Evaluation of an Ear Level Directional Microphone for Tympan

Given the near ubiquity of directional microphones in commercial hearing aids and the development of a smaller Tympan prototype that can be mounted on the head, it is desirable to find an easy, out-of-the box solution for achieving a directional response using a microphone placement that typifies a behind the ear device, including those styles commonly used for ‘open fits.’ I happened to have a mini boom microphone with a direct audio jack connection (i.e., no cord) laying around my lab; see picture below. I could not find any identifying markings to indicate the manufacturer, but I found similar microphones on the internet that are advertised for smart phonThis text will be hiddenes and portable recording equipment (e.g., video cameras). Some even have an articulated joint that could help with optimal positioning above the ear.


Test Box Measures
To represent a typical hearing aid fitting, the Tympan was programmed using the Tympan library code ‘WDRC_8BandComp_wExp,’ which is 8-channel compression with expansion. Unlike previous evaluations, for maximal clinical realism, gain was iteratively adjusted in the test box to meet DSL-adult prescriptive targets for the audiogram below, representing a mild-to-moderate sloping sensorineural hearing loss. Compression kneepoints, compression ratios, and expansion kneepoints were derived from recommendations of the DSL algorithm for this hearing loss. As in previous evaluations, Klipsch S4 earphones were used to deliver the sound. A Verifit (Etymonic Design Inc.) hearing aid analyzer was used to obtain the measurements. For the purposes of convenient microphone placement in the test box, an audio extension cord was used to decouple the microphone from the Tympan.


The figure below shows the fit to targets using the directional microphone.

The omni directional microphone used in previous evaluations (Sony ECM-CS10 Lapel Electret Microphone) was swapped with the directional microphone and the same measurements were obtained, without further adjusting gain, etc. The only change that needed to be made was to decrease the volume knob by 5 dB to accommodate what appears to be minor difference in sensitivity. As can be seen from the figure below, the overall frequency response is similar for the two microphones.

Equivalent Input Noise
Equivalent input noise (EIN) level was measured just as before using the ANSI S3.22 (2009) test sequence in the Verifit. First, the default settings for the ‘WDRC_8BandComp_wExp’ code was used to facilitate comparisons with earlier measurements:
{317.1666, 502.9734, 797.6319, 1264.9, 2005.9, 3181.1, 5044.7}, // cross frequencies (Hz)
{0.57, 0.57, 0.57, 0.57, 0.57, 0.57, 0.57, 0.57}, // compression ratio for low-SPL region (ie, the expander…values should be < 1.0)
{45.0, 45.0, 33.0, 32.0, 36.0, 34.0, 36.0, 40.0}, // expansion-end kneepoint
{20.f, 20.f, 25.f, 30.f, 30.f, 30.f, 30.f, 30.f}, // compression-start gain
{1.5f, 1.5f, 1.5f, 1.5f, 1.5f, 1.5f, 1.5f, 1.5f}, // compression ratio
{50.0, 50.0, 50.0, 50.0, 50.0, 50.0, 50.0, 50.0}, // compression-start kneepoint (input dB SPL)

Across a range of volume control settings, -15 to +15 dB, EIN was roughly constant at 32-34 dB SPL. Beyond these settings, EIN increased by a few dB, with a max EIN of 39 dB SPL at the -20 dB setting. When the expansion ratio was increased from 1:1.75 to 1:4 (0.25:1), EIN decreased by about 3 dB. Further increases in the expansion ratio did not reduce EIN further. Alternatively, expansion thresholds could be increased, but this comes at the risk of reducing audibility for soft speech. EIN with the directional microphone was slightly greater than previously recorded for the omni directional microphone. This is likely due to the slightly reduced sensitivity of the microphone. While directional microphones in hearing aids are generally associated with more internal noise, this phenomenon has more to do with amplifier-based equalization used to overcome the low-frequency roll off typical of a directional microphone frequency response.

EIN was then measured for both microphones using the compression and gain settings for the mild-to-moderate audiogram. Because DSL prescribes low compression kneepoints and even lower expansion kneepoints, the expansion kneepoints were increased to be equal to the compression kneepoints (i.e., effectively going from a 4-stage to a 3-stage nonlinear input-output function) and expansion thresholds were increased to be 10:1 in order to keep EIN to a minimum:
{0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1}, // compression ratio for low-SPL region (ie, the expander…values should be < 1.0)
{32.2, 27.4, 29.7, 29.5, 30.6, 30.5, 28.6, 34}, // expansion-end kneepoint
{0.f, 0.f, 10.f, 10.f, 10.f, 30.f, 35.f, 40.f}, // compression-start gain
{0.7, 0.9, 1.1, 1.1, 1.2, 1.3, 1.3, 1.7}, // compression ratio
{32.2, 27.4, 29.7, 29.5, 30.6, 30.5, 28.6, 34}, // compression-start kneepoint (input dB SPL)

