instrumentation - imperialaczaja/pg2008/instrumentationpart1.pdf · –the instrumentation messes...
TRANSCRIPT
Instrumentation
Chris Carr
23rd February 2009
Instrumentation
• Today:
– 10:00 to 11:30Introduction to Instrumentation
• Chris Carr
– 14:00 to 15:30Magnetometers
• Patrick Brown
• Friday:
– 14:00 to 15:00Problem Sheet session
• Chris and Patrick
24/02/2009 2
The Scientific Method
Question:Which Greek philosopher is principally responsible for modern science?
24/02/2009 3
The Scientific Method
24/02/2009 4
A ‘Generic’ Instrument
24/02/2009 5
A Real instrument…
Fluxgate Magnetometer Instrument for the ESA/CNSA ‘Double Star’ Mission
24/02/2009 6
Input Transducers
• Our sensor is an input transducer:A device to convert a measureable (physical quantity) into an electrical signal
24/02/2009 7
Input Transducers
• Sensor often separate from the electronics:
– Because of InterferenceExample: Space Magnetometer
– Because of access or ease of useExample: Ultrasound (medical or industrial)
– Because the environment is hazardous (to people or electronics)Example: Nuclear/Particle Physics and Space
24/02/2009 8
Transducer Behaviour
• The ideal sensor…
• …does not exist!
24/02/2009 9
Transducer Behaviour
• A more realistic model…
• Examples: temperature sensors, pressure sensors, force/acceleration sensors, etc.
24/02/2009 10
Transducer Behaviour
• Real-world sensor characteristics
• Non-linearity is a major problem (as we shall see)
24/02/2009 11
Transducer Behaviour
• Real-world sensor characteristics
• We can define a range over which the sensor is approximately linear
24/02/2009 12
Transducer Behaviour
• We want a linear response because
– ‘Gain’ or ‘sensitivity’ is constant
– And because linear systems allow formalised mathematical analysis
• To achieve linear operation
– Limit operating range to the linear region
– Operate the sensor with feedback (this afternoon’s lecture)
24/02/2009 13
Static Response
• These graphs gives us the static response
– Output for a fixed, constant input
• Example: mechanical scales
– f(m) is the response the system reaches a long time (in theory, ) after theinput changes
t
24/02/2009 14
Frequency Domain Behaviour
• But for most physics applications the physical measureable is a time-varying quantity
• If the static response of the system is linear then we characterise the linear system with a Bode plot
– Gain and Phase of output(relative to input)
24/02/2009 15
Bode Plot
• Specifies the Transfer Function
• Gives the bandwidth (B)
• Applicable to sensor, filter, amplifier or the whole instrument
B
24/02/2009 16
Signals
• The system (sensor, amplifier, etc…) has frequency dependent behaviour
• Need to be aware of how the frequency content of our signal is affected
• We need a “Fourier Understanding” of signals
• Recall the Fourier Transform
• And the inverse transform
24/02/2009 17
Example: Square Wave
• An infinite signal (over all time)
• Has a spectrum consisting of an infinite series of sinusoids
Signal f(t) Spectrum F( )
AngularFrequency
Time Domain Frequency Domain
Time t
Amplitude Amplitude
0
11,3 5
/ . .( ), ....
n Sin n on
24/02/2009 18
Other Waveforms
24/02/2009 19
Some Further Examples
• Infinite Sinusoid has a finite spectrum
• Conversely, a finite sinusoid has an infinite spectrum
24/02/2009 20
Implications
• Signals with an infinite Fourier representation require an infinite bandwidth to support them
• This is unphysical
– Sensors and electronics have finite bandwidths
– Real signals are of finite duration and hence have wide spectrum
• So:
– The instrumentation messes with the spectral content of your signal
– Fidelity of recorded signal always compromised
– Example: see problem sheet
24/02/2009 21
Transmission Lines
• Sensor signal usually small and weak
• Transmission Line
– Preserves signal integrity
• Shape, amplitude, phase etc
– Protects from external interference
• Noise pickup
• Example:Co-ax cable whichconnects aerial to TV
24/02/2009 22
Impedance
– A complex quantity. We just need the magnitude.
