What should be the criteria and priorities for packet network links? The following are my personal opinions with reasons on the subject in order of importance.
One of the most important criteria for a good link is a strong RF signal between stations. Errors rapidly increase if the receive signal is below 2 uv (assuming typical receiver sensitivity). So 2 uv should be considered the minimum, usable, receive signal strength. You should have 10 uv (14 dB fade margin) or even 20 uv (20 dB fade margin) if you want a reliable link. Note: 2 uv is typically S9 on the "s-meter" of fm voice radios so you should be able to put a 20 dB attenuator in line and still receive a full quieting S9+ signal.
Related to signal strength is the effect of site hardware on signal strength and reliability. Don't fall into the trap of using high gain antennas to achieve the signal level. Most practical high gain antennas are extremely vulnerable to environmental effects. A light coating of ice (or even rain sometimes) will kill most gamma or Tee matched beams. Most amateur grade antennas will break with the ice that falls off typical commercial tower installations. Use heavy duty low-Q commercial grade beams (for example: Sinclair 307EB = All welded aluminum, 7 element, 3 ft. long, UHF beam with sealed, folded dipole driven element ) and your site manager will love you. More important, your link will take a heavy ice coating and keep working.
If the signal strengths are not high enough with a typical 20-30 watt radio, don't bother using an amplifier. The few dB gained will not greatly improve the fade margin. If you need an amplifier, your path is too long and you made a mistake in your network plans.
Equally important to a good RF signal is the use of dedicated point to point links. The added collisions, divided or shared throughput and the slower parameters necessary on a multi-way link, even with a good RF path, make dedicated point to point links highly desirable. A dedicated point to point link at 1200 bps will beat a shared higher speed network link most of the time.
After a good rf signal and dedicated links, fast radios will have the most effect on link efficiency. Using fast radios on a poor rf signal path will not help much since they will spend most of their time waiting for FRACK timeouts and retrying. Fast radios are highly desirable on a multi-way to avoid "Key-up HTS" which is a serious cause of collisions on heavily used channels. Fast radios will speed up a dedicated link primarily because of reduced TXDelay. The amount of improvement on a dedicated link will however depend on the data rate used.
High speed data rate is last on my list because it is my opinion that unless you have a good rf signal, you will lose all the speed advantages on retries. Unless you have fast radios, your link will spend most of its time on TXDelay and turn-around times. What is the point of using 9600 bps if you need to use a 250 ms TXD to send a 15 ms ACK frame. Higher data rate should only be considered after you have optimized the signal strength and radio speeds.
If you are going to spend time designing a network, plan from the outset around good signal strengths and dedicated links. Don't expect miracles from long RF paths. They don't work!! (At least not very well.) If you start off with good signal levels, all the other items mentioned above can be changed later to upgrade a link for better throughput. Don't waste your time and money with high speed links until you have the good signal strengths and fast radios. It is difficult to change sites or add additional sites later to cure poor signal levels, so make good signal level your number one priority.
These are my considered opinions. I will welcome any constructive criticism/comments on the subject.
This article was excerpted (with minor modifications) from the Feb 25, 1996 issue of the NEDA Report distributed to all NEDA members.
Comments or criticism should be directed to the author by e-mail, or by packet at:
We welcome technical articles on networking topics for publication. Information should be freely disseminated and not hoarded so all can benefit from your knowledge.