>From daisy!uunet!ucsd.edu!tcp-group-request Tue Oct 25 10:50:26 1988 remote from osu-cis Received: by n8emr.UUCP (smail2.5) id AA12291; 25 Oct 88 10:50:26 EDT (Tue) Received: by osu-cis.cis.ohio-state.edu (smail2.5) id AA01107; 25 Oct 88 09:11:28 EDT (Tue) Received: by daisyvax (5.51/4.1.88.1) id AA15528; Tue, 25 Oct 88 00:47:56 PDT Received: from ucsd.edu by uunet.UU.NET (5.59/1.14) id AA26950; Tue, 25 Oct 88 01:09:40 EDT Received: by ucsd.edu (5.58/UCSD-1.0); Mon, 24 Oct 88 21:19:42 PDT id AA05785 for Received: from hp-sde.sde.hp.com by ucsd.edu (5.58/UCSD-1.0); Mon, 24 Oct 88 21:11:39 PDT id AA05545 for /usr/lib/sendmail -odq -oi -ftcp-group-request tcp-group-list Received: from hpnmd.HP.COM by hp-sde ; Mon, 24 Oct 88 12:48:45 pdt Received: from hpnmdlc1.HP.COM (hpsrgee) by hpnmd.HP.COM; Mon, 24 Oct 88 10:14:30 pdt Received: by hpnmdlc1.HP.COM; Mon, 24 Oct 88 10:12:03 pdt From: daisy!uunet!sde.hp.com!glenne%hpnmdlc1 Message-Id: <8810241712.AA22649@hpnmdlc1.HP.COM> Subject: whither layer 1? To: tcp-group@ucsd.edu Date: Mon, 24 Oct 88 10:11:54 PDT Cc: dougb%hpnmd@sde.hp.com X-Mailer: Elm [version 2.01 (alpha-release)] Band Choice Considerations for Improved Amateur Radio Digital Networking (Very Preliminary) Glenn Elmore N6GN 10-24-88 Propagation In general the vhf amateur bands are probably more alike in terms of their signal propagation than they are different. The vhf region is most suitable for broadcasting and receiving data from diverse (omnidirectional) sources. This is mainly due to the radiation patterns of physically practical antennas. Through perhaps 1200 MHz, omnidirectional antennas with reasonable physical capture area are available. In the higher microwave regions antennas with thin "pancake" patterns are more difficult/expensive to build and the resulting narrow vertical beam is at risk of being deflected by troposhperic circumstances with a consequential reduction in system margins. If system uptime of the same order as commercial services is required, say 99.9% or better, all bands will generally require line-of-sight (LOS) paths and in severe situations, such as those experiencing marine air masses, extra system margin will be necessary. As an example, obstructive fading produced by nightime movement of wet air may require shorter paths, even on vhf, to guarantee service. Generally, locations with well mixed air, like the Rockies are less likely to experience propagation abnormalities. Coastal locations and large flat areas with water sources are more likely to require special attention. Interference fading on a link will generally require increased C/N margin to guarantee performance. The Rayleigh distribution model probably serves as a useful guide to these margins. For typical fading at higher microwave 99% uptime requires about 18 dB excess C/N over LOS values, 99.5 about 22 dB and 99.9 about 28 dB excess. On many shorter, perhaps 30 mile or less, LOS paths these numbers may be unduly conservative. On such short links it may be possible to run with C/N only 3 to 6 dB above those necessary for the desired BER on LOS freespace paths. It may even be possible in some situations to have extra long paths which rely on tropospheric ducting to provide communication (it's not a bug it's a feature!). Whether such links which are typically only available a few hours or less a day can prove useful for periodic but not continuos service (like electronic mail transfer) is a question which remains to be answered. At 24 GHz and above, water vapor absorbtion is an item to consider, particularly on longer paths. At 10 and below it can probably be ignored when considering digital data links. Path links of sufficient length to make it noticable probably suffer from a much larger variations due to variations in the troposphere. 144 best suited for broadcast 220 | 433 | 900 | 1250 upper end of broadcast 2300 | 3400 | 5700 | 10000 | 24000 point-to-point Water vapor absorption significant on long paths Antennas Antenna selection and performance is crucial to choosing a band and equipment for digital data links. On transmit the antenna serves to couple the transmit power to the "ether" and to control the radiation pattern. On receive it is useful to consider the antenna as a "bucket" with a given aperture which "catches" flux. Using this picture instead of the more common antenna gain/path loss model helps to give insight into what happens as frequency is changed. Coupling of signal energy between the tranmitter output and the receiver input (with a minimum of interference or noise) is the goal. It is useful to consider the tranmit antenna as a mechanism for creating flux at the receiver's location and the receive antenna as a mechanism for recovering it. This aperture view of the antennas is especially useful since practical constraints for amateur link antennas are more likely to be physical size than gain. Antenna size is also a consideration because environmental concerns like