SlideShare a Scribd company logo
1 of 34
Download to read offline
Signal Frequency FSig, FSR, and FNyq
Summary
1/ 34
Nyquist frequency FNyq:= 0.5FSR, Historical upper limit
 FSR =16kHz
FSig
0.1 FSR
0.5 FSR=FNyq
0.9 FSR
November 15, 2020
Basic Pitfalls in Waveform Analysis
- Introduce DEEPS -
Yukio Okuda
sf.yukio@gmail.com
an independent, Atsugi, Japan
November 2020
Me = Industrial Soft  Hard +40 Years Introduction
3/ 34
Soft+Hard, 2 Years:
Mechanical Vibration of Mother Machines
Sensor Selection , Waveform Failure Analysis
Soft, 8 Years: OSS
Soft+Hard, 15 Years: Digital LSI Testing and Failure
Analysis
based on Data Analysis IEEE-ITC Papers
Soft, 17 Years: Developing LSI Design CAD tools
Soft, 6 Years: Developing Information Retrieval
Hard, 3 Years: Color TV Production
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Outline Introduction
4/ 34
Digitizing Error at lower than Nyquist Frequency FNyq
Digitize
(interpolate)
No-Repeatability
Pseudo Amplitude Modulation AM
FNyq
DCASE-2
DCASE-4
• Models
• Error signatures
Verified by DEEPS
Digitize Error Estimation (Prediction) by (Spectrum)
Improve
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Why Digitize Introduction
5/ 34
ADC mismatches for feature extractors ¬ Digitizing errors
ADC: Analog to Digital Converter
ADC
No error
Extractor
Feature
Analysis
Digitizing
Time
Time
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Outline of Models Models
6/ 34
Show errors at lower than Nyquist Frequency
Rarely reported
Improving applications is the next stage
– Low Sampling Rate
— # of Signal Repeats
Systematic Error
Shape Spectrum
Random Error
DEEPS
˜ Clock Skew
Show errors
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
What is ADC Models
7/ 34
Sampling at the predefined intervals of TS
Sampling Rate FSR := 1/TS
ADC = Sampling
Distortions at high frequency
Deterministic error ¬ Repeatable
Start time is uncontrollable
Random error ¬ No-Repeatable
H/W: designers-Clock Skew; users-Jitter
Repeatable :
the same data at the same condition
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Signal Frequency FSig, FSR, and FNyq
Models
8/ 34
Nyquist frequency FNyq:= 0.5FSR, Historical upper limit
 FSR =16kHz
FSig
0.1 FSR
0.5 FSR=FNyq
0.9 FSR
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Signal Repeats Models
9/ 34
Signal repeats change shapes  spectra
Pseudo Amplitude(AM) Modulation at FSig= 7.1 kHz
 16kHz sampled waves are up-converted to 160kHz
Genuine AM
? Boundary
FFT
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Un-Controllable Start Time Models
10/ 34
Errors between measurements  in a measurement
Errors depend on FSig ? Robustness
ADC
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Estimate Error Variance caused by Start Time Variance Models
11/ 34
At a Start Time variance of Clock offsets Co=[0, 0.3, 0.6]
Repeat number NR= 4, FSig= 6.1kHz, ∆ FSp := FSp− FSig
Wave
FFT
Spectrum
FSp
Estimation
∆ FSp
Amp.:Summarize,Normalize
Freq.:∆from6.1kHz
Heat Map
∆ FSp
6.1kHz
∆FSp
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Compare Error Variances of FSigs at NR= 4 Models
12/ 34
Clock offsets Co= 0, 0.1, 0.2, .., 0.9 Assume random uniform
FSig= 6, 6.1, 6.2, .., 7.5 kHz, NR= 4
FSig
– —
∆FSp
– 6.4kHz shows the own frequency
— 7.1kHz shows the own frequency
Others show:
Different one frequency
Different two or three frequencies
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Compare Error Variances of FSigs at NR= 8 Models
13/ 34
Clock offsets Co= 0, 0.1, 0.2, .., 0.9 Assume random uniform
FSig= 6, 6.1, 6.2, .., 7.5 kHz, NR= 8
FSig
– —˜
∆FSp
– 6.4kHz shows the own frequency
Same as at NR= 4
— 7.1kHz shows the own frequency
Same as at NR= 4
˜ 6.1kHz shows the own frequency
Three frequencies at NR= 4
Others show two or three frequencies
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
A Magic Signal of 6.4 kHz Models
14/ 34
Differently distorted waves yield a FSp of 6.4 kHz!
NR= 4
FFT
Spectrum
NR= 8
FFT
Spectrum
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Summary of Models Models
15/ 34
DEEPS estimates error signatures at a FSR of 16 kHz
Signals higher than 4 kHz are distorted
No error signatures
Except, the Signals of 6.4 kHz, 7.1 kHz
• Show the FSp of the FSig
• Signals of 7.1 kHz show AM modulation waves
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Outline of Verification Verification
16/ 34
A/B Testing is difficult for an independent
Challenge post-mortem analysis on DCASE Challenge Datasets
DCASE: Detection  Classification of Acoustic Scenes  Events
Task2 ĸ
• Obtained by the one tool Statistical analysis of spectra
• 13,000 nominal sounds
• 10 sec • FSR= 16 kHz • four device types
Task4 ĸ
• From Used by several Voice Recognition Activities
• 12,000 unlabeled sounds, non-controllable measurements
• 10 sec • FSR= 16 kHz, 44.1 kHz
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Averaged Spectra of Task2 Verification
17/ 34
Compare tails of spectra from 4kHz
– All 6.4kHz peaks — Slider/Valve 7.1kHz peaks
– –
–
–
—
—
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
How to locate error signals Verification
18/ 34
Apply Low-Pass Filter-Bank
Spectrum has no time information, i.e. Non-temporal
 Reversed signals show the same spectra
FFT
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Task2 Ex1: 6.4kHz Verification
19/ 34
Spectrum peaks around 6.4 kHz
 Ex1: slider/train/normal id 06 00000075
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Task2 Ex1: 7.1kHz Verification
20/ 34
Spectrum peaks around 7.1 kHz
 Ex1: slider/train/normal id 06 00000075
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Spectrum Endpoints of Task4 Verification
21/ 34
All sounds may include distortions
– 70% sounds are digitized by FSR=16 kHz, upconverted to 44.1 kHz
— 30% sounds are digitized by FSR=44.1 kHz
–
–
—
—
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
AM Signatures at a FSR of 44.1 kHz Verification
22/ 34
Peak count (2, 3, 4, 5) identifies AM signatures  FSig
19.9 kHz
— ˜ ™ š
˜ ™ ˜ ˜ ˜ ˜
— ˜
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Summary of Verification Verification
23/ 34
Demonstrated
1 Digitizing error at lower than FNyq
2 DEEPS estimations
3 sounds include distortions
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Discussion for Improvements Discussion
24/ 34
Application policies set Max FSig  Max Error
Feature error predictions set FSR
ADC + Waves
Waveform Variations
Controllable
Non-Controllable
Max FSig
Max Error
FSR
Correct
Feature Extraction
Error Prediction
FFT Statistics
Time Series
Wavelet Cepstrum
Human Ear Eq.
. . .
DA
DataAnalysis
Appli.
Science
Medical
Industry
Home
. . .
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
Feature Error Predictions by DEEPS Discussion
25/ 34
Feature extractor decide error from waveform variations
Wave Gen Drive Extractor
Features
FSig-N
Wave Gen Drive Extractor
Features
FSig-1
Sum
Indicator
FSig
Already FFT with spectrum
Challenging predictions: • Characteristics of waves
• References to Metrics • Parameters • Window size • FSig
dependancy • . . .
May be effective on non-controllable systems
Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38

