Practical examples from practitioners

Richard Tyler and Andrew Jellyman are both practitioners working in the field of acoustics. Richard Tyler works at AVI Ltd., and was consulted for the following section on Multi-channel systems. Andrew Jellyman works for Birmingham City Council and contributed the section on Tape recorders in acoustics.

Multi-channel systems

One of the main concerns is how and what to test on a multi-channel system.

Choosing which part of a multi-channel system requires testing is dependent upon both the intended purpose of the instrument and the method of operation. Some multi-channel systems use a single analogue-to-digital-converter (ADC), and multiplex between each input channel, others might consist of many independent channels.

In the case of a true multi-channel system, each channel would need to be tested. In the case of a multiplexed system it may be sufficient to test individual sensors and only one channel.

Often information on the inner workings is unknown to the calibration laboratory and the user, in which case all channels would need to be tested.

For a complete multi-channel system, checks are required on every microphone, every channel, linearity, weighting networks, L_eq integration and filters:

• Microphones are checked for their range, frequency response and linearity
• Testing of each channel requires implementation of BS-7580-1:1997 'Specification for the verification of sound level meters.'
• Weighting requires a check at the centre of the weighting network
• Integration requires a check on timing
• Filters require checks following guidance in IEC-61260

Filters

IEC 61260 'Electroacoustics - Octave-band and fractional-octave-band filters' specifies filters typically used in acoustical measurement systems. This standard (currently under revision) includes some guidance on testing, but there are no standard tests available.

 

It may be sufficient to perform tests using pure tones at the centre frequency of the filter bands. The bands chosen for testing would cover the operative range of the instrument, so that the 'skirts' of the filters are picked up in lower or higher filters. The test bands would therefore typically be 1 kHz and then at 2 further 1/3-octave bands (high and low frequency) on the outer edges of octave bands toward the limits of the instrument range.

Mix and matching

Mixing elements of systems is to be discouraged. A simple example is a sound level meter, used with a microphone not supplied with the sound level meter, and perhaps external recording equipment.

Issues arising include: mixing calibrators with different microphones where the correction factors are unknown; level compatibility problems; forward-feeding of signals (has it been filtered/amplified/attenuated); using microphones of different nominal sensitivity.

---

Tape recorders in acoustics - Good Practice

A S Jellyman, 21st March 2006

The example I am going to use is that of a 'nuisance' recorder commonly used by local authorities in the investigation of alleged noise nuisance.

These systems are typically produced by sound level meter manufacturers and comprise a sound level meter connected to a Digital Audio Recorder (DAT) or sometimes a Mini-Disc recorder. Modern systems may be incorporated into the sound level meter and use hard drive or solid state memory.

They also contain mains power supplies for long-term use and some form of external remote control for the recorder so that the case can be locked to prevent tampering.

The system I will describe is actually a bespoke unit developed in-house but the same principles apply to manufactured systems.

Addressing first the issue of fit for purpose lets first look at what we require from our system:

Local authorities are required to investigate complaints of alleged noise nuisance from members of the public. These complaints may be quite straightforward where the source is known and can be measured directly or assessed without the use of instruments. On the other hand, large numbers of complaints are made about sources of noise that occur randomly or that only occur when the alleged source feels that nobody in authority is around to witness it. In addition, complaints may also be received about undefined or unknown noises, which may in reality, not exist, but still must be investigated. It is these types of complaint that are often the most difficult and time consuming to resolve and where some form of tape recording is most useful [I have used the term 'Tape Recorder' as a generic term, it is equally applicable to other types of recording device such as hard drive or solid state].

To assess whether a noise is a 'nuisance' [In law, a 'nuisance' may be described as something that 'affects a persons use of enjoyment of their property'] a suitably qualified person [usually an Environmental Health Officer, or other suitably qualified enforcement officer] must witness the noise to make a judgment based on, amongst other things, loudness, time of day and duration. Where the noise cannot be predicted to occur at specific times, then some form of tape recording may be used.

