Introduction to electronic stethoscope

acoustics and signal processing

Version 1.00
Page 1 of 16
File name: Introduction to electronic stethoscope acoustics and signal processing

Author

Bjørn Knud Andersen

QMS00027, Version 4.00

Bang & Olufsen Medicom a/s

Struer: Gimsinglundvej 20 Copenhagen: Arne Jacobsens Allé 7 Phone +45 70 301 600 www.medicomnordic.com

DK-7600 Struer, Denmark DK-2860 Søborg, Denmark Fax +45 70 301 700
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

Table of Contents

1 Introduction ......................................................................................................... 3
2 Hardware key specifications ................................................................................... 3
2.1 General introduction to electronic auscultation .......................................................... 3
2.2 Auscultation sensor pickup of signal from the human thorax ....................................... 4
2.3 Introduction to typical auscultation signals and their perception................................... 6
2.4 Implications of A/D-conversion ............................................................................... 7
2.5 Electronic stethoscope key system features and specifications ..................................... 9
2.5.1 Frequency range and filtering ....................................................................... 9
2.5.2 Dynamic range, amplification and electro acoustic signal-to-noise ratio .............. 9
2.5.3 Amplification rate, volume control and maximum output limitation .................. 10
2.5.4 Handling noise and ergonomics .................................................................. 11
2.5.5 Ambient noise.......................................................................................... 13
2.5.6 Sampling frequency, resolution (bit depth) and compression .......................... 14
2.5.7 Hand-ear signal perception effects .............................................................. 14
2.6 Summary table ................................................................................................... 16
3 References ......................................................................................................... 16

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 2 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

1 Introduction
B&OM have worked within electronic auscultation for almost two decades and among other
achievements has developed and manufactured three generations 3M Littmann digital
stethoscopes. This has included e.g. inventing and implementing state-of-the-art auscultation
sensors as well as inventing and implementing real-time embedded software algorithms to
analyze for e.g. stethoscope use status and heart rate and to digitally record and wirelessly
transfer to third party devices e.g. using Bluetooth. Apart from this B&OM has been conducting
independent research and development within automated assisted auscultation diagnostics
both in the areas of cardiovascular and respiratory disease areas.
The present report provides a summary of the most ‘critical-to-design’ aspects influencing the
sensing of sound from the human (or animal) body as well as the signal processing applied to
help preserve signal integrity while amplifying, digitizing and reproducing such signals.

2 Hardware key specifications

Following a general overview introduction to auscultation and also electronic auscultation key
system features and specifications will be detailed both for analogue (or digital) systems and
furthermore for digital systems and devices. The focus will be strictly on features and
specifications directly influencing the audio signal and no in-depth assessment of design details
in-directly influencing such signals, e.g. general styling and ergonomics, materials selection
etc.

2.1 General introduction to electronic auscultation

The below basic figure illustrates the pickup and conversion of typical auscultation signals from
the human body. Here the sensor is illustrates as being a microphone type sensor mounted in
the bell of a conventional stethoscope effectively responding to relative deflection of its
diaphragm due to a combination of vibratory signals from the body and then counter acting
force either from a physician holding the bell (or alternatively/additionally gravitational force
acting on the sensor housing mass which acts to hold in position the housing for non-static
body signals).

Figure 1 Overview illustration of basic principles in electronic auscultation, from [1]

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 3 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

2.2 Auscultation sensor pickup of signal from the human thorax

In order to precisely express opinions on relative performance of various different sensor
systems it is relevant to assess the modelling of how auscultation sounds pickup may be
calculated based on given conditions. References [3] and [4] demonstrate models for both
coupling of sound out through the human thorax and also the conversion from the skin surface
by different sensor types.
The core of the model is the high-impedance velocity sound source acting via two distinct
signal paths across the accompanying thorax impedance, either via the sensor diaphragm or
via the transducer housing. In the following figure Vt is the thorax velocity, Mh is the sensor
housing mass, Mht, Rht and Cht denotes mass, resistance and compliance elements of thorax
impedance coupling to sensor housing, Mdt, Rdt and Cdt denotes mass, resistance and
compliance elements of thorax impedance coupling to sensor diaphragm and finally Cd which
represents the sensor diaphragm compliance.