More Related Content

What's hot

Design and implementation of sdr based qpsk transceiver using fpga
Design and implementation of sdr based qpsk transceiver using fpgaDesign and implementation of sdr based qpsk transceiver using fpga
Design and implementation of sdr based qpsk transceiver using fpgaTarik Kazaz
 
RFID based Smart Voting System
RFID based Smart Voting System RFID based Smart Voting System
RFID based Smart Voting System DEEPAK LODHA
 
A Glimpse into Developing Software-Defined Radio by Python
A Glimpse into Developing Software-Defined Radio by PythonA Glimpse into Developing Software-Defined Radio by Python
A Glimpse into Developing Software-Defined Radio by PythonAlbert Huang
 
RF Module Design - [Chapter 4] Transceiver Architecture
RF Module Design - [Chapter 4] Transceiver ArchitectureRF Module Design - [Chapter 4] Transceiver Architecture
RF Module Design - [Chapter 4] Transceiver ArchitectureSimen Li
 
SCA Next Part 1 - Software Defined Radio (SDR) Webcast Slides
SCA Next Part 1 - Software Defined Radio (SDR) Webcast SlidesSCA Next Part 1 - Software Defined Radio (SDR) Webcast Slides
SCA Next Part 1 - Software Defined Radio (SDR) Webcast SlidesADLINK Technology IoT
 
Software Defined Radio (SDR)
Software Defined Radio (SDR)Software Defined Radio (SDR)
Software Defined Radio (SDR)Drew Fustini
 
Poster receiver platform-080210
Poster receiver platform-080210Poster receiver platform-080210
Poster receiver platform-080210Bertalan EGED
 
Lab based ppt pluto-sdr_final
Lab based ppt pluto-sdr_finalLab based ppt pluto-sdr_final
Lab based ppt pluto-sdr_finalBhavna Singh
 
Instrumentation: Liquid and Gas Sensing - VE2013
Instrumentation: Liquid and Gas Sensing - VE2013Instrumentation: Liquid and Gas Sensing - VE2013
Instrumentation: Liquid and Gas Sensing - VE2013Analog Devices, Inc.
 
