The Art & Skill of Radio-Telegraphy
-Second Revised Edition-
William G. Pierpont N0HFF
Communication fails unless our message gets across and is understood. Weak signals and poor conditions during transmission (static, interference, fading) all contribute to partial failure to get through. In all of these conditions, telegraphic communication is vastly superior to voice because almost all its energy is effectively concentrated within a very narrow band. Yet it pays a cost for this by taking more time to communicate the same words. In addition, it too can suffer partial loss due to transmission conditions as well as from just plain accidental misunderstanding. How can we reduce these losses to a minimum? Let's focus on the originator's use of the words themselves (by "words" we include the use of abbreviations and Q-signals).
We rarely think much about how we speak when we are conversing. When we speak face to face we can generally tell whether we are being understood or not by feedback through the reactions and responses of the listener. But when our communication is remote, by voice over a wire or the radio, the visual clues to the hearer's understanding are missing. When the telegraph code is the link, auditory clues (tone of voice of a comment or reply, "uh-huh", "yeah", etc.) are also missing. Relatively awkward break-in is the only possible direct feedback while transmitting in code, and it is an ambiguous interruption, until the receiving operator explains his problem.
It is when we speak, whether face to face or by remote means, that most of us tend to use more words than the bare minimum necessary to be understood: this is called redundancy. The degree of redundancy varies from person to person and from situation to situation. Redundancy increases the context from which the hearer to may understand.
When we write we generally are much more careful of how we say things that are important than when we speak. We give more thought to the choice of words and the way we write them: we becomes more circumspect and precise in order to minimize the reader's possible misunderstanding of what we mean. Since we have no feedback at all, we generally tend to use more words than the minimum necessary in order to make up for that lack.
In telegraphic communication the tendency, largely because of the time required to transmit, is to eliminate every word which does not seem to be absolutely necessary. We abbreviate in various ways-- generally down to bare bones: the minimum required to express the thought. First we leave out words, and then we tend to abbreviate what is left as much as we think we dare to omit and still have it understandable. (This is especially true when paying on a per-word basis for transmission.)
What we have been saying is this: redundancy helps to insure adequate and more accurate communication. That is, we normally use more words and expressions than the bare minimum required to get our meaning across. Time, however, is a factor working against telegraphic communication. It is not as rapid as speech in terms of words per unit time. In order to balance the time factor against the intelligibility factor, the originator of a telegraphic message generally weighs more carefully exactly what words to use and how to put them together. If he is wise he will also consider the effect of possible mistakes or distortion during sending and receiving which might produce ambiguity.
What can we amateurs do to minimize misunderstanding or complete failure of our communications? One of the commonest things is simply to repeat each word or words, or the whole message. We may repeat only the most critical words or numbers two or three times. (Numbers are almost impossible to correct because there is no significant context to help out.)
Another form of repetition is to ask the receiving station to repeat the message back to the sender word by word. This nearly assures perfection. But this, like repeating each word as it is sent, requires at least twice the original time on the air.
Counting the words in a transmission has long been a common commercial practice, but is not generally used except for message type traffic. It does not assure complete accuracy (exact words and spelling).
We can often prevent misunderstanding by adding a word or two to a short communication. For example, to confirm a scheduled QSO later in the day, to say "CUL this afternoon," or "CUL in pm" instead of just "CUL" helps insure that the other operator knows that you mean today, and that you are not cancelling it (as he might assume otherwise due to some interference, etc.). When conditions are rapidly deteriorating this may be our only hope to get across before further communication becomes impossible.
A little forethought along these lines on the originator's part may help avoid unfortunate misunderstandings. Especially when we simply must get through, and conditions are very poor, we should choose our words and expressions carefully.
Here we ask "Will I be able to copy (or read) it?" and if I can't, "What is the problem?" -- "What can be done to improve the quality of this material I am receiving," or "What can be done to make sense out of this somewhat garbled transmission which is all I have?" -- "What is the nature of the problem?"
During the communication, speed of transmission is an important factor, one directly controlled by the sender. Both too fast and too slow sending can cause trouble in receiving -- here the receiving operator must tell the sender to slow down or speed up to meet the receiver's needs. Quite naturally, speed of transmission must set be within the receiving operator's capability.
It may be that the weighting of the dits is too light and I'm missing some of them. If so, can the sender make them a bit longer (heavier)? Maybe the sharpness of the pulses has been rounded off too much to remove "clicks" and the signals sound mushy. At higher speeds, perhaps the dits are too heavy and confusing the ear. These are things which the sender may be able to modify on the spot, but he must be told.
