Great Audio Concepts Of Our Time (Part 2)
Bamboozled by technical specs? David Mellor explains how to 'interpret' audio equipment specifications and also gives the lowdown on that little understood concept, Barkhausen noise.
Legend has it that when the first electric gramophone was demonstrated, listeners had difficulty in distinguishing a recording from live sound. It's hard to believe nowadays, but the fact is that every time something new comes on to the market it is accompanied by the accolade of 'Pure Perfect Sound' or some other such catchphrase. If you ever look at the back page of an audio equipment brochure there is usually a specification of how 'perfect' the sound should be. In a court of law, there is a requirement to tell the truth and the whole truth. The Trade Descriptions Act prevents specs actually lying, but there are a number of ways in which they can mislead. Let's have a look at some of the parameters which describe the 'goodness value' of a sound system, taking in not only what they mean, but also how they can be misrepresented.
As you know, sound is caused by vibrations in the air. If I could wave my hand backwards and forwards fast enough then it would produce a sound, but how fast would I have to wave it? I'm not going to attempt a world speed record, but if I could wave it back and forth 20 times a second then it would begin to produce a sound that my ear could respond to. The phrase 'times per second' is normally replaced by the word 'Hertz', so we would call this a frequency of 20 Hertz (20Hz). Hertz, by the way, is the name of the chap who discovered radio waves and possibly also rented out horseless carriages in his spare time!
If 20Hz is the lowest frequency we can hear, what is the highest? Well if you are aged under 10 years, then you can probably hear a frequency of 20,000Hz - 20 kilohertz (20kHz). If you are getting on a bit, then your hearing will deteriorate to the point where you can only reach 10kHz or so on a good day. For a fit, healthy person in his or her prime, the frequency range of the ear is about 20Hz to 17kHz, give or take.
From these figures it would seem that audio equipment needs to have a frequency range equal to or greater than this to give adequate sound quality. 20Hz to 20kHz is normally enough. This is not quite the whole story, however. What we desire from any item of sound equipment is that it will respond to all frequencies equally, from lowest to highest. In other words, a 'flat frequency response'. With modern technology it's normally easy to get middle frequencies flat but the lowest and highest frequencies will always get lost to a certain extent. By how much they are lost is the important question. If an equipment specification shows a frequency range of 20Hz to 20kHz, then it conveys little useful information. Figure 1 shows a graph of frequency response, which is much more meaningful.
By convention, frequency response is usually quoted between the two frequencies where the output level drops by 3dB (decibels - see Part 1 last month). You know that when you input a particular level at a particular frequency, you will get a certain response. Some manufacturers cheat by quoting a frequency response between the -6dB limits. This makes the figures look better, but the important point is that whatever the limits chosen, they should always be quoted within the spec. A frequency response specification should look something like this:
20Hz to 20kHz +0dB/-3dB
or this; 20Hz to 20kHz +/-1dB
rather than this; 20Hz to 20kHz
The first two examples give precise information on how the response may be expected to vary, the third gives no hard information at all.
Not in this case the person who supplies James Bond's gadgetry, but a measure of the 'sharpness' of a filter or equaliser.
Every so often during the perusal of manufacturers' literature comes a mention of 'low Q' or 'high Q'. It would be possible to go on for hours explaining 'Q' in all its ramifications, but I propose to take only the simplest case where it is used to describe an EQ (equaliser) circuit. Figure 2 shows the response of two different equalisers, each giving a 6dB boost at a frequency of 1 kHz. The difference is in the sharpness of the response. The first would be used for an overall change in sound quality, the second possibly for picking out an individual instrument in a mix.
'Q', which by the way is just a letter and doesn't stand for anything, is a measure of how 'sharp' the peak in the response is. A 'Q' of 0.3 would be a very broad peak covering a wide band of frequencies. A 'Q' of 10 would be an extremely narrow band. It's a case of horses for courses - sometimes you need a 'low Q', sometimes a 'high Q', to get the effect you're after. Often you don't get the choice on mixer equalisation sections.
As we are all well aware, noise is the enemy of music. No-one escapes, even CD owners buy a certain quantity of unwanted sonic rubbish along with their £9.99 worth of musical delight.
Let's consider first of all the noise in mixers. That's the hissing sound you hear when you push all the faders up to full. If you can't hear it, then there must be something wrong with your speakers (or your ears!).
Think back to school physics and you will remember that an electric current is simply the flow of electrons through a conductor. The audio signal is represented by variations in this flow. Imagine a situation where the general rate of electrical flow is low and an audio signal is superimposed on this. In this case, the electrons themselves can be regarded as tiny variations in the flow and, when the signal finally gets converted into sound, individual electrons will make their contribution heard. If the general rate of electrical flow is high, then the contribution of each individual electron is less, and the noise is lower. Note that the audio signal level has no effect on this noise level. The noise level depends on the quantity of electrons that carry the signal.