Multiband Transceivers - [Chapter 3] Basic Concept of Comm. Systems
Multiband Transceivers - [Chapter 3]  Basic Concept of Comm. SystemsMultiband Transceivers - [Chapter 3]  Basic Concept of Comm. Systems
Multiband Transceivers - [Chapter 3] Basic Concept of Comm. SystemsSimen Li
 
D Belver FEE for Trasgos
D Belver  FEE for TrasgosD Belver  FEE for Trasgos
D Belver FEE for TrasgosMiguel Morales
 
Why Is Expensive Equipment Necessary When A QRC WBT Will Do?
Why Is Expensive Equipment Necessary When A QRC WBT Will Do?Why Is Expensive Equipment Necessary When A QRC WBT Will Do?
Why Is Expensive Equipment Necessary When A QRC WBT Will Do?bluejeanstest1
 
Hardware Accelerated Software Defined Radio
Hardware Accelerated Software Defined Radio Hardware Accelerated Software Defined Radio
Hardware Accelerated Software Defined Radio Tarik Kazaz
 
The Building of Pulsed NQR/NMR Spectrometer
The Building of Pulsed NQR/NMR Spectrometer The Building of Pulsed NQR/NMR Spectrometer
The Building of Pulsed NQR/NMR Spectrometer IJECEIAES
 
Automatic Detection of Heart Disease Using Discreet Wavelet Transform and Art...
Automatic Detection of Heart Disease Using Discreet Wavelet Transform and Art...Automatic Detection of Heart Disease Using Discreet Wavelet Transform and Art...
Automatic Detection of Heart Disease Using Discreet Wavelet Transform and Art...Editor IJMTER
 
Software Defined Radio Engineering course sampler
Software Defined Radio Engineering course samplerSoftware Defined Radio Engineering course sampler
Software Defined Radio Engineering course samplerJim Jenkins
 

What's hot (20)

Software Defined Radio
Software Defined RadioSoftware Defined Radio
Software Defined Radio
 
Design and implementation of sdr based qpsk transceiver using fpga
Design and implementation of sdr based qpsk transceiver using fpgaDesign and implementation of sdr based qpsk transceiver using fpga
Design and implementation of sdr based qpsk transceiver using fpga
 
RFID based Smart Voting System
RFID based Smart Voting System RFID based Smart Voting System
RFID based Smart Voting System
 
A Glimpse into Developing Software-Defined Radio by Python
A Glimpse into Developing Software-Defined Radio by PythonA Glimpse into Developing Software-Defined Radio by Python
A Glimpse into Developing Software-Defined Radio by Python
 
RF Module Design - [Chapter 4] Transceiver Architecture
RF Module Design - [Chapter 4] Transceiver ArchitectureRF Module Design - [Chapter 4] Transceiver Architecture
RF Module Design - [Chapter 4] Transceiver Architecture
 
SCA Next Part 1 - Software Defined Radio (SDR) Webcast Slides
SCA Next Part 1 - Software Defined Radio (SDR) Webcast SlidesSCA Next Part 1 - Software Defined Radio (SDR) Webcast Slides
SCA Next Part 1 - Software Defined Radio (SDR) Webcast Slides
 
Software Defined Radio (SDR)
Software Defined Radio (SDR)Software Defined Radio (SDR)
Software Defined Radio (SDR)
 
Poster receiver platform-080210
Poster receiver platform-080210Poster receiver platform-080210
Poster receiver platform-080210
 
Lab based ppt pluto-sdr_final
Lab based ppt pluto-sdr_finalLab based ppt pluto-sdr_final
Lab based ppt pluto-sdr_final
 
Instrumentation: Liquid and Gas Sensing - VE2013
Instrumentation: Liquid and Gas Sensing - VE2013Instrumentation: Liquid and Gas Sensing - VE2013
Instrumentation: Liquid and Gas Sensing - VE2013
 
Lab based report
Lab based reportLab based report
Lab based report
 
Multiband Transceivers - [Chapter 3] Basic Concept of Comm. Systems
Multiband Transceivers - [Chapter 3]  Basic Concept of Comm. SystemsMultiband Transceivers - [Chapter 3]  Basic Concept of Comm. Systems
Multiband Transceivers - [Chapter 3] Basic Concept of Comm. Systems
 
D Belver FEE for Trasgos
D Belver  FEE for TrasgosD Belver  FEE for Trasgos
D Belver FEE for Trasgos
 
FM-RDS developments at CRC
FM-RDS developments at CRCFM-RDS developments at CRC
FM-RDS developments at CRC
 
Why Is Expensive Equipment Necessary When A QRC WBT Will Do?
Why Is Expensive Equipment Necessary When A QRC WBT Will Do?Why Is Expensive Equipment Necessary When A QRC WBT Will Do?
Why Is Expensive Equipment Necessary When A QRC WBT Will Do?
 
Hardware Accelerated Software Defined Radio
Hardware Accelerated Software Defined Radio Hardware Accelerated Software Defined Radio
Hardware Accelerated Software Defined Radio
 
The Building of Pulsed NQR/NMR Spectrometer
The Building of Pulsed NQR/NMR Spectrometer The Building of Pulsed NQR/NMR Spectrometer
The Building of Pulsed NQR/NMR Spectrometer
 
Gnu radio lecture
Gnu radio lectureGnu radio lecture
Gnu radio lecture
 
Automatic Detection of Heart Disease Using Discreet Wavelet Transform and Art...
Automatic Detection of Heart Disease Using Discreet Wavelet Transform and Art...Automatic Detection of Heart Disease Using Discreet Wavelet Transform and Art...
Automatic Detection of Heart Disease Using Discreet Wavelet Transform and Art...
 
