On the principles of operation of noise suppressors. Encyclopedia of radio electronics and electrical engineering

Encyclopedia of radio electronics and electrical engineering / Audio equipment

 Comments on the article

The issues of noise reduction in magnetic recording devices still attract the attention of radio amateurs and users of audio equipment. This is facilitated by the widespread use of imported tape recorders equipped with various noise reduction systems.

The lack of information is partly due to the fact that the operating instructions of the equipment do not contain information about the features of the implementation and use of embedded systems. As a result, all sorts of speculations are circulating, but audio quality problems remain.

First, it is worth noting that the word "noise suppressor" refers to two fundamentally different types of systems: one of them is designed to remove the noise already present in the phonogram (in English Denoiser), and the other - to prevent the accumulation of noise during transmission or recording of signals (Noise Reductor) . This ambiguity often gives rise to confusion and misunderstandings, therefore, to designate systems of the first type in a professional environment, it is customary to use the English word - "donozer" to distinguish them from systems of the second type. Well-known denoisers are dynamic filters (DNL, DNR, HUSH, "Mayak"), the principle of which is based on the simultaneous reduction of gain and signal and noise in a separate part of the spectrum (usually HF), where the useful signal can be neglected. Their advantage is their suitability for working with any signal source, and a serious drawback is the inevitable loss of some information.

Now denoisers are mainly used to "clean up" old (or technically unsuccessful) records. They are rarely used in household equipment, usually only as an aid: after all, to obtain the optimal result, manual or automatic adjustment to a specific phonogram is required. The professional denoiser can be made as a stand-alone device (analogue or digital) or as a program for a computer. Let's take the NoNoise software package from Sonic Solutions as an example. You can get an idea of ​​his work by listening to "The Beatles Live at the BBC" discs.

The highest quality analog denoiser known to the author was made by Orban. This five-band dynamic filter has a unique ability to analyze both level and signal type, preventing reverb sounds and low-pitched high-frequency percussion from being eaten away. Systems of the second type (Dolby, dbx, High-Corn, Super-D, etc.) process the signal twice: the first time before recording or transmission, and the second time during reception or playback. Therefore, they are also called complementary, in contrast to the systems of the first type, which received the conditional name of non-complementary. Since the work of complementary systems is based on the use of a combination of a compressor and a dynamic range expander, they are often called companders or simply companders (COMpressor + exPANDER).

Companders generally provide more noise reduction and less distortion in the music signal than denoisers. However, they impose certain requirements on the channel of reception-transmission (or recording-playback) and, as a result, are more "capricious" in application.

The main, but not disputable, idea on which all noise reduction systems (UWBs) are based, and not just compander ones. is the assumption that. that noise impairs the perception of only weak signals, and with a strong signal (high volume) it is not audible due to the effect of masking a weak sound by a stronger one.

Following this logic, there is no need for the noise level to be constant both in the absence and in the presence of a useful signal. That is, an increase in the absolute noise level with increasing signal level is considered acceptable and imperceptible by ear. This assumption opens the way to the construction of compander systems, in which the gains of both halves (compressor and expander) change depending on the signal level.

In practice, this means that weak signals are amplified before they are fed into the transmission channel (for example, to a tape recorder), while a strong signal passes without changes (or even attenuates). This operation is called compression (compression) of the dynamic range. At the other end of the channel, an inverse conversion is performed, as a result of which the signal is brought to the original level range, and noise is reduced with a weak signal.

Obviously, when implementing such a system, the dynamic range, measured as the ratio of the maximum transmitted signal to noise in the absence of a signal, can significantly exceed the same ratio measured for the transmission channel itself. It is clear that it is the first digit (as large) that appears as the value of the dynamic range in the UWB characteristics.

