Published September 2015

Study
on Detection of Audio Anomalies on CVR Recordings

Bureau d’Enquêtes et d’Analyses

pour la sécurité de l’aviation civile
Ministère de l’Ecologie, du Développement durable et de l’Energie
Table of Contents

GLOSSARY 3

1 - INTRODUCTION 4

2 - PURPOSE OF THE DOCUMENT 4

3 - METHODOLOGY FOR EVALUATION OF CVR RECORDINGS 5

4 - MAIN CVR ANOMALIES 7

4.1 No audio data on one or more tracks 8
4.2 Poor quality of cam recording 9
4.3 Interference 10
4.3.1 CAM recording pollution by aircraft power supply 10
4.3.2 CAM recording pollution by CVR power supply 13
4.3.3 CAM recording pollution by CVR GSM pulses 15
4.3.4 Audio popping on CAM track 16
4.4 CAM recording saturation 17
4.5 Erroneous memory management 19
4.6 Phase shift between tracks 20
4.7 Level imbalance between tracks 21
4.8 Playback speed anomaly on a SSCVR 22

This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission
Glossary

CAM Cockpit Area Microphone

Frequency Number of repetitions of a phenomenon within a particular period

Fundamental Lowest frequency of a periodic waveform

Lofargram (Lofar) Spectrum view (LOFAR) : graph with frequency on X-axis and time on Y-axis

Audio signal frequency components

Time (s)

Frequency (Hz)

H (n) Harmonic rank n

Harmonic family The whole of frequencies multiple of the fundamental

Sonagram (Sonar) Spectrum view (SONAR): graph with time on X-axis and frequency on Y-axis

Audio signal frequency components

Frequency (Hz)

Time (s)

SPL Source Power Level (sound analysis view): graph with frequency on X-axis and
level of energy on Y-axis

Audio signal frequency components

level (dB)

Frequency (Hz)

This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission
1 - INTRODUCTION
In recent years, numerous investigations conducted by various accident investigation
authorities (BEA, NTSB, AAIB, TSB, etc.) were slowed down due to the absence or poor
quality of cockpit voice recorder (CVR) data.
This equipment, which is not used daily by aircraft operators, is dedicated to safety
investigations with the sole objective of getting a better understanding of the event
for improvements in flight safety. The critical importance of the CVR is evident to
those involved in the world of aviation only after an accident. It then becomes
unacceptable to the aeronautical community to find unusable recordings on the CVR.
Since the early 90s, the BEA has been involved in the evaluation of the quality of
CVR channel audio recordings on behalf of French aircraft manufacturers. These
assessments associated with audio analysis in safety investigations have significantly
increased the BEA’s experience and improved its database of potential anomalies.
This document should enable stakeholders in the aviation industry to check, at
the very least, if regularly encountered anomalies are present or not on the CVR
recordings from their aircraft.

2 - PURPOSE OF THE DOCUMENT

This document was written to provide a general methodology for checking CVR
recordings (Chapter 3) and list the main anomalies encountered by the BEA during
work on CVR recordings, whether in the context of safety investigations or evaluations
of audio quality for the benefit of manufacturers and operators (Chapter 4). This
document should not be considered exhaustive.
European Regulation 965/ 201 ( PART CAT.GEN.MPA.195 CAT) makes it
mandatory for operators to conduct an annual inspection of CVR recordings.
This inspection mainly consists of checking that the recordings are audible and
intelligible. The methodology is described in chapters 3 and 4.1.
When installing or modifying a CVR system or adding a modification to the
cockpit that can affect this system, a thorough check of recordings is necessary. The
information provided in 4.1 and the following chapters describes the main anomalies
that have been encountered for which the absence should be verify.

