Gareth Pont, Design Director at Connevans Limited

Back to index of articles

Abstract

It is well known that some cochlear implant processors cause interference to personal fm systems, usually manifested as reduced operational range when using the personal fm system. The purpose of this study was to determine whether DSP (Digital Signal Processing) hearing aids also cause interference to personal fm systems, to quantify the nature of any such interference and to define practical remedies. The results showed that all but one manufacturer's DSP hearing aids tested produced interference and that some were much worse than others, suggesting that certain aids should not be prescribed if personal fm systems are expected to be used. The study also concluded that the interference would be detrimental to the audio quality of all models of personal fm system when the fm system was actually in use.

The objective of the paper is to make designers and users more aware of one of the common undesirable side effects of digital processes in order to encourage the production and prescription of better hearing aids that do not have the potential to impair the user's listening experience when used with fm systems.

Introduction

Connevans Limited designs and manufactures personal fm systems in the UK. Since mid 2002 we have been receiving occasional reports of interference to personal fm systems when used with certain digital hearing aids. The digital aids that have caused problems are all DSP aids as opposed to the 'digitally programmable analogue aids'. This study was undertaken in association with Manchester University and their MCHAS project in order to define the cause and effect of this interference. Practical tests have been carried out using all the DSP hearing aids on the NHS tender list (see Table 1) and the currently available personal fm systems listed in Table 2. The results show why it is difficult to trace and produce such interference on demand. There may be a reaction from some quarters with the suggestion that if DSP aids cause interference, then personal fm systems should not be used with them. These same sources may also suggest that DSP aids are now so good that personal fm systems are no longer necessary. Whilst it is not the purpose of this document to convert doubters to the benefits of personal fm systems, any hearing aid user or person who has compared the performance of any microphone close to the source of sound with the performance at a distance from the sound will know which one is by far the best. Personal fm systems take the sound from close to the source and relay it to the listener anywhere within radio range at that same Signal to background Noise Ratio (SNR).

There is obviously a danger that the message "DSP aids = problems with all fm systems" gets about. This is not the case and is firmly refuted. However it is true that the performance of different fm systems varies significantly and that the interfering potential of different DSP aids varies from none to considerable.

The study has resulted in the generation of a 'league table' of DSP aids ranked by the amount of interference that they produce. This table will be published in a more detailed paper by Manchester University later this year and is therefore omitted from this newsletter article. Whilst this omission will undoubtedly be a disappointment to readers, we hope that they will understand the sensitive nature of this information as far as the hearing aid and other personal fm system manufacturers are concerned. All the hearing aid manufacturers have been contacted with our findings and we are currently awaiting responses. It would be improper of us to publish this information until the manufacturers have had an opportunity to respond.

Practical Test Arrangement

The hearing aid with appropriate shoe and Direct Audio Input (DAI) lead under test are placed inside an electrically screened chamber as used for ElectroMagnetic Compatibility (EMC) testing of electronic products. The use of a screened chamber is mandatory in order to minimise the effect of external radio signals on the results. The chamber also contains Radio Frequency (RF) absorbing material to reduce internal reflections, very much like an acoustic anechoic chamber. The aid and DAI lead are placed on a non-conductive table suspended in the chamber and the DAI lead is connected to an interface connector in the side of the screened chamber. The interface connector is grounded to the metal wall of the chamber and a standard RF screened lead takes the signal from the chamber to an RF spectrum analyser. The test system is therefore effectively a closed box from which external influences are minimised and which emulates approximately the way the hearing aid would be worn on the body, but in a reproducible way. It is the radio frequency equivalent of a hearing aid test box (but better).

Validity of Tests

Control measurements are undertaken with a hearing aid, shoe and DAI lead in place, but with the aid switched off. These measurements show the inherent noise in the test system including any external radio signals that leak into the test environment due to imperfect screening. Providing the levels of noise and extraneous signals are at least 6dB below the interference that is being measured, then the test system itself is deemed to have no significant effect on the measured results. The test system meets this requirement.
Diagram of Practical Test Arrangement

Table 1. The DSP hearing aids tested in this study
  1. A&M Select
  2. Oticon Digifocus
  3. Oticon Spirit
  4. Oticon Spirit 700
  5. Philips Spaceline
  6. Phonak Aero
  7. Phonak Supero
  8. Siemens Prisma
  9. Siemens Prisma P
  10. Siemens Selectra
  11. Siemens Signia
  12. Starkey Gemini A-13
  13. Widex Senso P37
  14. Widex Senso P38
Table 2. Personal fm systems tested with the DSP hearing aids
  1. Connevans CRM220
  2. Connevans fmGenie
  3. Phonak Microvox
  4. Phonak MLx
  5. Phonic Ear 475
  6. Phonic Ear Solaris
  7. Sennheiser Microport 2013PLL
Discussion

Observing the conducted interference on the DAI lead over a wide radio frequency range, all except two of the DSP aids produced significant interference. However, the levels and characteristics varied. Over the frequency range used by personal fm systems in the UK (173 to 175MHz), all except two of the DSP aids produced levels of interference well above the noise floor of the measuring system.