Figure 2 Electrical equivalent of the sound transmission for physiological sound

signals to the auscultation transducer where (derived from [4])

According to [3] and [4] the thorax impedance elements may be estimated based on sensor
application surface area and application force (pressure). In example, for a Ø30 mm sensor
applied with 0.6 N force the parameters Rt=4.8 Ns/m, Mt=5.4 g and Ct=0.62 mm/N may be
used for calculating a representative thorax impedance. In order to apply two distinct thorax
impedances to either couple to housing or to the diaphragm it could be assumed that these
individually will yield the same value (accurate enough for enclosed examples):

1
Zt = Rt + jMt +
jCt
The influence from hand/arm holding the transducer housing may be relevant in some
situations, and the loading contribution from Zha may then be implemented by applying it in
series with Mht and Mh. A ‘typical’ holding force would best be represented by a resistive
element of ~40 Ns/m. Furthermore an inclusion of an enclosed air-cavity directly in connection
with the sensor diaphragm is implemented by adding this loading contribution in series with
Cd.
The force acting on the sensor diaphragm which controls its deflection and hence the
generation of electrical signal to convey to further processing may be calculated using the
following expression:

Zd (Zmht + Zmh)(Zcdt + Zrdt ) (Zmdt + Zd )

H=
Zd + Zmdt Zmht + Zmh + Zcht + Zrht + (Zcdt + Zrdt ) (Zmdt + Zd )
An interesting feature of the sensitivity towards physiological sound signals is the shunting of
the driving signal at low frequencies, in practice at frequencies below the resonance between
the housing mass and its underlying housing related thorax impedance component, which
directly affects sensitivity to low frequencies. Another interesting detail goes with the

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 4 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

resonance between the tissue and the diaphragm impedance which affects the high-end
frequency sensitivity.
The following figure demonstrates these effects (using arbitrary but realistic component
values) either where the sensor is positioned freely on the thorax or where the physician is
holding the sensor in place (applying more back-support to drive amplitude up at low
frequencies). Furthermore, a graph showing the influence from basic to twice as stiff sensor
diaphragm (half the resulting compliance) which increases sensitivity at high frequencies.

Figure 3 Simulation results for a realistic auscultation sensor system demonstrating

some low-frequency and high-frequency sensitivity behaviour depending
on hand-holding sensor on chest (left graph) and applying diaphragms of
various stiffness (right graph)

As shown through these examples, the effect of hand-holding the sensor onto the chest while
performing auscultation) has the positive effect (which is quite normal) of not only extending
the frequency range towards very low frequencies as the housing backing restricts the housing
from moving with the chest vibration below the resonance point but at the same time also
effectively dampen the noticeable low frequency resonance thereby typically improves overall
sound ‘clarity’.
Furthermore, the more ‘stiff’ diaphragm, which very well could be a result of optimizing an air-
cavity between the diaphragm and the microphone in the above illustration to just allow the
diaphragm deflection (smallest possible air-cavity), has the positive effect of increasing
sensitivity towards higher frequencies.
In short summary, when further down in this report the frequency range of a stethoscope or
sensor system is addressed it is of significance to understand how such performance is
achieved. E.g. as long as acoustic behaviour during auscultation is stable then filtering may be
applied whereby ‘any outer’ frequency response function can be achieved, however only if the
core sensor (and its application) is diligently optimized then this will also provide best possible
dynamic range and signal-to-noise performance. Furthermore, with respect to future sourcing,
collaborating or development of a novel auscultation sensor system for picking up delicate
respiratory signals it is therefore essential to not rely only on provided specification data but
combine such information carefully with an understanding of the core sensor principle to
secure high-quality data acquisition.
The simple simulation/modelling provided above is not intended to necessarily be highly
accurate but will be used in the following to point out the influence of specific design elements
affecting the performance of the stethoscope or auscultation sensor system.