This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission
Regulatory references:
ˆ ˆ Commission Regulation (EU) No 965/2012 - 5 October 2012 – Annex IV – Part-CAT;
ˆ ˆ Annex 6 to the International Convention on Civil Aviation (ICAO) Operation of
Aircraft;
ˆ ˆ EASA CM No.: EASA CM - AS – 001 Issue: 01 of 12 June 2012;
ˆ ˆ EASA Safety Information Bulletin n° 2009-28R1 “Flight Data Recorder and Cockpit
Voice Recorder Dormant Failures“. http://ad.easa.europa.eu/ad/2009-28R1
Technical references:
ˆ ˆ ED-55 “Minimum Operational Performance Specification for Flight Data Recorder
Systems“;
ˆ ˆ ED-56A “Minimum Operational Performance Specification for Cockpit Voice Recorder
Systems“;
ˆ ˆ ED-112 “Minimum Operational performance Specification for Crash Protected
Airborne Recorder Systems“ – see. appendices I-C and II-B Maintenance Practices.
Major accidents in which the CVR recording could not be recovered due to a
failure of the CVR system:
ˆ ˆ Auxiliary Power Unit Battery Fire Japan Airlines Boeing 787-8, JA829J Boston, http://www.ntsb.
(1)

gov/investigations/
Massachusetts January 7, 2013(1); AccidentReports/
Reports/AIR1401.pdf
ˆ ˆ Interim report on the accident on June 29, 2009 at sea off the coast of Moroni
(Comoros) to the Airbus A310-324 registered 7O-ADJ operated by Yemenia (2)
http://www.
Airways(2); bea.aero/
docspa/2009/7o-
ˆ ˆ Investigation Report on the Accident to Ethiopian 409 – Boeing 737-800 j090629e1/pdf/7o-
Registration ET-ANB at Beirut – Lebanon on 25th January 2010(3); j090629e1.pdf
http://www.bea.
(3)
ˆ ˆ Press release on 2 April 2014 on the accident to the McDonnell Douglas MD-83, aero/docspa/2010/
registered EC-LTV, on 24 July 2014 in the region of Gossi (Mali)(4). et-b100125.en/pdf/
et-b100125.en.pdf

3 - METHODOLOGY FOR EVALUATION OF CVR RECORDINGS

During an audio evaluation, it is necessary to work on the data of each track recorded http://www.bea.
(4)

aero/en/enquetes/
on the CVR (generally between 3 and 6 tracks) flight.ah.5017/
flight.ah.5017.php
For example, the following audio files are usually recovered:
ˆ ˆ A file with the recording of radio communications and left side pilot microphones
signal;
ˆ ˆ A file with the recording of radio communications and right side pilot microphones
signal;
ˆ ˆ A file with the recording of radio communications, third man (jump seat)
microphone signal, conversations with the cabin crew and the FSK signal when
recorded (see §4.8);
ˆ ˆ A file with the recording of a mix of the first 3 tracks;
ˆ ˆ A file with the recording of the signal from the Cockpit Area Microphone (CAM).

This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission
The number of recovered audio files and their duration varies depending on the type Refer to the
(5)

of aircraft(5). Component
Maintenance
The download of the CVR must generate, at least, the number of audio files expected Manual (CMM) of the
recorder, especially
for the duration defined by the recorder specification (currently more than 30 minutes the “Description and
or 2 hours). Operation“ part.

Some audio analysis software, fairly intuitive, is available (eg Audacity, Samplitude,
Soundforge, etc). Its use makes evaluation easier and more comfortable. This software
makes it possible to navigate through the audio file, to view the waveform and thus
to quickly identify some anomalies (eg absence of signals).

Figure 1 - samplitude audio analysis software interface

The first elements in an audio evaluation are as follows:

ˆ ˆ 1 - Once the audio files are opened with the software, ensure that they contain
audio data with no audio blanks (see § 4.1). The possibility of viewing the
waveform with audio analysis software (see figure 1) makes this operation easier.
ˆ ˆ 2 - It is then necessary to synchronize audio files with each other, which allows
them to be replayed together. A quick listening to the tracks, first individually
then simultaneously, makes it possible to quickly identify some anomalies such
as full or partial absence of audio data on some tracks (see § 4.1), phase shifts
(see § 4.6) or time stretching (see § 4.8).
ˆ ˆ 3 - Listening to the whole file makes it possible to check the quality of the recording.
It is especially important to check if crew members’ speech is intelligible and if
noises and warnings are audible.

This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission
When the signals do not show obvious anomalies, (see § 4.1 and § 4.2), it is acceptable
to quickly browse the files, though passing slowly over high noise level areas (take
off, landing…) or in low noise level areas to check the actual presence of audio data.
If some anomalies are present on the recordings, a more thorough and detailed check
should be conducted.

