Event Type: Audio

Event Type: Audio
Test Mode	Severity	Event	Description
Non-Referenced	Warning	Low Volume Alarm	Warn the user that the volume level of the detected audio is below the best range for performing meaningful audio analysis. Alarm is initialized when volume level above the “Measurement ThresholdThe volume threshold above which useful audio analysis is possible.” level is detected. Alarm is activated when the detected volume drops below the “Measurement Threshold” level for 10 consecutive 0.5 sec measurement intervals.
Non-Referenced	Warning	Clipping	Reports the detection of suspected distortion that occurs when the amplitude of a signal exceeds a digital systems ability to represent it accurately. Clipping is a type of amplitude distortion. The system reports a Clipping event when consecutive samples at the maximum value that can be represented by the digital system have been detected. Note that the maximum value that can be represented is different depending on the number of bits per sample (i.e. bits of resolution) of the audio stream. The system limits the number of reported Clipping events to typically 10 to 20 per sec.
Non-Referenced	Warning	High Volume Alarm	Warn the user that the volume level of the detected audio is above the best range for performing meaningful audio analysis (i.e. above a level where the audio will likely become distorted). Alarm is activated when the detected audio volume is continuously above the high volume thresholdHigh Volume Threshold for speech: - 6dBFS High Volume Threshold for music: -12 dBFS (see Figure 2) for 10 consecutive 0.5 sec measurement intervals (i.e. 5 sec total). The event will not be repeated again until the detected volume level drops below the high volume threshold for 10 more consecutive 0.5 sec measurement connections.
Non-Referenced	Warning	Dropout	Reports the detection of an unusual brief silence period where the brief silence is preceded and followed by “normal” audio levels. A typical definition of Dropout is the short dramatic loss of volume typically caused by lost digital information. Root causes include transmission system errors resulting in lost data packets, transmission channel reconfigurations, bad sections of memory, processor overloads that temporarily interrupt the flow of information, and so on.
Non-Referenced	Warning	Glitch	Extremely large sample-to-sample audio amplitude transitionsGlitch sample-to-sample audio amplitude transitons: Speech: greater than 40 dB change Music: greater than 90 dB change that have little probability of occurring within natural speech or music. Such dramatic changes would typically happen only in situations of dropped samples.
Referenced	Info	TestID Found	Occurs when a valid Test IDA “Test ID” is three digits minimum in length, representing a dot notation “N.vv” Test Identifier. The Value ‘N’ may be any length >= 1 indicating a specific test number, and “vv” represents a two digit version. Each digit is represented by a tone between 200 and 290 Hz, and is followed either by a 1 kHz delimiter tone or a 400 Hz Test ID terminator. The digit ‘0’ is represented by 200 Hz, the digit ‘1’ by 210 Hz, and so on, up to the digit ‘9’ represented by 290 Hz. has been recognized. A valid Test ID must meet the level, frequency, duration, and delimiter requirements. If any of these parameters do not match, the process is terminated and is reset to the initial conditions. Until a Test ID is successfully recognized, the system will continue to operate in Non Referenced Mode; therefore, no events related to false starts are reported. This is because for arbitrary audio there is no expectation of any Test ID.
Referenced	Warning	Test Script Not Found	Occurs if a valid Test ID was found , but the script for that Test ID was not found. The system reverts to Non-Referenced Mode if this happens. This event should not occur if using a valid Reference Audio file provided by Frontline.
Referenced	Error	Invalid Test Script	This event is generated when an error occurs while accessing information in a script. This event should not occur if using a valid reference audio file provided by Frontline.
Referenced	Error	Synchronization Lost	Generated when after a successful TestID recognition the system encounters unexpected frequencies or durations of audio segments while analyzing a received Reference Audio file. If this situation occurs, the internal segment tracking logic attempts to look forward and/or backward in the test script to determine if the currently measured characteristics are consistent with the previous or next segment of the script. If there is a match, the internal segment pointer is advanced or retarded appropriately, the Synchronization Lost event is not generated, and the audio analysis continues. However, if a match cannot be found, the system declares itself out of sync and generates the Synchronization Lost Event, terminates any active test script, and reverts to Non-Referenced Mode.
Referenced	Error	Unexpected Frequency	Reported when a measured frequency deviates from an expected frequency by a specific percentage (determined by the negotiated parameters of the over-the-air audio stream). The system knows the Reference Audio file that is being played on the Source DUT; therefore, the system knows which frequencies (tones) to expect at a given time.
Referenced	Error	Unexpected Level	Reported when the measured level at the start of a tone segment is not within tolerance. The tolerance is dependent on sample rate and bits per sample, but it generally is +/- 3 dB for speech and +/11 dB for music. The system knows the Reference Audio file that is being played on the Source DUT; therefore, the system knows which amplitude level to expect at a given time.
Referenced	Error	Unexpected Duration	Reported when a tone segment of the Reference Audio file is shorter or longer than expectedThe amount that a measured duration must deviate from the programmed duration of a tone segment before the system declares this event varies, depending on the negotiated over-the-air audio stream specific parameters, but it is generally in the range of 5% to 10%. Note that this event will result in an attempt to resynchronize if the measured duration is greater than expected.. The system knows the Reference Audio file that is being played on the Source DUT and therefore knows how long a specific tone segment should last. If either a change of amplitude or frequency arrives either before or after that programmed duration, then the change is by definition unexpected. This type of audio impairment can be caused by lost or corrupted data, repeated data, faulty packet loss concealment algorithms, etc.
Referenced	Error	Amplitude Fluctuations	Reported if the system detects unexpected amplitude changes over a given interval. The test tones in Frontline’s Reference Audio files have a fixed amplitude level over their duration. Therefore, if the corresponding audio levels received over the air by the system fluctuatesThe system calculates amplitude fluctuations as: (Max Level – Min Level) / (Max Level + Min Level) * 100 more than a specified level (this level is based on the received audio stream parameters), then the system generates an Amplitude Fluctuations event.
Referenced	Error	Unexpected Phase Change	Provides a fine-grained indication of lost or repeated energy. The system knows when a specific tone should be expected. During this interval, the system checks that the measured average frequency is the same as the expected frequency. If this is correct, the system will continue to monitor the instantaneous frequency. If the instantaneous frequency deviates sufficiently from the current average frequency, the frequency measurement state machine will reset and begin re-measuring. Typically, the outcome is the discovery of the next scripted (expected) frequency. However, another outcome can be that the same frequency as the previous average frequency is rediscovered, and this is reported as an Unexpected Phase Change event. Such phase changes are an indicator of losses of signal that do not result in amplitude dropouts, or signal substitution (repetition) of previous audio energy due to things such as “packet loss concealment” tactics.
Referenced	Error	Excess Noise	The Excess Noise event is reported when energy sufficiently above the “Silence Threshold” is detected during programmed segments of silence. Excess noise can indicate a poor analog audio chain with an inherently poor noise floor, glitches occurring during silence intervals, or codecs that do not transition to silence instantaneously.
Referenced	Error	Clipping	Reports the detection of suspected distortion that occurs when the amplitude of a signal exceeds a digital systems ability to represent it accurately. Clipping is a type of amplitude distortion. The system reports a Clipping event when consecutive samples at the maximum value that can be represented by the digital system have been detected. Note that the maximum value that can be represented is different depending on the number of bits per sample (i.e. bits of resolution) of the audio stream. The system limits the number of reported Clipping events to typically 10 to 20 per sec.
Referenced	Error	CVSD HF Level Too High	Reported when a CVSD encoded audio stream is detected and there is high frequency energy above 4 kHz that is greater than -20 dBFS.
Referenced	Info	End of Test Event	Reported to indicate that the system has completed processing a test script for a Reference Audio file, and that the system has exited Reference Mode. This event is generated when the elapsed time from the start of test is equal to or greater than the scripted duration of a test. It is reached when the number of samples processed equals the number of samples associated with the test duration.