However, it characterizes, rather, the range of acceptable levels of the input signal, while the signal-to-noise ratio in the presence of a signal (ie, the instantaneous signal-to-noise ratio) is determined primarily by the characteristics of the transmission channel itself. Without additional measures, such as special frequency equalization, the use of multiband systems, or a dynamic frequency response equalizer, the signal-to-noise ratio in the presence of a signal cannot exceed that of an unsquelched channel. Simply put, if the noise in the channel is heard even at the maximum signal level, there is no gain from using companding. As unfortunate as it may seem, this is exactly what happens in most cases. It is connected with the fact that the widespread assumption that any loud sound makes inaudible (masks) any weak sounds, including noise, is generally not true.

Experts in psychoacoustics (the science of human perception of sounds) established the fact many decades ago that the masking phenomenon operates only in a limited frequency range, mainly near the frequency of a loud (masking) signal. This is most clearly reflected by the so-called "masking curves" (Fig. 1, 2), from which, in particular, it follows that in the presence of narrow-band sounds with a loudness of up to 90 ... 95 phon2, human hearing at a number of frequencies is still able to distinguish sounds , which are near the threshold of hearing in the absence of a masking signal. And only an increase in volume above about 95 phon leads to a reflex decrease in sensitivity, protecting the ear from damage.

Rice. 1. Pure tone masking curves by narrowband noises with center frequencies 0,25; 1 and 4 kHz

Rice. 2 Control tone masking curves (depending on its frequency and level) by another pure tone with a frequency of 1,2 kHz and a sound pressure level of 80 dB. From Licklider (1951), Fletcher (1929), Wegel and Lane. The interference frequency is shown in red, the masking curve in green. A tone below this curve is not heard. A test tone with level and frequency falling within the shaded areas creates beats; if this tone has a level above that outlined by the blue dashed line, then a mixture of several tones (intermodulation products in the ears) will be heard. When the level of the test tone is in the range from yellow to blue line, noise, test tone and difference tone are heard, and if the level lies in the range from green line to yellow - noise and tone of the difference frequency. To the left of the green line, noise and test tone are heard separately.

Thus, the human ear has a kind of dynamic range compressor that allows it to work with signals in the dynamic range of about 130 dB, with a simultaneously perceived (instantaneous) dynamic range of about 90 dB. Therefore, if in the presence of a signal, noise and distortion do not exceed the absolute threshold of audibility or - 90 dB relative to the maximum signal level (taking into account uneven hearing sensitivity), then neither noise nor distortion will be heard under any conditions (and signals). However, these conditions are not provided even by most amplifiers, not to mention tape recorders.

Therefore, another approach is more realistic: it is necessary to take measures to ensure that when playing various signals, the spectra of noise and distortion products of the sound transmission system would go as far as possible below the masking curves of these signals. In particular, for intermodulation distortion products, this means that the formation of difference tones during processing of high-frequency signals, as well as sum tones from low-frequency signals, is highly undesirable. At the same time, harmonic distortion of fundamental tones may well be -50 dB and go unnoticed.

As for noises, the nature of their perception is different from that of "organized" sounds. The ability of human hearing to perceive noise depends on the spectrum and the rate of change of the useful signal, and the permissible signal-to-noise ratio in the presence of a signal with a level of 85 ... 95 dB (relative to the hearing threshold) is from 40 ... signal, up to about 45...75 dB for pure tones, especially at the edges of the audio frequency band. On average, it is 85 ... 50 dB.

Based on this, we can say that in magnetic recording compander noise suppressors in most cases work "on the verge of a foul". Even if the compressor and expander are perfectly matched, if the record-playback channel has a signal-to-noise ratio in the presence of a maximum signal of less than 80 dB, there may be situations when noise will still be heard.

The relative noise level in the recording-playback channels of analog tape recorders, even in the absence of a signal, as a rule, does not reach -80 dB. Appearing in the descriptions of some household tape recorders (for example, Tandberg SE-20), this value was achieved through the use of non-standard frequency correction, but with a loss of overload capacity at higher frequencies.