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 5 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

2.3 Introduction to typical auscultation signals and their perception

Auscultation signals from the heart, lungs etc. are generated covering a relative wide
frequency range, e.g. from a few cycles per second (i.e. a person’s pulse rate) up to several
thousand Hz (i.e. ~5000 Hz for wheeze and tracheal sounds). While coupling through the body
skin and tissue etc. these sounds are however subject to low-pass filtering (i.e. in accordance
with the above transmission model) whereby higher pitched sounds are attenuated
significantly more compared to lower frequency auscultation signal components.
The lowest frequency components are typically related to cardiac activity generating signals
which are of relatively high amplitude (and consequently simple to assess using any type of
stethoscope) whereas the highest frequency components are related to respiratory activity and
of much weaker amplitude. The below table (borrowed from [2]) compares a range of
respiratory sounds based on frequency contents and duration.
In the auscultation situation, the high-amplitude and lower frequency cardiac signals are often
listened to simultaneously with lower-amplitude and higher frequency respiratory sounds which
adds complexity in performing auscultation on humans as the human ear may not be able to
pick up the finer details from respiratory and cardiac murmurs while in presence of much
coarser heart beats. These masking effects occur both in the frequency domain, where
simultaneously presented sounds will be perceived only by the dominating component, but also
in the temporal domain, whereby e.g. a coarse component will dominate over a subsequent (or
even preceding) finer component.
To compensate, a stethoscope will typically have several distinct filters whereby it can select
individual frequency areas. Filtering may be mechanical, e.g. bell/diaphragm or light/hard
application force use modes of a conventional stethoscope, or by means of filtering in
electronic stethoscope. Even if such filters may be labeled for use with lung or heart sounds
etc. their benefit is actually that one such filter/setting will better attenuate the loud low-
frequency signal components thereby reducing masking effects deteriorating audibility and
perception of the higher-frequency and lower amplitude signal components.

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 6 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

Figure 4 Overview of respiratory auscultation signals from [2]

Masking phenomena is however merely a problem of the human auditory perception, for digital
signal processing algorithms the collective assessment of the entire frequency range and
dynamic range is not an issue. However before being able to make available the full signal to
e.g. a digital signal processing algorithm, is has to transform from the analogue domain and
into the digital (A/D-conversion) and here the ‘extreme’ dynamics of the auscultation signals
also complicates registration that does not critically deteriorate the signal qualities.

2.4 Implications of A/D-conversion

Typical A/D-converters operate with a finite resolution whereby only the portion of the signal
inside e.g. a 16 bit wide dynamic range is registered, and below the least significant bit (LSB)
only the quantization noise will later be heard. Since the signal dynamics is attenuated out
through the body an overall amplitude decline towards higher frequencies can be assumed
which then allows pre-emphasis (frequency dependent weighting) of the signal to, in analogue
domain, amplify the higher frequency contents prior to digitization.

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 7 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

The below figure demonstrates the situation where a normal respiratory auscultation recording
is registered either without (left side) or with 1st order pre-emphasis of the higher frequency
components.

Figure 5 Demonstration of frequency spectrum for a normal respiratory

auscultation signal to be A/D-converted with the indicated high and low
dynamic range boundaries. Left side shows normal (neutral) conditions
and right side plot shows improved situation with 1st order pre-emphasis
towards higher frequencies.