Taking a very simple case, that of a barking dog, it may be sufficient to be able to tell that the dog did in fact bark, and did so for a significant period of time, so time and duration are important considerations. The precise level of noise may be irrelevant other than that it was significantly louder than the ambient noise.

An equally common type of complaint is that of loud music. In this case, level and frequency response become more significant. Level because it may be necessary to describe to a court of law just how much louder the music was than the prevailing ambient noise and frequency response so that we are able to demonstrate that the perceived character of the music (e.g. loud bass) had not been influenced by the recording process. In this situation, the equipment (amplifier and loudspeakers or headphones) used to replay the recording and the listening environment can also play a significant part in the assessment. Once again, time and duration of recording are also important.

Finally, there are complaints of industrial or commercial noise nuisance. In this case, a sound level meter is often used to 'log' required parameters such as L_A90, L_Aeq etc., for later assessment against standards. The tape recorder is then used to determine that the sound level meter data is indeed that which is being complained of. Level may not be so important in this case though the recordings may be useful if further analysis becomes necessary at a later date. An example of this might be frequency analysis.

In the above three examples, 'fitness for purpose' has a very different meaning.

In the first case, a very simple criterion could be applied, that of the ability to record clearly and at a fixed level i.e. without automatic volume control.

In the second case, level becomes significant, as does frequency response. However, absolute level is less of an issue than change in level. For this reason, differential linearity is important. Recording range is less of an issue than the ability to record down to the low ambient noise levels encountered in domestic dwellings, particularly when investigating 'unknown' noises. If noises are loud enough to cause instrument overloads, this is usually sufficient description of a nuisance in itself!

In the third case, one could apply either of the previous two criteria. My personal preference would be the latter as this gives scope for later analysis of the recordings. However, it is more important to ensure that the recording system is able to cope with the higher noise levels likely to be encountered, particularly when the microphone is sited outdoors, than the potentially very low levels of noise inside a dwelling.

In reality, the last two cases may be amalgamated in one system where the differing noise levels are catered for in two or more sensitivity ranges.

One final issue that is often overlooked is that of the equipment used for replay, the amplifier and loudspeakers (or headphones) used to reproduce the recordings and the environment in which the recordings are analysed. Here, fitness for purpose would include an overall frequency response that does not unduly colour the replayed signal (this is mainly the amplifier and loudspeakers together with the room acoustics) and a listening environment that is acoustically quiet enough and free from other distractions.

Testing the system

A typical sound level meter will be verified against a standard, which was devised to test the functions of that instrument which was itself designed and manufactured to meet a specific standard. Many of these tests are not applicable to a tape recorder. So, what can we check/test?

In the first case above, often the only tests that may be carried out are functional tests. These might include the use of a pair of headphones to establish that the recordings are reasonably free from hiss, hum, radio interference etc.

In cases two and three, more detailed and scientific tests can be performed. In the authors' laboratory, all tape recording systems [this is taken to mean the whole system from microphone to readout] are subject to the following:

1. Dynamic range and linearity
2. Attenuator accuracy
3. Long-term stability
4. Frequency response

Dynamic range and linearity tests are performed using a signal generator and calibrated step-attenuator, which is altered in steps of 10 dB. The tests are carried out over the whole of the instruments range from just below overload to the instrument noise-floor. Recordings of the test signals are made and then played back through the separate analysis system. This checks the entire system (excluding the microphone, but including the pre-amp) in the way in which it is routinely used. The results are displayed on a measuring amplifier that forms part of the analysis system and also plotted out on a graphic level recorder, which also forms part of the same analysis system. These plots are retained on file as part of the calibration records. The criteria used are "does a 10 dB change in input at the microphone pre-amp result in a 10 dB change in indicated output?", "is the noise floor at least 10 dB below that of the microphones published specification?", and "has the noise floor changed significantly since the last calibration check?". We expect to see a 10 dB change in level down to within 20 dB of the noise floor. At 10 dB above the noise floor we would not expect an error of more than 2 dB. This is at or around the systems lower limit and as such we consider it insignificant.

