Microphone recording quality is concerned with the performance of the microphones and the quality of speech, when used in an ideal (quiet) environment. It shows how natural, neutral, extended and intelligible speech would be as heard by the listener.
It is part of our series of tests of the microphone of headphones.
Recording quality score is comprised of five tests: low-frequency extension (LFE), frequency response standard deviation (FR. Std. Dev.), high-frequency extension (HFE), harmonic distortion (Weighted THD), and gain.
When it matters
A microphone with a good recording quality ensures that the person listening to you hears a full, clear, and easily understandable speech. Therefore, it is important whenever a good quality of speech transmission is needed. It is worth noting again that recording quality doesn't take into account the noise handling capability of the microphones, and the results are only valid in a quiet environment. For noise, we test this separately here.
Our isolation tests are performed in a 6' x 3' x 3' isolation box. The box is partially treated with bass traps and absorbers, and an M-Audio BX8 Carbon speaker is placed at one end of the box with the HMS (Head Measurement System) being positioned at the other end. The speaker is positioned at an angle in order to create a balance of direct and diffuse sound.
The target for our recording quality test is the response of a measurement microphone placed 5cm away from our dummy head’s mouth calibrated to 94dB SPL, and based on that, we calculate low-frequency extension, frequency response, high frequency extension, harmonic distortion, and gain.
The signal for the LFE, FR, HFE, and gain tests is a pink noise (random noise with equal energy per octave) limited between 20Hz - 20KHz. The reason for using a pink noise instead of a swept sine is to circumvent the effects of the automatic gain control (AGC) algorithm that most the active/Bluetooth headphones use. For the Weighted THD test, a swept sine is used since it is better at catching the distortion artifacts produced by the AGC algorithm, compared to a pink noise.
Each device is tested 5 times and the results shown here are the average of 5 passes. For headphones with an in-line microphone, the microphone is fixed at 0/90/180 degree angles relative to the speaker. For the last 2 passes, the microphone is allowed to dangle freely in order to simulate the random positions the microphone could end up in real-life situations. For the headphones that have the microphone built in the ear cup, a re-seat is performed before each pass. The headsets that come with boom microphones are measured with the boom in the optimum position for the first pass, and for the subsequent 4 passes, the microphone is moved up/down/front/back relative to the optimum position.
Recorded speech is the recording of a reference track captured using the headphone's microphone. It can be used in comparison with other microphone recordings on the website, or it can be compared against our reference recording captured using our reference microphone.
Reference Recorded Speech:
LFE is the lowest frequency at which the microphone response reaches -3dB of our target response. It can be considered as a metric for how deep and full the voice recording will be with the microphone. The lower the LFE the better, however, values below 150Hz should be considered good. It should also be noted that LFE has very little effect on the intelligibility of speech and even a microphone with an LFE of 500Hz could still be very well understood (even though it will sound very thin).
A number of factors, such as the microphone's polar pattern and proximity to the source, affect the low-frequency extension of microphones. However, as a rule of thumb, the closer the microphone to the mouth (as is the case with boom mics) the better the LFE tends to be.
Frequency Response Standard Deviation
Frequency response is calculated between LFE and HFE, ignoring the rest of the spectrum. It is scored by calculating the amount of deviation between the microphone's response and our target response. FR Std. Dev. can be considered as a metric for the neutrality and spectral balanced of the recorded speech. The effects of FR on speech intelligibility is more than HFE, but less consequential than LFE.
HFE is the highest frequency at which the microphone response reaches -3dB of our target response. It can be considered as a metric for clarity, presence, and intelligibility of speech, since the frequency range of 1KHz-8KHz has the most effect on speech comprehensibility. The higher the HFE the better, however, values above 8KHz should be considered good. Compared to the other tests present in the recording quality section at the moment, HFE has the highest correlation with speech quality and intelligibility.
A number of factors, such as the microphone's polar pattern and proximity to the source, as well as the connection protocol, affect the HFE performance of microphone. Generally, wired headphones or wireless headset with their own proprietary protocol tend to do better at HFE, compared to Bluetooth devices. This is due to the fact that Bluetooth Hands-Free Profile (HFP) supports audio only up to 4KHz, as opposed to the other protocols that can go as high as 20KHz.
Weighted Total Harmonic Distortion
Harmonic distortion for microphones is captured using a swept sine, as opposed to a pink noise used for the other tests. A swept sine signal can expose the artifacts caused by the automatic gain control (AGC) of the headphones better. The final value is calculated using a weighted filter that gives more importance to high-frequency distortion since humans are more sensitive to harmonic distortion at higher frequencies. However, it should be noted that THD (total harmonic distortion), compared to LFE, FR, and HFE, doesn't have a high correlation with perceived recording quality, even after applying a weighting filter.
Our target SPL (sound pressure level) for our microphone tests is 94dB SPL. However, the gain on all microphones can be adjusted either on the headphone or on the device they are connected to. The gain value is acquired by setting the input level of the microphone to the maximum and recording the level. It is reported relative to the 94dB reference, therefore a gain of 10dB means that at the highest setting, the recorded SPL was 104dB SPL. As a rule of thumb, Bluetooth headphones tend to get the highest gain value because of their dedicated internal microphone amplifier.
What is not included
A few elements that you could care about are not included in the score at the moment:
- Evaluation of the codec quality (PEAQ)
- Evaluation of speech quality (POLQA)
If you feel there is an item missing that should be included, please let us know in the Q&A section.