With these settings, EIN was measured at 40 and 44 dB SPL for the omni and directional microphones, respectively (recall that volume control was set at -5 dB for the omni directional microphone when fit to prescriptive targets). Later measurements using the Frye Fonix 7000 hearing aid analyzer, with Tympan programmed with flat 20 dB gain (no compression or expansion) using the ‘BasicGain’ library code yielded 46.0 and 43.9 dB EIN for the omni and directional microphones, respectively. It is clear that EIN is not greatly affected by the choice between these two particular microphones and that higher expansion kneepoints will be needed in order to keep internal noise from being audible for users with mild-to-moderate hearing loss. This also means that for DSL-based prescriptions that compression kneepoints will also need to be subsequently increased.

Total harmonic distortion (TDH) was measured with Tymapn programmed to prescriptive targets for both microphones using both the Audioscan Verifit and the Frye Fonix hearing aid analyzers. The former measured TDH at 1/3 octave frequencies from 250 – 4000 Hz and the latter at the ANSI test frequencies (500, 800, 1600 Hz). TDH was measured to be 0-2% for both microphones and hearing aid analyzers.

Directional Test
The Directional Function Test in Verifit was used to collect a gross measure of the front-to-side ratio for both microphone arrangements. Speech and noise were set at 60, 70, and 80 dB SPL (i.e., 0 dB SNR). Per the user’s manual:

“The unique Verifit directional test presents over 1000 tones simultaneously at different frequencies from both test chamber speakers. The tones are individually controlled in real time to produce a precise spectrum from each speaker at the test box reference microphone. The coupler SPL is then analyzed into two real-time response curves, labeled L (left) and R (right) to indicate which speaker generated the curve. Since the curves are generated simultaneously, their difference is not influenced by electronic processing such as compression or noise reduction but reflects only the effect produced by the directional microphones in the instrument. The real-time nature of the test allows the operation of adaptive directional systems to be readily visualized. . . . . At the end of each presentation of the speech passage, the Verifit momentarily presents its unique 1000 tone directional test signal to capture the directional behavior. . . Because the Verifit test chamber is small and anechoic only at higher frequencies, results are not expected to agree with data taken in large anechoic chambers.”

The microphones were orientated so that they faced the left speaker, therefore, the curves denoted “L” represent the front speaker and the curves denoted “R” indicate the side speaker. The first figure below was conducted with the omni microphone, which shows inconsistent differences between the front and side measurements across frequency. That is, what appears to be a small positive directional benefit (higher output for the front compared to the side speaker) at some frequencies is negative at other frequencies. For the 80-dB input levels, there appears to be a 5-10 dB directional benefit across most frequencies. The second figure below was conducted with the directional microphone. In contrast to the first figure, the directional benefit is consistent across frequency and is approximately 10 dB at the 60- and 70-dB input levels. At the 80-dB input level, the directional benefit is about 10 dB greater than what was measured using the omni microphone.

Omni Directional Microphone:

Directional Microphone:

Directional On Ear (KEMAR) Measures
The full directional response of the directional microphone was tested used a KEMAR in a medium-sized sound booth. The Audioscan Verifit was used to collect the measurements with KEMAR being rotated 10 degrees azimuth between measurements. The test signal consisted of a pink noise at 60 dB SPL. Typically, this level would be low to evoke the directional features in a hearing aid, however, since directionality in this case was fixed in the microphone and was not determined by a software-determined threshold level, presentation level was not a concern. Output level at each angle was computed using 1/3-octave wide filters with a frequency resolution of 1/12th octave between 200 and 8000 Hz, thereby generating 65 measurement frequencies at each of 36 measurement angles.

Problems with Feedback
Initially, measurements were obtained using the WDRC code set to prescriptive targets. However, for the purposes of evaluating only the directional response properties of the microphone this is incorrect since WDRC will reduce the differences in level at the output of the microphone. Nonetheless, it is worth commenting on the challenges that immediately arose when trying to obtain ‘on-ear’ measures using an ear level microphone and a device that output as much as 30-40 dB of gain > 2000 Hz. In order to eliminate feedback, two solutions worked. One was to cover the concha of the manikin’s ear with putty (see picture), which is not practical in human trials. The other was to use a disposable foam plug (www.complyfoam.com), as pictured below. Eventually, this method of coupling the earphone to the ear will probably be desirable since disposable plugs can help with infection control. They also come in different sizes and amounts of venting, thereby providing comfort and retention.

image image

Ultimately, the Tympan was programmed with Basic Gain using the conventional earphone coupling (see picture below). Using the volume control knob, the maximum gain that could be obtained before feedback was around 27 dB. Measurements were obtained with gain set to 20 dB across frequency.