A common characteristic of all the 'noisy' DSP hearing aids tested with one exception was that the interference consisted of obvious harmonics of the Digital Signal Processor (DSP) clock used in the aid. This clock typically seems to be running at around 1MHz and is unstable, changing frequency slightly as a function of time, battery voltage, temperature and power demand. A small change in clock frequency makes no difference to the efficiency of the DSP process but makes large changes to the high harmonic frequencies produced. For example a frequency change of 10kHz is only 1% for the 1MHz clock but the 174th harmonic at around 174MHz will move by 1.74MHz. This means that the frequency of the interference moves around, essentially at random, over the whole of the frequency band used by the personal fm systems. This is why it is so hard for a user to track down what is causing the random interference. It only happens when the interference frequency is at the same frequency as the personal fm receiver is tuned to and is only obvious when the personal fm transmitter is off or the received signal level is low. It is suspected that the one 'noisy' aid that has a different characteristic to the others uses a spread spectrum technique on the processor clock for reducing the peak levels of interference at any particular frequency. However this simply spreads out the interference so that it affects more radio channels at any one time. The interference from all the DSP aids typically sounds like 'white noise' to the user.

Why do DSP hearing aids produce this interference?

All digital processing systems use a master clock to make the internal processes work. It is well known in the electronics and computer industry that as the clock frequencies for computers have risen, the potential for interfering with other systems has also risen. Since 1996 it has been law within Europe that the interference produced by electronic systems must be controlled. Many techniques have been employed to reduce this interference but many of the common ones are difficult to employ in a physically small device such as a hearing aid. One of the most effective techniques for reducing interference used by designers of digital systems involves slowing the transition rates or edges of the clock signals inside the processor chips and this reduces the levels of high frequency harmonics enormously. This technique is widely used these days and is eminently suitable for implementation in a hearing aid DSP. It is most unfortunate for users of personal fm systems that the majority of designers of DSP hearing aids have apparently overlooked this important and fundamental design aspect.

Recalling that the DAI lead connecting the hearing aid to the personal fm system is the aerial for the personal fm system, it is obvious that any interference generated by a hearing aid connected to the DAI lead will travel directly along the DAI lead into the personal fm system. How the personal fm system then reacts to that interference is therefore a critical question.

Some background to Personal fm Systems

Personal fm systems are low power licence exempt one way transmitter-receiver systems and are intended to be used over short ranges up to a few tens of metres to transmit a speaker's voice directly to the hearing aid of a hearing impaired user. The licence exempt status means that the transmitters are only permitted to transmit at low power, typically one milliwatt or less. This also means that to obtain good (i.e. noise free) audio performance from a transmitter-receiver pair over useful distances the receivers must be sensitive. Having a sensitive receiver to provide good audio performance necessarily means that the receiver is also sensitive to interference if it occurs in the same frequency band that the personal fm system is tuned to.

Squelch circuits

The worst case scenario for external interference to affect a personal fm receiver is when there is no deliberate transmission from the associated transmitter. The receiver is then 'open' to any signal that comes along at the frequency it is tuned to. To prevent the normal background noise present in all electronic systems from bothering the user a circuit known as a "squelch" is employed. All personal fm systems use a squelch circuit to cut off the audio output when the audio signal becomes noisy (i.e. the Signal to Noise Ratio or SNR is below a certain value). The squelch circuit may be considered as a sound gate that is shut when the received signal is noisy (such as at excessively long range) and open when the received signal is clear. The design and setting of this squelch circuit determines how much wanted signal or interference is required before the gate opens.

If we observe from the practical measurements (see Figures 2 and 3) that the interference from the DSP hearing aid is typically a band of noise occupying many times the receiving bandwidth of the receiver that slides around in centre frequency as a function of time, there will be many occasions when the noise is not at the same frequency as the receiver is tuned to, but some occasions when the noise is at the same frequency. Let us first take the case when there is no deliberate transmission on the receiver channel (i.e. the transmitter is off). If the level of that interference is such that it can open the squelch gate, then the receiver will, at random, produce loud audio noise when the receiver frequency and the interference frequency coincide. At other times the audio will be cut off by the squelch circuit. If we now look at the case when the transmitter is on and the receiver is in range then the squelch gate would be open, allowing the transmitted audio to be heard. What happens this time depends on the absolute and relative levels of the wanted transmitted signal versus the interference. When fm receivers are in receipt of strong radio signals they are operating in a mode known as "fully quieting". This means that the signal to noise ratio is as good as it can get and interference will be suppressed. This is important because it means that even if no sound is transmitted, the presence of the strong radio signal should prevent interference from being heard. Below a particular received signal level (which is dependent on the individual receiver design) then the signal to noise ratio begins to gradually degrade and the effects of external interference start to become noticeable.

