I've convinced myself after 2-3 years of working with my current (unspread) 900 MHz radios and typical users that getting small groups to cooperate enough to implement a local area amateur network with near-LOS paths which don't have multipath that significantly degrades higher speed data links is pretty unlikely. Having said this I have to also emphasize that simply removing multipath in any manner; with SS, DSP or whatever, is really a band-aid over a bad situation. The very conditions which cause multipath to be an issue are also the ones which add 20 dB or more of additional pathloss to a link. This very rapidly kills the economy of moderate datarates over significant distances because much more expensive RF hardware is required. Hams can't afford the kind of hardware which gets highspeed data across poor quality paths with high excess losses. However, neither can they always provide perfect paths. I believe we need a solution somewhere in the middle and that this solution must be very low cost. Low cost can allow a single "bad" path to be broken up into a couple of good ones with only an incremental increase in link cost. L3TNC is meant to address this need.
Briefly, L3TNC RF is a direct conversion 24 cm radio system with all carriers, clocks and data running fully synchronous to a master VCXO. A pilot tone is SSB modulated onto the RF carrier, nominally at 1265 MHz. The receiver on the far end of the link phaselocks it's own master VCXO to this pilot. Because of the the full synchronism, no tracking loop is needed after acquistion; system phaselock also guarantees lock to the carrier, spreading code and data clocks.
This RF portion is (yet to be) joined with a digital controller which performs a significant number of functions in software which might otherwise have to be done with hardware. Among these are, receive gain control, transmit power control and SS signal acquisition. In addition this controller performs layer 2 and layer 3 functions; link management and routing.
The two sections; RF (up to about the 1 watt level) and control, together are to occupy a housing of similar size to existing multi-mode packet controllers, thus the "layer 3 TNC" title. The combination definitely has got to be inexpensive and self contained to be viable in the amateur context. External connections are 12V power, antenna and digital interface to a host computer; nominally a bi-directional parallel interface but with Ethernet possible at increased cost. The availability of the Motorola 68EN302 controller may make it possible to have inexpensive Ethernet as a standard capability.
The system can be run either with spreading on or off because the data/information path is separate from the frequency/spreading control. Simple BPSK may be used for data but the radio can actually accomodate a generallity of modulations; QPSK, DSB or even SSB are possible since each end has both a synchronous carrier oscillator and I/Q modulators/demodulator. Within its information bandwidth, any waveform (represented by an I and a Q component) can be supported.
For spreading, so far I've only used the shortest, 7 bit, PN sequence currently allowed under US amateur regulations. At 21 dB, its process gain/multipath rejection seems adequate. This is so because I'm targetting only paths which are moderately "dirty". These are the ones within the limits of acceptable incremental pathloss. My measurements and the literature seem to show that normally the unwanted path components are several dB down from the desired (most direct) path. A short spreading code has the advantage of decreased acquisition time. Also, though the 23 cm amateur band supports a nice contiguous chunk of spectrum; 60 MHz, the higher data speeds require higher information bandwidth and thus limit the maximum processing gain.
I'm still considering using a second randomizer on the data side to make sure the information bandwidth has more uniform spectral density. The 7 bit sequence at the higher rate should then spread everything out over the 32 MHz between the first minima. The particular modulation method chosen will affect this too. Spreading the pilot, which currently has a line spectrum and significant energy, is still a concern. How well the lines associated with it will coexist with existing narrowband communications in the band is not yet clear. It may be that spreading will have to be disabled at mountaintop/backbone sites. Fortunately this type of site probably has the highest quality path and the least need for the channel equalization provided by SS. Also, directional antennas are a crucial part of such a system and can greatly reduce interference on both transmit and receive and allow very considerable spacial frequency reuse.
I'm striving to make construction/manufacture simple and cheap. Everything is being done with printed circuit board techniques. My first breadboard used a low phasenoise phaselocked 1265 MHz carrier oscillator with a high Q coaxial resonator. .
See my original 1988 Ham Radio Magazine series(Feb/June/Oct issues) for a coaxial resonator oscillator running at 1010 MHz if you are interested. These Phase Noise plots are generated with the coaxial resonator oscillator and the microstrip mixers. I've achieved Considerably better phasenoise, particularly close to the carrier, with a newer design. This plot is an p;d one showing the coaxial oscillator and using only simple loop filtering with the MC145149.
I have since designed a simple oscillator using stripline techniques between layers of multilayer PC board. With relatively inexpensive board materials I'm getting pretty good performance. I'll plot some samples of this as I have time. Also, I've built numerous filters using similar stripline resonators and I plan to use one of these in a linear phase bandpass filter. Good uniformity and accuracy has been readily achieved without tuning.
Because the radio converts directly to and from baseband, all the signal handling after the RF board can be done with opamps and low frequency devices. I expect to do the next turn of the image reject mixer (I/Q), power dividers and 90 degree hybrid to be implemented in stripline instead of microstrip as they are now. Most of the RF amplifier circuits are common to receive and transmit, including the power control/gain control circuits. It appears this will allow me to turn around large sections of the gain circuits between transmit and receive, thus saving board space and parts cost. I have not yet breadboarded all the small signal and adjustable gain stages. However, they are similar to those used in the existing 900 MHz radios. The PA and switching circuits, including a 'post amplifier/preamplifier' at the 25 watt level have all been built and tested. THe PA allows for improving a given link's C/N budget by 10-15 dB as well as for a more remote antenna without suffering from increased feedline losses. In general though, throwing more power (beyond 1 watt) at a non-ideal link is not the desired solution. I hope that adding an additional L3TNC to break a bad path into two shorter good ones will be more attractive, even economically.
A considerable amount of the system economy is due to using a CPU to set/maintain proper gain/power settings, acquire initial spreading PN clock phase and do other radio housekeeping that would otherwise have to be done in hardware. The 68302 looks like a good candidate but none of the digital side, other than simple h/w controllers for the phaselock circuits and circuits to spread and recover the DS signal has yet been fully designed or implemented.
A lot of help is necessary if the goal of this project is to become a reality. My hope is that I might get to a point that I can inexpensively build and make available a few pairs of these radios which may then be used by other amateurs to further develop FEC, layer 2 and layer 3 algorithms. Whether or not this is a realistic hope remains to be seen. If you are interested in helping, please let me know. Perhaps TAPR or some other existing group may become involved as well. I'm making progress slowly on this whole experiment but I could certainly use help. If you are interested, please indicate.
Glenn n6gn
glenn@santarosa.ampr.org
May 1996
