|Elliott Sound Products||Acoustic Feedback & Frequency Shifting|
With Special Reference To Frequency Shifting
Copyright © 2017 - Phil Allison & Rod Elliott
In this article, the term PA refers to speech amplification systems employing microphones, amplifiers and loudspeakers used in auditoriums and churches to address an audience or congregation. The industry name for a high powered system used for musical performances is 'Sound Reinforcement' or 'SR'. While much of the information supplied is applicable to both, the emphasis here is about voice systems.
Acoustic feedback (aka the 'Larsen Effect') is especially troublesome when non-professional speakers use the system, as they tend to have poor microphone technique, and will often speak more quietly if they hear their own voice through the PA system's loudspeakers. If the person controlling the sound then increases the gain, a vicious cycle is started, and feedback is almost inevitable. Since feedback is usually preceded by 'ringing' (an unrelated tone that starts and stops with sound excitation) which usually dies away before it becomes a full 'howl', you do get some warning, but it's usually too late.
This article was written by Phil Allison, with additional material by Rod Elliott (indicated by dark grey text and ended with ).
|Please note that in this article, the person speaking is referred to as the 'speaker' or 'talker', while the loudspeaker is referred to as the 'loudspeaker'. This is done to ensure there is no ambiguity between the shortened version of loudspeaker and the person speaking into the microphone.|
Anyone who has been part of an audience while a PA system was being used is likely heard that piercing squeal called 'acoustic feedback'. The problem has been with us since the very first systems were installed and has not gone away. Acoustic feedback is a natural phenomenon, inherent in situations where the people using the PA system occupy the same room with the audience.
Electronics has made it possible to record sound for later reproduction and also to transmit sound from one place to another, in both cases it is possible to reproduce the original sound at deafening levels if you have enough amplification equipment. However, making a sound become louder in the same location where it originates is very different, as it attempts to break the laws of nature.
"But that is what PA systems do", I hear you say. The voices of people speaking through a PA system are much louder than their unamplified voices, and singers in rock bands can achieve deafening levels from the kind of SR systems usually employed. So what is the real story?
Try this simple test - have someone speak in a loud voice directly into your ear, from less than an inch distance. I bet you will find this very unpleasant and quickly pull your head away because the peak SPL involved is around 110 to 120 dB. A good singer's voice might be 15dB louder again. The test simulates what the diaphragm of a microphone is often subjected to during PA or SR system use.
The combination of using a close microphone and speaking loudly make it possible to generate quite high sound levels some distance away in the body of a room – but not nearly as high as at the microphone itself.
In a voice PA system, the sound heard by most of the audience is similar to normal speech level, which peaks at about 85dB SPL note 1. Close to the loudspeakers it will be far more, but back at the microphone position it is normally no more than this. If the speaker at the microphone can deliver 95dB SPL peaks with their voice, it will be well above the sound level arriving back and should render the PA system free of feedback problems.
The unfortunate fact is that most people, when speaking into a PA system, lower their voice and back off from the microphone as soon as they hear themselves being amplified. Turning up the gain for such users is no solution, because the same person then speaks even more quietly and/ or backs off further. If the system begins to feed back, the person stops talking altogether. Experienced and professional users of PA systems have long known they must not let themselves be distracted by hearing their own voice and do the opposite in order to avoid acoustic feedback.
Note 1: Many sources quote normal speech level as being 65 to 70dB SPL, but invariably fail to describe how the measurement was done. The figure of 85dB is the peak, measured at 1 metre. and based on actual tests with a 'Rode SPL-1' sound meter connected to a scope displaying instantaneous peaks levels.
The discrepancy is easily accounted for from the fact that speech has a 14 dB peak to average ratio, and SPL meters are commonly average responding and set to give an A-Weighted value.
Although it's not common to describe it this way, you can think of a PA system that's experiencing feedback as an oscillator, much like the many oscillators that we use in electronics. While a 'true' oscillator has a frequency determining network to fix a specific frequency, a PA system can have many different frequencies that are often just below the point of instability. In this context, the 'system' consists of the microphone, preamp and power amps, the loudspeaker(s) and the room. They cannot be separated, because they all play a part in the overall response.
This can be seen as a 'closed loop' feedback system, with the acoustic path completing the loop. The frequency at which the system oscillates is determined by amplitude and phase. At frequencies where the feedback is out of phase with the microphone's output, the system is stable, but when the signals are in phase it will be prone to oscillation if the overall system gain is high enough. In all typical rooms, the phase is random - it varies with frequency and is highly dependent on the reverberation and standing wave characteristics of the room. The operator usually has control of only one thing - gain. Equalisers can be used (specifically notch filters to reduce the gain at feedback frequencies), but this is rarely useful if the talker is moving around with a hand held mic, because phase and overall frequency response depend on the relative positions of the mic and loudspeaker(s).