When comparing to the digitization upper and lower boundaries it is obvious that with under
normal (neutral) conditions then the high-frequency signal components lies very close to lower
boundary and hence are represented in the digital domain with fairly poor SNR, assuming a
firm 100 dB dynamic range ranging from -80 to +20 dB (which even is slightly higher than
typical 16 bit precision of 96 dB), and even some details around 700 Hz appears to fall entirely
below the lower limit. If however applying pre-emphasis to the analogue signal and then
digitizing, then these same higher frequency signal components are ‘lifted up higher’ into the
dynamic range whereby the SNR is preserved (e.g. the 700 Hz details now appear to be
represented with ~42 dB distance to the lower limit).
For later playback of the pre-emphasized and recorded signal, obviously the weighting has to
be reversed for a true sound comparable to the original analogue signal and this may be done
by applying analogue filtering after the final D/A-conversion.
The concept of pre-emphasizing the picked-up physiological signal prior to digitization is
particularly useful when developing an auscultation system for general purposes, e.g. both
low-frequency cardiac sounds and higher-pitched respiratory components and may be of
secondary relevance if narrowing the frequency range significantly. In other words, the
concept may not be critical for initial indications related e.g. ‘only’ with IPF (if its signal
components exist only in a narrow frequency range). However later expansion of therapeutic
scope might stress performance such that optimal digital representation will only be
accomplished by applying some sort of frequency weighting/shaping to preserve sufficient
upper/lower calculation margin in the digital domain.

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 8 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

2.5 Electronic stethoscope key system features and specifications

2.5.1 Frequency range and filtering

In accordance with the above an electronic stethoscope, or an auscultation sensor, should
ideally provide sensitivity towards sounds ranging from at least 20 Hz and e.g. as high up as
5000 Hz representing overlap between the human hearing range and the presence of body
sounds of relevance.
Within this span, e.g. for the electronic stethoscope intended to directly reproduce sound into
the user’s ears, filtering should preferably be provided to effectively separate low-frequency
components from higher-frequency details thereby limit masking phenomena. A good start
here would be e.g. one filter covering the range from 20-200 Hz and another filter covering the
range from 200 Hz and upwards.
Even better however it would be to target a filter directly to a frequency range of particular
interest (if such knowledge is available) thereby effectively excluding most other sounds and
leaving only the key information carrying signal. Such an approach could easily be imagined
for a solution directed towards IPF diagnosis e.g. focusing a filter sharply around the 500-900
Hz range. Of course, if not listening to the signal but rather sending to external analysis for
diagnosis the filtering is not directly relevant, as signal processing will be designed to extract
exactly the signal needed.

2.5.2 Dynamic range, amplification and electro acoustic signal-to-noise ratio

The dynamics in auscultation signals is very large, both within individual signals (cardiac heart-
beat and murmurs, low frequencies and high frequencies) but also due to the face that the
sound source in the practical setting may not be controlled. E.g. with a very slim person sound
from heart and lungs may travel to the skin surface with minimal attenuation while for an
obese individual there can be extensive damping of interior sounds prior to becoming available
for pickup using a stethoscope or another auscultation sensor.
Although in laboratory settings humans are capable of exploiting up to approximately 120-130
dB dynamic range, range between threshold of hearing and threshold of pain, we are typically
only in practice able to ‘activate’ 90-100 dB for ‘mixed’ signals. This range is not at all far from
the dynamics to be anticipated during auscultation and although filtering and amplification may
be employed to ‘extract’ weaker sounds then a physician may typically not know in advance
where exactly to look meaning that the stethoscope shall provide the entire spectrum and/or
dynamics. While amplification may increase the audible level of sound generally, so to expand
the dynamic range as it lifts lower level sounds into the audible range, it will similarly increase
the level of noise (not carrying information) whereby the result may not necessarily be positive
as the signal-to-noise ratio now becomes critical to whether this particular sound can be
interpreted and a diagnosis thereby extracted. Moreover, noise present may not necessarily be
associated with the same filter band but could be due to the electro acoustic system generally
causing broadband noise to be generated across the frequency spectrum and hence also
potentially in regions where the human hearing is even more sensitive and hence the noise is
perceived relatively louder.