Moreover, in the presence of a signal, the noise level in an analog tape recorder always increases, amounting to a value from -35 to -60 dB at the nominal signal level. This increased noise is caused by the presence of a signal and is roughly proportional to the signal strength. That is why it is called modulation noise.

When recording a pure tone with a nominal level, the modulation noise spectrum on a good-quality tape recorder consists of two components: relatively narrow sidebands caused by parasitic amplitude and frequency modulation of the recorded signal, and broadband noise exceeding the pause noise level by 10 ... 25 dB, depending on signal frequency and tape quality.

The sidebands, unless their total level does not exceed -40...-46 dB, with their small width (less than 5...8% of the central frequency), are almost never audible, since they are under the corresponding masking curve (Fig. 3a and 3b).

Rice. 3a. Modulation noise in a cassette recorder at nominal speed and 70 µs correction. The black line is the frequency dependence of the hearing threshold of one-third octave noise, the blue line is the frequency dependence of the one-third octave noise of a pause (*0 dB "on tape, i.e. 250 mWb / m corresponds to a level of 60 dB sound pressure), the red line is the signal spectrum of 315 Hz with 0 dB level and the green line is the masking curve As can be seen, at frequencies above 2 kHz, there is no masking and the increase in high-frequency noise in the presence of a low-frequency signal becomes clearly audible.

Rice. 3b. The same, but the signal frequency is 3150 Hz. It can be seen that the low-frequency part of the broadband component of the modulation noise, which is not masked by the useful signal, is the most audible. This is the main reason for the "turbidity" and "dirt" in the sound of cymbals, bells, as well as high notes of violins.

The broadband component, when playing pure tones, is quite often heard (in the form of “dirty” sound) even on a studio master tape recorder, since its overall level is rarely below -50 dB relative to the signal level. Unfortunately, there are only two ways to reduce the level of the wideband modulation noise component: improving the quality of the tapes and increasing the width of the recording tracks (each doubling gives a gain of only 3 dB).

Modulation noise causes a lot of trouble: every blow on the keys of the piano is accompanied by crackling, as if they were laid with paper, the bass pipes of the organ are very hissing, string instruments begin to resemble wind instruments, “sand is pouring” from the cymbals, etc. By the way, the main reason for audible differences when When using different types of magnetic tapes, it is precisely the difference in the amount of intermodulation distortion and in the level (as well as the frequency dependence) of the modulation noise.

The only way to reduce the visibility of broadband noise growth in the presence of a signal - the so-called "breathing" (breathing) or "pumping" (pumping) - is to introduce frequency equalization into the recorded signal so that the inverse frequency equalization during playback attenuates the parts of the noise spectrum that are not masked by the wanted signal (Fig. 4).

Rice. 4a. The degree of dependence of modulation noise can be reduced if, during playback, attenuate those parts of the spectrum where it is not masked by the signal. So for low-frequency signals (up to 400 Hz), the role of such an adaptive filter can be performed by the Dolby-B noise suppressor expander. Its frequency response is shown at the top of the figure, and the resulting signal spectrum is shown as an orange line.

Rice. 4b. The same, but the useful signal frequency is 3150 Hz. It can be seen that the required frequency response of the filter, which reduces the visibility of the modulation noise, turns out to be completely different - with a "step".

This frequency correction can be done in several ways. The first and most obvious is the division of the signal spectrum into separate bands, each of which has its own compander. Due to this, the presence of a strong signal in one of the bands does not lead to the appearance of noise in the others. It has long been established that to ensure an acceptable quality of operation of such a system, four to seven bands are required, which greatly complicates the design of the noise suppressor and makes its operation critical to the accuracy of the frequency response of the recording-playback channel. Thus, four-band Dolby-A built according to this principle requires tuning the frequency response of the tape recorder with an error of no more than ± 0,3 ... 0,5 dB.

