Rob's web

Narrow-band F.M. for voice communication

Home - Techniek - Electronica - Radiotechniek - Radio amateur bladen - QST - Narrow-band F.M. for voice communication

A comparison of N.F.M. and A.M. in amateur work.

The advantages of frequency modulation are well known, but how does narrow-band f.m. stack up with other forms of modulation in the matter of readability when the going is rough? W1EYM supplies some answers, based on carefully-conducted intelligibility tests, which indicate that n.f.m. is at its best under the borderline conditions so frequently encountered in amateur work.

An increasing interest has been displayed by the amateur fraternity in the application of frequency modulation to amateur voice communication. Any of us who have followed the commercial development of wide-band frequency modulation are aware of its tremendous advantages over other systems of modulation where a high order of noise reduction is desirable, such as in broadcast and mobile communication services. This high-noise-reduction advantage of wide-band f.m. does not obtain unless the peak frequency deviation of the transmitted signal exceeds the highest modulation frequency by several times. Most wide-band f.m. systems use a peak deviation ratio of 5. Thus mobile f.m. equipment usually is designed for a peak deviation of ±15 kc. with a 3-kc. upper limit on the audio modulation range, while broadcast practice dictates a 75-kc. peak deviation and a 15-kc. audio modulation range. Both systems are wide-band f.m. and have the same deviation ratio.

Both systems maintain their noise-reduction advantages over amplitude modulation down to a critical value of signal input, called the threshold of improvement, which occurs when the signal and noise voltages are comparable. Below the threshold of improvement, an a.m. system will, for the same carrier power, produce a usable signal which would be lost in a wide-band system operating under identical noise and carrier-power conditions. What this boils down to is that w.f.m. will do a far better noise-reduction job than any other system if sufficient channel width is available and if its use is confined to service areas where the signal is stronger than receiver and external noise.

The amateur, however, never considers his sta tion as having a "service area" and is mainly interested in obtaining the maximum range of intelligible communication on any band for a given expenditure in equipment or parts. Crowded conditions on most amateur bands preclude the use of "sufficient channel width" for effective w.f.m. Is, then, a.m. the optimum system for voice modulation for amateur work? The author believes not, and his belief is based on a theoretical comparison of narrow-band f.m. and a.m., and a series of tests conducted on the 6-meter band with the able assistance of W1JKC, W2BYM and W2JCR.

Narrow-band f.m. may be defined as a system of frequency modulation wherein the peak deviation is limited to a value equal to or less than the maximum modulation frequency. Thus a signal which has a maximum modulation frequency range of 15 kc. and swings a maximum of ±15 kc. is, by definition, still narrow-band f.m., while a signal which has a 3-kc. upper limit on modulation frequency range but has a peak swing of ±15 kc. is wide-band f.m., though both occupy the same channel width. A more specific definjyrtion of n.f.m. for amateur voice communication depends upon the choice of upper modulation-frequency limit. Assuming that a high order of speech intelligibility is desirable, the upper frequency limit will fall between 3000 and 5000 cycles. The peak deviation of the transmitted signal should therefore be limited to the particular upper limit chosen, the choice of which is beyond the scope of this article.(1) If, for the sake of description, 5 kc. is chosen as the maximum desirable modulation frequency, the maximum peak deviation must be limited to ±5 kc. or less.

The question then arises whether the peak deviation should be limited to something less than the maximum modulation frequency in the interests of conserving bandwidth. Recent articles in QST have indicated that the second and higher-order sidebands, which fall outside a range of the carrier frequency ± the highest modulation frequency, assume appreciable proportions when the modulation index is unity. In order to interpret this statement in terms of the amount of adjacent-channel interference created by a voice-modulated n.f.m. transmitter, we must remember that the energy distribution in speech is such that the amplitudes of the high-frequency components are small indeed as compared to the lower-frequency components. In other words, a transmitter designed for a maximum modulation frequency of 5 kc. and a peak deviation of ±5 kc. will never, when voice-modulated, be deviating ±5 kc. at a 5-kc. rate, when the peak deviation has been limited by the gain-control setting. This is especially true when the high frequencies are not pre-emphasized in transmission, which omission is advocated by the author in the case of n.f.m. voice transmission.

It should be remembered that if the peak deviation is limited to something less than the highest modulating frequency, a loss in signal-to-noise ratio occurs which is proportional to the reduction in deviation. The receiver i.f. bandwidth must be equal to twice the highest modulation frequency. This bandwidth establishes the noise level. If we only swing the carrier one-half the receiver bandwidth, we take a 6-db. loss in signal-to-noise ratio (hiss noise, impulse noise, or th), and many advantages of the system dispear.