The requirement for oscillation is simply that there is sufficient positive feedback to make the system unstable. In a PA system, positive feedback can occur at any frequency, determined by the frequency response of the loudspeaker and microphone, the time delay between the two, the nature of the room (standing waves, reflective surfaces etc.) and the system gain. When the loop gain (the total gain of the entire system including the room) exceeds unity, feedback will occur. Imagine a microphone, amplifier and loudspeaker as shown below. Only a few reflections have been included, but in most rooms the effects are chaotic, with potentially hundreds of different paths between the loudspeaker and microphone. Most will have a level that's well below the feedback threshold, but there will be one or more that can provide enough level at the mic diaphragm to cause feedback.
Figure 1 - PA System Feedback Conditions
With the gain structure shown in the drawing, the SPL from the loudspeaker at the microphone diaphragm will exceed 74dB at many frequencies, the loop gain of the system is above unity, and feedback is assured. The only condition is that the loudspeaker can provide a level at the mic position that ensures that the overall gain is more than 1. The frequency is indeterminate - it depends on too many different factors. It should be apparent from the above that the reference SPL at the microphone is not some fixed value. The signal from the loudspeaker at the mic position will always be greater than the speech level because the system has excessive gain.
Many community halls and other venues use ceiling loudspeakers, and as often as not some will be positioned almost directly above where most talkers will stand. This means that the signal from the loudspeaker has a direct path to the microphone, which when added to the multiple indirect paths ensures that feedback is not only likely, it's almost a certainty. System gain will always be very limited before feedback happens.
We can look at an example that may help you to understand the system's gain structure. The required loudspeaker output is 98dB SPL at 1 metre, based on a loudspeaker that's 94dB/ 1W/ 1m at a power of 2.65W ...
Mic: 2mV/ Pa 200µV at 74 dB speech level (1 metre), (-74 dBV) 1.4mV at 90 dB speech level (-57 dBV) Total Electrical Gain 23,000 (87 dB) 3,300 (70 dB) Speaker SPL 98 dB @ 1m, 92 dB @ 2m, 86 dB @ 4m 98 dB @ 1m, 92 dB @ 2m, 86 dB @ 4m System #1 - Unstable System #2 - Stable ... Maybe
Note that all SPL and voltage levels are average. System #1 will feed back! Even if the loudspeaker's signal path is 4 metres to get back to the microphone, the returned SPL at the mic will be 86dB. This is louder than the speech level allowed for by the amount of gain applied, and feedback is guaranteed. The gain must be reduced, and the talker will have to get a lot closer to the mic, and/ or speak louder to get the same SPL from the loudspeaker(s). Note that this doesn't account for the delayed reflected sound paths that send additional loudspeaker energy back to the microphone. This effect is shown in the next section.
If the talker's mouth is around 25mm from the diaphragm, the level will be closer to 90dB SPL, and the mic's output will then be around 1.4mV. The gain can now be reduced to a total of 3,300 meaning that the mic preamp gain will be a little over 14 instead of 100 (assuming that the other gains are fixed).
With the mic at 4 metres from the loudspeaker, the level at the mic diaphragm from the loudspeaker will still be 86dB. However, this is now lower than the speech level and the system should be stable. However, this relies on the room being well damped so that at 4 metres from the loudspeaker, the level really will be 86dB SPL. In most cases it will be more ! A single narrow band peak in the system's overall response (microphone, loudspeaker and room) is all that's needed for feedback to start.
The aim of any operator is to ensure that the loudspeaker's SPL at the microphone will always be less than that which will create feedback. This means that sound picked up by the mic from the loudspeaker is at a lower level than that produced by the talker. This is not always possible, and feedback will occur. In many cases, the only solution is to lower your expectations regarding the SPL in the audience area. A reduced SPL simply means that less gain is needed.
A room's acoustic properties have a major influence on the amount of sound reaching the microphone position from the loudspeakers. In the study of acoustics, locations in a room where the direct and reverberant sound levels from a source are the same are said to be at the 'critical distance' [ 1 ].
In nearly all rooms without significant acoustic treatment, the critical distance is only a metre or two away from the loudspeakers, and beyond this distance the famous inverse square law no longer applies. This is a benefit, as it makes it possible to fill a room with sound from a modest size PA system, but the drawback is in how it aggravates the problem of acoustic feedback.
Another problem is how sound waves in a reverberant field arrive back at the microphone position from any and all directions, having bounced off the walls, floor or ceiling first – largely negating any benefit from using directional loudspeakers and microphones to minimise the feedback issue. In such a room, moving the microphone or loudspeaker positions has little effect on the gain setting that results in feedback.