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 9 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

Figure 6 Human hearing levels across frequency

To reflect the relevance of taking into account the frequency dependent threshold of hearing,
usually signal-to-noise ratio will often be expressed in terms of maximum undistorted (e.g.
below 2% total harmonic distortion for a pure tone played back) output level’s elevation over
the system A-weighted noise level.
The higher SNR the better performance but generally a number higher than 75-80 dB will
provide acceptable user listening comfort and details audibility. Again, if not listening to the
signal but rather sending to external analysis, the SNR performance is not directly relevant as
long as signal processing is able to extract relevant, e.g. IPF, signal components. Instead, the
‘fundamental’ pick-up and digitization of this signal (as discussed above) are the key features
to pursue.

2.5.3 Amplification rate, volume control and maximum output limitation

As briefly touched upon previously, the use of amplification may be relevant to pull-up weaker
sounds into the dynamic range of the human hearing and typically amplification in an
electronic stethoscope is implemented together with a volume control so that the user may
vary amplification flexibly depending on signal strength, e.g. depending on different type of
signals, different types of patients or even different listening sites within one particular patient.
A typical approach to amplification and volume control is to have the lowest volume setting
represent amplification similar to a conventional (mechanical) stethoscope and then have e.g.
8 steps of ~3 dB each (clearly audible increment) thereby reaching a total of 24 dB
amplification at highest volume setting.
Naturally any applied filtering will also influence the stated amplification rate meaning that
such 24 dB amplification should be referenced to possibly a single selected frequency for the
conventional stethoscope, e.g. 100 Hz, where typically the purely acoustical amplification is
relatively high (the tubing length will facilitate this acoustical amplification with the
conventional stethoscope). If comparing to other areas in the frequency range, e.g. at higher
frequencies where the conventional stethoscope has a much poorer amplification, the stated
amplification rate of the electronic stethoscope could of course be much higher (but the
number would be practically useless unless this selected frequency is of particular interest in
the application).

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 10 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

While applying amplification and then changing input signal may result in sudden increase in
signal amplitude which may be associated with signal distortion due to amplitude restrictions
and/or uncomfortable or even hazardous signal levels (which also may be further pronounced
due to presence of distortion). For that reason it may be desirable to incorporate signal
compression so that e.g. towards higher level (absolute level when compared to the electronic
stethoscope capabilities) the amplification rate is automatically lowered. Also it may be
desirable to incorporate a hard upper limit for sound reproduction, such that under no
circumstances will the acoustical signal reproduced by the stethoscope system exceed e.g. 140
dB SPL (impulse measurement). As much as such an upper limit might sound very loud indeed
it is worth noting that conventional stethoscopes are perfectly capable of producing peak levels
well exceeding 150 dB SPL (e.g. tapping on the diaphragm or rapidly placing it on the chest of
a patient).

Figure 7 Illustration of some native approaches to dynamic output limitation

The above figure shows the native features of such limiting either by a ‘soft clipping’ to lower
amplification rate for signals exceeding a predefined threshold or entirely capping the signal
amplitude for such higher levels.
For the ‘soft clipping’ approach many different implementations would be possible, e.g.
optimizing for lowest resulting distortion or for least risk of exceeding a ‘hard’ threshold. In
general terms, the sooner (lower levels) such limiting begins the least audible the result will
be, however non-linear amplification may possibly influence the character of certain
auscultation sounds which then could be critical to clinical utilization (i.e. compared to
conventional mechanical stethoscope).
Furthermore such a clipping feature may be common to the entire signal, e.g. based on the
highest amplitude present, or subject to individual clipping strategies for signals in different
frequency ranges, e.g. a soft compression and/or hard limiting feature for frequencies in the
20-200 Hz range but no compression or limitation for signals beyond 200 Hz. There are no
rules or standards for such details, and all comes down to the specific requirements rising from
perceived operator listening sound quality and no such output limitation would influence any
remote signal analysis.