wind, birds,snow, ice and vandalism need to be reckoned with. On the vhf bands, a physically reasonable antenna (bigger than a computer but smaller than a hamshack) may have an effective aperture (bucket size) somewhat bigger than its physical size. This is generally the case with lower gain antennas. For high gain antennas physical and electrical sizes are inclined to be similar. Commercial omnidirectional antennas with up to about 10 dB gain are available (not necessarily cheap) through 1200 MHz. Parasitic arrays are available and probably fairly cost effective in this same region. Above 20-25 dB gain, reflector antennas with a single, or a small number of fed elements become attractive and will usually have better aperture efficiency. 24 GHz wins by being far and away the most gain/$. However there is probably an upper limit to useful gains on longer microwave paths. Although upper-microwave antennas can be made to have good performance with narrow beams and low sidelobes relatively easily, tropospheric variations can cause the apparant location of the other end of a link to move and eventually contribute to significant downtime. Until this point is reached, higher frequencies win over lower at 20*log(F) for constant antenna physical aperture; the receive antenna intercepts the same area and the transmit antenna focusses more flux into a given area as the frequency goes up. 144 commercial omnis and small yagis 220 " 433 " 900 " +UHF TV parabolas 1250 commercial omni/loop yagis/reflector antennas 2300 reflector type antennas above 2300 3400 small TVRO antennas cost effective, $200? for 6-8 feet 5700 1-6 foot dishes, approx $1/inch of diameter 10000 " (upper end of dish size may be limited on longer paths due to tropospheric abnormalities. Larger dishes may still be useful for short very high speed links) 24000 1-3 foot dish Transmit Power/Cost Transmit power is of less interest by itself than is transmit ERP, which is the desired result as mentioned under antennas. But the transmit chain does impact modulation types and signalling bandwidths and therfore must be considered separately. In considering transmit power I am making a guess of the costs of the transmit chain from 0 dBm up. In some situations there is a great difference between the cost of linear and Class C power amplifiers. This may require using different modulation techniques with a resulting difference in required S/N to recover the information, but for the purposes of comparison just the cost of "raw power" is considered here. The Gunn diode transceivers when used alone have no signal frequency amplification at all and must be used with frequency modulation. The following estimates are of component cost for the transmit chain. 144 approx $50 @ 10W, $250 @ 100W linear 220 approx $60 @ 10W, $300 @ 100W class C 433 approx $60 @ 10W, $400 @ 100W linear 900 approx $1 0.2W linear, $10 @ 10W class C 1250 approx $35 @ 2W linear, $100 @ 20W linear 2300 approx $20 at 0.2W linear, $50 @ 3W linear 3400 approx $60 at 0.1W linear 5700 approx $60 at 0.1W linear 10000 approx $100 at 0.1W class AB,C $40 at .02W (xcver,T/R switch included) 24000 $50 at .005W (xcver,T/R switch included) Transmit ERP/cost This is a difficult section to define because there is an almost infinite variety of combinations to consider. As noted, modulation mode and volume electronics makes a great deal of difference in cost and performance. ERP (shown in dBm) is just the sum of transmit power and antenna gain for the particular case, no feedline loss is included. Band tx pwr/ant ERP $cost _________________________________________________________________ 144 could be either linear or class C but spectrum limited a) 10W brick 6 dB omni antenna +46 dBm $100 b) 10W brick 10 dB yagi +50 dBm $150 c) 100W 10 dB yagi +60 dBm $350 220 could be either linear or class C but spectrum limited a)10W brick 6 dB omni antenna +46 dBm $130 b)10W brick 12 dB yagi +52 dBm $150 433 could be either linear or class C but spectrum limited a) 10W brick 9 dB omni antenna +49 dBm $130 b) 10W brick 15 dB yagi +55 dBm $150 900 Japanese personal radio band HW *very* cheap Class C else "homebrew" linear amps, fast PIN T/R switch is very cheap a) 10W 6 dB omni antenna (class C) +46 dBm $30 b) 10W 20dB yagi +60 dBm $100 1250 Mitsubishi 2W linear driver($35) and 20W output($55) PIN switch ($12), 4-6 MHz of digital bandwidth a) 20W 9 dB omni antenna +52 dBm $175 b) 20W 22dB antenna +65 dBm $175 c) 20W 7' Channelmaster w/ screen, 27 dB +70 dBm $220 2300 linear, "homebrew" , lots of bandwidth, a) .1 watt 4' dish 27 dB +47 dBm $100 b) 1 watt 6' TVRO dish 30 dB +60 dBm $350 3400 small and TVRO antennas cost effective a) .1 watt 4' dish 30 dB +50 dBm $100 b) 1 watt 6' TVRO dish 33 dB +63 dBm $250 5700 a) .1 watt 4' dish 35 dB +55 dBm $100 10000 a) .1 watt 4' dish 40 dB linear +60 dBm $150 b) .01 watt 4' dish 40 dB Gunn xcvr, FSK +50 dBm $90 c) .25 watt 6' dish 43 dB linear +69 dBm $225 24000 