jabusch@uiucdcsb.UUCP (01/29/87)
The problem with most commercial fiber systems is that they are equivalent to baseband coaxial systems. That is, they transmit at a given light frequency (actually a range) and everyone on the fiber uses that range. In order to get multiple stations to transmit near-simultaneously, each station is in effect running TDM (Time Division Multiplexing). In other words, CSMA/CD, in effect. More normal systems that provide video, such as CATV, run broadband networks, each station using a 6MHz range for the transmission of video. This allows continuous broadcast of video in each of many frequency ranges, with receivers tuned to catch the channel desired. In order to support video on fiber, now, you would need buffering at the receiver to store the incoming data, which would then have to be transmitted at higher frequency rates than typically possible from most video equipment. The buffering is necessary because the receiver would continue to re-create the analog signal from its digital source (fiber) at the 6MHz rate to provide decent video. Since the channel is shared between multiple transmitters (TDM), the basic data transfer rate would have to be fast enough to guarantee that no matter how many transmitters are on a given fiber, enough packets would arrive at the receiver in a given time frame (forgive me for not doing the appropriate math here) to keep the receiver buffer from emptying. There are a lot of factors to consider here. One possible aid is to buffer at the transmit end, so that packets would always be ready to send when the channel is open for transmission. This implies that the receiver is always behind by some "lag" time which is proportionally (or perhaps exponentially) related to the number of stations on the network. Add to this problem the fact that there just aren't too many A-D (Analog to Digital) Converters out there than can keep up with the 6MHz bandwidth needed for the video broadcast. There are people working at several fiber vendors out there now, trying to solve these problems. One proven method for turning a fiber into a broadband system is to transmit at different freqencies of light. This is expensive to do, but technology evolves and costs decrease. I know of at least one lab that has done this (right here, the Information Engineering Laboratory, University of Illinois, Urbana, Illinois). It was referred to at the time as "Spectrum Sampling". There is a Master's thesis on it, written by Apostolos Dollas, who can now receive e-mail as: dollas%mcnc.csnet@RELAY.CS.NET. There is, of course, much that I haven't even begun to discuss here, and which I will receive many flames for, if all continues with the SOP on this net. I hope that I've helped a little. John Jabusch p.s. don't bother to send me e-mail. I am between hosts, and even though I can post notes from this site, I can't receive mail here. I will have another mailing address by the end of the week.
rvk@tiger.UUCP.UUCP (02/08/87)
/* Written 11:57 pm Jan 22, 1987 by mdr@reed.UUCP in tiger.UUCP:comp.dcom.lans */ /* ---------- "What can you do with an existing fi" ---------- */ How can you set up a fiber optic LAN so that it will later be easy to use the installed fiber for other services such as video, voice traffic or whatever? We have recently installed a fiber optic "Ethernet" throughout the campus using FiberCom equipment (I'm not sure what type of fiber we're using - I can find out). I've heard some thoughts that by going with a fiber based network, we will later be able to fully exploit the potential bandwidth of the glass. Has anyone retrofitted an in-place special purpose fiber network to carry additional traffic, probably using equipment unrelated to the original? Even if the cable is multimode, wouldn't the FiberCom equipment have to specifically forward those "channels" of the cable it was not using if the whole thing were to work with independently operating equipment? I am not a fiber wizard, as you might be able to tell, but would be interested in people's (informed?) thoughts on what might be possible. Mike Rutenberg -- Reed College -- Portland, Oregon -- 503/774-9192 /* End of text from tiger.UUCP:comp.dcom.lans */ I use a lot of fiber at the AT&T Oklahoma City factory for local networking. Its a terrific medium. Unfortunately, vendors are just now starting to discover how to make use of the bandwidth. The limitations of your fiber net are determined by the electronics plugged into the ends of it. Unless you use separate fiber, you are limited to what ever Fibercom can transmit over the fiber. Local network fibers are typically 100 micron or 62.5 micron, multi or graded index, normally used with LED going up to 200 Mega bps. Fibercom probably has a fused STAR somewhere that makes the fiber a light bus which allows it to handle ethernet/IEEE 802.3. I don't know if anyone is trying to transmit full motion video over ethernet, but its an interesting idea. Uncompressed video generates about 90 Mega bps of data so it could really load a network down. Artel Communications has a 4 X 25 Mega bps (100 Mbps) fiber ring with ethernet interface boxes. These cost about $9,000 per 8 port box. Optical Data Systems of Richardson Texas is a small company who will do about anything you want. They do fiber right in my opinion. They build "dumb" boxes and use the fact that bandwidth is cheap on fiber. For example; their 32 port asynch mux goes 20 Mbps on the fiber because it doesn't try to figure out when a byte has been received, or even a bit. It simply pumps through voltage level values 256,000 per second for each data lead. AT&T is just now discovering this gaping bandwidth conduit for its datacom products and I occasionaly get funny phone calls like "What would you do with 50 or 100 Mbps if you had it?" I answer "Anything I want." Only a product manager can ask such dumb question. No one ever asks "How fast should a computer go?" but for local networks, the idea is still novel. Off-the-shelf 200 Mbps LEDs have existed for quite a while so 100 to 200 Mbps local networks are starting to appear. Naturally the smaller companies make them first. I know "broadband networks have 250 to 400 Mbps bandwidth." True, but have you ever talked, seriously, to a broadband vendor...big surprise, after all the hot air disapates, you find that "well, rf modems that go over 1 Mbps are really expensive, so most bb networks actually go around 1/4 Mbps!!! In the last year or two, 10 Mbps "ethermodems" have come out so people can run ethernet over broadband. Just ask for pricing on a 250 Mbps rf modem and watch the salesman start back-pedaling! Yeah, I know, your can use lower speed rf modems on separate frequency bands. All I can say is, as a person who has to live with local networks every day, I do not want to try and troubleshoot "harmonics" problems and other analogue impairments. I may have problems with my electronics, like everyone, but when I go to bed at night, I sleep like a baby because I know that my fiber net will not break, degrade, go out of tune, transmit or receive RFI, and etc. I am putting in a fiber based ethernet. I will put in a fiber based MAP network next year and I already have a fiber based DATAKIT asynch data switch network (over 4000 ports). Some other big names in fiber electronics are Canoga-Perkins, Fibronics , Optelcom and, of course, AT&T. Incidentally, AT&T is introducing a metro area video conferencing service called VIVID. It requires that your area be fibered already since it only compresses the video signal down to 45 Mbps. Send me unix mail if you want more fiber, fiber product or VIVID info. Ray Kellogg ihnp4!ocsmb!rvk