2.5.4 Handling noise and ergonomics

Having addressed several core relevant electro acoustical features of an optimally designed
electronic stethoscope, there are still design aspects which hold significant influence onto final
real life performance. I.e. the pick-up of handling noise is one vital area which holds potential,
if the device is designed poorly, to ruin many technically optimized features. Having said that

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 11 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

however it is very difficult to specify and quantify how exactly a device should be designed to
ensure minimal disturbance from handling noises as this will depend much on the specific use
situation being faced.
The sources to handling noise come from a long list of interactions between device, operator
and patient and the following table will highlight some of these while attaching brief comments
on means to optimize design.

Figure 8 Illustration of three main sources to handling noise during auscultation

a) Obviously one of the most central sources to handling noise is scratching directly at the
interface between the stethoscope diaphragm and the patient skin where the main body
of auscultation sound is assessed. Not only friction of larger surfaces here but also the
presence of any edges etc. will produce noise. In the presence of such movement over
the patient to stethoscope interface one of the most direct ways to limit generation
diaphragm scratching noise is to minimize the friction e.g. by selecting optimal
materials and/or surface treatments, coatings etc.
b) Another type of input to handling noise is scratching in the operator to stethoscope
interface while holding and operating the stethoscope. Part of this is associated with
e.g. finger scratching on the surface of the stethoscope structure thereby due to friction
generating noise which can travel through the material (or even through air) to be
picked-up by the sensor. Again, in the presence of such operator handling stethoscope
surface one of the most direct ways to limit generation diaphragm scratching noise is to
minimize the friction e.g. by selecting optimal materials and/or surface treatments,
coatings etc. Furthermore however, the pick-up of such noise in the sensor can be
minimized through designing the device mechanical structure for minimal transduction,
e.g. using very stiff or dampened materials and possibly integrating vibration ‘traps’
(e.g. a physical node over which a structural wave cannot transmit its energy).
c) Another significant part of operator handling noise is associated vibration which pushes
the stethoscope towards the patient skin whereby relative vibratory signals are
generated directly on the diaphragm for pick-up by the stethoscope sensor. There are
no ‘easy fixes’ to eliminate such noise generation. However, as typically these noises
are associated with very low-frequency components originating from operator muscle
tremor (predominantly below 100 Hz and most energy below 20 Hz), such noise may be
handled through amplification and filtering, e.g. during subsequent processing in the
stethoscope (before presenting to the operator) or not at all if ‘only’ sending for remote
signal analysis where any noise will not disturb the listening experience). If it can be
determined that such filtering will be performed because no relevant physiological
information will ever be present here, analog filtering may be integrated directly in the
sensor thereby possibly relaxing dynamic range requirements for digital representation
of the signal (which again may result in a more efficient and lower-cost system).
On top of specifically suggested mitigation approaches the device styling and ergonomics plays
an important role in limiting the handling noise contributions, the more easy and natural it is

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 12 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

for the operator to hold and control the stethoscope, the better he will be able to do so in a
controlled and calm manner. Moreover, if further silence is required it may be solved e.g. by
strapping the sensor head to the patient thereby completely eliminating the operator influence
during auscultation.
Testing and quantification and comparison of performance is highly challenging for handling
noises as styling and ergonomics plays a big role, meaning that comparison is ‘only’ possible if
narrowing down the isolated effect of single design elements, e.g. different types of coating
applied in various positions on the stethoscope. In addition, also the rate of amplification and
filtering approach plays a significant part as in some systems frequency areas of significant
noise contribution might not be amplified similarly.

2.5.5 Ambient noise

While handling noise may be seen upon as a ‘necessary evil’, the influence from ambient noise
may be considered as being more ‘voluntary’ as the auscultation always could be moved to
sufficiently quiet surroundings. Nonetheless, in a clinical setting there may be other (e.g. life
supporting) devices making noise apart from fans, air-conditioners etc. all which potentially
may influence the auscultation signal thereby disturbing the accuracy (or speed) of diagnosis.
In accordance with [4] the influence from air-borne ambient noise present in the stethoscope
surroundings during auscultation may be modelled by its resulting force implied onto the
sensor housing structure, whereby it is picked up by the stethoscope sensor following
movement of the sensor into the patient’s skin, similar as for handling noise type ‘c’ above. In
addition however there will be minor contributions coming via the tissue and/or in the interface
between the sensor and the patient’s skin. The following figure roughly illustrates the situation.

