Bandwidth versus Keying Speed Revisited

(Theory and Experimental Data in Harmony)

In late May my eHam article "Bandwidth versus Keying Speed" (http://www.eham.net/articles/16649) pointed out that the FCC's definition of occupied bandwidth is equivalent to the 99% power bandwidth. Using the concept of Fourier series to model the output signal from a typical CW transmitter, my calculations demonstrated that the occupied/power bandwidth is a function of both the keying speed and the rise/fall characteristics of the keying envelope. Table 1 from my original article, which is reproduced below, lists the 99.1% power bandwidths for sinusoidal keying (5-ms rise and 5-ms fall times) and square-wave keying at 2.4 and 30 words per minute.

Table 1. 99.1% power bandwidth for CW/ASK transmitter for periodic signaling

Although the theoretical results presented in Table 1 are consistent with the professional electrical engineering literature, as well as the latest editions of The ARRL Handbook, a number of hams criticized my article's lack of experimental data on typical CW transmitters. Fortunately, my school recently purchased two Agilent Technologies N9020A signal analyzers (at a cost of approximately $26,000 each) and I now have occupied bandwidth data as a function of keying speed for four different transmitters.

The occupied bandwidths for four CW transmitters I've tested can be seen at http://www.arrl.org/sections/?sect=LA (Figures 6 through 15). These measurements are all consistent with the fact that the occupied bandwidth does vary with the keying speed, even in the case of constant rise/fall characteristics of the RF envelope. In other words, my measurements are consistent with the calculated results I presented in my original eHam article. The transmitters I tested were operating into a dummy load at 18.1 MHz with an output power of approximately 50 W. In all cases, the transmitters were sending a string of "dots" (dits) with a 50% duty cycle.

For example, Figures 6, 7, and 8 show that the occupied bandwidth (the 99% power bandwidth) for my Kenwood 940 transceiver is about 60 Hz, 190 Hz, and 260 Hz at 6, 30, and 53 words per minutes (wpm), respectively. (I observed the RF envelope on an oscilloscope to verify that the rise/fall characteristics are essentially constant at all the keying speeds used in my tests.) Measurements of the occupied bandwidths at similar speeds as the 940 are also presented for my Kenwood 2000, Ten Tec Corsair, and Ten Tec Orion II. The occupied bandwidths are noticeably smaller for my Orion II (Figures 14 and 15) compared to my other three rigs because the rise/fall times on the Orion were set to 8 ms (the normal setting I use on the air), whereas the rise/fall times on the other rigs are closer to 5 ms and are not readily adjustable. When I set the rise/fall times on the Orion II to 5 ms, the occupied bandwidths track the other three rigs rather closely.

As I tried to point out in my original article and follow-up comments, the occupied/power bandwidth is not the same concept as the "key-click" bandwidth. However, if the occupied bandwidth of a CW signal is "reasonably" close to the necessary bandwidth as calculated by the simple equation in Part 2 of the FCC's Rules, there should be little or no key-click interference on adjacent frequencies. For example, since the necessary bandwidth is 120 Hz at 30 wpm (for a fading circuit where K = 5), sinusoidal keying at 5-ms rise/fall times (with an occupied bandwidth of about 150 Hz at 30 wpm) should generate little key-click interference, whereas the key clicks generated by square-wave keying (with an occupied bandwidth of about 525 Hz at 30 wpm!) will be very objectionable for many kHz up and down the band.

In conclusion, I hope the fact that my measurements of occupied bandwidth using a state-of-the-art spectrum analyzer will convince practically everyone that the Fourier series models I assume in my original eHam article are correct. The occupied bandwidths of typical CW transmitters do, in fact, vary with the keying speed.

73, K5MC