Speech processing
| Speech processing | |
|---|---|
| Also called | Speech processor, compressor, COMP or PROC |
| Used with | SSB voice, and sometimes AM |
| What it does | Raises a voice signal's average power without raising its peak power |
| Main methods | Audio compression, audio clipping, RF clipping, DSP processing |
| In use | |
| Most useful for | Weak signals, DX pileups and contests |
| Best turned off for | Strong local contacts and all digital modes |
| Set after | Microphone gain and ALC, with the processor off |
| Risks of overuse | Harsh audio, louder background noise, splatter |
Speech processing is the treatment of a voice signal in a transmitter to raise its average power without raising its peak power. Ordinary speech consists of a few loud peaks with much quieter sounds between them, and a single-sideband transmitter's output is limited by the peaks. A speech processor narrows the gap between the loud and quiet parts, so more of the transmitter's capacity is in use more of the time. The result sounds louder and easier to understand at the receiving end, especially when the signal is weak or competing with noise. Too much processing has the opposite effect, making a voice harsh, distorted and wider than it should be.
Average and peak power
An SSB transmitter's output follows the voice. Its maximum, the peak envelope power (PEP), is set by what the final amplifier can deliver cleanly, and the ALC circuit holds the peaks to that level. The average power depends on the shape of the speech. Vowels and stressed syllables produce brief peaks, while many consonants, which carry much of the meaning of a word, are far quieter. As a result, the average power of unprocessed speech is only a small fraction of the PEP.
The relationship between the two is often called the average-to-peak ratio, and processing raises it. A 100-watt transmitter still peaks at 100 watts, but its average power goes up, and the quieter parts of each word are lifted out of the noise where a distant listener can hear them. The other operator may see little change on the S-meter, which follows the peaks, yet find the voice much easier to copy. Hams call this extra intelligibility "talk power."
Compression and clipping
There are several ways to do the job.
- Audio compression works like a fast automatic volume control on the microphone audio. When the voice is loud, the gain is reduced; when it is soft, the gain is increased. Because the waveform is not cut off, a compressor adds little distortion, but its effect on average power is relatively modest.
- Audio clipping amplifies the audio and then cuts off, or clips, everything above a fixed level. Clipping raises the average power more strongly, but it creates harmonics of the audio, heard as distortion. A filter after the clipper removes the products above the speech range, but those that fall within it remain.
- RF clipping does the clipping after the audio has been turned into a sideband signal at the transmitter's intermediate frequency, and then passes the result through a sideband filter. Many of the distortion products fall outside the filter's passband and are removed, so RF clipping produces less audible distortion than audio clipping for the same degree of processing.
Most modern transceivers carry out processing with digital signal processing (DSP), combining compression, clipping and filtering in software. The front-panel control is usually labeled COMP or PROC, with a level setting, and many radios have a meter that shows the amount of compression in decibels.
Splatter and other risks
Processing is easy to overdo. An overprocessed signal sounds gravelly and fatiguing, and sounds in the room, such as a computer fan, an amplifier blower, traffic or a television, are raised along with the voice and fill every pause. When heavy processing is combined with high microphone gain and constant ALC action, the signal also produces splatter, distortion products that spread onto neighboring frequencies and interfere with other stations. FCC rules forbid emissions outside the necessary bandwidth that cause splatter interference on adjacent frequencies, so a splattering signal is the transmitting operator's problem to fix; see Harmonics and spurious emissions.
No setting replaces good microphone technique. Speaking across the microphone at a steady distance and a normal volume matters as much as any processor.
Setting a processor
- Start with the processor off. Set the output power, then adjust the microphone gain so the ALC meter moves on voice peaks within the range the manual recommends.
- Turn the processor on at a low level and increase it gradually. If the radio has a compression meter, the manual usually suggests a range. Recheck the ALC reading, since processing raises the average audio level.
- Listen to the result. Ask a station nearby for an honest report, or listen to the signal on a WebSDR some distance away. Transmitting into a dummy load while watching a separate receiver with a panadapter shows whether the signal stays within its normal width.
- Use it when it helps. Moderate processing is valuable in DX pileups, contests and poor conditions. For a strong local ragchew, a natural, unprocessed voice is more pleasant to listen to.
- Turn it off for data. Processing must be off for FT8 and other digital modes. Many radios bypass it automatically when a data mode is selected, but it is worth checking.