PA systems sound far better in rooms with minimal reverberation. Purpose designed auditoriums (like cinemas) minimise sound reflections by the use of large amounts of absorbent materials and by avoiding having parallel walls and parallel floors and ceilings. The average public hall and most churches have the exact opposite, hence exhibit massive reverberation and as a result are very hostile environments for a PA system.
Figure 2 - ≈10Hz Spaced Frequency Peaks Caused By Standing Waves [ 6 ]
The loudspeaker signal picked up by the microphone will always be delayed. The delay is determined by room dimensions (as are standing waves), and may vary between perhaps 5ms (a distance of 1.7 metres), up to 100ms (34.5 metres) or more. In any given room, you will usually have (something close to) both examples, as well as many other intermediate delay times, as determined by the dimensions of the room itself. The above graph shows the measured response in a 'typical' room. The response is primarily due to standing waves.
Shorter delay times increase the frequency spacing, but also mean that the feedback builds up faster. The way sound behaves in a room is very complex, and while it would be nice to be able to explain it all in a few simple charts or graphs, it's impossible to do so. To makes matters worse, every room and loudspeaker system is different, and moving the microphone or loudspeaker(s) even a small distance can change everything - usually radically. However, it rarely provides an improvement.
Directional microphones like cardioid and super-cardioid types help greatly, but not for the obvious reason. While both types discriminate against sound waves arriving from the rear of the microphone, this has no benefit unless the rear is carefully aimed at the loudspeaker. Where there is more than one loudspeaker in the room, this becomes impossible, and is also impossible when the microphone is being hand held. See polar response graphs ...
Figure 3 - Shure SM58 Polar Response [Original]
The way a directional microphone actually helps depends on something called the 'proximity effect' - the name given to an increase in mid and low frequency sensitivity when the sound source is close to the microphone. At a distances under 25mm (1 inch), the increase can be 20dB at low frequencies and about 10 dB in the middle of the voice range. An omni-directional microphone has no such effect and provides no benefit. See response graph ...
Figure 4 -Directional Microphone Proximity Effect [Original]
The increased sensitivity does not exacerbate feedback as it only applies to close sound sources, so does not involve the PA system's loudspeakers. If having extra low frequency content in the speaker's voice is a problem, turning down the bass tone control on the mixer channel in use will compensate. Doing so automatically increases the feedback margin for low and mid frequencies.
Many speakers dislike fixed microphone locations, preferring to move about and also have their hands free. By far the best solution for them is to use a head worn, miniature, cardioid microphone. This has every possible benefit - the microphone is always in the same spot, very close to the speaker's mouth (but just out of breath and pop noise range) and goes wherever they go.
The electret microphone capsules used have very high sound quality, better than typical dynamic microphones used in most voice PA systems – see an example [ here ].
The microphone's signal is normally transmitted via UHF radio link to the PA system avoiding problems with trailing cables. A mute button on the transmitter allows the speaker to have a silent conversation or cough if need be.
Sometimes a PA system must be used in a highly reverberant room, the people who will use it are not experienced professionals, head worn microphones simply cannot be employed, the allowable loudspeaker and microphone positions are not ideal and still the system has to perform well and be free of feedback without the benefit of an expert operator. An impossible task ?
There is an electronic device that can come to the rescue here, one that has been around for over fifty years but has lately fallen into disuse. Known by various names the device will do the following ...
That device is an 'Audio Frequency Shifter'. What it does is unique and in most respects superior to other methods of improving the feedback threshold of a PA system, plus can be used in addition to graphic equaliser or adjustable notch filter devices.
The technique is called frequency shifting and the device employed is also known as a 'Howl Round Stabiliser'. Such units were developed and used by the BBC around 1960, based on experiments by M. R. Schroeder and employed many valves (vacuum tubes) and radio frequency techniques [ 3 ].
By the mid 1970s, advances in analogue microchips called 'analogue multipliers' made similar or better performing units far cheaper to build, plus kept all signal processing down at audio frequencies.
What a frequency shifter does is move the audio band upwards by adding a few Hz to every incoming frequency, removing the possibility of a reverberating tone circulating in the room being directly reinforced by the same tone also coming from the loudspeakers. This strikes directly at the root cause of acoustic feedback in reverberant spaces.
With a 5Hz shift, an input signal of 500Hz changes to 505Hz while an input signal of 1000Hz changes to 1005Hz and so on. Applied to a speaking voice or recorded music, it is very difficult to hear that there has been any change.