In the light of the above statements, the thor submits the following as a trial balloon for teur standards for n.f.m. transmission at 50 Mc. and above:

  1. Maximum modulation frequency: 5 kc.
  2. Maximum peak deviation during voice or keyed-tone modulation: ±5 kc.
  3. Deviation vs. modulation-frequency characteristic essentially flat from lowest to highest modulation frequency. (No deliberate pre-emphasis of high modulation frequencies.)

Many of the advantages of n.f.m. over a.m. have been well covered in recent issues of QST. We know that an n.f.m. transmitter requires less equipment and consumes less power for the same carrier power, because of the elimination of the amplitude-modulation components. We know that the r.f. voltages in the final stage are less than when the a.m. is used, thereby simplifying insulation problems. The possibilities are remote of unwanted detection resulting in cross-talk in nearby b.c. receivers, telephone circuits, etc. Many such circuits have nonlinear elements which will detect a strong a.m. signal. Fortunately for f.m., detectors for it have to be designed and seldom occur by accident. These advantages are all fine; but what happens to the range, signal-to-noise ratio, and intelligibility of a voice communication circuit when n.f.m. is substituted for a.m. with the same carrier power and the same audio-frequency range?

Theory dictates that the signal-to-randomnoise ratio of an n.f.m. system with a peak deviation ratio of 1 is 4.77 dB better than an a.m. system of equivalent carrier power and the same audio-frequency range, when the signal is comparable to or stronger than the noise. On impulse noise n.f.m. shows a 6-db. improvement under the same conditions. M. G. Crosby has indicated that at any detectable signal strength n.f.m. showed higher intelligibility than a.m. for the same received carrier.(2) The sum total of these advantages led the author to put the system to a practical test on the 6-meter band and the results obtained have more than repaid the time and effort involved.

The first series of tests was conducted over a 12-mile path between W1EYM and W1JKC. In this case the f.m. and a.m. signals were transmitted by W1JKC and observed at W1EYM. A breadboard transmitter, the exciter portion of which is shown in Fig. 1, was constructed which provided a 10-watt carrier at 52.8 Mc. Crystal control was used for both a.m. and f.m. N.f.m. was produced by phase modulation of the output of a 3.3-Mc. crystal oscillator. The audio input circuit was equalized to give essentially fiat deviation from 500 to 5000 cycles. The crystal frequency was multiplied 16 times, by one quadrupler and two doublers, this order of multiplication allowing a peak deviation of 5 kc. on speech without noticeable distortion. For a.m. transmission, the final stage was plate- and screen-modulated by a push-pull 6L6 modulator. The radiated power was adjusted by varying the final plate and screen voltage and varying the antenna coupling to the final stage.

Fig 1
Fig. 1. Schematic diagram of the phase-modulated exciter used at W1EYM. Output of the portion shown is on 13.2 Mc., requiring a further frequency multiplication of four times for 52.8-Mc. operation.

The receiver layout at W1EYM may best be explained by reference to the block diagram, Fig. 2. It will be noted that the variable-bandwidth feature of the Super-Pro was not available in the n.f.m. channel where the i.f. selectivity was permanently aligned for a 10-kc. bandwidth.

Fig 2
Fig. 2. Block diagram of the f.m.-a.m. receiving set-up used in the readability tests. The 456-kc. i.f. unit in the n.f.m. channel uses a double-tuned input transformer, over-coupled, a double-tuned interstage transformer set for optimum coupling, and a single-tuned circuit between the second i.f. stage and the limiter.

When W1JKC first came on the air with maximum power and optimum antenna coupling, the signal, measured by signal-generator substitution, was 20 microvolts at the input of the 6-meter converter (S9+ on a report basis). The improved signal-to-noise ratio of n.f.m. over a.m. was readily apparent, though naturally not as striking as a comparison with wide-band f.m. W1JKC then dropped the signal level by decoupling the transmitting antenna and lowering the final-stage power input. As soon as the signal was attenuated to the point where a.m. reception was noisy and the speech intelligibility suffered, a shift to n.f.m. would increase the signal-to-noise ratio and restore the intelligibility. The signal was taken down to the point where a.m. speech was just audible but unintelligible. Even then the intelligibility of the n.f.m. held up. Up to this point the i.f. bandwidths in both channels were approximately the same. (The Super-Pro selectivity control set for 10-kc. bandwidth.) The bandwidth on the Super-Pro was now closed up to maximum selectivity (in our case about 4 kc.). The increase in signal-to-noise ratio on a.m. was thereby improved but the loss of high frequencies reduced the intelligibility. Without delving into questionable quantitative data, it was readily apparent that n.f.m. produced an entirely intelligible voice signal with a considerably-weaker signal input to the receiver. It should be emphasized that these results did not obtain when the n.f.m. was detected by detuning the a.m. channel.

