2. The recognized
standard in audio test
Since 1985 AP has provided audio test
and measurement instruments to R&D
and production lines for every type of
audio device.
AP is the number one maker of audio
analyzers in the world.
Who is Audio Precision
3. Who uses AP?
" Audio Precision analyzers are used by makers
of all types of audio technology
! Pro audio
! A/V receivers
! Broadcast
! Smartphones
! Semiconductors
! Loudspeakers
! Microphones
! Automotive head units
! Blu-ray players
! MP3 players
! Bluetooth headsets
! Tablets and PCs
! Hearing aids
! Military
4. Monitoring vs. Measurement
4
Monitoring
" Displays key characteristics of actual program material in real time, such
as
! Level
! Channel activity
! Codec status
! Synchronization
Measurement
" Problem identification
! Noise, Distortion
! Frequency Response
! Glitches
" Proactive maintenance or used as needed
" Insures quality
…reveals very little of signal quality.
5. Measurement
5
" Studio maintenance
! Regular servicing
! Troubleshooting audio chain
! Evaluating new equipment
" On air broadcast quality of service
! Ensure broadcast quality across the service area outside the studio
6. Equipment maintenance
6
" Need the right interfaces & supported audio formats to evaluate
individual components
! AES/EBU interface
! High quality analog I/O
! SDI – can use an embedder / de-embedder if properly characterized
! Dolby E -- embed/deembed or “bit test”
9. In the studio
9
" Audio quality measurements
! Analog and digital
! Level
! THD+N
! Frequency response
! Phase
! Crosstalk
! Signal to noise ratio
! MOL
! Many more
" Signal chain and transport issues
! Encoding / decoding
! Digital and analog converters
! “Transparent” streaming of digital audio
! Bit truncation
! The transport itself (jitter)
16. Coded audio
16
" In television systems, multichannel streams may be coded in Dolby E
for studio transport, and Dolby Digital for emission.
" In such cases, the APx stimulus audio signal can be encoded and
passed down the chain of interest, and then can be measured after
decoding.
! Bit-accurate measurements will not provide useful measurements when passed
through an encode/decode cycle, which by design is not bit-accurate.
! Sine waves in coded audio
18. Radio Data System (RDS)
18
Narrow band signal with a 57 kHz sub-
carrier (3 x 19 kHz pilot tone)
30
Hz
15
kHz
23
kHz
38
kHz
53
kHz
Stereo Audio
(Left – Right)
19 kHz
Stereo
Pilot
(10%)
Mono Audio
(Left + Right)
58.7
kHz
RDS
(3%)
57
kHz
Our tech support engineers have developed a utility
to test FM RDS signals.
21. Mul=tone
" Single signal with multiple tones, typically 3,5,7 or 31 tones
" Extremely fast: 20+ results in <1 seconds
" Ideal for over the air broadcast and encoded formats
22. Mul=tone
22
The multitone is acquired and analyzed by FFT, providing many results.
The blue line represents a frequency response result aSer analysis.
The DUT has imposed its response on the mul=tone frequencies.
25. " FFTs (Fast Fourier Transform)
! The engine of modern audio analyzers
! Can derive most other measurements mathematically
" Look for
! 24-bit resolution
! Large max length 500K to 1M+ points
Today’s FFTs
24 bit APx monitor8-bit oscilloscope
27. Today’s FFTs
" Wide bandwidth option:
DC to 1MHz bandwidth @ 1.2M points = 2.38 Hz bin width
" Ideal for Class D amp design
! See switching frequencies
28. Mul=channel capability
" Multichannel analyzers can be
useful for maintenance in
broadcast and pro audio
environments.
" 16x faster than testing one
channel at a time
" One to many crosstalk
" See more channels on one screen
APx I/O varia2ons
• 2 x 2
• 2 x 4
• 2 x 8
• 8 x 8
• 8 x 16
30. UI, repor=ng and sharing data
" Analyzer should be easy to use
! Intuitive interface & real time monitors
! Easy automation with no coding
" Sharing data is as important as the research
! Share projects with limits and audio files with other departments anywhere
! Send results embedded in a project to compare or troubleshoot
! Database import / export
" Need fast, clear and easy reports
! Automatically generate PDF
! Customized Word documents
31. Calibra=on
" Similar, but very different
! Adjustment
! Calibration
! Accredited Calibration
" Traceability
! an unbroken chain of comparisons
! measurement uncertainty
! documentation
! competence
! reference to SI units
! calibration intervals
33. Bluetooth Case study: Hands-free Car Kit
" Engineer had a reference design for a Bluetooth car kit with HFP 1.6
(including the new wideband voice feature).
" His observation:
“When I’m connected to other Bluetooth devices in HFP, the audio I receive
from them sounds good.”