Figure 9 Entry paths for ambient noise into a stethoscope sensor system

Unless dedicated technology is available, as e.g. demonstrated in [4] for a passive approach,
the options to reduce ambient noise through design are limited as a major contributor is
associated with the in-coupling of noise to the stethoscope sensor housing structure. Even so
supplementary parameters such as stethoscope mass (the higher the better), closeness of
interface to patient skin (the tighter the better) may prove significant in the practical setting.
For a radical approach one could consider acoustic insulation of the sensor head (applying a
solid shell around the entire auscultation site) but such solutions may not be commercially
viable (e.g. due to size and handling) and no marketed devices are identified demonstrating
such principles.
On top of isolated mechanical and acoustical approaches however, amplification and filtering
may help to attenuate unwanted ambient noises and then of course there exist active noise
cancelling approaches to cancel such ambient noise, e.g. based on the auscultation sensor
input combined with an alternative stethoscope microphone not subject to body sounds.

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 13 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

2.5.6 Sampling frequency, resolution (bit depth) and compression

In extension of the above recommendations for basic audio specifications the use in digital
systems would imply at least a sample rate of at least ~11 kHz in order to allow a 0.9xNyquist
frequency (half the sampling rate which represents the highest theoretical signal
representation with a given sampling frequency) of approximately at 5000 Hz. To the extent
that the system upper frequency handling limit is reduced, e.g. due to deliberately
disregarding such high-frequency information, the sampling frequency may be reduced
similarly.
Sampling data resolution (data digitized at the sensor level and also used for reproduction to
the physician’s ears) should be at least 16 bits in order to allow a signal-to-noise ratio of 96 dB
however in order to ensure similar accuracy on processed signal the internal calculation bit
depth shall preferably be at least 24-32 bits in order to limit the accumulated effects from
truncation errors etc. (all depending on complexity of calculations to be performed).
Generally, any form of signal data compression should be done only in strict agreement with
an understanding of the physical nature of such signals as otherwise there is a risk of altering
or even destroying and cancelling vital information. This fundamental principle also applies for
auscultation signals, which, due to the fact that such data may represent a multitude of vastly
different body processes and function, should never be compressed without clear clinical
governing evidence. Specifically related to the iLaennec scope, the fine crackles associated
with IPF disease appear very different from typical speech, music etc. for which signals most
‘known’ compression algorithms will preserve well.
From this perspective alone it is not recommended to impose compression algorithms onto
recorded IPF sound recordings e.g. to save transmission time/bandwidth. If optimized
transmission of such recordings becomes essential however, is would likely be possible to
develop special compression approaches which could be validated to preserve IPF components
(and probably the understanding for how to accomplish such compression will be closely
related to the understanding of how to accurately detect IPF signals in auscultation
recordings).

2.5.7 Hand-ear signal perception effects

When introducing digital stethoscope solutions the processes of analog-digital and digital –
analog conversion an all processing taking place in the digital domain, will typically cause
delayed transmission of the acoustic signal picked up by the sensor to the time of presentation
to the ear as compared to an analog transmission system.

Figure 10 Basic configuration of a digitized audio system

The reason for such delay lies in the fact, that for robust and efficient software execution the
digital data will typically not be processed sample by sample but in fact processed in package
structure whereby (at minimum) the delay of ‘filling’ one package will be induced onto the
audio stream. Moreover, if the signal processing requires any minimum number of samples in

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 14 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

order to analyze specific frequency contents etc. this may in itself tie up a much higher
number of samples before reproducing over the stethoscope speaker.