Software Defined Radio Engineering course sampler
Software Defined Radio Engineering course samplerSoftware Defined Radio Engineering course sampler
Software Defined Radio Engineering course sampler
 

Similar to Basic Pitfalls in Waveform Analysis

Synchronous Time / Frequency Domain Measurements Using a Digital Oscilloscope...
Synchronous Time / Frequency Domain Measurements Using a Digital Oscilloscope...Synchronous Time / Frequency Domain Measurements Using a Digital Oscilloscope...
Synchronous Time / Frequency Domain Measurements Using a Digital Oscilloscope...Rohde & Schwarz North America
 
Spectral Analysis at the Limit - Applications in Radio Astronomy
Spectral Analysis at the Limit - Applications in Radio AstronomySpectral Analysis at the Limit - Applications in Radio Astronomy
Spectral Analysis at the Limit - Applications in Radio AstronomyHochschule für Technik FHNW
 
[Best]Chromatic Tuner Project Final Report
[Best]Chromatic Tuner Project Final Report[Best]Chromatic Tuner Project Final Report
[Best]Chromatic Tuner Project Final ReportNicholas Ambrosio
 
RFID Masterclass 2015
RFID Masterclass 2015RFID Masterclass 2015
RFID Masterclass 2015CNRFID
 
Tesis Maestria - Presentacion Final
Tesis Maestria - Presentacion FinalTesis Maestria - Presentacion Final
Tesis Maestria - Presentacion FinalBruno Espinoza
 
IRJET- FPGA based Processor for Feature Detection in Ultra-Wide Band Radar
IRJET- FPGA based Processor for Feature Detection in Ultra-Wide Band RadarIRJET- FPGA based Processor for Feature Detection in Ultra-Wide Band Radar
IRJET- FPGA based Processor for Feature Detection in Ultra-Wide Band RadarIRJET Journal
 
FIR_Filters_with_FPGA
FIR_Filters_with_FPGAFIR_Filters_with_FPGA
FIR_Filters_with_FPGAIrvn Rynning
 
DickeyS_presentation_2015_3_26_2_1
DickeyS_presentation_2015_3_26_2_1DickeyS_presentation_2015_3_26_2_1
DickeyS_presentation_2015_3_26_2_1Sergey Dickey
 
Updated! Debugging EMI Problems Using a Digital Oscilloscope
Updated! Debugging EMI Problems Using a Digital OscilloscopeUpdated! Debugging EMI Problems Using a Digital Oscilloscope
Updated! Debugging EMI Problems Using a Digital OscilloscopeRohde & Schwarz North America
 
Digital Implementation of Costas Loop with Carrier Recovery
Digital Implementation of Costas Loop with Carrier RecoveryDigital Implementation of Costas Loop with Carrier Recovery
Digital Implementation of Costas Loop with Carrier RecoveryIJERD Editor
 
Advanced lock in amplifier for detection of phase transitions in liquid crystals
Advanced lock in amplifier for detection of phase transitions in liquid crystalsAdvanced lock in amplifier for detection of phase transitions in liquid crystals
Advanced lock in amplifier for detection of phase transitions in liquid crystalsIAEME Publication
 
Introduction to OFDM.ppt
Introduction to  OFDM.pptIntroduction to  OFDM.ppt
Introduction to OFDM.pptStefan Oprea
 
A 45Gb/s PAM-4 Transmitter Delivering 1.3Vppd Output Swing with 1V supply in ...
A 45Gb/s PAM-4 Transmitter Delivering 1.3Vppd Output Swing with 1V supply in ...A 45Gb/s PAM-4 Transmitter Delivering 1.3Vppd Output Swing with 1V supply in ...
A 45Gb/s PAM-4 Transmitter Delivering 1.3Vppd Output Swing with 1V supply in ...aiclab
 
Getting started with digital modes
Getting started with digital modesGetting started with digital modes
Getting started with digital modesskutaboot
 
305575138-VIBRASI-SE-3-Level2.pdf
305575138-VIBRASI-SE-3-Level2.pdf305575138-VIBRASI-SE-3-Level2.pdf
305575138-VIBRASI-SE-3-Level2.pdfSoufianeAychour1
 
anritsu mp1800a Анализатор качества сигнала BER
anritsu mp1800a Анализатор качества сигнала BERanritsu mp1800a Анализатор качества сигнала BER
anritsu mp1800a Анализатор качества сигнала BERonlinestuden
 

Similar to Basic Pitfalls in Waveform Analysis (20)

Synchronous Time / Frequency Domain Measurements Using a Digital Oscilloscope...
Synchronous Time / Frequency Domain Measurements Using a Digital Oscilloscope...Synchronous Time / Frequency Domain Measurements Using a Digital Oscilloscope...
Synchronous Time / Frequency Domain Measurements Using a Digital Oscilloscope...
 
