1
Developer Techniques Sessions Physical Measurements and Signal Processing Control Systems Logging and Networking 2
Abstract This session covers the technologies and configuration of a physical measurement system including high speed signal acquisition and embedded analysis and signal processing. Systems like this are often used in performance and machine monitoring to predict and diagnose faults. 3
Physical Measurements & Signal Processing Application Have you had unexpected machine failures? How would you improve your machine to prevent failures? A monitoring system may have predicted this failure! 4
Physical Measurement Subsystem Transducer & SigCon ADC Analysis Software 5
Traditional Implementation Computer Sensors Cable Cable Signal Conditioning DAQ Device Terminal Block MUX DAC ADC Digital Counter Computer I/O RTSI 6
Modern Implementation Sensors Ethernet USB Standard cables and data buses Direct sensor connectivity Integrated signal conditioning and ADC Sensors 7
Technology Shifts & Benefits Analog Optocouplers Multiplexed Architecture Low Resolution ADCs Basic Processing Digital MEMS Isolation Multi-ADC Architecture High Resolution ADCs Advanced Analysis Lower cost & smaller size Mixed-signal inputs Better measurement quality Improved decision-making 8
Inside a Modern I/O Module Direct Connectivity Screw terminals, BNC, D-Sub connectors for industrial sensors and actuators Isolation Barrier Safety, noise immunity, commonmode rejection Built-in Signal Conditioning Amplification, anti-aliasing filters, bridge completion, CJC, sensor power, IEPE, TEDS Advanced ADCs 24-bit, delta-sigma, simultaneous sampling, up to 800 ks/s, NISTtraceable calibration 9
Case Study: Machine Monitoring System Transducer & SigCon ADC Analysis Software 10
Transducers and Signal Conditioning Phenomena Transducer Signal Conditioning Temperature Proximity Vibration Force and Pressure Position and Displacement Thermocouples Resistive Temperature Devices (RTDs) Thermistors Limit Switches Proximity Switches Accelerometers Proximity Probes Strain gages Load Cells Potentiometers Linear voltage differential transformer (LVDTs) Optical Encoders Amplification, Linearization, Cold- Junction Compensation, Current Excitation Power Current Excitation, AC-Coupling, Eddy Current Power Voltage Excitation, Bridge Completion, Linearization RMS Voltage Excitation Fluid Flow Rotational Flowmeters Excitation, Filtering 11
Thermocouple Operation Due to the Seebeck Effect, dissimilar metals in contact produce a voltage proportional to the temperature Vo + - Machine Monitoring Tip: Motor temperature can be a simple way to evaluate at machine performance 12
Thermocouple Output Output is in mv range The voltage vs. temperature relationship is nonlinear over large temperature ranges 13
Demo: System Configuration & Thermocouple Measurement Demo 1 14
Why Use Signal Conditioning? Signal Conditioning Noisy, Low-Level Signal Improve signals for better measurement quality Power or excite sensors Read sensor information TEDS User and system safety Filtered, Amplified Signal 15
Isolation Signal Conditioning for Safety New digital MEMS isolators replace analog optocouplers with chip-scale transformers: Increased data rates and bandwidth Smaller size Lower power consumption Lower cost 2300 Vrms Transient 16
Accelerometer Operation In an accelerometer a seismic mass is coupled to a piezoeletric crystal. The force generated by the mass is caused by its momentum, which opposes the applied acceleration The force, and resultant output signal, are proportional to acceleration according to Newton s Law, F=mA The acceleration sensing element is installed into a protective housing An amplifier is included to amplify small signals to the mv-v range Machine Monitoring Tip: Accelerometers are often mounted to bearing housings and gear boxes to monitor vibrations 17
Accelerometer Output When exposed to vibration, the accelerometer generates an analog output signal, which is proportional to the acceleration of the applied vibration. 27 VDC 24 VDC 21 VDC AC Signal (Acceleration) DC Offset due to IEPE 0 VDC IEPE (Integrated Electronic Piezoelectric): Accelerometer power for internal amplifier at 2-20 ma constant current; 18-30 V DC 18
Demo: Accelerometer Measurement Demo 2 19
Analog-to-Digital Converter Performance & Price Decreasing Price of ADCs 16-bit ADCs cost 85% less than 10 years ago Increasing ADC Performance 24-bit resolution Up to 102 db dynamic range Integrated anti-aliasing filters Simultaneous sampling architectures with multiple ADCs Price in USD $60 $50 $40 $30 $20 $10 $0 12-bit ADC 16-bit ADC 18-bit ADC NI Price/IO Ch 1990 1992 1994 1996 1998 2000 2002 2004 Year 20
Lower Prices Enable Multi-ADC Systems Single ADC & MUX System Relays <500 S/s ADC Data Transfer < 1.8 MB/s (GPIB) Multi-ADC System ADC ADC ADC ADC ADC 3.2 MS/sADC ADC ADC Timing & Bus Controller Data Transfer > 60 MB/s (USB 2.0) 21
ADC Performance Improvement Resolution = # of bits the ADC uses to represent a signal Determines how many different voltage changes can be measured: # of levels = 2 resolution = 2 16 = 65,536 levels Code width is the smallest signal change your system can detect Smaller Code Width = more precise representation of your signal code width = range amplification * 2 resolution = 20 V 1 * 2 16 = 305 µv 16-bit ADC = 20 V 1 * 2 24 = 1.19 µv 24-bit ADC 22