Now that we can see that random variations in electricity cause noise in mixers, it's not a big jump to think of the individual magnetic particles in tape as a source of noise. Look at it like this - any audio signal is carried by a medium, random variations in the consistency of that medium, whether electricity, magnetism or grooves in a record, cause noise. Sad but true.
The next concept we have to think of is signal-to-noise ratio (S/N). This gives us a measure of how noisy a piece of sound equipment is. Suppose that I feed the equipment a signal which is as loud as it can cope with, and measure the level of that signal with my trusty voltmeter (calibrated in dBu, where 0dBu = 0.775 volts RMS). A typical professional mixer can cope with a maximum level of around +20dBu. This is 20 decibels above my standard reference level of 0dBu (see last issue for more about reference levels). Next, I remove the signal and without changing any settings measure the level of the noise. This would typically be around —80dBu, or 80 decibels below my reference level. By subtracting the two values (+20dB and -80dB), I can say that my mixer has a signal-to-noise ratio of 100dB. This is a simple way of measuring noise that works well in the case of purely electronic systems where the noise level is constant.
Sometimes, noise is not constant. A good example is the tape recorder, where in addition to a certain amount of signal-independent noise, there is a certain amount of modulation noise. This is noise that is only present when there is a signal present, and comes and goes with it. Last month I promised to tell you what Barkhausen noise is, and this is it. I found the term in an old reference manual and I've been trying to popularise it ever since! Noise reduction systems, while reducing the signal-independent noise level, do not always have a beneficial effect on modulation noise. Sometimes they can actually make it more objectionable. Let's see how you can measure the noise level of a tape recorder that uses noise reduction.
Method 1: This is exactly the same procedure as I used with the mixer. Measure the maximum signal level; eg. +10dBu. Remove the signal and measure the noise; eg. —70dBu. Subtracting the two gives a signal-to-noise ratio of 80dB. Looks good!
Method 2: Instead of removing the test signal, use notch filters to remove it (and any distortion components - more on this later) from the output of the tape recorder. This time, we are measuring the noise level that is present when a signal is present, not just during the gaps in the music. A typical result would be —50dBu. Assuming the same maximum signal level of +10dBu, the signal-to-noise ratio is thus reduced to 60dB.
If you were a manufacturer, which method would you choose - the one that gives 80dB signal-to-noise or the one that gives 60dB? Method 2 would be tough to carry out in practice but Method 1 certainly does not tell the whole story. I am reminded of an advert I saw for a 'HiFi sound track' video cassette recorder which claimed to have a better signal-to-noise ratio than a certain popular semi-professional reel-to-reel tape recorder. The video was found to perform better of course, but what was not mentioned was the powerful noise reduction system required for it to do this. If a noise reduction system is used during signal-to-noise measurements, then this should be clearly shown in the spec sheet. To leave out this important fact is simply lying.
That's one way of 'cooking' the figures. Another, more honest, method is to 'weight' the figures. What this means is that the measurements are taken in a way that mimics the way the ear would respond to the noise. The snag here is that there are several weighting methods and it is difficult to compare measurements from one weighting system to another. Look out for these in noise specs: 'unweighted' or '20Hz-20kHz band' both mean that an honest reading was taken. 'A-weighted' or 'CCIR-weighted' are more likely to mean 'these are both correct readings but we chose the one that gave the best result'. In this business, it's easy to get cynical.
When talking about signal-to-noise ratios, there is another term which often crops up - dynamic range. I'm not altogether happy about manufacturers who talk about a 'dynamic range of 100dB'. If the noise level of a piece of equipment is -80dBu, then it's possible for musical signals to lie below this level and still be quite audible. Does the manufacturer intend these signals to be included in his estimate of dynamic range? 'Signal-to-noise ratio' is a more precise term and I think we should stick to it. There is enough confusion.
Before I leave the topic of noise, I ought to discuss the concept of equivalent input noise or EIN. This often crops up when a manufacturer specifies the noise performance of the microphone inputs of his mixer. An example spec might be 'EIN at 70dB gain: —125dBu (200 ohm source)'. This means that the gain control on the mixer was set to 70dB and the noise measured at the output of the mic amp - in this case the measurement would be -55dBu. When the set amount of gain is subtracted from this we get the amount of noise that would have to be present at the input of a noiseless mic amp to give the same result. The '200 ohm source' bit is necessary to make the measurement meaningful. If the EIN figure does not give the source impedance, then I am afraid the measurement is useless. Sometimes a 600 ohm source is used which makes the figure look one dB or two less good. Perhaps it is giving the game away to say that the reason a gain of 70dB is quoted is because mic amps normally give their optimum EIN figures at a fairly high gain. The lower the gain at which a manufacturer dare quote his EIN figure, the better the mic input circuit.