a) .005 watt 4' dish 47 dB Gunn xcvr, FSK +54 dBm $100 Receiver System NF Since all antennas are looking at a terrestrial horizon, I will assume that there is a 300 deg. K noise floor limit. Through 433 MHz, QRN may make things worse than this. Only the radar gun transceivers without signal frequency amplification need show a significant degradation from this level. Even at 24 GHz a 3-5 dB noise figure is possible in systems with signal frequency amplification. I assume 14 dB NF for the transceivers. Receiver construction will add some cost to a station but for linear systems, low power stages which can be turned around for receive are already present as is the LO and mixer. The radar gun transceivers also have the mixer diode and T/R switch built-in. As a result, incrementing cost by a few percent is a crude approximation of total tx/rx/switch hardware cost for all examples. In any event, data recovery may easily swamp the cost of the rf portion of a receiver. Data BW This area is such a political and legal mess I don't know what to do with it. In addition to this, performance expectations are ill-defined and probably something of a moving target. For the moment lets bound speeds on the lower side with 10K baud and the upper end with 10 Mbaud. For non-synthesized systems such as the radar transceivers there is probably a lower bandwidth/ data rate limit which must be used due to phase noise and frequency drift of the sources. _______________________________________________________________________________ Some scenarios: To compare all of the above. I'll consider a couple of (likely?) paths. Let path #1 be a 15 mile entirely LOS path. This might be between two well located tcp/ip home-qth nodes, between a home qth gateway and a high level backbone or between two moderately close backbones (maybe high speed). Let path #2 be 100 miles and part of some kind of long haul network path. Assume it is marginally LOS, like between two current high level AX25 digis. Path Loss,dB Path #1 Path #2 144 103 120 220 107 124 433 113 130 900 119 136 1250 122 139 2300 128 145 3400 131 148 5700 135 152 10000 140 157 24000 148+4=152 (H20) 165+24=189 Assume tx ant=rx ant, 10 dB of fading margin on path #1 and 20 dB margin on path #2. Let the noise floor be KTB+4dB (+14dB for xcvrs) or -130 dBm in 10KHz,-120dBm in 100KHz. For component cost I am just incrementing transmit-chain+antenna cost by 20 percent. There is no question that these numbers represent relatively "clean" paths. For non-LOS and long paths more than the above margins will be required. The lower frequencies will probably be somewhat better than higher ones for non-LOS situations, in part bacause propagation over hills and through vegetation will not incur as much additional path loss. However some kinds of extended propagation, like troposcatter, will not be strongly influenced by frequency. Component cost, [Carrier/Noise,dB] for [bandwidth, Hertz] ----------------------------------------------------------------------- $Cost Path #1 Path #2 Note ----------------------------------------------------------------------- 144 MHz a) $120 66 for 20K 39 for 20K broadcast b) $180 70 for 20K 43 for 20K c) $420 80 for 20K 53 for 20K 220 MHz a) $155 55 for 100K 28 for 100K broadcast b) $180 67 for 100K 40 for 100K 433 MHz a) $155 55 for 100K 28 for 100K broadcast b) $180 67 for 100K 40 for 100K 900 MHz Class C a) $40 33 for 1M 16 for 100K broadcast b) $120 67 for 100K 40 for 100K 1250 MHz a) $220 39 for 1M 12 for 1M broadcast b) $220 65 for 1M 38 for 1M c) $265 65 for 10M 38 for 10M ( I wish we still had 1215-1240 MHz!) 2300 MHz a) $120 36 for 10M 19 for 1M b) $420 52 for 10M 25 for 10M 3400 MHz TVRO hardware might help a) $120 39 for 10M 12 for 10M b) $300 58 for 10M 31 for 10M 5700 MHz a) $120 45 for 10M 18 for 10M 10000 MHz a) $180 50 for 10M 23 for 10M b) $110 30 for 10M 13 for 1M radar gun xcvr module c) $275 59 for 10M 32 for 10M 24000 MHz a) $120 29 for 10M 12 for 10K radar gun xcvr module I tried to pick bandwidths for the above scenarios which leave enough C/N to enable proper detection of the information (assuming 10 dB thresholds). This is just a paper excercise to help get us thinking. Any real installation is going to require partcicular study to properly configure. I think I remember from a previous estimate that the WA4DSY modem wants almost 20 dB of C/N to get a high BER . It seems to me that 900 (a), all of 1250 and 10000 (a) and (b) are particularly of interest. 900(a) probably requires FSK or PSK and 10000(b)must be FSK. Can someone tell me the "right" modulation scheme for such a system? Please don't unduly flame my numbers, none of these scenarios is complete and my cost estimates are crude at best. I'm presenting this in hopes of making more concrete our dreams of having a 'real' network. Please let me know your comments and interest. I apologize if this tome is longer than should be posted to the group. Glenn Elmore N6GN 10-24-88