The short answer, as Steve Ford likes to say, based on the Cohen paper, is that the "necessary bandwidth" appears to be "roughly" twice the frequency shift, although an exact calculation is obviously very complicated.
More importantly, with regards to the amateur radio service is the summary statement, "The necessary bandwidth is the minimum emission bandwidth required for an acceptable quality of service." It has already been concluded, after many months (even years!) of debate, that radio amateurs are "amateurs" and not "professionals" and do not have either the ability or the means to measure "necessary bandwidth" of their signals. Their communications are casual "amateur"communications and not "professional" communications. If the "necessary bandwidth is the minimum emission bandwidth required for an acceptable quality of service" were to be codified into the radio amateur service regulations, it would also be necessary to also define what "acceptable" quality is, in particular for the radio amateur service. That definition will obviously be different for casual conversation, DX exchanges, and contest exchanges, than it is for commercial or quasi-commercial "messaging" services. It will probably fall somewhere between PSK31 and MFSK16 or WSJT bandwidths, which provide "casual" communications quality in exchange for the higher bit rates needed for sending long messages. Even narrow bandwith modes, like PSK31, can be utilized to reduce the error rate to zero through the use of ARQ. It is just that the throughput is half that of the non-ARQ use of the mode, but that is generally "acceptable" for casual communications. What would NOT be acceptable is using a 150 KHz-wide signal on a band that is only 350 KHz wide merely in order to achieve faster throughput for two dominating stations at the expense of hundreds of others. Should 150 KHz-wide signals start being used on 20m, for example, it would not take very long for the FCC regulations to be changed (or re-interpreted) to protect the "casual" communications use of the 20m band. To infer that using "low power" would make that acceptable ignores the fact that "low power" to someone distant is "high power" to someone close by. The BPL debacle should have made that clear by now. The regulations already require that the minimum power necessary for communicatons be used, and if a similar requirement were made for emitted bandwidth, it could easily stifle innovation (at least with regard to using wider, or spread-spectrum modes), and not promote it. We might all then wind up having to be content with PSK31 plus ARQ for our casual communications! Better not ask for something you may not want! I agree that the regulations do not "specifically" limit bandwidth on the HF bands, but that does not mean this could not easily happen if there are enough abuses to justify it. It is true that the regulations have not kept up with technology, but the intent to protect casual communications is still there, and that intent could be codified if it becomes necessary. However, we may not be happy with the end result, especially considering the extremely minor interest in digital messaging or using digital modes other than PSK31, CW, and RTTY. With the advent of satphones, cell phones, and the Internet, the relevance of amateur radio as anything more than a hobby activity is rapidly diminishing and we can expect future regulatory changes to further support the hobby interests rather than quasi-commercial interests in amateur radio. 73, Skip KH6TY