Unfortunately, any item of sound equipment 'bends' or 'distorts' the sound waveform to a greater or lesser extent. This produces, from any input frequency, extra unwanted frequencies. Usually, distortion is measured as a percentage. For a mixer or an amplifier, anything less than 0.1% is normally considered quite adequate, although once again it's the poor old tape recorder that lets us down with distortion figures of anything up to 1% and above.
Distortion normally comes in two varieties: harmonic distortion and intermodulation distortion. Looking at the harmonic kind first, suppose you input a 1 kHz tone into a system. From the output you will get not only that 1 kHz tone but also a measure of 2kHz, 3kHz, 4kHz etc, etc. In fact, harmonic distortion always comes in integral multiples of the incoming frequency - a bit like musical harmonics in fact (hence the name). This is why distortion is sometimes desirable as an effect - it enhances musical qualities. Valve amps tend to emphasise the even-numbered harmonics; transistors and tape recorders beef up mainly the odd ones. Perhaps this is an explanation for the popularity of the 'valve' sound.
'Intermodulation distortion' is not so musical in its intent. This is where two frequencies combine together in such a way as to create extra frequencies which are not musically related. For instance, if you input two frequencies - 1,000Hz and 1,100Hz - then intermodulation will produce sum and difference frequencies of 2,100Hz and 100Hz. Unfortunately, both types of distortion tend to come hand in hand. It's a tough world, and usually a single distortion figure is adequate to cover both types.
While distortion is fairly well under control these days, crosstalk is becoming of increasing importance as manufacturers struggle to cut equipment costs. Crosstalk is easily defined as a leakage of signal from one signal path to another. For instance, if you have cymbals or hi-hat on one channel of your mixer and you find they are leaking through to the adjacent channel, then you have a crosstalk problem.
Crosstalk is usually worst at high frequencies - these find it relatively easy to jump across the gap between wires and circuit board tracks. Crosstalk is measured in decibels: typical channel-to-channel figures might be -80dB at 1 kHz, -60dB at 10kHz. Note that the frequency of the measurement should be mentioned, and the manufacturer who does not give the 10kHz figure probably has something to hide.
The most inconvenient manifestations of crosstalk are in connection with (you guessed it!) tape recorders. Multitrack recorders with narrow tape track widths, such as the Tascam and Fostex ranges, are most at risk. With a professional full-width machine, you should be able to bounce one track onto the adjacent track with no problems. Reduce the track width - and as a consequence, the 'guard band' between tracks - and crosstalk actually inside the tape head can create a feedback loop and you will get a nasty whistling sound on your monitors.
Timecode, such as SMPTE/EBU, can also be a problem because it's chock-full of troublesome high frequencies which just love to leak out and spoil your recording. My advice is to make sure that timecode never gets inside your mixer. It is usually possible to invent a wiring scheme which will avoid this.
Nothing to do with Max Headroom, who was last seen stuck under a low bridge. I have already mentioned the concept of operating level last month, which is the 'round about' preferred level in a studio. This can be 0dBu or -10dBu, or in fact any level you want it to be. Let's take 0dBu as an example.
If you were a BBC type, then your operating level would be 0dBu and your peak level +8dBu. This means that if you wanted to put a zero level tone on tape then you would do it at 0dBu. If you wanted to blow up Tony Blackburn's transmitter you would do it at a sniff over +8dBu!
In the studio, you would probably line up equipment with limited dynamic range, such as tape recorders and digital effects, so that they would overload just above +8dBu. Your mixer, however, would still have the ability to go up to +20dBu if required. This 12dB difference is your headroom.
Operating a mixer requires all sorts of juggling about with levels and gains so it is useful to have a margin of error. In fact, I could rephrase that - it is essential that you have a margin of error. Some domestic recording equipment makes economies in the power supply department with a negative effect on available headroom. I hope this will not become a trend, but when checking the spec of a mixer you should look for a maximum output level of +20dBu, or at least +10dBu if you work to the -10dBu standard.
Now you know something about most of the major concepts involved in sound recording, so the next time a problem arises in the studio you can say 'Aha! - sounds like we're running out of headroom!' or some other such meaningful phrase. Alternatively, when something goes wrong in a studio you have hired and the engineer offers an explanation which falls short of the standards of knowledge required, you'll know whether you would be justified in asking for some free time - or a refund.
Buying equipment should pose fewer problems as you will now be able to make comparisons between manufacturers' specifications — or at least see when they are not telling the whole story. Reputable manufacturers are not ashamed of their specs, they have something to boast about and want to tell the world. If the spec sheet you read looks as though it is trying to hide something, it may well be that there is something to hide. Watch out!
Feature by David Mellor
Previous article in this issue:
Next article in this issue:
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!