• Longitudinal Compression Waves, e.g. sound
• EM Waves
• Electrical waves (signals) in a co-axial cable
– Where L is the inductance per metre, C the capacitance per metre
24/02/2009 23
Boundaries
• At the boundary between 2 media with different impedances, we get reflections
• Amplitude reflection coefficient KR
• The physics is the same:Mechanical Waves, Optical EM Waves, Electrical signals in a cable…
24/02/2009 24
Optics Example
• At air/glass interface we get 4% reflection (96% transmission)
• 10 interfaces gives 0.9610 = 66% light transmitted!
• In optics we use anti-reflective coatings
• In electronics we can directly change the impedances to minimisesignalloss
24/02/2009 25
Impedance Matching
• Z1 = Z2 means KR = 0
• Impedance matching is the condition for Minimum reflectionandMaximum Signal Power Transfer
• Example: see problem sheet
24/02/2009 26
Noise
• Have seen how sensors and electronics can change our signal characteristics
• And how signal power can be lost through impedance mis-matches
• A further consideration is the noise introduced into our measurements
• Generally, we need to worry about 3 sources
– Thermal Noise
– Shot Noise
– Flicker Noise
24/02/2009 27
Thermal Noise
• Arises from the random thermal movement of free-electrons in a conductor
• Is therefore a function of temperature
• The RMS noise voltage measured with an instrument bandwidth B is
• Where
24/02/2009 28
Shot Noise
• Arises from statistical fluctuations in charge carriers
• Wherever we have a current crossing a junction
• Typically in semiconductors and vacuum tubes
Photomultiplier Tube
24/02/2009 29
Flicker Noise (aka 1/f noise)
• Source not understood
• Fundamental property of all measurement systems
• Noise power is inversely proportional to frequency
• Contrast to Thermal & Shot noise
– Have power constant with frequency
24/02/2009 30
Composite Noise Spectrum
0
0.5
1
1.5
2
2.5
3
3.5
0 0.5 1 1.5 2 2.5 3 3.5
Log(Power)
log(f)
log (1/f)
log (shot)
log (thermal)
log (total)
24/02/2009 31
Consequences
• To reduce noise in our measurements
– Avoid low-frequencies
– Reduce temperature
• Of sensor and, often, electronics such as pre-amplifiers
– Reduce measurement bandwidth
• A filter is effectively a bandwidth reducing device
Pass-BandFilter
24/02/2009 32
Systems Analysis
• Mathematical framework to
– Model complex sensors and instruments
– Predict behaviour
– Extract measureable m from the output f(m)
– Test stability of systems
• Example of the “block diagram”approach used insystems analysis
24/02/2009 33
First Order System
• Simple Example: Mercury Thermometer
• A first order differential equation
• Characteristic of many sensors and electrical systems
Thermometer temperature
Initial: θ0
At time t: θ(t)
Heat Bath
temperature: θR
Input u(t)
Time t Time t
Output x(t)
t = 0 t = 0 t =
63.2%
Steady State
Initial Value
24/02/2009 34
Second Order System• Damped Harmonic Oscillator
• 2nd order D.E.
• Many mechanical andelectrical systems
• A physical representation:
Mass m
Damping Term
SpringConstant K
Scale
Pointer
Deflection x(t)
Time t
Time t
Time t
Time t
Input y(t)
Output x(t)
Output x(t)
Output x(t)
(b) Zero damping
Infinite oscillation
(c) Moderate damping
Oscillation decays to steady state
(d) Heavy Damping
No oscillation
Zero damping
Infinite oscillation
(a) Step Input
A mass is hung on the balance
24/02/2009 35
These are
Linear Time-Invariant Systems
• Frequency Preservation
– ω is not changed by passing through the system
• Linear Superposition
– For each frequency component in the signal only the amplitude and phase are modified
24/02/2009 36
We can build and analyse complex systems
• Where Gn is the Transfer Function for each part of the system
1. May be the sensor
2. Could be an amplifier
3. Could choose to be an integrator… etc…
• N.B. For Transfer Function, remember the Bode Plot
G1(s) G2(s) G3(s)Y(s) X(s)
1 2 3
G(s)
24/02/2009 37
Writing the Equations
• For each part we can write a DE of the form
• Then we would have a series of DE’s we could solve…
• But this rapidly becomes very hard
• Use the Laplace Transform to simplify the problem
24/02/2009 38
Laplace Transform very briefly…
• Transform all signals and transfer functions into the s-domain
• Differential equations become linear algebraic
G1(s) G2(s) G3(s)Y(s) X(s)
1 2 3
G(s)
24/02/2009 39