David Strange unlocks the mysteries of VU's and PPM's

Thumb through almost any specification sheet or piece of sales literature for a mixer or tape recorder and you will find lots of talk about frequency response, distortion and signal-to-noise ratio, but only a word or so concerning level meters. This is not surprising, you may say; after all, the meters don't have anything to do with the signal chain in the machine, so why be too worried?

It is true that the signal does not pass through the meters, but meters can affect the performance of any system as a whole. Most of the performance figures claimed for the system will have been measured under the most favourable conditions. Signal-to-noise ratio, for instance, will have been measured using every bit of dynamic range. Distortion, on the other hand, might be tested at lower levels, and level will also have a bearing on frequency response when level-dependent filters are involved in noise reduction systems.

Generally the meters used on equipment will be described as "Peak reading" or "VU" and even "Peak reading VU" or "PPM". With those descriptions the meters are glossed over and seemingly more important data is included with more detail. Meters, however, are very important as we shall see, before looking at particular meters in detail, it is a good idea to take a look at the studio signal itself in order to specify some important parameters that should be taken into account when meters are designed.

Why A Meter?

Fundamentally, a level meter is required in order that the best modulation level consistent with low distortion and low noise can be maintained. In any system whether it be tape, telephone line, radio transmitter or amplifier there is inherent noise consisting of hiss, mains hum or interference. The noise, if the system has been designed correctly, will be very low, but even so the signal must be kept high so that the noise is as unnoticeable as possible. The catch is that if the signal is allowed to become too large, although the signal-to-noise ratio may be satisfactory, distortion will onset due to clipping.

There are then two extremes between which the signal has to be steered to give a good technical quality, the noise at low level and the limits of amplifiers at high level (see Fig 1). There are also artistic reasons for being able to define level accurately and anyone who has tried to splice a piece of music to dialogue will understand how wrong it can sound if the relative levels are not correct.

Ballistics

Simply defining sound signals in terms of electrical level is not the whole story; take for instance the two sections of wave form in Fig 2. Displayed as they are it would be a mistake to say that the section on the left would be louder than the section on the right simply by virtue of its electrical amplitude. Once a signal reduces in duration to below 100 milliseconds the ear begins to make an assessment of its volume using time as well as level. In other words, the ear, at this point, integrates the sounds and makes its assessment of volume using the energy in the sound packet. A level meter therefore should be capable of reflecting the ear's perception of these 'pulsive' sounds in its response times.

Rectification

The electrical sound signal is obviously an AC waveform and requires rectification before being applied to a voltmeter. The waveform can be very asymmetric and the two halves of a cycle can vary in amplitude by 50%. To truly reflect the total amplitude of the signal therefore, full wave rectification is required. The rectifier should be capable of operating over the full range of audio frequencies with equal accuracy.

Decibels

Lastly the calibration units for the meter scale require definition and they should really reflect the ear's perception of volume. When the signal is steady and not pulsed, the relationship between the ear's perception and the real level is logarithmic (see Fig 3). For example, two level changes of 10mV to 100mV and then 100mV to 1000mV are interpreted by the ear as being equal, even though the actual range covered by the second change is much greater than the first. Put another way, the ear perceives as equal changes that are, in fact equal ratio changes (10:1 in each case).

Practical Meters

The next step is to look at practical meter systems and draw some conclusions. We shall take two meters, the VU and Peak Programme Meter.

The VU Meter

Virtually all VU meters (see Fig 4) sold are not VU meters at all but cheap frauds! This might sound a rather sweeping statement, but it's true, because to incorporate real VU meters in a piece of equipment is very expensive at £20 or £30 apiece. Clearly the expense of £60 in your average stereo tape machine is unjustified and so imitations are used.

