Monday, September 17, 2012

The Science of Music -- Part Seven

Noise! Who wants to retire to the country? No traffic, no neighbors, no noise!! I always wanted a place in the country. It was so I could make some noise. I remember a time, over thirty years ago. We had a garage band -- a real garage band -- we were playing in the garage. We had the garage door open to let in fresh air. We were hitting our highs and hitting our lows, and the drummer was providing the transients. I looked up and saw my neighbor’s porch light -- the neighbor across the street. The light was going on and off and on and off. It was about 4:00 in the afternoon on a Saturday. I don’t think anyone was sleeping. But I think my neighbor was signaling that he (or she) would like us to turn it down ... or just quit altogether.

I don’t suppose we sounded that good. Like a lot of bands with more equipment than talent, we were much louder than we were melodious -- at least yours truly. We had the amps turned up to be louder than the drums, and the drummer was bangin’ hard to be louder than the guitars. No-one could hear the singers -- P.A.’s are usually less capable than guitar amps.

I guess our music was just noise to the neighbor. That’s when I decided I needed to live in the country ... way out in the country. Never happened. We just closed the garage door, cut down the volume, drank more beer, and that was the end of that ... and the band, as far as that is concerned.

Noise -- one man’s treasure is another man’s garbage. One man’s meat is another’s poison. Well, in music and amplifiers and recording, there is noise. It can be the sound of a big truck driving by during the capture of a great vocal, or it can be added by the equipment. That’s what this is all about: Noise.

Noise, in general, refers to unwanted sound, often loud, like that noisy person at the restaurant table next to yours talking on their cell phone. In audio systems, noise is the low-level hiss or buzz that intrudes on quiet passages that is usually the problem. All recordings will contain some background noise that was picked up by microphones, such as the rumble of air conditioning, or the shuffling of an audience, but in addition to that, every piece of equipment which the recorded signal subsequently passes through will add a certain amount of electronic noise, which ideally should be so low as to contribute insignificantly to what is heard. These internal noises generated by the equipment can include hum, buzz, and hiss.

In addition, there can be “interference,” which are signals from other sources being amplified. These could be radio or TV or other sources of electronic radiation. Hum could be classified as interference since it’s source is from an outside agency: the power lines, which could include the fluorescent lamps in the room or the power supply in the amplifier itself.

Hopefully, the noise in the audio system is relatively low level, and, for that reason, it tends to cause the most problems when listening to quiet or “soft” passages of  music.

Many classification of noise are based on the frequency of the noise. For example, hum and buzz refer to low frequency noise, while hiss is mostly high frequencies.

Measurements of noise can be made by comparing the power of the desired signals to the power of the noise. Power is related to voltage; it is proportional to the square of the voltage or signal amplitude. The most common and useful measurement of noise is a ratio between the signal and the noise called, quite naturally, the “signal to noise ratio” or S/N. It is calculated by the Power(Signal) / Power(Noise) or [(Amplitude(Signal) / Amplitude(Noise)] quantity squared where amplitude is in volts.

The S/N is usually given in decibels or db and the formula is 20 log (base 10) (A signal / A noise), but enough of that math stuff. The higher the signal to noise ratio, the better. If the amplitude of the signal is small (quiet), then the S/N will be worse -- assuming noise is constant which is typically true.

Electronic noise exists in all circuits and devices as a result of thermal noise, referred to as Johnson Noise. This noise is caused by random variations in current or voltage created by the random movement of charge carriers (usually electrons) carrying the current as they are jolted around by thermal energy. Thermal noise can be reduced by reducing the temperature of the circuit. Noise limits the minimum signal level that any radio receiver can usefully respond to, because there will always be a small but significant amount of thermal noise arising in its input circuits. This is why radio telescopes, which search for very low levels of signal from space, use low noise "front-end" amplifier circuits cooled with liquid nitrogen.

Home satellite television systems place a special, low-noise amplifier right in the dish antenna, but no liquid nitrogen! As satellite signals have increased in amplitude, the dishes got smaller and the low noise preamp was less critical. Still Dish Network and Direct TV locate a preamp in the antenna before sending the signal down the long cable to the set top box, thereby improving the signal to noise ratio.

Since noise generated in the first stage of any amplifier system is amplified by all the following stages, reducing noise requires the first amplifier circuit contribute the minimum amount of noise. This first stage is often called the “preamp.” Long cables from a weak signal source such as a microphone or phono cartridge can contribute noise and interference. A good solution is to install a preamp very near the signal source. Then the noise added by the long cable to the amplifier won’t affect the "S/N" ratio as much because the preamp boosted the "S." Some modern microphones have built in preamps exactly for this reason. I've mentioned before that some turntables have a low noise preamp with RIAA equalization. That way the cables from the turntable to the amplifier can be longer without increasing the total S/N. Otherwise, keep the cable from turntable to amplifier three feet long or even shorter.