Demo: 16-bit vs. 24-bit ADCs 16 bits 24 bits 12 bits Demo 3 23
Effect of ADC Resolution High resolution ADCs allow you to detect both strong and weak signal components at the same time. 24-bit 16-bit Measurement 90 db 140 db Detect Low-Level Signals Machine Monitoring Tip: 24-bit ADCs have the dynamic range to detect small vibrations in machines sooner so that repairs can be scheduled 24
Analysis & Signal Processing Software Filter Out Unwanted Components Frequency Determination Order Analysis Signal Processing Function Pure Sinusoid Signal Processing Function 500 Hz Input Signal Signal Processing Function 1x, 2x, Magnitude 25
Filtering Lowpass Filter Time Domain Time Domain Frequency Domain Removes noise Blocks unwanted frequencies Analog & digital implementations Frequency Domain 26
Ideal Filters Lowpass and Highpass Filters 27
Practical (Non-ideal) Filters Lowpass Filter Gain (db) Bode Plot Ripple Passband Frequency f c Corner Rolloff Passband frequencies the filter lets pass Ripple - filter s affect on the signal s amplitude Corner frequency where the filter begins blocking the signal Rolloff how sharply the filter cuts off unwanted frequencies 28
Comparing Implementations Digital Filters (Software) Software programmable Easier to implement Stable and predictable Do not change with time or manufacturing changes Analog Filters (Hardware) Set functionality Signal is filtered before ADC Less processor intensive Useful for anti-aliasing 29
Digital Filter Classification Designing digital filters involves making compromises to emphasize characteristics you want over characteristics you do not want IIR - Infinite Impulse Response Significantly faster Can achieve the same level of attenuation as FIR filters but with far fewer coefficients More efficient filtering operation Non-linear phase response Useful for applications that do not require phase information, such as signal monitoring applications Generally use less hardware or less code and thus are usually less expensive. FIR - Finite Impulse Response Generally slower and more processor intensive Linear phase response Often required in applications with band pass filtering Machine Monitoring Tip: Antialiasing filters make sure that measured signals do not contain false components 30
Choosing Filters Answer the following questions to select a filter for an application: Does the analysis require a linear-phase response? Can the analysis tolerate ripples? Does the analysis require a narrow transition band? 31
Types of IIR Filters Butterworth 32
Types of IIR Filters Chebyshev 33
Types of IIR Filters Chebyshev II 34
Types of IIR Filters Elliptic 35
Types of IIR Filters Bessel 36
Comparison of IIR Filters 37
Demo: Digital (Software) Filter Demo 4 38
Spectral Analysis Discrete Fourier Transform (DFT) Converts discrete time domain signals into the frequency domain DFT 500 Hz 39
Spectral Leakage A signal that is not sampled for a integer number of cycles will contain high frequency discontinuities causing spectral leakage Spectral leakage distorts the frequency measurement because the high frequency energy at the discontinuity is spread over the frequency spectrum Block size contains integer number of cycles (3) No leakage in FFT Block size contains non-integer number of cycles (3.3) Spectral leakage (spectrum spreading) visible in FFT 40
Reducing Spectral Leakage Windowing allows us to emphasize portions of a signal while de-emphasizing others Windowing allows us to minimize the amplitude of the discontinuity Reduces spectral leakage Machine Monitoring Tip: Frequency analysis is useful for dissecting vibration signals 41
Demo: Frequency Analysis Demo 5 42
What is Order Analysis? Many noise and vibration signal components are directly related to running speed: Imbalance, misalignment, gear mesh, bearing defects, loose coupling Order analysis normalizes the measurements to the rotating speed to better dissect these signal components 43
Relationship of Orders and Faults We can diagnose machine faults by knowing the order: Imbalance Low Order Misalignment Loose Coupling Valve Noise Bearing Defects / Wear Blade Pass Frequency Gear Mesh High Order 44
Using Order Analysis on Machinery 3,300 RPM = 55 RPS 55 1 st order 45 220 4 th order 385 7 th order Hz
Power Spectrum Rotating Speed: 60 Hz (3600 RPM) Rotating Speed: 50 Hz (3000 RPM) Frequency components shift with speed change 46
Order Spectrum Rotating Speed: 60 Hz (3600 RPM) Rotating Speed: 50 Hz (3000 RPM) Order components remain fixed with speed change 47
Power Spectrum to Order Spectrum Spectrum of the fan housing vibration, with the fan running at 600 RPM (10 Hz) 1X 2X 600 RPM (10 Hz) frequency (Hz).5X 3X 4X 5.5X 10 20 30 40 55 1 2 3 4 5.5 frequency (orders) Express frequency in terms of multiples of the rotating speed Machine Monitoring Tip: Order analysis is especially useful for machines with varying speeds and run-up, coast-down tests 48
Understanding Order Analysis Time Domain sec FFT Frequency Domain Hz Resampling Angular Domain deg FFT Order Domain order 49
Resampling for Changing Machine Speed Data is converted to angular domain from time domain with the use of a tachometer (speed) signal Machine speed is increasing 50
Demo: Order Analysis Demo 6 51
Summary There are 3 main components of a physical measurement system Monitoring machines can prevent failures Transducer & SigCon ADC Analysis Software 52