4 - MAIN CVR ANOMALIES

The main anomalies identified by the BEA are:
ˆ ˆ No audio data on one or more tracks;
ˆ ˆ Poor quality of CAM recording;
ˆ ˆ Interference;
„„CAM recording pollution by aircraft power supply
„„CAM recording pollution by CVR power supply
„„CAM recording pollution by GSM pulses
„„Audio popping on CAM track
ˆ ˆ CAM recording saturation;
ˆ ˆ Erroneous memory management;
ˆ ˆ Phase shift between tracks;
ˆ ˆ Level imbalance between tracks;
ˆ ˆ Playback speed anomaly on an SSCVR.

In the rest of the document, each anomaly is described by:

ˆ ˆ the impact on the CVR recording (anomaly characteristics);
ˆ ˆ the methodology used to highlight it and check its presence on the recordings;
ˆ ˆ non-respect of MOPS specifications from Eurocae documents.

This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission
4.1 No audio data on one or more tracks
This defect is characterized by a lack of useful audio data on one or more tracks on
the CVR. It is immediately identifiable when listening and by observing the waveform
(see Figure 2): only noise associated with aircraft power supply is audible.
This anomaly is the most negative for any work to be carried out on the cockpit voice
recorder because it makes a valuable source of information for understanding the
event unavailable to the safety investigation.
Reference specifications :
ˆ ˆ ED 56A, §1.4.3.a

ˆ ˆ ED 112 §I-1.3.3.a

Figure 2 – No audio data – Waveform overview

No useful audio data

Figure 2 - no audio data - waveform overview

This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission
 4.2 Poor quality of cam recording
This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission

This anomaly can be caused by an intermittent malfunction of the CAM amplification box (Control Unit) or by poor electrical wiring throughout the CAM
installation. The overall recording of the CAM signal may be of a very poor quality.
On listening, this defect leads to random variations in signal level: on some portions of the CAM track, the signal may have a very low level, or even non-
existent – thus no audio signal (see figure 3).
This defect results in a loss of useful audio data on the CAM track (warnings, crew exchanges).
Reference specifications :
ˆ ˆ ED 56A, §1.4.3.a;
ˆ ˆ ED 112 §I-1.3.3.a.

Low amplitude
signal
No signal recorded Poor quality signal: useful audio data are
embedded in wide band noise
Figure 3 - poor recording quality CAM track - waveform overview
9
4.3 Interference
Most of the interference detected in a variety of work carried out on CVR recordings
has an electromagnetic origin and is due to:
ˆˆ CAM recording pollution by aircraft power supply (pages 9, 10);
ˆˆ CAM recording pollution by CVR power supply (pages 11, 12);
ˆˆ CAM recording pollution by GSM pulses (page 13);
ˆˆ Audio popping on CAM track (page 14).
Reference specifications:
ˆ ˆ ED-56A, §6.2.5 -- ED 112, §2-5.3.5 b (effects of interference from other equipment);
ˆ ˆ ED-56A, §1.4.3.a. -- ED 112, § §I-1.3.3.a (definition of the purpose of a CVR and its
installation);
ˆ ˆ ED-56A, §A.1.3.2 -- ED 112, §I-A.3.2 (no electrical interference).

4.3.1 CAM recording pollution by aircraft power supply

This pollution is characterized by the presence of a continuous parasite “sound” on
the CAM track. The cases studied at the BEA showed that this sound was coming from
the aircraft power supply. Listening is then very uncomfortable. This phenomenon
appeared only on the cockpit area microphone track.
The presence of this anomaly on the recording can be confirmed by spectrum
analysis (see Figure 5). This method reveals the presence of a harmonic family with
a fundamental frequency moving around 400 Hz (between 340 and 490 Hz – aircraft
power supply has a variable frequency for the cases studied). It appears that the
greater the harmonic richness, the more severe the pollution will be and render any
listening work difficult.
ˆ ˆ Remark 1:
„„On some CVRs, the frequency causing the pollution is stable. It is then tolerated
because it provides a useful time frame for the investigation (see ED 56A,
§A.1.3.2, NOTE 1 or ED 112, § I-A.3.2.).
ˆ ˆ Remark 2:
„„Audio signal listening and waveform observation make detection of this
anomaly possible but are not sufficient to prove its presence (see Figure 4). The
use of spectrum analysis is essential.
ˆ ˆ Remark 3:
„„It is possible to use signal processing tools to filter this frequency and its
harmonics. However, this technique may be difficult to implement because
the fundamental frequency of the noise varies during flight. In addition, such
filtering should be done with care and precision because if the pollution
covers elements of the acoustic signature of the aircraft, removing it entirely
(fundamental and each of its harmonics) can cause the loss of information
essential to the investigation.