…
“But the audio they receive when I speak into the microphone sounds bad.”
34. Step 1: PESQ measurement
" Connect APx analog generator to his DUT and analyze speech signal
received over Bluetooth.
! PESQ: Perceptual Evaluation of Speech Quality
! Compares source and DUT for time alignment
! Uses speech samples
! Calculates MOS result
" Mean opinion score = 1.22
! Terrible!
! With wide-band voice,
MOS should be 4.0 to 4.5
MOS Quality Impairment
5 Excellent Impercep=ble
4 Good Percep=ble but not annoying
3 Fair Slightly annoying
2 Poor Annoying
1 Bad Very annoying
36. Step 3: Diagnos=c Tests
" Generate some pure sine signals through the DUT and observe the
results.
" In the following FFT spectrum plots, you can clearly see something
wrong.
40. Observa=ons from FFT spectra
" DUT has significant distortion due to a spurious tone at:
ftone = 8 kHz – fsignal
" Level of spurious tone increases as the frequency of the signal
" Note: 8 kHz = sample rate /2
Conclusion:
Clearly a DSP problem.
41. Bluetooth in – analog out?
" In Part 1, we looked at the signal transmitted by his DUT with analog
in and Bluetooth out.
" Next, we look at the opposite direction: Bluetooth in and analog out
(the signal from the phone calling the car kit).
Customer’s observation:
“In this direction, the audio sounds good.”
42. Step 4: PESQ measurement
" Connect APx Bluetooth generator to his DUT and analyze speech
signal received at analog input.
! Mean opinion score = 3.40
! Not bad, but should be better!
! With wide-band voice, MOS should be 4.0 to 4.5
45. Observa=on from Frequency Response
" Signal from DUT is being low-pass filtered at 4 kHz instead of 8 kHz
" This filter is designed for narrow band voice, not wide-band.
" Results in less clear sounding speech and a lower Mean Opinion
Score.
Conclusion:
Another DSP problem (wrong filter).
46. Conclusions …
" In ~ 1.5 hours, we were able to find and diagnose some serious
problems with this Bluetooth device.
" Using the “hello, …hello, … can you hear me now?” method, you
would never be able to find the problems or offer clues to fix.
47. For More Informa=on
47
" Free books:
! Measurement Techniques for Digital Audio
# ap.com/display/file/17
! The Audio Measurement Handbook
# ap.com/display/file/24
" Thomas Williams: tomw@ap.com
48. Why tes=ng is important
" How much is an hour of dead air?
" Do you want to be told about an error by a concerned listener?
Audio analyzer or sound card?
49. Audio analyzer or sound card?
" Trust & reliability
! Believe your own results
! Others believe your results
! Calibrated & traceable
Is the error you see
caused by your product
or your “analyzer”?
50. Audio analyzer or sound card?
" Productivity & sophisticated measurements
! No wasted time
! Built in measurements & proven algorithms
! Easy to share test data with others
! Professional technical support
Who do you call when
there’s a problem?
51. Common mistakes: grounding
" Most common mistake we see
" Star grounding is best
" Use heavy gauge and always check your grounds
The resistance in each leg of
the chain puts the devices at
different ground potentials, and
is not as effective as star
grounding.
52. Common mistakes: Class D filtering
" Class D switching amplifiers have unique problem of out of band noise
" Can cause inaccurate measurements, or even damage analyzer inputs
" Simple filter stops the problem
55. Common mistakes: Connec=ons
" Sounds simple, but it happens a lot
" Crossed cables
" Bad cables
" Analyzer set for wrong connections
" Always check your connections
" Use Loopback mode to confirm settings
56. 2700 Series
Most advanced analyzer for R&D
• Vanishingly low residual noise and THD
+N: —115 dB @ 2.0 Vrms
• True Dual Domain – analog and digital
• Generate and analyze a wide range of
waveforms
• API automation
• LabVIEW integration
• Chip-level connectivity with PSIA
• User-defined sweeps, switcher
support up to 192 channels
The world’s highest performance audio analyzer
THD+N of AP 2700 (Black) compared
with 5 other audio analyzers
57. APx Series
High performance for R&D
• Up to -110 dB THD+N
• Test Bluetooth, HDMI, I2S
• 1 million point FFT analyzer with
24 bit resolution from DC to 1MHz
• Multichannel and high bandwidth options
Multiple interfaces
• HDMI, Bluetooth, PDM
• AES/EBU, I2S, S/PDIF
Intuitive & easy to use
• One-click measurements
• Real-time monitors
• Automated reports
Fast, simple automation for production
• Up to 21 measurements in 1.2 seconds
• Automation without coding or use
the .NET API and LabVIEW driver
• Lockable projects with limits
Connectivity, Flexibility and Intuitive Operation