Speech
Processing in Ham Radio
Speech processing, as it relates to amateur radio, consists mainly of techniques that increase the mean modulation level that drives our ssb transmitters.
All voice modes can benefit, but, since ssb contains no carrier or second sideband it has the most to gain. We might even say that modulation is everything in this mode. At the least, it’s important to get as much in there as you can for a strong signal within our 3khz or so. You can bump this up to VERY important when operating at QRP levels. The real trick is to do it without generating audio harmonics that distort it to the point of unintelligibility or have us transmitting them up and down the band outside our desired 3khz. Therein, lies the rub!
Quick
‘n’ Easy:
One of the simplest and most widely used methods is to simply drop out some of the bass tones and increase the higher tones in our voice signals, since this results in very little loss of intelligibility and a small increase in perceived ssb power. This can be done relatively simply with a well chosen capacitor in the audio chain and results in something akin to the difference between Heil’s type 4 and type 5 mic elements. There are, however, more rewarding (and more difficult) techniques available to us.
The
Mean Machines:
There are two techniques that result in far more mean modulation that I’ll discuss here.
First, let’s be clear on exactly what we want to accomplish.
The very nature of language requires that we modulate our speech, so the level of our voices are constantly going up and down in volume as we pronounce each of our words. Ideally, we would like to have our meager 5 watts actually be 5 watts all the time! Well, you can see the dilemma. We really can’t have our cake and eat it too… or can we?
Suppose we’re able to build a device that will increase the gain (volume) for the quieter parts of speech, decrease the gain for the louder parts, and do this automatically and proportionally for all levels in between. Such devices are readily available in the world of integrated circuits (or as finished goods) and are known as audio compressors. They’re the automatic gain controls of the live microphone audio business. They do a good job of evening out the peaks and valleys of audio to a large extent, but, introduce a few snags as well.
There is a finite time that is taken to first detect, then respond to a signal by changing the gain in the appropriate fashion. Another condition exists when the compressor must either remain at this gain setting or return to the default level. Which it ought to do next cannot be foreseen (are we between syllables, between words, at the end of a sentence, …?), so must rely on us to set it at an arbitrary “time to let go” which is a compromise at best and a disaster at worst. These are known as the compressor’s attack and decay times and result in a distortion known as “breathing” in which noises without any intelligence content are introduced into the audio spectrum and ultimately wastes transmitter power.
Another, far more troublesome problem, comes with compressing audio specifically for ssb transmission. The very nature of constantly changing the original audio signal introduces harmonic distortion. This means that even though we restricted the input audio bandwidth to 3khz, harmonics will be generated up the audio band to 6khz for the 2nd harmonics, 9khz for the 3rd, and so on. These can be very difficult to get rid of well enough for ssb work. Besides, it’s just another problem to deal with that the compressor generated and was never there in the beginning! If we don’t get them out of there they will appear as part of the ssb output which will now be many khz wide causing “splatter” up or down the band.
I don’t mean to imply that compressors are no good. A well-designed compressor for ssb work that is gentle in its effects and has filtered its output sufficiently is a great little addition to the shack and will raise the mean modulation and ssb power level considerably. I do mean to imply it’s not an easy task. Most are designed to be used in, either compressing live rock music where these distortions can hardly be heard over others, or in compressing pop music and commercials AFTER recording when very sophisticated techniques are available to deal with shortcomings.
Let’s consider what happens when we amplify the audio signal to a level where even the lowest volume parts of speech are, say able to produce an ssb output at ˝ full power. In order to avoid over modulating the ssb generator and power amp (poof!), we clip the audio peaks off at around 90% modulation. This won’t over modulate the rig, but it’ll have some big time splatter to put on the air! I mean this is going to be way worse than the compressor was above. Hold on to this clipping idea for a moment and take a different approach with it.
If we superhetrodyne the original audio (3khz wide) with an RF local oscillator using a balanced mixer, we’ll get a dsb signal at RF. We, then, amplify and clip this dsb modulation in the way we did above. What results is an awful bit of radio witchery that will, among other things, take up 6khz of the band and have splatter across bands at the harmonics of the local oscillator!! Now comes the magic… Put this jumble of frequencies through an ssb filter designed to pass only the 3khz above (or below) the local oscillator frequency and we get a beautifully clean ssb signal with a high mean modulation. The last thing to do is demodulate this ssb down to audio again and… voila! We have compressed audio at high mean levels and without harmonics over 3khz. Use a potentiometer to reduce the signal to a mic level again and feed it to your radio!
Now, that’s pretty cool! Whoever figured this one out deserves to be acknowledged for his/her contribution to the world of radio communications. If anyone can point me in the right direction, I’d be happy to do so here.
Tony Wallace, VE3TNW, bravhart@cogeco.ca