10

This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission
This interference causes listening discomfort which decreases the performance of
the person responsible for carrying out the transcription work. In addition, when this
pollution is severe, audio data present on the CAM track are hardly usable, especially
conversations between crew members. However, it should be emphasized that
regarding listening comfort, ED112 does not mention any restrictions or limitations
on the level of the recorded signal as long as it does not exceed the recording
capability of the whole CVR installation.

Polluted part of the

CAM recording

Figure 4 - CAM recording pollution - waveform overview

11

This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission
 Propeller acoustic signature
This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission

Frequency (Hz)

Time (s)

Lofargram of the polluted CAM recording

Fundamental frequency of the aircraft power
supply (interference origin) Harmonics H2 to H8
SPL of the polluted CAM recording
Level (dB)

Frequency (Hz)

Figure 5 - CAM recording pollution - spectrum view (Lofar et SPL)

12
4.3.2 CAM recording pollution by CVR power supply
This type of pollution can be described as a high level wideband noise (continuous
whistle) whose centre frequency fluctuates slightly according to the flight phase. The
width of this band is approximately 300Hz. This phenomenon appears when the CVR
starts and disappears when the usual power supply of the device is cut off (transition
to RIPS(6)). This interference is generated by the electric power supplied by the CVR to (6)
RIPS : Recorder
the preamplification module (Control Unit). Independent
Power Supply.
This anomaly is easily detectable because it makes listening work quite uncomfortable.
The presence of this type of pollution can be confirmed by a spectrum analysis (see
Figure 7).
ˆ ˆ Remark 1:
„„Like the previous defect, audio signal listening and waveform observation
make possible the detection of this anomaly but are not sufficient to prove its
presence (see Figure 6). The use of spectrum analysis is essential.
ˆ ˆ Remark 2:
„„Depending on the value of the centre frequency, it may be possible to remove
this noise by using a notch filter. However, such filtering should be done
with care because if the pollution covers over some elements of the aircraft’s
acoustic signature, removing it can cause a loss of information essential to the
investigation.

Polluted part of the

CAM recording

Figure 6 - CAM recording pollution - waveform overview

13

This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission
 Frequency (Hz)
This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission

Transition to RIPS Main power OFF

Acoustic signature of
Time (s)

the interference

CVR ON

Lofargram of the polluted CAM signal

Figure 7 - CAM recording pollution - Spectrum view (Lofar)
14
 4.3.3 CAM recording pollution by CVR GSM pulses
This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission

This anomaly is characterized by the presence of interference on the CAM track. Depending on the level of the transmission, such interference may cause
poor perception or even total masking of cockpit conversations. This defect can be detected by listening to the audio recording. A spectrum analysis will
confirm its presence (see Figure 8).
Frequency (Hz)
Time (s)

Fundamental frequency Harmonics H2 to H32

of GSM pulses

Figure 8 - CAM recording pollution by GSM pulses - spectrum view (Lofar)

Remark: the interference generated by GSM pulses depends on the frequency used (eg, 450 MHz, 900 MHz, 1800 MHz, 2.4 GHz). This pulse is transmitted every 4.6 ms. This periodic
(7)

transmission produces, on the CVR acquisition installation, an audible signal whose fundamental frequency is 217Hz and some harmonics.

GSM pulse

Figure 9 - CAM recording pollution by GSM pulses - waveform overview

(7)
GSM: Global System for Mobile communication.
15
 4.3.4 Audio popping on CAM track
This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission

This problem appears when listening as a succession of “pops“ on the CAM track. It is due to the sensitivity of the directional microphone or the Control
Unit to electrostatic discharges (radiation). This event is predominant and easily identifiable on the waveform (Figure 10): It is similar to a succession of
peaks, reaching the electric saturation point of the microphone.
The defect becomes apparent with a spectrum analysis (see Figure 11): it may look like a series of «peaks» with variable levels of energy. The frequency
of occurrence of these peaks is irregular: they may appear quasi-periodically or completely erratically.