The real VU, and we shall only talk about the real VU, was originally designed in the USA to make measurements on telephones lines. One of the problems with the VU is that it is relatively slow in its response, taking 0.3 seconds to reach 0VU when a signal is applied. This leads to a certain amount of interpretation having to be placed on the indication by the observer. In fact, the meter has to be observed for a fairly long period of time, 10 seconds or so, to be sure of the reading. There is however very little overshoot of the meter needle due to special care taken in the design to obtain damping.

The damping of the meter is dependent on the source impedance from which it is driven. This is due to the meter, as in all meter movements, generating its own back EMF as its coil moves within the pole-pieces of its own magnet. By careful selection of a series resistance, the current caused by the back EMF is used to damp the meter's accelerations. An external resistor of 3600 ohms is normally specified and the line impedance across which the meter is placed is usually 600 ohms.

The meter has an internal rectifier which is chosen to have a frequency response of plus or minus 0.2dB from 35Hz to 10kHz and plus or minus 0.5dB from 25Hz to 16kHz, and to be free from threshold effects at minimum indication.

The scaling of the meter is in decibels directly, but this does lead to some tradeoffs which are not immediately obvious. (See Fig 5. [This seems to have not been published.])

The scale is effectively divided into three parts, a red overload region, a middle region covering only 5dB and a very cramped region covering 15dB down to infinity. The overload region is presumably, if that is what it means, a no-go part of the scale and being so large is a waste of scale space. The central section of 5dB is also a wasted space because audio is never that consistent to remain in the area and fully justify the resolution of 1dB scale marks. Audio is however consistently in the lower third of the scale, and this is where resolution is lacking.

The non-linear scale also leads to a rather odd behaviour of the needle in response to signals. As the needle rises up the scale it has to cover more scale for less increase in volume and so there seems to be little correlation between what the eyes see and the ears perceive. The overall effect of this is a certain loss of visual feedback to the person monitoring the sound.

Conclusions

The main problem with the VU is that, with its relatively slow speed, it fails to correctly read sharp percussive sounds and speech. When, therefore, this type of material is allowed to modulate at 100% (0VU) it will distort. A more reasonable reproduction on sharp sounds is obtained when the indication is around 50% (-5dB).

Arguably then, the VU meter is not the best level indicating device, however, since it does not require an external amplifier it is relatively cheap, even at £30. There are of course light column versions that might be cheaper, but be careful that they have the right speed of response. Many light column VU scaled meters are in fact peak reading with rise times that are far too fast for any audio work.

The crucial specification to look for is the ANSI standard C16.5-R, the characteristics of which are listed below:

1. Full wave rectification must be used — this ensures that the true amplitude of the signal is registered in the indication.

2. The sensitivity of a free standing instrument, not incorporated into equipment should be 0VU indicated when a sine wave of 1.228V RMS (+4dB) at 1 kHz is applied.

3. The frequency response should not vary by more than plus or minus 0.2dB over the range 35Hz to 10kHz and not more than plus or minus 0.5dB over the range 25Hz to 16kHz from that given at 1kHz.

4. The needle should reach 0dB in 0.3 plus or minus 10% seconds when a 0dB signal in the range of 35Hz to 10kHz is suddenly applied. The over-swing of the needle should be less than 1.5 % past 0dB.

5. The return of the needle to rest should take 0.3 seconds when the signal is removed.

The PPM

The PPM, unlike the VU, does not rely entirely on the ballistics of the meter movement for its reaction times to a signal. Instead the PPM system uses a special amplifier which processes the signal before it is applied to the meter movement. A British Standards PPM meter and drive amplifier will cost somewhere around £65, and a stereo pair is definitely not for your average cassette recorder!

The PPM has its roots, like the VU, back in telecommunications, and its forerunner was designed in the nineteen-twenties or thirties by the GPO. Later it was picked up by the BBC and modified and developed over a number of years into its present form. It is now used, without exception, throughout the broadcasting industry studios in preference to any other meter.