Spectral Analysis at the Limit - Applications in Radio Astronomy
Spectral Analysis at the Limit - Applications in Radio AstronomySpectral Analysis at the Limit - Applications in Radio Astronomy
Spectral Analysis at the Limit - Applications in Radio Astronomy
 
[Best]Chromatic Tuner Project Final Report
[Best]Chromatic Tuner Project Final Report[Best]Chromatic Tuner Project Final Report
[Best]Chromatic Tuner Project Final Report
 
RFID Masterclass 2015
RFID Masterclass 2015RFID Masterclass 2015
RFID Masterclass 2015
 
Neural source-filter waveform model
Neural source-filter waveform modelNeural source-filter waveform model
Neural source-filter waveform model
 
Tesis Maestria - Presentacion Final
Tesis Maestria - Presentacion FinalTesis Maestria - Presentacion Final
Tesis Maestria - Presentacion Final
 
Lecture_1.ppt
Lecture_1.pptLecture_1.ppt
Lecture_1.ppt
 
Lecture_1.ppt
Lecture_1.pptLecture_1.ppt
Lecture_1.ppt
 
IRJET- FPGA based Processor for Feature Detection in Ultra-Wide Band Radar
IRJET- FPGA based Processor for Feature Detection in Ultra-Wide Band RadarIRJET- FPGA based Processor for Feature Detection in Ultra-Wide Band Radar
IRJET- FPGA based Processor for Feature Detection in Ultra-Wide Band Radar
 
FIR_Filters_with_FPGA
FIR_Filters_with_FPGAFIR_Filters_with_FPGA
FIR_Filters_with_FPGA
 
DickeyS_presentation_2015_3_26_2_1
DickeyS_presentation_2015_3_26_2_1DickeyS_presentation_2015_3_26_2_1
DickeyS_presentation_2015_3_26_2_1
 
Updated! Debugging EMI Problems Using a Digital Oscilloscope
Updated! Debugging EMI Problems Using a Digital OscilloscopeUpdated! Debugging EMI Problems Using a Digital Oscilloscope
Updated! Debugging EMI Problems Using a Digital Oscilloscope
 
Digital Implementation of Costas Loop with Carrier Recovery
Digital Implementation of Costas Loop with Carrier RecoveryDigital Implementation of Costas Loop with Carrier Recovery
Digital Implementation of Costas Loop with Carrier Recovery
 
Advanced lock in amplifier for detection of phase transitions in liquid crystals
Advanced lock in amplifier for detection of phase transitions in liquid crystalsAdvanced lock in amplifier for detection of phase transitions in liquid crystals
Advanced lock in amplifier for detection of phase transitions in liquid crystals
 
Introduction to OFDM.ppt
Introduction to  OFDM.pptIntroduction to  OFDM.ppt
Introduction to OFDM.ppt
 
A 45Gb/s PAM-4 Transmitter Delivering 1.3Vppd Output Swing with 1V supply in ...
A 45Gb/s PAM-4 Transmitter Delivering 1.3Vppd Output Swing with 1V supply in ...A 45Gb/s PAM-4 Transmitter Delivering 1.3Vppd Output Swing with 1V supply in ...
A 45Gb/s PAM-4 Transmitter Delivering 1.3Vppd Output Swing with 1V supply in ...
 
Getting started with digital modes
Getting started with digital modesGetting started with digital modes
Getting started with digital modes
 
305575138-VIBRASI-SE-3-Level2.pdf
305575138-VIBRASI-SE-3-Level2.pdf305575138-VIBRASI-SE-3-Level2.pdf
305575138-VIBRASI-SE-3-Level2.pdf
 
NT1065 LE DS v2.04
NT1065 LE DS v2.04NT1065 LE DS v2.04
NT1065 LE DS v2.04
 
anritsu mp1800a Анализатор качества сигнала BER
anritsu mp1800a Анализатор качества сигнала BERanritsu mp1800a Анализатор качества сигнала BER
anritsu mp1800a Анализатор качества сигнала BER
 

Recently uploaded

Customer Service Analytics - Make Sense of All Your Data.pptx
Customer Service Analytics - Make Sense of All Your Data.pptxCustomer Service Analytics - Make Sense of All Your Data.pptx
Customer Service Analytics - Make Sense of All Your Data.pptxEmmanuel Dauda
 
Digi Khata Problem along complete plan.pptx
Digi Khata Problem along complete plan.pptxDigi Khata Problem along complete plan.pptx
Digi Khata Problem along complete plan.pptxTanveerAhmed817946
 
100-Concepts-of-AI by Anupama Kate .pptx
100-Concepts-of-AI by Anupama Kate .pptx100-Concepts-of-AI by Anupama Kate .pptx
100-Concepts-of-AI by Anupama Kate .pptxAnupama Kate
 
Kantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdf
Kantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdfKantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdf
Kantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdfSocial Samosa
 
VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...
VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...
VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...Suhani Kapoor
 
Call Girls In Mahipalpur O9654467111 Escorts Service
Call Girls In Mahipalpur O9654467111  Escorts ServiceCall Girls In Mahipalpur O9654467111  Escorts Service
Call Girls In Mahipalpur O9654467111 Escorts ServiceSapana Sha
 
Dubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls DubaiDubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls Dubaihf8803863
 
FESE Capital Markets Fact Sheet 2024 Q1.pdf
FESE Capital Markets Fact Sheet 2024 Q1.pdfFESE Capital Markets Fact Sheet 2024 Q1.pdf
FESE Capital Markets Fact Sheet 2024 Q1.pdfMarinCaroMartnezBerg
 
Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝soniya singh
 
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130Suhani Kapoor
 
Best VIP Call Girls Noida Sector 39 Call Me: 8448380779
Best VIP Call Girls Noida Sector 39 Call Me: 8448380779Best VIP Call Girls Noida Sector 39 Call Me: 8448380779
Best VIP Call Girls Noida Sector 39 Call Me: 8448380779Delhi Call girls
 