Figure 11 Basic configuration for a buffered digital audio system reading multiple
samples for processing

While under normal (e.g. music audio) any added delay may not represent any problem it
might when performing auscultation as the audio pathway is not the only transmission of
signal from the patient body to the physician’s ears. When resting his and on the patient chest
the physician may feel a heartbeat which then is mentally registered, the associated sound
from that heartbeat will however only arrive following the processing delay as illustrated
below.

Figure 12 Illustration of ‘simultaneously’ perception by audio and tactile auscultation

signal

In rough terms the human brain will not be able to isolate two such signals, thereby
disconnecting the tactile event from the parallel audio signal, for delays below ~30-50 ms and
it should always be ensured that delays are below 100 ms as longer periods not only will be
noticeable but also disturbing to the physician. Such limited delays should normally be easy to
maintain in confined and integrated digital stethoscope systems.
However, if instead of relying on wired connection between system peripherals e.g. the
connection between sensor and DSP, the DSP and headset or both are to be implemented by
wireless interface (e.g. Bluetooth) then yet additional delay builds onto the resulting overall
delay from body to ear. In that case it should preferably be investigated clinically if the
resulting usability (validated safe and efficacious use) is still maintained and comparable to
e.g. a normal conventional stethoscope.
In supplement to these aspects, if based on the signal present inside a digital confined
stethoscope system enabling a wireless link to an outside device, e.g. a mobile App, for visual
graphical signal representation then effects similar to the tactile-audio issue may again
become relevant. In this situation however, the delay over wireless will likely be closer to the
direct sound reproduction through the stethoscope and hence disturbing delay might not be
excessive or even disturbing to the effective use.

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 15 of 16
Andersen BK CONFIDENTIAL
Introduction to electronic stethoscope acoustics and signal processing

2.6 Summary table

The following table summarizes the above established core (audio) specifications for a well-
functioning stethoscope and/or auscultation sensor system. No data have been entered
regarding handling noise as such performance is tightly linked with the specific ergonomic and
also audio filtering design. Nor is any indications entered to represent ambient noise as such
performance is only available for designs that holds specific technology which has certain
strengths and limitations.

Index Parameter Target Comment

Suggested range to cover all possible

auscultation signals (limited ranges typical for
1 Frequency response Min 20-5000 Hz
devices, e.g. 20-2000 Hz or even 20-1000
Hz)
Suggested as ~0.9 x Nyquist for up to 5000
2 Sampling frequency E.g. ~11000 Hz Hz (can be reduced in line with reduction of
overall frequency range handling)
Suggested to allow unmasked assessment of
3 Filtering E.g. 200 Hz HP-filter high frequency area (more filters will allow
more specific listening modes)
4 Amplification E.g. 24 dB Suggested gain over conventional stethoscope
Suggested @ THD max 2 % and A-weighted
5 Signal to noise (SNR) Min 75 dB
noise
6 Output limitation Max 140 dB SPL Suggested peak level to ensure user safety
7 Volume control Min 8 steps of 3 dB Suggested for optimal user convenience
Data representation
8 Min 16 bits (minimum to support recommended SNR)
(external and storage)
Data representation (suggested minimum to still allow for simple
9 Min 24 bits
(internal) audio processing algorithms
10 Maximum audio delay Max 50 ms Suggested to ensure user acceptance

Table 1 Stethoscope system target specifications

3 References
[1] Kaniusas E.: ‘Fundamentals of Biosignals’ - Biomedical signals and sensors I, Biomedical
and Medical Physics, Biomedical Engineering, Springer 2012
[2] Bohadana A. et al: ‘Fundamentals of Lung auscultation’ - The new England Journal of
Medicine 370;8, February 20, 2014
[3] Vermarien H. and Vollenhoven E.: ‘The recording of heart vibrations: a problem of
vibration measurement on soft tissue’. Medical & Biological Engineering & Computing,
1984.
[4] Andersen BK.: ‘Transducer for Bioacoustic signals’ – Patent US7593534B2

File name: BUA1921d Introduction to electronic stethoscopes acoustics and signal processing Page 16 of 16