There is a multitude of apps for Smartphones that can turn them into sound level meters, recording devices, playback devices, and even allow them to be coupled with external devices for hearing aids via Bluetooth or other wireless protocols.

However, each step in the recording/playback/control pathways can add some error to the final measured result. Some of these errors are small and are well-discussed in the literature such as the realm of usefulness of a Smartphone as a sound level meter; other sources of error are less discussed in the literature.

Starting at the “beginning”, Smartphones have microphones that are quite amazing for their size and whose production quality is quite high given the sheer numbers that need to be manufactured. They are MEMS microphones and that stands for MicroElectroMechanicalSystems.  These are from the cellphone industry and many have built in A/D converters. MEMS microphones use silicon instead of Teflon; an element that can get hotter before it starts to lose electrons.  MEMS microphones still lose electrons though which is why the use a charge pump to replace the lost electrons on the microphone diaphragm.

Electrons are replaced on the microphone diaphragm to maintain the function. Courtesy of www.comsol.com

So far so good.  But many of the modern Smartphones have more than one microphone.  The reason is the same reason as why modern hearing aids also have more than one microphone: to allow the microphone system to be directional in the sense that it helps to reject unwanted noise if the noise is coming from the rear direction.  Although this noise-rejection strategy can be very useful, it may be problematic if the noise is what you want to measure.  

For example, if one wants to use an app that turns the Smartphone into a sound level meter, then this can pose problems if the Smartphone is not held properly or aimed appropriately at the noise source.

While use of a MEMS microphone has added substantially to the quality of modern Smartphones- you can leave the Smartphone in your car on a hot summer’s day with temperatures hovering around 40 degrees Celsius- there are limitations with the signal to noise ratio (SNR).

The SNR is the difference in decibels (dB) between the quietest sound that can be transduced by the microphone and the loudest one.   The SNR for modern MEMS microphones is on the order of 60 dB.  The difference between the softest speech sound (‘th’ as in ‘thin”) and the loudest sound (‘a’ as in ‘father’) is about 35 dB.   A MEMS microphone that can handle a 60 dB range is therefore quite adequate for speech.   It may be a different story for inputs that have a greater dynamic range such as music.   Some form of compression or AGC would be required to match the large dynamic range of music to the 60 dB SNR capability of a MEMS microphone.

And to further complicate things, many microphones have a frequency response that have been optimized for phone communication-after all, these are telephones.  The issue arises if one wants to measure sound sources that are not speech or speech-like.  This many include using the Smartphone as a noise measuring device or as a recorder of music.

There is even a YouTube video (actually I am sure that there are many) describing how to connect an external microphone to a Smartphone.

The frequency region of concern consist of the bass notes below middle C on the piano keyboard.  Having said this, app developers can use  an “Application Programming Interfaces” or API that are provided by the Smartphone manufacturers on their operating systems (OS).  These APIs can assist to provide a broader frequency response as well as disable certain features such as compression or AGC that would normally be implemented.  Not all developers use these APIs with the result that two seemingly similar apps may function quite differently.

A strategy that some app developers have used is to utilize an external microphone that is connected directly to the Smartphone.  This obviates many of the concerns about using the internal microphone(s) and some of the design parameters, such as compression and directionality, which may have posed a problem.

Not all Smartphone microphones are created equal.  Actually they are, but the subsequent hardware and software design decisions can create quite different animals.

In part 2 of this blog series, the potentially deleterious issue of time delay will be discussed.

In part 1 of this blog series, some characteristics of the occlusion effect were discussed.  This refers to low frequency voice or musical energy (below 500 Hz) that is transduced (typically through the boney portion of the ear canal) and unless the ear is blocked up, is not heard- the low frequency sound energy escapes out of the unoccluded ear canal.   However, when blocked by a hearing aid, an in-ear monitor, or an earplug as in the case of a musicians’ earplug, then this low frequency energy is trapped in the era canal and transmitted onwards to be heard.

This effect only occurs for sounds with significant energy that is in the 100-400 Hz region and therefore is restricted to bass instruments and for the vowels [i] as in ‘beat’ and [u] as in ‘boot’.

Of course, one great strategy would be to play the piccolo or learn to speak a new language with no high vowels, but alas, all languages of the world have the vowels [i] and [u]… even Klingon!

Courtesy of www.iemreview.com

This is something that can clinically be measured and takes about 15 seconds.  In the following case, this was of a trombonist who was referred to me to see if I could resolve the occlusion effect even though his ER15 musicians’ earplugs had been remade with a 25 foot long earmold!

The solution ended up being that I simply drilled a 1.4 mm diameter vent through his hearing protection.  This did compromise the low frequency attenuation by about 4-5 dB but at least he was wearing the earplugs and could successfully use them while practicing and during live performances.

So, here is the strategy:

1.       Calibrate the real ear measurement device in the normal fashion.

2.       Disable the loudspeaker and the reference microphone of the real ear measurement device.  Various manufacturers do this in different ways but you can check the owner’s manual.  For Audioscan turn to the stimulus level to “0 dB”.  For Frye turn the stimulus level to “off”.  In both cases, this turns out the real ear measurement device’s loudspeaker and disables the reference microphone.   Of course, the person who we are working with shouldn’t move much during this measurement since the reference microphone is now disabled.

3.       Have the person utter and sustain the vowel [i] as in ‘beat’ for the length of the sweep (Audioscan) or the FFT acquisition (Frye) WITHOUT any hearing protection in place.   This will trace out the formant (resonant) structure of the first formant (F1) for [i] which is around 250 Hz. 

4.       Then do number 3 again with the hearing protector (or turned off in-ear monitor or turned off hearing aid) in place.

5.       This difference (measured at 250 Hz) is an estimate of the occlusion effect.

6.       Finally, verify that your clinical resolution of this problem has actually worked.

Other than performing a pair of real ear measurement results (unoccluded and occluded) you can ask your client to say the two vowels [i] as in ‘beat’ and [a] as in ‘father’.  If the [i] is louder than the [a] then there is still significant occlusion effect that still needs to be addressed.  If the [i] and [a] are similar in loudness, then the occlusion effect has been resolved. 

This would work as well with [u] as in ‘boot’ instead of the [i] but both of these two high vowels have the same low frequency first formant (at or below 250 Hz).

When I first began working with musicians in the 1980s after the invention of the ER15 musicians’ earplugs in 1988, I made the earmold bores as long as the person could tolerate- which wasn’t very often.  Today, whenever I have a vocalist, a brass player, or a single reeded instrument (clarinet or saxophone) that was being played, I routinely order it from the earmold laboratory with a vent.   This does compromise the low frequency attenuation somewhat but inevitably ensures wearability of these hearing protectors.