Any room with smooth, parallel surfaces will support 'standing waves' - single frequency sounds with wavelengths mathematically related to the dimensions of the room. The frequencies involved typically start at around 10Hz and then every multiple of that number up to the limit of the audio range (see Figure 2 for an example). If you carry out a very slow frequency sweep of the room using a loudspeaker and a sound level meter spaced well apart, the result is a pattern of intensity peaks at about 10Hz intervals. These peaks and corresponding dips in-between are centred about the average level by around plus and minus 10dB.
When a PA system in such a room suffers acoustic feedback, the howl frequency will coincide with one of the peaks, typically the strongest one – which explains why the frequency is very steady and repeatable. However, most people will have noticed that the frequency often changes when the microphone is moved, and sometimes the distance can be fairly small (around 300mm or so may be enough).
The PA system is then supplying energy at one of these standing wave frequencies, sustaining the oscillation and quickly raising the sound level to the full available output of the system. When the gain is reduced, the howling soon stops but the sound quality of speech may still be affected by ringing at the same frequency.
Adding a frequency shift of about 5Hz into the amplification loop defeats oscillation at any standing wave frequency, because the sound leaving the speakers is not the same frequency as that picked up by the microphone, and so cannot reinforce the oscillation.
Since frequency shifters operate with unity gain and have an essentially flat frequency response, installation at audio line level is simple. Once installed in-line with the output signal from the mixer to its associated amplifier, a user puts it into bypass mode and then increases the microphone gain control until the first signs of feedback are heard. In this condition, any PA system is quite unusable and attempts to approach the microphone and speak will be greeted with severe ringing and possibly loud howling noises.
When the device is switched out of bypass, the same PA system instantly becomes stable and usable, any tendency to ringing or howling having gone. The microphone gain can be increased by several dB and the PA system is still fine. On first encounter, most people find this quite magical.
If the gain is increased too far, instead of loud howling a mild warbling tone is generated, accompanying speech. Simply backing off the gain by about 2dB makes this effect disappear. The only control a frequency shifter normally has is a switch to vary the number of Hz the audio band is shifted by – in most rooms a shift of 4 to 5Hz is optimum. A smaller shift is better for large auditoriums with reverberation times of more than 2 seconds, where the standing wave response peaks are more closely spaced.
Of great benefit is that when a head worn or hand held microphone gets too close to one of the loudspeakers, the user is immediately alerted by hearing the warbling sound and only needs to move away. The PA system does not howl, there is no risk of damage to loudspeakers or to an audience's equanimity.
|Please Note: A frequency shifter for this purpose uses analogue circuit techniques and must not to be confused with the now commonplace 'digital pitch shifter'. A pitch shifter changes incoming frequencies by a fixed ratio, such as an octave or several semitones. Using the latter will not provide any real benefit in relation to acoustic feedback.|
Frequency shifters used to be available, although they were never actually common despite the significant advantages they offer. Today, it's almost impossible to get one. They have always been a niche product, seemingly known to relatively few people, and therefore never became 'main stream'. Today we have all-singing, all-dancing automatic feedback 'eliminators' that may or may not work, and will always colour the speech quality because they rely on narrow notch filters controlled by clever electronics. This seems to be the preferred approach, because a DSP (digital signal processor) can - so we are told - do everything we'll ever need.
There are also frequency shifters that are used for effects during performances or recording, often as a 'plug-in' for digital audio workstations. These are usually not suitable, because they are designed for comparatively large frequency shifts and to intentionally remove harmonic relationships within the sound being processed. While it might be possible to use one to provide a 4-5Hz shift, it would be a very expensive addition to most PA systems used in community halls and/ or churches.
A fully proven design for a 5Hz frequency shifter will be described as a contributed project article in the near future. A printed circuit board will be made available for purchase if there is sufficient interest. The design features very low noise and THD, has balanced audio input and output and requires only a small AC transformer to supply power. Only commonly available components are specified and it should be possible to fully assemble a board for under US$150.
There are many references to frequency shifters, but very few available schematics, and none that use parts that are available now (as opposed to 40-odd years ago). There has been a lot of academic work (as demonstrated by the references above), but for reasons that are rather puzzling, the devices themselves have all but vanished from sale. The last known commercial version is made in the UK and although it's still shown as a current product, it may or may not be available in reality. The website is obscure, rarely shows up in searches, and seems to be poorly organised.
|Copyright Notice. This article, including but not limited to all text and diagrams, is the intellectual property of Phil Allison & Rod Elliott, and is Copyright © 2017. Reproduction or re-publication by any means whatsoever, whether electronic, mechanical or electro- mechanical, is strictly prohibited under International Copyright laws. The author (Phil Allison) and editor (Rod Elliott) grant the reader the right to use this information for personal use only, and further allows that one (1) copy may be made for reference. Commercial use is prohibited without express written authorisation from Phil Allison and Rod Elliott.|