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改atducpo
 
Full night 🥵 Call Girls Delhi New Friends Colony {9711199171} Sanya Reddy ✌️o...
Full night 🥵 Call Girls Delhi New Friends Colony {9711199171} Sanya Reddy ✌️o...Full night 🥵 Call Girls Delhi New Friends Colony {9711199171} Sanya Reddy ✌️o...
Full night 🥵 Call Girls Delhi New Friends Colony {9711199171} Sanya Reddy ✌️o...shivangimorya083
 
Industrialised data - the key to AI success.pdf
Industrialised data - the key to AI success.pdfIndustrialised data - the key to AI success.pdf
Industrialised data - the key to AI success.pdfLars Albertsson
 
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...Suhani Kapoor
 
Invezz.com - Grow your wealth with trading signals
Invezz.com - Grow your wealth with trading signalsInvezz.com - Grow your wealth with trading signals
Invezz.com - Grow your wealth with trading signalsInvezz1
 
Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...
Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...
Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...Sapana Sha
 
Market Analysis in the 5 Largest Economic Countries in Southeast Asia.pdf
Market Analysis in the 5 Largest Economic Countries in Southeast Asia.pdfMarket Analysis in the 5 Largest Economic Countries in Southeast Asia.pdf
Market Analysis in the 5 Largest Economic Countries in Southeast Asia.pdfRachmat Ramadhan H
 
VIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service Amravati
VIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service AmravatiVIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service Amravati
VIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service AmravatiSuhani Kapoor
 

Recently uploaded (20)

Customer Service Analytics - Make Sense of All Your Data.pptx
Customer Service Analytics - Make Sense of All Your Data.pptxCustomer Service Analytics - Make Sense of All Your Data.pptx
Customer Service Analytics - Make Sense of All Your Data.pptx
 
Digi Khata Problem along complete plan.pptx
Digi Khata Problem along complete plan.pptxDigi Khata Problem along complete plan.pptx
Digi Khata Problem along complete plan.pptx
 
100-Concepts-of-AI by Anupama Kate .pptx
100-Concepts-of-AI by Anupama Kate .pptx100-Concepts-of-AI by Anupama Kate .pptx
100-Concepts-of-AI by Anupama Kate .pptx
 
Kantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdf
Kantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdfKantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdf
Kantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdf
 
VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...
VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...
VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...
 
Call Girls In Mahipalpur O9654467111 Escorts Service
Call Girls In Mahipalpur O9654467111  Escorts ServiceCall Girls In Mahipalpur O9654467111  Escorts Service
Call Girls In Mahipalpur O9654467111 Escorts Service
 
Dubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls DubaiDubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls Dubai
 
FESE Capital Markets Fact Sheet 2024 Q1.pdf
FESE Capital Markets Fact Sheet 2024 Q1.pdfFESE Capital Markets Fact Sheet 2024 Q1.pdf
FESE Capital Markets Fact Sheet 2024 Q1.pdf
 
Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝
 
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
 
Best VIP Call Girls Noida Sector 39 Call Me: 8448380779
Best VIP Call Girls Noida Sector 39 Call Me: 8448380779Best VIP Call Girls Noida Sector 39 Call Me: 8448380779
Best VIP Call Girls Noida Sector 39 Call Me: 8448380779
 
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
 
Full night 🥵 Call Girls Delhi New Friends Colony {9711199171} Sanya Reddy ✌️o...
Full night 🥵 Call Girls Delhi New Friends Colony {9711199171} Sanya Reddy ✌️o...Full night 🥵 Call Girls Delhi New Friends Colony {9711199171} Sanya Reddy ✌️o...
Full night 🥵 Call Girls Delhi New Friends Colony {9711199171} Sanya Reddy ✌️o...
 
Industrialised data - the key to AI success.pdf
Industrialised data - the key to AI success.pdfIndustrialised data - the key to AI success.pdf
Industrialised data - the key to AI success.pdf
 
Delhi 99530 vip 56974 Genuine Escort Service Call Girls in Kishangarh
Delhi 99530 vip 56974 Genuine Escort Service Call Girls in  KishangarhDelhi 99530 vip 56974 Genuine Escort Service Call Girls in  Kishangarh
Delhi 99530 vip 56974 Genuine Escort Service Call Girls in Kishangarh
 
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
 
Invezz.com - Grow your wealth with trading signals
Invezz.com - Grow your wealth with trading signalsInvezz.com - Grow your wealth with trading signals
Invezz.com - Grow your wealth with trading signals
 
Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...
Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...
Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...
 