The exciter unit was then reinstalled at W1EYM and the n.f.m. receiver was shipped to W2BYM at Lakehurst, N. J., 95 miles airline from Fairfield. W2BYM used a DM-36 converter, revised for 6-meter operation, and a National HRO for the a.m. channel. To make a long story short, the transmissions to W2BYM further proved that n.f.m. outperformed a.m. in providing a highly intelligible signal over a greater percentage of each transmission time. Fading occurred during most transmissions, which condition is average over this path.

In an attempt to evaluate the readability provided by each system, lists of ten disconnected words were read on a.m. and n.f.m., changing lists for each transmission. An average of several tests indicated that n.f.m. was approximately 2 to 1 up on a.m. Words like "code" and "cold" were easily identified by W2BYM on n.f.m. Several transmissions on n.f.m. gave perfect "copy" but on a.m. the "copy" varied from 40 per cent to 60 per cent correct. It should be remembered that all these tests were carried on with the same power input to the final stage, the same carrier frequency, the same microphone and speech amplifier, and the same audio-frequency range.

The author's Super-Pro has now been converted for n.f.m. or a.m. reception by the addition of a "black box" mounted on the rear chassis skirt. A 6J5 cathode follower is coupled to the Super-Pro second-detector diode plate (Fig. 3), which in turn drives a 6SJ7 limiter and 6116 discriminator. The output of the discriminator is tied to the Super-Pro audio system through a switch which allows a choice of n.f.m. or a.m. audio. This system performs just as well as the n.f.m. i.f. strip previously described and has the added feature of variable i.f. bandwidth provided in the Super-Pro. Tests with W2JCR in New York, who has recently made provisions for n.f.m. transmission, gave further confirmation of the improved grade of reception provided by n.f.m.

Fig 3
Fig. 3. Schematic diagram of an n.f.m. adapter for use with a communications receiver. The cathode follower allows coupling to the second detector without undue capacitive loading.

As a conclusion to this article, I am going to ask and attempt to answer several questions which I am sure might come up if the reader were sitting across my desk from me:

Q) Do you feel that the advantages of n.f.m. will hold for the lower-frequency amateur 'phone bands as well as 6 meters?

A) Not without adequate tests on each band. It may well be that where crowding is at its worst, the peak deviation must be limited to something less than 3 kc. I would also like you to remember that early wide-band f.m. over long paths, where multipath transmission obtained, showed higher distortion than a.m. This may not hold for n.f.m. where the significant sidebands occupy approximately the same frequency range as in a.m. Only practical tests will tell the story.

Q) How about receiver and transmitter frequency-stability requirements?

A) The best transmission and reception techniques in use on 6 meters for a.m. are adequate for n.f.m. work on this band. Crystal control followed by phase modulation is ideal but not essential. A stable VFO, which may be reactance-modulated, will give adequate stability if carefully designed. Ample deviation may be obtained by grid modulating the VFO. Most well-designed v.h.f. converters have ample stability for successful operation on n.f.m.

Q) Isn't an a.m. receiver easier to design and align than an n.f.m. set?

A) If you are comparing an n.f.m. set with a superregenerative detector plus audio, the answer is yes. If, however, you are comparing a converter plus an n.f.m. i.f. and audio strip, the answer is that there is little if any added complication. The i.f. must be well aligned in either case. In n.f.m. the discriminator adjustment may be made by ear on any signal source.

The advantages offered by n.f.m. are cheap indeed in terms of equipment and power. In terms of effort, the attainment of these advantages calls for the best technical know-how that the serious amateur can bring to bear on the problem.


  1. On the frequencies below 50 Mc. n.f.m. can be used in bands where a.m. is already entrenched only if it occupies the same channel width as that used by an a.m. signal, or less. This points to the desirability of limiting the audio pass-band to 3000 cycles. For 50 Mc. and higher, where QRM is not an important factor, this limit could be raised to 5000 cycles without harmful effects. - Ed.
  2. "Band width and readability in F.M.," M. G. Crosby, RCA Review, January, 1941; also QST, March, 1941.

Nathaniel Bishop, W1EYM.