Let us take an example of a personal fm receiver that claims to have a sensitivity specification of 0.5uV (or ?113dBm or -6dBuV in alternative units) for 26dB SNR and let us further suppose that the squelch circuit cuts off the audio at a SNR of 26dB. Then the squelch circuit would normally open at a radio signal level above 0.5uV. Now suppose that the wanted transmitted signal is received at a level of 70uV (-70dBm or 37dBuV) and that a particular DSP aid is connected to the DAI lead producing some in-band interference at 26dBuV. The wanted signal level is higher than the squelch level therefore the squelch gate is open and the audio can come through. However, the interference is only 11dB (= 37 - 26dBuV) below the wanted signal so the user actually hears the audio plus the interference at a very bad SNR of 11dB instead of the 40dB or more that could otherwise be expected with a 'quiet' aid. It can be seen that this SNR result was arrived at without using the sensitivity specification of the personal fm receiver.

The important result is that if the squelch is opened by the wanted signal, then the actual SNR is determined only by the difference between the levels of the interference and the wanted signal although when the receiver has a strong enough signal to be "fully quieting" the interference will be further suppressed.

Let us take one more example of another personal fm receiver that is much less sensitive than the earlier example, such as the Phonak MLx. The sensitivity specification for this device is in a different format to the others because it does not have a conventional aerial. It is specified as >16dB SNR for an electric field strength of 3mV/m. A practical measurement shows that the squelch opens at a conducted level in the region of 25uV (28dBuV). Comparing this value with the 26dBuV interference level from the example DSP aid, it looks as though the MLx could well respond to the interference. However, in practice the low SNR and audio bandwidth available from this particular style of personal fm receiver will tend to mask the effects of most interference in any case. This result might seem to indicate that the use of an MLx is to be preferred over other personal fm systems on the basis of its apparent immunity to interference. It should be noted however that this apparent immunity is simply a result of the very low radio sensitivity of this style of receiver. Such low radio sensitivity results in much poorer SNR performance even over short ranges than is possible with conventional body worn personal fm systems. In other words, what the user hears through an MLx style receiver has much more 'white noise' added to it by the receiver than would occur with a good body worn receiver. Immunity to interference outside the radio channel that a MLx receiver is tuned to is also much worse than in a conventional body worn receiver. Additionally, there are currently other compatibility issues between MLx receivers and hearing aids not manufactured by Phonak.

Is it possible to stop the interference travelling down the DAI lead?

This is actually very difficult because the interference travels along the DAI lead in "common mode" i.e. the interference exists equally on all the conductors in the cable. The difficulty with filtering common mode interference is that the filter circuit has to work against some electrical reference potential such as 'earth' or a 'large' electrically conductive area. These do not exist between a hearing aid and the personal fm system. In order for a filter to work it must offer a low series impedance to the wanted signal (the audio) and a high series impedance to the unwanted signal (all the RF interference), usually also short circuiting the unwanted RF interference to 'ground'. Firstly there is no 'ground' so the RF cannot be shorted out and secondly the aerial input impedance to personal fm systems is usually high. This means that the series impedance required in all of the DAI lead conductors would need to be very high. If the filtering elements are physically very small (as they would need to be for aesthetic reasons) then any high frequency interference can simply leak across the filtering elements, reducing their expected advantage. Another problem is that there is a limit to the impedance that may be used in the DAI cable because a high value will adversely affect the audio frequency response. A further issue is that the more effective interference filter designs will prevent the use of remote hearing aid microphone muting. In practice, DSP aids tend to use Wide Dynamic Range Compression (WDRC) so hearing aid microphone muting is generally regarded as being less necessary and some DSP aids cannot respond to the mute control signal in any case.

In spite of the difficulties described, some filtering is possible. Initial designs suggest that interference attenuation in the region of 6 to 12dB should be possible in an aesthetically pleasing package. Further work is ongoing in this area. Whilst 6 to 12dB may not sound much, it could be enough to make the difference between a usable and unusable DSP hearing aid/personal fm system combination, especially in conjunction with the other remedies suggested at the end of this article.