Market Analysis in the 5 Largest Economic Countries in Southeast Asia.pdf
Market Analysis in the 5 Largest Economic Countries in Southeast Asia.pdfMarket Analysis in the 5 Largest Economic Countries in Southeast Asia.pdf
Market Analysis in the 5 Largest Economic Countries in Southeast Asia.pdf
 
VIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service Amravati
VIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service AmravatiVIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service Amravati
VIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service Amravati
 

Basic Pitfalls in Waveform Analysis

  • 1. Signal Frequency FSig, FSR, and FNyq Summary 1/ 34 Nyquist frequency FNyq:= 0.5FSR, Historical upper limit FSR =16kHz FSig 0.1 FSR 0.5 FSR=FNyq 0.9 FSR November 15, 2020
  • 2. Basic Pitfalls in Waveform Analysis - Introduce DEEPS - Yukio Okuda sf.yukio@gmail.com an independent, Atsugi, Japan November 2020
  • 3. Me = Industrial Soft Hard +40 Years Introduction 3/ 34 Soft+Hard, 2 Years: Mechanical Vibration of Mother Machines
  • 4. Sensor Selection , Waveform Failure Analysis Soft, 8 Years: OSS Soft+Hard, 15 Years: Digital LSI Testing and Failure Analysis
  • 5. based on Data Analysis IEEE-ITC Papers Soft, 17 Years: Developing LSI Design CAD tools Soft, 6 Years: Developing Information Retrieval Hard, 3 Years: Color TV Production Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 6. Outline Introduction 4/ 34 Digitizing Error at lower than Nyquist Frequency FNyq Digitize (interpolate) No-Repeatability Pseudo Amplitude Modulation AM FNyq DCASE-2 DCASE-4 • Models • Error signatures Verified by DEEPS Digitize Error Estimation (Prediction) by (Spectrum) Improve Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 7. Why Digitize Introduction 5/ 34 ADC mismatches for feature extractors ¬ Digitizing errors
  • 8. ADC: Analog to Digital Converter ADC No error Extractor Feature Analysis Digitizing Time Time Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 9. Outline of Models Models 6/ 34 Show errors at lower than Nyquist Frequency Rarely reported Improving applications is the next stage – Low Sampling Rate — # of Signal Repeats Systematic Error Shape Spectrum Random Error DEEPS ˜ Clock Skew Show errors Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 10. What is ADC Models 7/ 34 Sampling at the predefined intervals of TS
  • 11. Sampling Rate FSR := 1/TS ADC = Sampling Distortions at high frequency Deterministic error ¬ Repeatable Start time is uncontrollable Random error ¬ No-Repeatable
  • 13. Repeatable : the same data at the same condition Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 14. Signal Frequency FSig, FSR, and FNyq Models 8/ 34 Nyquist frequency FNyq:= 0.5FSR, Historical upper limit FSR =16kHz FSig 0.1 FSR 0.5 FSR=FNyq 0.9 FSR Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 15. Signal Repeats Models 9/ 34 Signal repeats change shapes spectra
  • 16. Pseudo Amplitude(AM) Modulation at FSig= 7.1 kHz 16kHz sampled waves are up-converted to 160kHz
  • 17. Genuine AM ? Boundary FFT Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 18. Un-Controllable Start Time Models 10/ 34 Errors between measurements in a measurement Errors depend on FSig ? Robustness ADC Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 19. Estimate Error Variance caused by Start Time Variance Models 11/ 34 At a Start Time variance of Clock offsets Co=[0, 0.3, 0.6] Repeat number NR= 4, FSig= 6.1kHz, ∆ FSp := FSp− FSig Wave FFT Spectrum FSp Estimation ∆ FSp Amp.:Summarize,Normalize Freq.:∆from6.1kHz Heat Map ∆ FSp 6.1kHz ∆FSp Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 20. Compare Error Variances of FSigs at NR= 4 Models 12/ 34 Clock offsets Co= 0, 0.1, 0.2, .., 0.9 Assume random uniform FSig= 6, 6.1, 6.2, .., 7.5 kHz, NR= 4 FSig – — ∆FSp – 6.4kHz shows the own frequency — 7.1kHz shows the own frequency Others show: Different one frequency Different two or three frequencies Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 21. Compare Error Variances of FSigs at NR= 8 Models 13/ 34 Clock offsets Co= 0, 0.1, 0.2, .., 0.9 Assume random uniform FSig= 6, 6.1, 6.2, .., 7.5 kHz, NR= 8 FSig – —˜ ∆FSp – 6.4kHz shows the own frequency Same as at NR= 4 — 7.1kHz shows the own frequency Same as at NR= 4 ˜ 6.1kHz shows the own frequency Three frequencies at NR= 4 Others show two or three frequencies Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 22. A Magic Signal of 6.4 kHz Models 14/ 34 Differently distorted waves yield a FSp of 6.4 kHz! NR= 4 FFT Spectrum NR= 8 FFT Spectrum Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 23. Summary of Models Models 15/ 34 DEEPS estimates error signatures at a FSR of 16 kHz Signals higher than 4 kHz are distorted No error signatures Except, the Signals of 6.4 kHz, 7.1 kHz • Show the FSp of the FSig • Signals of 7.1 kHz show AM modulation waves Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 24. Outline of Verification Verification 16/ 34 A/B Testing is difficult for an independent Challenge post-mortem analysis on DCASE Challenge Datasets