Audio « pop »

Figure 10 - audio popping on CAM recording - waveform overview

Frequency (Hz)

Time (s)
Figure 11 - audio popping on CAM recording - spectrum view (Sonar)
16
 4.4 CAM recording saturation
This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission

This saturation is due to an excessive sensitivity of the CAM to low frequency signals. This saturation is acoustic, not electrical, because observation of
the waveform indicates that the signal does not reach the maximum amplitude allowed by the CAM (no clipping on the waveform) (see Figure 12). When
listening, the recording is indeed saturated and all relevant audio signals (conversations, warnings, callouts) are covered by the low frequency noise.

Maximum
amplitude allowed

CAM signal saturated

Figure 12 - CAM recording saturation - waveform overview

A spectrum analysis makes it possible to reveal this anomaly. In the example below (Figure 13), a harmonic family with fundamental frequency of 80 Hz
(probably the acoustic signature of the aircraft propeller) is present. An SPL shows that this low-frequency harmonic family has a clearly predominant
level compared to the general level of the CAM signal. This high level of energy is the cause of the acoustic saturation.
This interference causes listening discomfort which lowers the performance of the person responsible for carrying out the transcription work. However,
it should be emphasized that ED112 does not mention any restrictions or limitations on the level of the recorded signal as long as it does not exceed the
recording capability of the whole CVR installation. However, in an evaluation of CVR recordings, it is essential to draw attention to this problem when it
occurs.
17
 Harmonic family causing the
This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission

acoustic saturation

Frequency (Hz)

Time (s)
Lofargram of the saturated CAM signal
Frequency (Hz)
20 dB

Signal
overage

Level (dB)
Energy level of each
harmonic

SPL on saturated CAM signal – Zoom on low frequencies

Figure 13 - CAM recording saturation - spectrum views (Lofar and SPL)
18
 4.5 Erroneous memory management
This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission

This anomaly is characterized by a deformation of the audio signal. In some parts of the recording, the signal becomes unintelligible and includes
“electronic noises“. It may happen that some parts of the signal have very low amplitude and sometimes a total lack of audio data (audio blanks).
This problem can be quickly identified by observing the waveform of the CVR recording (see Figure 14), observation confirmed by listening to the audio
data. Generally, the waveform presents “blanks“, significant differences in amplitude on the same track (alternation of saturated signal and low amplitude
signal). This defect often makes CVR recordings unusable, causing partial or total loss of audio data.
Reference specifications:
Deformation of the
ˆ ˆ ED 56A, §3.2.1 and §A2.3 waveform /
ˆ ˆ ED 112, §I-2.1.3. “electronic noises”

Audio blanks

Low amplitude signal Signal saturation

Figure 14 - erroneous memory management - waveform overview

19
4.6 Phase shift between tracks
It sometimes happens that some audio signals (crew conversations, warnings, ATC
communications, etc.) are not recorded with the same phase on each track of the CVR.
This phase shift is particularly troublesome when it is 180°(opposite phase signals).
In practice, the same sound event (ATC message, warnings, etc.) will have weakened
or even canceled amplitude if it is recorded on a mixed track on the CVR. The effect
of this anomaly also appears when the tracks containing phase shifted signals are
synchronized and played back simultaneously.
However, as the multiple signals received by the AMU have different amplitude, the
result of phase shift will mostly be a strong attenuation of the signal.
This anomaly can be detected in several ways:
ˆ ˆ if the CVR has a mixed track, observation of the waveform and listening to the
audio content on that track will reveal the defect immediately, especially if the
other tracks have ‘‘normal’’ waveforms;
ˆ ˆ if the CVR does not contain a mixed track, listening simultaneously to all the CVR
tracks once synchronized will reveal the anomaly: signals present on both tracks
will have low amplitude. A comparison of each track’s waveforms will confirm
the «diagnosis». Better detection is possible by enlarging the track waveforms
where a noticeable event is recorded (ATC messages, warnings, transient high
amplitude signals). The phase shift will be more visible at this location.
This anomaly is similar to a loss of audio data and is detrimental to any analytical
work on audio recordings, particularly if some of the information useful for the
investigation can be found only on the mixed track.
Reference specifications :
ˆ ˆ ED 56A, §1.4.3.a
ˆ ˆ ED 112, §I-1.3.3.a

180° phase

Zoom on waveform – 180° phase shift

Figure 15 - audio signals in opposite phase - waveform overview

20

This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission
 4.7 Level imbalance between tracks
This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission

This anomaly is characterized by a level difference between the input signals, particularly with regard to the level of ATC communications. In the
following example (Figure 16), the ATC message is strongly attenuated on the second track (circled in red), and is almost inaudible.