The second, simpler method is to use a fixed frequency equalization network, chosen in such a way that for most signals a frequency response is provided that is close to optimal for suppressing wideband noise. The quality of the operation of such a system depends very much on the competent selection of correction characteristics. A similar approach is used in most broadband companders (High Com, ADRS, dbx, etc.). Unfortunately, the correction characteristics in the wideband companders known to the author are far from optimal.

The third way is to use companders with adaptive frequency response, which automatically adapt to the spectrum of the input signal. This approach (in combination with a chain with a fixed frequency response) is implemented in the Dolby-S / SR system. The nature of the change in the frequency characteristics of the compressor is shown in fig. 5. A system with an adaptive frequency response, as a rule, perfectly processes single pure tones and monophonic instruments, but, alas, the possibilities of adaptation are limited on a real signal. So, in the Dolby-S / SR system, in the presence of wideband signals, the “stretching” of the mid frequencies during recording stops. During playback, this leads to "breakthroughs" of noise and distortion in the frequency range from about 500 ... 800 Hz to 2 ... 4 kHz ("unnatural middle").

Rice. Fig. 5. Frequency response family of the Dolby-SR encoder at different signal levels at 200 Hz (brown lines) and at 3 kHz (blue lines). The use of inverse frequency characteristics during playback can significantly reduce the visibility of both conventional, tech and modulation noise.

Of course, combinations of these methods are also possible.

All the methods discussed above assume that the time and level characteristics of the compressor and expander are the same, and the recording-playback channel does not distort the structure of the signals. In practice, you can't count on this, so tracking errors inevitably occur in compander systems. Their influence on the final signal strongly depends on the structure of the system, but mainly comes down to two points: to the distortion of the processes of rise and fall of sounds, which changes their timbre, and to the appearance of operation noise (clicks and pops).

The main reason for the appearance of clicks and pops is, for example, the following fact. When the compressor reacts quickly to a signal level jump (for example, when clapping your hands), all frequencies in the band processed by the compressor are attenuated simultaneously. Due to phase shifts, components of different frequencies arrive at the expander with a time delay, but are processed simultaneously. As a result, impulse errors appear in the output signal and, accordingly, actuation clicks (see Fig. 6a and 6b).

Rice. 6a. Influence of phase distortion in the compressed signal transmission channel, From top to bottom: input signal waveform (damped mid-frequency oscillation superimposed on a weaker low-frequency signal), compressor output signal waveform, phase-distorted signal waveform, expanded signal waveform. A large error and interference are visible.

Rice. 6b. "Eating" the attack of a musical signal during expansion due to phase distortions of the compressed signal (sequence of oscillograms, as in the previous figure). A blowout is also visible when the compressor is activated.

As for errors in signal level, most often they arise due to errors in the frequency response or the transmission coefficient of the recording-playback channel. Another cause of errors is the parasitic amplitude modulation of the signal in the recording-playback channel.

And finally, at low signal levels, the penetration of various interferences into the control circuit of a compressor or expander is a problem. To reduce the penetration of radio-frequency (and infra-low-frequency) interference, the compander inputs must have band-pass filters that cut off signals with frequencies outside the audio frequency band. The absence of such a filter often leads to the inoperability of the noise suppressor in real conditions.

It is precisely because of the circumstances listed above that the sound of a tape recorder equipped with any of the well-known companders will not be free from problems. Unfortunately, the perfect (or nearly flawless) compander denoiser does not exist today. Moreover, in connection with the development of digital technologies, the main attention of UWB developers is turned to the creation of denoisers. However, work to improve the companders is ongoing at the present time. Successful developments include, for example, a compander in the audio channel of a VHS-HiFi video recording system.

Nevertheless, Dolby-B / C is still used in mass cassette recorders, less often Dolby-S or dbx. Therefore, each time, before pressing the button, it is worth considering whether it is necessary to use this compander for this recording? And if the original recording on a CD is of average quality, and a tape recorder.

Author: S. Ageev, Moscow

See other articles Section Audio equipment.

Read and write useful comments on this article.

