Technical notes index
Principles of audio noise reduction
Principles of audio noise reduction
<<previous page       page 5     next page>>

            CONTENTS

1. INTRODUCTION
      Distortion
      Frequency response
      Pitch errors
      Noise

2. FIXED PITCH NOISE

3. IMPULSIVE NOISE

4. RANDOM NOISE (HISS)

5. HISS REDUCTION
      Examples
      Companders

6. NOISE REDUCTION EXAMPLES

Hiss reduction: Examples

Simple hiss such as tape hiss isn't usually at too high a level and is fairly easy to reduce, depending on the material - speech is easier than music, and music with quiet passages can sometimes be quite difficult to deal with. The object is to reduce the hiss as much as possible without having any significant effect on the original sound. Even removing a low level of hiss is worthwhile because the difference between a slight hiss and no audible hiss can be surprisingly effective. In the case of higher hiss levels, which can be quite distracting, even a reduction which still leaves some audible hiss is worth while.

The following example is an off-air tape recording accidentally made at rather a low level (not by me!), so that when the level is brought up there is a fair amount of hiss.


[tapehiss1.mp3]

Louder passages often mask the hiss, so that in effect the noise reduction program doesn't have to do anything. When setting up a process it's important to test on a quiet section, balancing the noise level against any adverse effects on the recording: generally aiming for a slightest possible effect on the original sound produces reasonable results, particularly if a direct comparison isn't going to be made. In this quieter passage it's still been possible to effect a considerable reduction in hiss with only a very slight effect on the piano and almost none on the percussion: the result is much easier to listen to.


[tapehiss2.mp3]

78s can benefit from the same technique, even with the very much higher noise level, though again it's very dependent on the material. The example used earlier to demonstrate the removal of crackle shows just what can be possible: even though the speech has been recorded at quite a low level it's responded very well to processing (music would have been more difficult and a higher level of noise would probably have to be accepted). This extract starts with the unprocessed recording and then changes to the version with the crackle and hiss removed.


[companions.mp3]

Of course you can't put back what wasn't there in the first place: this 1898 recording by Albert Chevalier has so much noise it's practicably impossible to listen to: however though the noise can be reduced spectacularly you still can't understand what he's singing because the detail was never there in the first place.


[chevalier1.mp3]

Companders (compresser-expanders)

Compressor-expanders are a special case (not relevant to normal audio processing) designed to compress the audio before recording and expand it on playback so as to minimize tape hiss: one might regard them as noise prevention rather than noise reduction. They were developed to deal specifically with tape hiss, initially in professional studios where Dolby A provided a very considerable reduction in hiss levels. Dolby B, the domestic version, became ubiquitous on cassette recorders - indeed without it the inherent hiss level would have prevented them from being much more than low quality devices for use in cars and personal stereos.

The simplest system is DBX, which applies 2:1 compression over most of the dynamic range: unfortunately this method can cause the hiss to be heard 'pumping'. Dolby B affects only the higher frequencies below a certain threshold, compressing - raising the level - as the original audio drops below this until there has been about a 10dB increase in the quietest levels. (The actual curve used is a little more complicated, but that's the basic principle). On playback the process is exactly reversed (requiring accurate level alignment to get the threshold in the right place) and reduces tape hiss by 10dB with remarkable effectiveness, having practicably no detectable effect on the actual audio; Dolby A used four bands and heavier compression, and Dolby C was a more advanced application of the same basic principle for domestic use. Later developments in the professional field, such as Dolby Spectral, work in a more complex fashion with more frequency bands, and give much greater noise reduction, but the basic principle is the same. Dolby S was developed as the domestic version of Dolby Spectral and was quite remarkably effective on cassette recordings; however the entire technique has largely been rendered obsolete by digital recording.

More examples

On the final page are two complete 78 sides to demonstrate the results of careful transfer and noise reduction.


<<previous page       page 5     next page>>

© Roger Wilmut.