Options for High-Speed Packet: 2010 Edition

ve6vq's picture

(With apologies to Barry McLarnon, VE3JF who wrote the original Options for Medium to High-Speed Packet in 1993)

Periodically, certain well known and notorious members (you know who you are) of WETNET have gathered to discuss the if, why and how packet radio activity might be made to grow, particularly in the Pacific Northwest. The arguments regarding "if" and "why" are far too controversial to continue discussing here, but let it suffice it to say that there seem to be two opposing views:

  1. If you build it, they will come
  2. Nobody will ever need or use it again

Since the above assertions are more or less unprovable without performing some kind of experiment, let's will focus more on the "how" aspect. Let's assume some basic engineering requirements:

  • Node cost of less than $500 for end-nodes
  • NLOS range of at least 8 km (5 miles)
    @ 10 watts with isotropic antenna @ 10 meters HAAT @ 440 Mhz
  • Data throughput of at least 100 Kbps
  • Built-in RF-aware link layer management (FEC, rate adaptation, etc.)
  • Able to act as part of a mesh-connected network
  • Easy to install and configure for non-specialist users
  • Compatibility with any existing amateur packet radio systems is not mandatory (or perhaps desirable)

Let us also ignore (for the time being) the more or less open-ended discussions about which bands such operation might take place on.

Today, there are a few systems that could be proposed as the basis of a high-speed (i.e. 10 or more times faster than 9600 baud) packet network meeting the requirements listed above:

Of these, the Icom entry is disqualified, because at $700-$1000/node it is too expensive and it lacks any link layer FEC. The DD portion of the system has not been engineered with realistic RF links in mind (it implements an extremely simple Ethernet bridge protocol), and the focus of the overall system seems to be more on concurrent digital voice and low speed packet rather than a general purpose, high speed packet radio system.

The ARRL HSMM initiative made impressive progress until the HSMM committee imploded. It is not clear where their work ended up. The OFDM modem they developed was so flexible that it probably should be considered a proof-of-concept, and would likely have cost considerably more than $500/node unless many were manufactured.

The 802.11 mesh networking work in Texas and elsewhere is continuing, and has shown some progress but it is severely hampered by non-exclusive access to the 2.4 GHz band, and literally millions of other devices cause severe interference on 2.4 Ghz and 900 Mhz and other shared bands. A number of groups have attempted to make transverters that allow the use of standard, off-the-shelf 802.11 hardware on other amateur-only bands. This approach might be quite promising, as most users are very familiar with the installation and configuration of 802.11 gear, at least on the 2.4 and 5 Ghz shared bands. The biggest challenge with this approach is finding amateur-only frequencies where the needed bandwidth (22 Mhz) is legal.

Bdale Garbee's model rocket telemetry system makes use of one of a growing class of single chip RF modems that are generally designed for industrial SCADA and telemetry applications, using FSK and related modulation schemes intended to "fit" into conventional NBFM radio channels. Many of these do not qualify based on our speed requirements, but some do, including chips from TI and others that offer data rates up into the hundreds of kilobits per second. The challenge with these designs is that the power output of these devices is usually very low, in the milliwatt range, and needs to be amplified considerably to provide practical range. They also typically do not provide any built-in support for link-layer functions such as FEC/ARQ etc. There is a considerable amount of both hardware and software work needed to produce a practical solution here.

Another consideration here is the role of TAPR, both in the past and currently. During the original packet "boom", TAPR acted as a technology "incubator"; Proving that the technology was available and could be manufactured and marketed, and then licensing it to commercial firms who then marketed it further. Considering the current efforts of TAPR with the HPSDR project as an example, this appears to be a case where the technology is available but there does not appear to be a critical mass of market demand.

However, one of the differences between the last packet radio boom and today is the presence of very low cost manufacturing capability in Asia. UHF radios with 10-20 Kbps FSK modems can be bought online from Chinese manufacturers for less that $50/node today.

Another interesting shift in the environment since the late 1980's is the rise of the open-source paradigm. This approach has proven very successful in facilitating the development of very complex software projects by many volunteers spread all over the world. Whether it could be applied equally well to hardware and/or RF projects is an open question. For approaches that require hardware and/or RF development (virtually all of them), there is a significant barrier with respect to finding the NRE (non-recoverable engineering) funding to make them go. The open source development model does not address this. The TAPR 900 Mhz spread spectrum radio project was ultimately killed by the speed of commercial product development versus volunteer amateur radio projects (as a number of other amateur projects have been).

Since the late 1980's Moore's law has held, and now it is possible to at least slightly isolate such projects from changes in the hardware and RF components marketplace, simply because so much more can be done using high speed digital hardware and software instead of analog RF hardware. For example, much of the proof-of-concept OFDM modems developed by Stephensen and Champa were implemented in VHDL software using FPGA hardware. It is much less likely that similar FPGA hardware would be obsoleted soon, since it is very general purpose hardware that is being used in many projects in conjunction with project specific software. The advent of lower-cost SDR techniques as shown by the HPSDR project should make it feasible to build a packet radio hardware "block" that could be programmed to implement any number of possible packet radio schemes using a combination of VHDL and DSP programming. This dramatically lowers the entry cost for a new design, and at least in theory, a new system could be deployed via a simple software update.

D-Star <-> COTS FSK modems

This is a very good point, and it effectively reduces the number of options mentioned in the article from 4 to 3. The TI and SiLabs RF FSK radio-on-a-chip products also produce FSK signaling in the same speed range as the D-Star. Perhaps one of these is used in the D-Star anyhow. Regardless, the list of items mentioned here is about the same as those needed to convert a TI or SiLabs chip into a workable system. If it were inter-operable with the D-Star system, that would be good, but the point about the off-the-shelf D-Star system is the cost per bit-per-second. If a D-Star clone could be made at a significantly lower cost than today's (Icom) prices, it would be a viable option but it seems to me that it's far too much money for far too little speed.

Larry, VE6VQ/W7 PGP Sig: 616D 4E52 CF1F 3FEC FFFB F11B 7DB9 C79A EA7E B25B