The philosophy behind the PPM says that because the audio signal is not perfectly symmetrical sine waves, but a spiky and random wave form, anything other than a peak measurement is not a valid measurement. In practice any sounds of the same level will give the same indication on a PPM so long as they last longer than 100ms. This means that an interpretation of the reading with an ear to the type of sound is not required as with the VU meter.

Operation

A basic system diagram of the British Standard PPM is shown in Fig 6 where the signal is first applied to a perfect diode full wave rectifier to obtain a unidirectional voltage. The rectifier is usually configured using two operational amplifiers and Fig 8 shows a typical arrangement in detail. IC1 has unity gain during the positive half cycle and x2 gain in the negative half cycle of the input signal. The output of the first stage of the rectifier is recovered from the junction of R2 and D1 and differentially summed with the unprocessed input by IC2. The result is a 1:1 unidirectional version of the original input signal.

A transistor Q1 is used to obtain a low output resistance from the rectifier whilst D4 isolates the output from the time constant capacitor. Both Q1 and D4 are placed in the feedback path of IC2 to remove their forward volts drop. IC1 and IC2 are required to be high slew rate operational amplifiers so that a frequency response of plus or minus 0.3dB is obtained from 16Hz to 16kHz.

The next crucial part of the PPM is the time constant circuit Fig 7. Sounds of sufficient duration will cause C2 to be charged up to the voltage on the input side of R8, while shorter ones will be over before C2 has time to charge fully. The voltage across C2 is therefore a measure of the duration as well as amplitude of the incoming signal. In this way a response from the meter, called the attack time, is obtained which is analagous to the ear's response. The voltage on C2 is leaked away by R4 and the combination determines the recovery time. The recovery time is made quite long, as compared to the attack time, and this reduces eye fatigue when observations are being made.

The voltage from the time constant circuit is next applied through the output amplifier to the meter movement. The meter movement is calibrated in seven very nearly linear increments of 4dB each and the range covered is 24dB. A linear scale, rather than a logarithmic one, is possible because of the way in which the output amplifier processes the voltage.

Processing is accomplished by steadily reducing the gain of the output amplifier as the voltage input to it increases. This is done by progressively turning on transistors in the feedback loop (see Fig 9). The bases of each of the transistors are held at particular fixed voltages so that when the output from the amplifier is low they are biased off. As the output from the amplifier increases and the base-to-emitter voltage becomes sufficient, each transistor conducts and adds its emitter resistor to the amplifier's total feedback. The resultant piecewise reductions in gain produce an approximation to the logarithmic level response of the ear (see Fig 10)

Specification

BS 5428 and IEC 268 parts 10 and 10A.
Frequency Response: 30Hz to 16kHz + 0.3dB
Accuracy:

Dynamic Response: Attack is measured by applying 5kHz tone bursts of various specified periods at an amplitude of +8dB.

Decay is measured by recording the time taken for the signal to fall between Mark 7 and Mark 1 when the signal is removed. This should be 2.8 plus or minus 0.3 seconds.

Conclusions

The PPM is the "Rolls Royce" of meters; its responses are very similar to those of the human ear and so it provides an excellent feedback to the eyes. There are no problems of interpretation as with a VU meter and its indications may be taken as absolutely correct at all times. The degree to which levels may be matched is very high even when they are as dissimilar as speech and music.

Each meter amplifier is matched and aligned to its own meter movement during manufacture and this precision, allied with superb performance, enables confident handling of signals even close to the maximum headroom of equipment. When serious recording is undertaken it is most worthwhile to invest in Peak Programme Meters for level-monitoring.

Finally a word about matching PPM readings to those of VU's. Calibration of the PPM against an existing VU meter — in a tape machine say — is quite simple. A 1 kHz tone should be used to make the VU meter indicate 0VU and then simultaneously on the same signal, the PPM should be adjusted to read Mark 5 or +4dB. This calibration having been carried out, the VU meter should be ignored and the signals monitored on the PPM alone (signals should be allowed to peak up to +8dB or Mark 6).