Clipping

The number of consecutive samples needed to qualify as a clipping event depends on both sample rate and number of bits per sample. Table 1 specifies the number of consecutive samples at the maximum value level that will generate a Clipping event.

Clipping Event Thresholds
Consecutive Samples	Sample Rate, Samples/sec	Resolution, bits
3	8000	16
5	16000	16
11	41000	16
2	64000	16
12	48000	16
24	96000	16

Clipping Event Thresholds
Consecutive Samples	Sample Rate, Samples/sec	Resolution, bits
3	8000	16
5	16000	16
11	41000	16
2	64000	16
12	48000	16
24	96000	16

Dropout

Dropout events are reported when the average audio level (RMS) is initially above the Measurement Threshold, then falls below the Silence Threshold, and then quickly rises above the Measurement Threshold again ). This approach largely disqualifies the natural inter-syllable silence and pauses that occur in natural speech, but will detect gaps caused by dropped data. Note that the system does not report dropouts that begin at very low energy levels.

Audio Dropout description

Dropout: Measurement and Silence Threshold

Glitch

The Glitch event is reported whenever an extremely large sample to sample amplitude transition occurs that has little or no probability of occurring within natural speech or music. As illustration, back to back +N, -N, …, +N, -N values (where N is any non-zero number), represents energy at the Nyquist frequency, or ½ the sample rate. Neither speech nor music contain average energy levels at this frequency more the 20 dB below nominal. However, moderately large sample to sample changes in amplitude do occur, and these naturally limit how sensitive this measure can be configured.

The system uses back to back transition levels of 90 dB for music and 40 dB for speech as the threshold for reporting the Glitch event.

Such dramatic changes would typically happen only in the face of dropped samples, and serve as an additional means of detecting gross abnormalities

Also see these related topics...