There are several other sources of noise in electronic circuits such as shot noise, seen in very low-level signals where the finite number of energy-carrying particles becomes significant, or flicker noise (1/f noise) in semiconductor devices.

There is a special type of noise used to measure and adjust audio systems. It is called “white noise.” White noise is a random signal with a flat power spectral density. In other words, the signal contains equal power within a fixed bandwidth at any center frequency.

White noise draws its name from white light in which the power spectral density of the light is distributed over the visible band in a similar manner. White noise is a statistical phenomenon, since it is random. But it follows statistical expectations. White noise has the quality of being independent and identically distributed. If the white noise follows certain characteristics and values for mean and standard deviation, including being normally distributed, it is called a Gaussian white noise.

Oh wait, I said no more math. OK, white noise is just noise that is “equal power” at all audio frequency bands. There are white noise generators which generate this special signal. It is quite useful for testing and adjusting audio equipment since it can be considered a signal with, basically, all the audio frequencies at once. Recall that graphic or multi-band equalizers worked with individual frequency bands, so white noise is good for adjusting the equalizer settings -- either full octave or one-third octave. To see how that is done, we’re going to enter the “frequency domain.”

Up until now we’ve viewed waveforms in what is called the “time domain.” Our charts and graphs (or oscilloscope screens) showed time along the x-axis and amplitude on the y-axis. We saw waveforms as cycles of crests and troughs. We’ve spoken about converting these periods (since the measure is time) or equivalent wavelengths into component frequencies using Fourier analysis. But it is also possible to display the frequencies on a chart with an x-axis of frequencies, say from 20 Hz to 20 kHz (or higher). The y-axis remains amplitude.

There are instruments, called spectrum analyzers, which are similar to oscilloscopes, only they show the frequency domain instead of the time domain. A pure sine wave of 1 kHz would show up on a frequency domain chart or a spectrum analyzer as a single spike at the 1 kHz point. A square wave would show 1 kHz and 3 kHz and 5 kHz and 7 kHz, etc. Each harmonic spike would be smaller.

Square Wave Spectrum

White noise on a spectrum analyzer shows a constant set of frequencies all across the band.

White Noise Spectrum

You can use a white noise source, a graphic equalizer, a calibrated microphone, and an audio frequency spectrum analyzer to adjust a sound system to compensate for auditorium acoustics.  I’ve done this often. The spectrum analyzer I use has a built in white noise generator. I connect the generator output to the mixer board and P.A. system at the church. I put a microphone in the center of the auditorium and connect it to the input of the analyzer. With the white noise playing through the speakers and captured by the special, very flat-frequency response microphone, the white noise shows on the spectrum analyzer. Peaks in the analyzer display represent room resonances increasing the amplitude, and valleys are areas of frequency absorption or frequency distortion.

I then adjust the graphic analyzer to attenuate the peaks and boost the valleys. After that tune up has been performed, you can turn the P.A. volume up much higher without any feedback or squeal and the sound is clearer and less muddy. Our church also has delay lines to synchronize the phase of the speakers along the wall. I set them by using pulses fed to the speakers and viewed through the microphone on an oscilloscope. A tune up of a sound system in a large auditorium, or in your music room, can do wonders for sound clarity and reducing feedback. It can also compensate for the furnishings in the listening room.

In addition to white noise, there is also something called “pink noise.” Pink noise or 1/f noise (sometimes also called flicker noise) is a signal with a frequency spectrum such that the power spectral density (energy or power per Hz) is inversely proportional to the frequency -- it decreases as frequency increases. In pink noise, each octave carries an equal amount of noise power. The name arises from being intermediate between white noise (1/ƒ0) and red noise (1/ƒ2) which is commonly known as Brownian noise.

Most hiss is actually Johnson noise from the thermal activity in the electronics, but, as explained in a previous article, it can also come from irregularities and dust in record grooves or tape hiss from the magnetic particles on a tape.

Noise problems can be made worse if the audio system has impedance mismatches. Certain cables, called low impedance cables, contribute less noise and interference than high impedance cables. But the cable impedance must match the source impedance. Impedance matching transformers can be used to match the connection to improve noise performance. Also, cable shielding and even the grounding of the equipment can all be selected in such a way as to reduce noise. The best cables have two, balanced conductors (twisted pair) surrounded by a third electrical shield. This is sometimes called twin-ax as opposed to co-ax. The characteristic impedance of this microphone cable is 300 ohms.

