you just as easily may get lots of errors and collisions from the reflected signals at your tap. +--------------- | ...and I get a few a week for a really good price (free), but they | are all used. I also have enough rg8, rg213, and rg214 to run around | the inside of my house a few dozen times... +--------------- As long as it really is 50-ohm. [RG-8/U for thick, RG-58/CU for thin. Don't use RG-58/AU, I think it's 53 ohms.] Note that since "collision detect" is generated by measuring *voltage*, and transceivers drive constant-*current* into the cable, 75-ohm cable may very well cause a collision (with yourself) every time (depending on the C.D. threshold); 93-ohm cable *will* collide solidly! [This assumes the cable is longer than a bit-time, about 20 meters. If the total length of the net is one or two meters, you can get away with almost anything!] -Rob ----- Rob Warnock, MS-9U/510 rpw3@sgi.com rpw3@pei.com Silicon Graphics, Inc. (415)335-1673 Protocol Engines, Inc. 2011 N. Shoreline Blvd. Mountain View, CA 94039-7311