Conclusion

In response to the question of the title, "is there a problem?" the short answer is "yes, in most cases". The interference problem is caused by the digital processor chip in the hearing aid. The interfering effects are made worse by the fact that the processor clock frequencies are not stable resulting in unpredictable, random bursts of 'white noise' interference from the attached personal fm system. The hearing aid manufacturers could have eliminated this interference for all practical purposes by following established 'low EMI' (ElectroMagnetic Interference) processor chip design procedures but only one appears to have done so. Interestingly enough, the manufacturer that has produced the 'quiet' DSP aids, also happens to manufacture personal fm systems. Is that a coincidence?

In practice, the effects of the interference may be minimised by keeping the personal fm transmitter on when the receiver is in use and muting the transmitter microphone when transmitted sound is temporarily not needed. This also means that where there is a known interference problem the receiver should be switched off when it is not needed. The stronger the received signal, the more any interference will be suppressed. It may therefore be beneficial to operate the transmitter at a higher power level, and/or reduce the distance between transmitter and receiver, where these are possible. The interference may also be minimised by prescribing 'low EMI noise' aids and avoiding the use of 'high EMI noise' aids for use with personal fm systems. Radio frequency filtered DAI leads will also provide some benefit when they become available.

A further alternative that will eliminate the problem is the use of an inductive neckloop with the personal fm system in conjunction with the 'T' or 'M-T' setting on the hearing aid. We know that this method of wearing is already preferred by some users such as teenagers who prefer not to draw attention to their disability. However, there are limitations on the audio quality attainable by this method that could outweigh any potential advantages in respect of radio interference. As users will know, telecoils have a common habit of picking up all kinds of interference from electrical equipment anyway.

Whilst the response to interference of different personal fm receivers varies when there is no deliberate signal transmission, all personal fm receivers can be affected by this interference when the system is operating normally with an audio signal being transmitted.

In the longer term, pressure on the designers of DSP hearing aids is required to encourage the manufacture of low conducted EMI products. Users and prescribers should request conducted EMI information from manufacturers even though at this point in time it is unlikely that any of the manufacturers have this information. As far as the NHS is concerned, this requirement for low EMI is very likely to form part of the tender requirements for the renewal of paediatric DSP hearing aid provision contracts, so this will eventually force the issue into the open and will encourage the design of personal fm system compatible low pollution DSP hearing aids

Postscript

Whilst this discussion and measurements are in respect of DSP hearing aids, similar arguments apply to the case of interference from cochlear implant processors. There are differences in the form and effect of cochlear processor interference compared with that from DSP hearing aids but the ways of improving the situation are similar. A further article on 'realizing the best performance from a cochlear processor and personal fm system' will be published separately.

The views expressed in this paper are those of the author, supported by Manchester University, and are based on measurements of a small sample of available DSP hearing aids. Further samples of as many types as possible continue to be tested as opportunities arise. Interested parties are welcome to contact the author by e-mail at gareth@connevans.com to discuss the issues raised herein or to provide further subjective evidence in support of or challenging our findings. Please just remember not to ask for the league table!

Conducted Radio Frequency noise from DSP hearing aids: Graphical Results


Common test conditions:

Hearing aid and DAI lead laid out straight in RF screened chamber
DAI lead length 40cm
DAI conductors both connected to centre pin of chamber sense cable via a capacitor
Measurement bandwidth 10kHz (to match typical personal fm systems)


Figure 1: Typical wideband peak noise spectrum for one of the highest interference DSP aids. The lower line is the noise floor of the measuring system. The spikes are commercial radio stations such as Radio 4 etc. that leak into the test system.


Figure 2: Zoomed in peak background noise spectrum for the same high interference DSP aid as in Figure 1 at a frequency of around 174MHz where the UK personal fm systems operate. This trace was collected for five seconds. Each peak is a band of noise produced by the aid. Observe that the larger peaks are spaced by about 1MHz - the clock frequency in the processor. This particular peak was at 174.125MHz at the time this record was taken. If an attached personal fm receiver was tuned to 174.125MHz, it might well have produced audio noise, but if it was tuned to another frequency where there is no peak, it wouldn't produce any noise.


Figure 3: Peak background noise spectrum for the same high interference DSP aid and over the same frequency range as in Figure 2. This trace was collected for three minutes. Observe that the peak in Figure 2 has moved to both the left (lower frequencies) and right (higher frequencies) during this measuring period. The top of the rectangular parts only look like this because the spectrum analyzer is holding the peak of the signal as it moves about. In reality the signal is like Figure 2 but with the peaks moving to different frequencies. The plot is shown this way to make it obvious what is happening.