In Chapter 14 "The Ear" we have discussed some of the things which can be done to help, especially the use of filters. Here we look at the filter requirements for an audio filter. We want a filter which will separate the desired signal and still keep it intelligible. At this point we are not concerned with any of the radio frequencies of the signal as it passes through the receiver, but only with the audio beat signal which is output.
That audio signal consists of
The minimum basic telegraphic element is the "dit, an "on" signal lasting a given length of time in seconds. For example, a 10 baud rate of signaling means that there are ten basic telegraphic elements per second (or 5 cps or Hertz), and each element lasts 1/10 of a second, the reciprocal of the baud rate. Obviously, to perceive a dit or a dah requires silence both before and after it. The minimum element of silence (space) is also equal to one dit. One dit followed by one element of space constitutes a square wave two telegraphic elements long and may be called one "cycle," by analogy with a cycle of sinusoidal wave. (This is expressed symbolically in Chapter 28 by "10".) A continuous series of dits would then for a given length of time have twice as many bauds as cycles per second. A sequence of 25 such dits and spaces (10101010..., 50 elements) in one second would thus correspond to a frequency of 25 Hertz, 50 bauds. It is in this sense that we compare these two frequencies (audio frequency and telegraphic keying frequency).
For a filter the two predominant factors for intelligibility are passband width and center frequency of the beat note. (The actual shape of the filter's frequency-amplitude response curve is also of importance but for other reasons: see Chapter 24 and engineering manuals.)
There must be enough audio cycles to fill in the keying pulse shape of the smallest code element, the dit, in such a way that all code elements begin and end clearly and are therefore properly timed. That means that the audio center frequency (pitch of the beat note) must be high enough to preserve the square wave shape closely. A mathematical (Fourier) analysis shows that the center audio frequency needs to be about 7 times the telegraphic cycle rate to give the best shape of telegraphic pulses.
A square wave frequency related to words-per-minute, and the duration of one telegraphic unit can be worked out for English using the data in Chapter 28 as follows:-
For standard English text, there are 49.38 elements per word. This is only 1% less than the standard 50 elements used as today's standard word, so we shall use the 50 element standard here.
If this 50 element word is, for example, assumed to be sent in one second, it will be at the rate of 50 bauds, or 25 Hertz, (cps square wave equivalent). For this example there will then be 60 words in one minute -- 60 wpm, a high speed. Using this to convert wpm to bauds we multiply (wpm) by 60/50, that is by 1.2. Since the duration of one basic telegraphic element is the reciprocal of the baud rate, in this case it will be 1/50 second.
Now to determine the minimum audio frequency needed to fill in the telegraphic square wave shape well and give really high quality audio code signals, the following factors must be taken into account:
For our 60 wpm example above, this means an audio frequency of 50 x 7 = 350 Hertz for best quality of code pulses. Thus it can be seen that, except for extremely high speed transmissions, there will be no problem, since the typical values of beat frequency are in the 400 - 1000 Hz. range.
The minimum bandwidth will be concerned with signal stability and intelligibility limits. If the bandwidth is too narrow the signal may drift out and be hard to find again. If it is too wide the risk of random noise and interfering signals increases. The rise-fall time of a filter to square wave input should not exceed about half a dit length. Working through the arithmetic for 6 dB down shows that the minimum bandwidth for Standard English should not be less than about 1.33 x (wpm). This is well below the bandwidth needed for signal stability, so there is no problem here for normal CW use.
Finally, if your copy doesn't seem to make good sense, and there is no way to verify it, see the end of Chapter 8 "Copying" for suggestions.
Signal required for CW with 5% character errors is 20 dB below that of double-sideband a.m. A good operator with CW at 15 wpm in presence of thermal noise, a signal to noise ratio (in one kHz bandwidth) of -1 dB is required for 10% character errors and +1 dB for 1% character errors. This latter is 22 dB below double sideband order-wire quality. However, 17 dB below double-sideband a.m. for CW was chosen to account for differences between operators.
Thus: CW needs at 0 dB compare with SSB needs at +14 dB (room for improvement)
DSB needs at +17 dB (5 dB Difference in operators!)
Reference: Power relationships and operator factor: (QST Fe 1967 p 46, US Army Rept):
The Art &Skill of Radio-Telegraphy
©William G. Pierpont N0HFF
This page last updated August 02, 1998
Modifications and compile by Thom LaCosta - K3HRN - December 2004