<< Back

Latest news of science and technology, new electronics:

Machine for thinning flowers in gardens 02.05.2024

In modern agriculture, technological progress is developing aimed at increasing the efficiency of plant care processes. The innovative Florix flower thinning machine was presented in Italy, designed to optimize the harvesting stage. This tool is equipped with mobile arms, allowing it to be easily adapted to the needs of the garden. The operator can adjust the speed of the thin wires by controlling them from the tractor cab using a joystick. This approach significantly increases the efficiency of the flower thinning process, providing the possibility of individual adjustment to the specific conditions of the garden, as well as the variety and type of fruit grown in it. After testing the Florix machine for two years on various types of fruit, the results were very encouraging. Farmers such as Filiberto Montanari, who has used a Florix machine for several years, have reported a significant reduction in the time and labor required to thin flowers. ... >>

Advanced Infrared Microscope 02.05.2024

Microscopes play an important role in scientific research, allowing scientists to delve into structures and processes invisible to the eye. However, various microscopy methods have their limitations, and among them was the limitation of resolution when using the infrared range. But the latest achievements of Japanese researchers from the University of Tokyo open up new prospects for studying the microworld. Scientists from the University of Tokyo have unveiled a new microscope that will revolutionize the capabilities of infrared microscopy. This advanced instrument allows you to see the internal structures of living bacteria with amazing clarity on the nanometer scale. Typically, mid-infrared microscopes are limited by low resolution, but the latest development from Japanese researchers overcomes these limitations. According to scientists, the developed microscope allows creating images with a resolution of up to 120 nanometers, which is 30 times higher than the resolution of traditional microscopes. ... >>

Air trap for insects 01.05.2024

Agriculture is one of the key sectors of the economy, and pest control is an integral part of this process. A team of scientists from the Indian Council of Agricultural Research-Central Potato Research Institute (ICAR-CPRI), Shimla, has come up with an innovative solution to this problem - a wind-powered insect air trap. This device addresses the shortcomings of traditional pest control methods by providing real-time insect population data. The trap is powered entirely by wind energy, making it an environmentally friendly solution that requires no power. Its unique design allows monitoring of both harmful and beneficial insects, providing a complete overview of the population in any agricultural area. “By assessing target pests at the right time, we can take necessary measures to control both pests and diseases,” says Kapil ... >>

Random news from the Archive Sweet malaria vaccine 08.11.2002 All attempts to create a vaccine against malaria end in failure, as the causative agent of the disease hides inside human red blood cells and liver cells, where the immune system cannot reach it. Therefore, Australian doctors decided to create a vaccine not against malarial plasmodium, but against the toxin that it produces. It turned out that this poison is a complex sugar. Australian researchers were able to synthesize a similar molecule in the laboratory, but without the toxic properties. After its introduction to mice, the immune system of animals begins to produce antibodies to the malarial toxin, neutralizing it. After that, the body copes with the pathogen, devoid of the main weapon. But before the appearance of such a vaccine for humans, it will take another ten years.

Other interesting news:

▪ 30" HDTV LCD Panel

▪ big data transfer record

▪ Cold for fast heating

▪ The world's oceans accumulate mercury and release it into the atmosphere

▪ Diving breathing tube

News feed of science and technology, new electronics

Interesting materials of the Free Technical Library:

▪ section of the site Children's scientific laboratory. Article selection

▪ article It is good where we are not. Popular expression

▪ article What is gum? Detailed answer

▪ article Canoe, like the Indians. Travel Tips

▪ article Imitation pearls. Simple recipes and tips

▪ article Penetrating ball. Focus Secret

Leave your comment on this article:

All languages ​​of this page

Arabic	Hebrew	Polish
Bengali	Hindi	Portuguese
Chinese	Italian	Spanish
English	Japanese	Turkish
French	Korean	Ukrainian
German	Malay	Vietnamese

Home page | Library | Articles | Website map | Site Reviews

www.diagram.com.ua
2000-2024