Attenuator accuracy. One of the design criteria for our systems was ease and simplicity of use. In the days of analogue tape, it was often necessary to alter recording levels to suit the circumstances 'on the fly'. Modern digital recording systems with their wide dynamic range have made it possible to reduce the need for range adjustment and in our case we have just two ranges which are achieved using a 20 dB switched attenuator. This is always switched in for calibration when using a 94 dB nominal calibrator and may either be left in when measuring outdoors or switched out for greater sensitivity when measuring indoors. The criterion used is "does it provide a 20 dB change in level (within the resolution of the graphic level recorder - less than ± 0.5 dB)?". Again this is determined using the measuring amplifier and plotted on the graphic level recorder for records.

Long-term stability is checked in two ways. The microphone sensitivity is measured; again using the measuring amplifier referred to above, and is compared to its calibration chart and previous records. Other than when microphones have been physically damaged or obviously non-operational, we have rarely detected a significant drift (> 0.5 dB) here. The long-term 'level' stability is difficult to measure in a calibration situation as our definition of long-term is several days (a fit for purpose issue also). Also, it would be difficult in our laboratory to create environmental effects such as extremes of temperature. However, our operational procedures require that calibration signals are recorded at the start and end of each tape and these signals are then used to initially calibrate the analysis system and to verify at the end of the analysis that no significant change in recording level has taken place. Our criteria for significant is "less than 1 dB" and experience has shown that this is perfectly achievable over extended periods of time.

Frequency response is measured in two ways. The microphone is checked using a 'multi-function' acoustic calibrator and is tested at spot frequencies of 1 kHz (reference) and 31.5 Hz and 12.5 kHz to detect issues such as pin-holed diaphragms. As with long-term stability above, other than when there is obvious physical damage or the microphone exhibits noise, crackle etc. we have not found a significant problem here. The rest of the system is checked electronically using a signal generator at spot frequencies and once more the recordings are played back on exactly the same equipment chain that is used for normal analysis. Digital recording systems are capable of very wide flat frequency responses and are generally very stable with time. Analogue recording systems need more care and different criteria, which will depend on tape speed, recording and replay head wear and electrical and mechanical alignment. The criterion we use for digital tape is "less than 1 dB, or the graphic level recorders resolution" as appropriate. Plots are obtained and retained on record. Our analogue tape recorders, which are still used in certain circumstances, are calibrated 'out of house' so I make no further comment here.

Finally, I have referred to other instruments such as graphic level recorders, measuring amplifiers, step-attenuators, signal generators etc. Where relevant, these are subject to separate calibration, which is not covered here.

General comments on the use of tape recorders in environmental noise measurements

All of the above comments and calibration systems are based on approximately 30 years of personal experience of using tape recorders from analogue to digital and cheap to very costly professional on an almost daily basis. Indeed, at a meeting held a few years ago at the NPL at which many well known independent consultants were present, we made the statement that we (Birmingham City Council) probably make more environmental noise measurements in a year than anyone else in this country, a statement that was not disputed at the time! A large proportion of these measurements involve the use of tape recorders in some way.

It is my personal experience that many of the problems attributed to tape recording of noise are more to do with the way equipment is set up (often 'off the shelf' systems leave a lot to be desired), lack of understanding on the part of the user and a 'perceived inaccuracy' which is often just that - perceived not real. These factors have often been exploited by consultants acting for clients being prosecuted by their Local Authority for alleged noise nuisance.

I have, over many years, taken side-by-side measurements using sound level meters and 'tape recording systems' and have never found discrepancies in noise level that could be described as environmentally significant. Indeed, in most cases the difference in level has often been no more than 1 dB, a figure just as likely to be seen from two sound level meters in the same situation.

Some of the 'perceptual errors' seem to have more to do with the Watergate scandal that brought down US President Nixon than real errors in noise level (tape recorded conversations were eventually discredited after it was proved that they had been edited to alter the context of what was said)! This is certainly the case within certain elements in the legal profession.

Testing standards >>