Shown below are the polar plots for the directional microphone that were obtained in two conditions. The first is with the reference microphone (contained in the black unit below the ear in the pictures above) set to equalize the presentation level between measurements, that is, with concurrent equalization. This has the effect of eliminating the natural directivity that comes with factors related to head shadow. While not a pure measure of the microphone directivity that would be obtained in a typical free field measurement, it provides a rough estimate of the microphone’s directional properties. The second condition was obtained with fixed equalization, obtained at 0 degrees. This condition shows the combined directivity of the microphone and its placement on the ear (e.g., head shadow).

The polar plots below show the directivity of representative samples of the 65 frequency-specific measurements (different colors). To facilitate comparison, each frequency was normalized so that the angle with maximum output was set to 20 dB. The closer the response is to the center of the plot, the greater the attenuation. The first plot shows that the frequencies between 1000 – 4000 Hz have the greatest directional benefit in this particular test environment and setup, with around 10-15 dB of reduction for sounds in the rear hemi-field between 120-240 degrees. While the maximum output was generally at 0 degrees for most of the frequencies, output level was similar across the front hemi-field +/- 60 degrees. This pattern is consistent with a cardioid directional microphone response. With fixed equalization, the nulls and lobes across frequency are shifted by about -60 degrees, so that lobes are closer to the aided ear and the nulls are closer to the unaided ear. In addition, the directivity at 8000 Hz is significantly increased. All of these effects are expected based on head shadow.

The articulation index weighted directivity index (AI-DI) was computed by normalizing the calculation relative to the lobe of each polar plot. The computed AI-DI was 9.1 dB with concurrent equalization and 8.2 dB with fixed equalization. This is suggests the microphone under test may be a 3rd order directional microphone.

A generic directional microphone (‘mini boom’) with a direct audio jack connection that can be connected to a head worn device was evaluated against a previously tested omni directional microphone (Sony ECM-CS10 Lapel Electret Microphone). The directional microphone under test was estimated to have about 5 dB less sensitivity than the Sony microphone. Using the same default WDRC settings, internal noise (equivalent input noise, EIN) was slightly greater than was measured previously for the Sony microphone. Increasing the expansion ratios successfully reduced EIN by about 3 dB. The 8-channel WDRC was then fine-tuned to prescriptive output targets for a typical sloping mild to moderate sensorineural hearing loss. Due to the increased gain and low expansion kneepoints, EIN was significantly higher for both microphones. Low distortion was noted for both microphones. The directional properties of the mini boom microphone was tested in a test box and on a KEMAR. Using the front-side ratio in the test box, the directional benefit was estimated to be around 10 dB compared to the Sony microphone at each input level. When the microphone was placed above the ear on KEMAR, feedback immediately became a problem when the Tympan was programmed for the sloping mild to moderate hearing loss. Feedback was eliminated when a foam tip was used to seal the earphone in the ear canal. For evaluation, the Tympan was programmed with Basic Gain fixed at 20 dB across frequency using the conventional earphone coupling. Polar plots were generated with 10-degree precision at 65 analysis frequencies. When presentation level was controlled to remove the effects of head shadow, 1000 – 4000 Hz had greatest directional benefit, with around 10-15 dB of reduction for sounds in the rear hemi-field compared to the front hemi-field. When the effects of head shadow were included in the measurements, the nulls and lobes across frequency were shifted by about -60 degrees, so that lobes were closer to the aided ear and the nulls were closer to the unaided ear. In addition, the directivity at 8000 Hz was increased significantly. It is hypothesized that the microphone under test is a 3rd order directional microphone with a cardioid response.

We are also looking into alternative mics for tympan but have found that plugging into the mic input does not mute the built in mics. What are we doing wrong?

We find bad responses when using the built in mics. I suspect this is from acoustic resonance within the housing. We removed the housing and got smooth responses. Acoustic shielding between the mics and the outside world seems to be needed.
Arthur Boothroyd

I tried and failed to replicate having both mic inputs active at the same time.

During the trials, I commented out the line that selected the built in mic:

// audioHardware.inputSelect(TYMPAN_INPUT_ON_BOARD_MIC); // use the on board microphones

audioHardware.inputSelect(TYMPAN_INPUT_JACK_AS_MIC); // use the microphone jack - defaults to mic bias 2.5V

When both lines were uncommented the mic input that was active was whichever line was called last.