  • 25. DCASE: Detection Classification of Acoustic Scenes Events Task2 ĸ • Obtained by the one tool Statistical analysis of spectra • 13,000 nominal sounds • 10 sec • FSR= 16 kHz • four device types Task4 ĸ • From Used by several Voice Recognition Activities • 12,000 unlabeled sounds, non-controllable measurements • 10 sec • FSR= 16 kHz, 44.1 kHz Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 26. Averaged Spectra of Task2 Verification 17/ 34 Compare tails of spectra from 4kHz – All 6.4kHz peaks — Slider/Valve 7.1kHz peaks – – – – — — Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 27. How to locate error signals Verification 18/ 34 Apply Low-Pass Filter-Bank Spectrum has no time information, i.e. Non-temporal Reversed signals show the same spectra FFT Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 28. Task2 Ex1: 6.4kHz Verification 19/ 34 Spectrum peaks around 6.4 kHz Ex1: slider/train/normal id 06 00000075 Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 29. Task2 Ex1: 7.1kHz Verification 20/ 34 Spectrum peaks around 7.1 kHz Ex1: slider/train/normal id 06 00000075 Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 30. Spectrum Endpoints of Task4 Verification 21/ 34 All sounds may include distortions – 70% sounds are digitized by FSR=16 kHz, upconverted to 44.1 kHz — 30% sounds are digitized by FSR=44.1 kHz – – — — Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 31. AM Signatures at a FSR of 44.1 kHz Verification 22/ 34 Peak count (2, 3, 4, 5) identifies AM signatures FSig 19.9 kHz — ˜ ™ š ˜ ™ ˜ ˜ ˜ ˜ — ˜ Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 32. Summary of Verification Verification 23/ 34 Demonstrated 1 Digitizing error at lower than FNyq 2 DEEPS estimations 3 sounds include distortions Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 33. Discussion for Improvements Discussion 24/ 34 Application policies set Max FSig Max Error Feature error predictions set FSR ADC + Waves Waveform Variations Controllable Non-Controllable Max FSig Max Error FSR Correct Feature Extraction Error Prediction FFT Statistics Time Series Wavelet Cepstrum Human Ear Eq. . . . DA DataAnalysis Appli. Science Medical Industry Home . . . Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 34. Feature Error Predictions by DEEPS Discussion 25/ 34 Feature extractor decide error from waveform variations Wave Gen Drive Extractor Features FSig-N Wave Gen Drive Extractor Features FSig-1 Sum Indicator FSig Already FFT with spectrum Challenging predictions: • Characteristics of waves • References to Metrics • Parameters • Window size • FSig dependancy • . . . May be effective on non-controllable systems Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 35. Error Metrics for FFT: REsp Discussion 26/ 34 Relative FSp Error REsp := (FSp– FSig)/ FSig Co= 0, 0.1, 0.2, .., 0.9, FSig= 1, 1.1, .. 7.9, NR= 4 FSig REsp Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 36. Error Metrics for FFT: FSp Discussion 27/ 34 Ghost FSp, FSig of 4.2, 4.3, 4.4 kHz show FSp of 4.0, 4.3, 4.6 kHz Co= 0, 0.1, 0.2, .., 0.9, FSig= 1, 1.1, .. 7.9, NR= 4 FSig FSp Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 37. Aliasing Drops at near FNyq Discussion 28/ 34 Unremovable error signals ¬ Reject FSig ≥ FNyq by H/W filters Co= 0, 0.1, 0.2, .., 0.9, FSig= 8.0, 8.5, 15.5, NR= 20 FSig FSp FNyq Drops FSp Spectrum Task2-Ex1 Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 38. Waves are Noises or Signals Discussion 29/ 34 Amplitude Ratio RAmp := AmpFSig /Amp Task2-Ex1, high pass filter of 7kHz ¬ Highly destorted signals Amplitude absRAmp Time Time Spectrum is a poor indicator of waveform amplitudes Waves with different amplitudes show the same spectrum amplitude FFT Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 39. Basic H/W Requirements Discussion 30/ 34 MUST-1 Reject FSig ≥ FNyq by – Mics or — Low-pass filters(LPFs) MUST-2 Apply ADCs with enough high FSR Option Reject FSig Max-FSig If need – Mics or ™ digital LPFs, Not˜LPFs MaxFSig – Low-Pass Filter MaxFSig —˜ ADC Low-Pass Filter ™ MaxFSig Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 40. Low-Pass Filter Distortions Discussion 31/ 34 Phase Shifts of analogue low-pass filter(LPF) cause distortions Phase Shifts depend on Circuit Frequency By Brews ohare ĸ FFT ¬ Spectrum + Phase Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 41. Audio Recording Discussion 32/ 34 Low FSR standard 1982 CD(Compact Disc) FSR= 44.1 kHz ¬ Challenging at ’80 • Poor Human Hearing, Ear Frequency Curve 1984 PC-AT Intel 80286 clock= 6, 8 MHz 2003 Pro audio standard AES3: 88.2, 96, 176.4, 192 kHz Mics of up to 20kHz ¬ 90% FNyq of FSR= 44.1 kHz Ear Frequency Curve ¬ Voice recognition unique features EarGain FSigɀ ĸ Pre-emphasis Gammatone Spectrograms Cepstrum Mel Spectrogram MFCCs Chroma Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 42. Conclusion 33/ 34 DEEPS demonstrated Existence of digitizing error at lower than FNyq sounds include distortions Indicated Apply FSRs obtained from Max FSig Max Error Feature error predictions Check Audio recording for Data Analysis Y. Okuda Basic Pitfalls in Waveform Analysis PyData Global 2020 talk-38
  • 43. Thank you, the supporting staffs of DCASE, PyData Global, Your Questions Comments