Same ATC message

recorded on both tracks

Figure 16 - level imbalance - waveform overview

This defect is detected by observing the waveform of the signal on each track. It can be confirmed by listening to each track separately, the difference in
level can then be easily detected.
Reference specifications:
ˆ ˆ ED-56A - §3.2.5
ˆ ˆ ED112-§ I-3.2.2 (RSL < 3dB)
21
4.8 Playback speed anomaly on a SSCVR
This anomaly has only been encountered once at the BEA. During normal playback,
the audio data downloaded replayed slower than the original data (audio content
slowed down). This defect was due to a problem during the signal sampling.
This fault has no influence on the quality of audio data but on the precision of the
retrieved audio content. It is noticeable only when the audio must be synchronized
with other information (FDR, radar data, for example). Then, resampling is necessary
(patching coefficient used on the playback speed).
This anomaly can be detected and highlighted in two ways:
ˆ ˆ Inconsistency of the FSK (8) signal if it is recorded on the CVR. If the interval of (8)
FSK: Frequency
4 seconds is not respected (uncertainty included), the data downloaded can be Shift Keying. Beep
recorded every 4s
suspected to not be replaying at the same speed as the data originally recorded. (+/- 2ms) on one
For the measurement to be significant, measure the interval between two distant track of the CVR
FSK peaks on the CVR track and compare this interval with the theoretical value. (usually on the Public
Address track and
Example (see Figure 17): There is theoretically 96 s (24 four-second gaps) between which encodes UTC
time on 32 bits. This
25 FSK peaks. If the measured value of 25 peaks is not equal to 96 s, then the fault is signal is recorded on
present on the CVR. ATR, Airbus (except
A380 and A350) and
ˆ ˆ Inability to synchronize the CVR and the FDR using events recorded both on the Fokker 50 CVR ‘s).
CVR and the FDR (VHF keying, warnings).

96 secs between 25 FSK peaks

FSK signal

FSK Signal – Zoom on waveform

Figure 17 - FSK signal - waveform overview

22

This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission
Case of time stretching
This anomaly has similar effects to a playback speed problem. The main difference
is that audio tracks can’t be synchronized with each other due to the sampling rate
varying over time. Therefore, the application of a patching coefficient does not
correct the problem and thus does not allow synchronized tracks to be obtained.
This anomaly presents several major problems:
ˆ ˆ when retrieving audio data stored on the CVR, there is a loss of time correlation.
Thus, the CVR is not compliant with ED-56A, §2.6.3 and ED 112, §I-2.1.3;
ˆ ˆ the audio content does not precisely reproduce the data originally recorded;

ˆ ˆ after resampling, the amount of recorded audio data is lower than the declared
recording capacity of the recorder (eg 1h 56 min audio data instead of 2 hours).
The CVR is not functioning to the performance specifications it should have,
given the recorder class to which it belongs (see ED-56A, §1.4.2 and §2.6.7 - ED
112, §I-1.3.2 et §I-2.1.5);
ˆ ˆ synchronization between the audio data from the CVR and flight parameters
recorded in the FDR is impossible. The CVR is not compliant with ED-56A §2.6.4
and §3.4.2 -- ED 112, §I-2.1.4 and §2-1.11;
ˆ ˆ no detection of this anomaly is possible if no FSK signal is recorded on the CVR,
if the FDR is not available, or if the alteration of audio (pilot voice distortion for
example) is not obvious (except in the case of time stretching).

23

This document is the property of the BEA and cannot be copied or reproduced, even partially,without prior written permission
 Bureau d’Enquêtes et d’Analyses
pour la sécurité de l’aviation civile

10 rue de Paris
Zone Sud - Bâtiment 153
Aéroport du Bourget
93352 Le Bourget Cedex - France
T : +33 1 49 92 72 00 - F : +33 1 49 92 72 03
www.bea.aero