Since noise causes the greatest listening problems with low level signals, many systems of noise reduction compensate by increasing the amplitude of soft signals. The trick is to increase this amplitude only for weak or low signals, and don’t increase the loud signals during recording. Then, in playback, if the system recognizes low level signals, it can attenuate them, thereby attenuating the noise. With loud signals, no change is made. This is also called "compression / expansion" and the expansion actually lowers the noise level. It is similar to the pre-emphasis in the RIAA equalization, but much more sophisticated and complicated.

This is really a nonlinear amplifier design and special care must be taken to avoid introduction of distortion. The Dolby and dbx brands are most recognized for this dynamic noise control, and they were in very common use in high quality tape recording systems. This is because magnetic recording tape introduced hiss due to magnetic issues on the tape surface. These noise reduction techniques can also be used with FM radio broadcasts.

Ray Dolby originally developed what is called “Dolby A” for professional use in studios and professional recording. The “Dolby B” system was later developed in conjunction with Henry Kloss for use with home equipment. Dolby B, in my opinion, made cassette tapes a viable "reasonably high" fidelity recording medium.

There are techniques to reduce noise from open microphones, called noise gates. They basically shut off the microphone if someone isn't actually singing into it. Muting unused inputs and proper adjustment of all gain controls in an amplification chain is also an important techniques to both reduce noise and to reduce distortion from over driven inputs. Just some miscellaneous examples of sophisticated solutions to the noise problem. Best advice: treat noise at its source, don't try to mask it -- unless that's the last option available. Also remember, you can improve the S/N by either raising the "S" or lowering the "N" or both.

Modern digital systems allow very powerful methods to reduce noise. With the recording software I use (Pro Tools and Adobe Audition), I can sample some output without any signal to create a noise profile. That recording is "pure noise." I then use it as a noise sample to my software which analyzes the noise digitally and then uses this analysis results to reduce noise in the signal section of the recording. It is an amazing process, and I once used it to recover an interview that, after doing the recording, I learned that the audio had very bad artifacts and problems that can be called noise. After using the software programmed with a sample of the sound with no speech, the algorithms produced a marvelous result, and I was able to use the badly recorded interview for a web program. I consider the result nothing less than a miracle!

Noise can also be reduced by filtering. For example, there can be possible interference from the 19 kHz sub carrier used in FM stereo broadcasts. It can modulate with the AC bias used in tape players and produce audible tones. High quality tape recorders often contained a low-pass filter to remove the FM sub carrier before recording. Since FM only contains audio frequencies up to 15 kHz, it does not reduce the signal to filter out all frequencies above 15 kHz, but it will reduce noise -- or, in this case, a frequency that can interact and produce noise.

A signal with a lot of high frequency noise or hiss might sound better with a low-pass filter set at 10 kHz or even 5 kHz. The music would have less fidelity, but possibly still sound better without the high frequency noise. There are many computer programs available designed specifically to remove the clicks and pops from records. The repetitive nature of the noise from a bad scratch on a record can be removed. It may be that a momentary gap in all the sound would be less of a problem that a loud pop or click from a scratch. I've done well using a DAW (Digital Audio Workstation) to find a sudden noisy peak in the music and just cancel it out with a short drop in the gain for the few milliseconds the noise spike occurs.

Other types of noise and interference can be removed with smart use of equipment. I used to live by an Air Force base in Great Falls, Montana. The nearby, and very high power radar would make a buzz in all the electronic gear each time the antenna turned in our direction. Better grounding of equipment chassis and shielded cables would have fixed this problem, but the simple (and cheap) equipment I used back then (I was only 19) did not provide adequate shielding. My audio amplifier was very inexpensive and used high impedance microphone cables which were more sensitive to the noise. Sometimes it just costs more to do it right. Or, as Judge Milian says, "Sometimes the cheap comes out expensive."

Troubleshooting noise problems can be quite a detective tale. In the case of the Air Force base, it didn’t take Sherlock Holmes to figure out the source of the noise, but I couldn’t afford the fix. Today, I’m better equipped to eliminate the noise.

One audio engineer I knew back in the 70’s told me there was no noise problem he couldn’t solve with the proper used of the D-filter. I asked what the D-filter was. He said "Dinero." Anyone want to fund the construction of my new studio. It will have a sand floated floor to reduce ground vibration and a Faraday shield to prevent radio interference and sound proofing of all the walls and ceiling with acoustic treatments covering the walls. I won’t have to move to the country, I’ll just need the Dinero.

Originally written on Feb. 21, 2012 during a visit to my Dad's home in Hillsboro, Oregon and posted on Facebook. During my two week visit with my dad, I wrote an article a day. I started with a long series on the Science of Photography which had thirteen individual articles. I then started this series on the Science of Music. It isn't finished and I have a lot more to say. I hope to add to this series in the future.

No comments:

Post a Comment