##Electroacoustic Evaluation (Flat Gain, No Compression)
To further evaluate the frequency response of the Tympan unit, the code (‘BasicGain’) was set to amplify the broadband signal by 20 dB without frequency shaping or compression of any sort. Gain was further adjusted by another -10 to +20 dB via the volume control knob. Three tests were conducted. First, the response to 90-dB SPL broadband pink noise was obtained for -10, 0, +10, and +20 dB volume settings. Second, the response to 90-dB SPL tones was obtained for the same volume settings using the MPO signal in the Speechmap test. Finally, total harmonic distortion measures were obtained for the same volume settings for 90-dB SPL tones to estimate the levels and frequencies that were in saturation.
One thing that is apparent from the pink noise and MPO data is that the basic frequency response does not change as the volume is changed from -10 to +10 dB. This indicates that the changes to the frequency response that were observed with the 8-channel WDRC code was due to the multichannel output limiting compression which prevented the high-frequency channels from reaching saturation levels. Another thing that is apparent from the data is that there are peaks and valleys in the frequency response that are similar to those observed with the 8-channel WDRC code. This suggests that resonance is responsible for peaks in the frequency response and not channel summation, which was previously proposed to be an explanation. Finally, obvious changes in the frequency responses can be seen for the +20 dB volume setting, which is a strong indication that output levels were at saturation. The output levels are not much different from the levels for the +10 dB volume setting; therefore, it is possible that saturation levels were also reached for this latter setting at some frequencies. The distortion data shows that there was minimal distortion for the two lowest volume settings, which indicates that the front-end components (microphone, ADC, etc.) are not in saturation with these high input levels. However, as suggested earlier, significant distortion is present for the +10 dB volume setting at frequencies corresponding to some of the peaks in the frequency response, 400, 1000, and 2000 Hz. Substantial distortion is seen at almost every frequency with the +20 dB volume setting, especially at 1000, 2000, and 4000 Hz. Previously, with the 8-channel WDRC code, distortion was present at only the low frequencies. The current data provide further evidence that the low frequencies were in saturation while the high frequencies were in compression limiting.
##Electroacoustic Evaluation for Other Earphones
Out of curiosity, the previous tests were also conducted using two other sets of earphones,
Roland CS-10EM earphones and Apple EarPods MD827LL/A earphones. It should be noted that the Roland earphones have built-in, ear-level microphones. These microphones were not used for the present measurements. The Apple earphones were carefully puttied onto the coupler so that the putty did not cover the front and rear earphone ports, which sort of acts like a bass reflex on subwoofers and some desktop speakers. In actuality, the lower port on the Apple earphones influences the low frequencies while the earphone ports influence the mid frequencies (it was a very interesting to see how the frequency response dramatically changed as I selectively covered the ports). It should be noted that in everyday use, the front earphone port is designed to be blocked by the wearer’s concha.
As can be seen from the figures, the frequency response of the Roland earphones is very peaky, with a 20-dB peak-valley difference in some places. Interestingly, the frequencies of these peaks are similar to those for the Klipsch earphones, which makes me wonder even more about their source. Compared to the Klipsch earphones, the Roland earphones have much greater distortion at the -10 dB volume setting. The distortion is progressively greater at higher volume settings, especially for frequencies ≤ 1000 Hz. These results highlight the importance of carefully choosing the right earphone for this project.
The Apple earphones also have a somewhat peaky frequency response, which as noted before, changes dramatically as the earphone ports are selectively blocked. Again, the frequencies of these peaks are not much different from the other two earphones. The frequency response for the tones shows a much better high-frequency response compared to the other two earphones. Distortion at the -10 and 0 dB volume setting are comparable to the Klipsch earphones, but it is higher at the +10 and +20 dB volume settings. Interestingly, for the +10 dB volume setting, where significant distortion is present only for some frequencies, the greatest distortion is seen at the same frequencies (400, 1000, and 1600 Hz) across the three earphones.
#Electroacoustic Summary
Electroacoustic evaluation of the Tympan unit programmed with 8-channel, wide dynamic range compression (WDRC) was conducted using a commercial hearing aid analyzer. In my expert opinion, the Tympan unit performed like an actual hearing aid: (1) throughput delay was measured to be 5.7 ms; (2) input/output curves clearly demonstrated the action of the WDRC, output limiting compression to control maximum output levels, and volume changes; (3) WDRC attack and release times equaled the nominal values in the software; (4) minimal distortion was evident at most volume settings; (5) internal noise levels were modest and comparable to today’s premium devices; (6) the frequency response was consistent across most input levels and was relatively flat with predictable peaks that probably could be smoothed away; (7) the microphone seemed to demonstrate adequate dynamic range, although the rising frequency response will need to be calibrated out before the signal is processed; (8) the amplifier was able to achieve at least 55 dB of gain with minimal distortion; (9) the Tympan unit with Klipsch earphones seems to be capable of accommodating hearing losses up to 70 and maybe 75 dB HL; and (10) real-ear measurements using a probe microphone were easily performed.
#References
Alexander, J. M. (2016). “Hearing aid delay and current drain in modern devices,” Canadian Audiologist, 3(4).
Alexander, J. M., and Masterson, K. M. (2015). “Effects of WDRC release time and number of channels on output SNR and speech recognition,” Ear and Hearing, 36, e35-e49.
ANSI (2003). “S3.22: Specification of Hearing Aid Characteristics,” American National Standards Institute.
Audette (2017a). “Calibrating my earphones with tympan,” Open Audio Blog, http://openaudio.blogspot.com/2017/05/calibrating-my-earphones-with-tympan.html. Retrieved on June 16, 2017.
Audette (2017b). “Calibrating microphones with tympan,” Open Audio Blog, http://openaudio.blogspot.com/2017/05/calibrating-microphones-with-tympan.html. Retrieved on June 16, 2017.
Audioscan (2016). “Verifit Users Guide v3.14, June 2016,” Audioscan, A Division of Etymonic Design Inc.
Byrne, D., Dillon, H., Tran, K., Arlinger, S., Wilbraham, K., Cox, R., et al. (1994). “An international comparison of long-term average speech spectra,” Journal of the Acoustical Society of America, 96, 2108–2120.
Holder, J. T., Picou, E. M., Gruenwald, J. M., and Ricketts, T. A. (2016). “Do modern hearing aids meet ANSI standards?” Jounal of the American Academy of Audiology, 27, 619-627.