You are on page 1of 23

9/16/2015

Measurement Systems- More basic concepts


Walter Elson
And Webster

HW#1
Problems 1.3, 1.4, 1.9,
and 2.1
Due- a week from now!

Measurement Systems- More basic


concepts
Scientific method- a hypothesis is tested by
experiment to determine its validity.

Problem
statement

Review
prior work

State
hypothesis

Perform
experiments

Design further
experiments
More
experiments
necessary

Analyze
data

Final
conclusions
Problem
solved

9/16/2015

Measurement Systems- More basic


concepts
The physician obtains the history, examines
the patient, performs tests to determine the
diagnosis and prescribes treatment.
Chief
complaint

Obtain
history

List the
differential
diagnosis

Examination
and tests

Treatment
and
evaluation

Select further
tests

Use data
to narrow the
diagnosis

Final
diagnosis

More than
one likely

Only one
likely

Measurement Systems- More basic concepts

A typical measurement system uses sensors to measure


the variable, has signal processing and display, and may
provide feedback.
Outputs

Measurand

Sensor

Signal
conditioning

Feedback
Effector

Signal
processing

Data
storage

Data
displays

Data
communication

9/16/2015

Measurement

Range

Frequency, Hz

Method

Blood flow

1 to 300 mL/s

0 to 20

Electromagnetic or
ultrasonic

Blood pressure

0 to 400 mmHg

0 to 50

Cuff or strain gage

Cardiac output

4 to 25 L/min

0 to 20

Fick, dye dilution

Electrocardiography

0.5 to 4 mV

0.05 to 150

Skin electrodes

Electroencephalography

5 to 300 V

0.5 to 150

Scalp electrodes

Electromyography

0.1 to 5 mV

0 to 10000

Needle electrodes

Electroretinography

0 to 900 V

0 to 50

Contact lens electrodes

pH

3 to 13 pH units

0 to 1

pH electrode

pCO2

40 to 100 mmHg 0 to 2

pCO2 electrode

pO2

30 to 100 mmHg 0 to 2

pO2 electrode

Pneumotachography

0 to 600 L/min

0 to 40

Pneumotachometer

Respiratory rate

2 to 50
breaths/min

0.1 to 10

Impedance

Temperature

32 to 40 C

0 to 0.1

Thermistor

Table 1.4 Common medical measurands.

Specification

Value

Pressure range

30 to +300 mmHg

Overpressure without damage

400 to +4000 mmHg

Maximum unbalance

75 mmHg

Linearity and hysteresis

2% of reading or 1 mmHg

Risk current at 120 V

10 A

Defibrillator withstand

360 J into 50

Table 1.5 Sensor specifications for a blood pressure sensor are


determined by a committee composed of individuals from academia,
industry, hospitals, and government.

9/16/2015

Measurement Systems- More basic


concepts

Amplitude

Amplitude

a limited number of
amplitude values

Time

Time

(a)

(b)

Amplitude

Amplitude

Figure 1.8 (a) Analog signals can have any amplitude value. (b) Digital signals have
a limited number of amplitude values.

Time
(a)

Time
(b)

Figure 1.13 (a) Continuous signals have values at every instant of time. (b)
Discrete-time signals are sampled periodically and do not provide values between
these sampling times.

9/16/2015

Specification

Value

Input signal dynamic range

5 mV

Dc offset voltage

300 mV

Slew rate

320 mV/s

Frequency response

0.05 to 150 Hz

Input impedance at 10 Hz

2.5 M

Dc lead current

0.1

Return time after lead switch

1s

Overload voltage without damage

5000 V

Risk current at 120 V

10

Observe
Committee

Table 1.6 Specification values for an electrocardiograph are agreed upon by a


committee.

Laboratory test

Typical value

Hemoglobin

13.5 to 18 g/dL

Hematocrit

40 to 54%

Erythrocyte count

4.6 to 6.2 106/ L

Leukocyte count

4500 to 11000/ L

Differential count

Neutrophil 35 to 71%
Band 0 to 6%
Lymphocyte 1 to 10%
Monocyte 1 to 10%
Eosinophil 0 to 4%
Basophil 0 to 2%

Table 1.7 Complete blood count for a male subject.

9/16/2015

1.9 Performance CharacteristicsQuantifying the quality of measurement


A. Static Instrument characteristics:
1) Accuracy
2) Precision
3) Resolution
4) Sensitivity
B. Dynamic Instrument characteristics:
All are functions of frequency.

A. Static Instrument characteristics


1. Accuracy
With this

Accuracy is the capacity of a measuring instrument to give RESULTS


definition

close to the TRUE VALUE of the measured quantity

Accuracy is measured by the absolute and relative errors

- smaller
is better

ABSOLUTE ERROR RESULT - TRUE VALUE

Notes:

1) True value can be a varying function of time and/or frequency


2) True value need to be connected also with the National Institute of
Standards Technology (NIST).

9/16/2015

A. Static Instrument characteristics


1. Accuracy

(a)

(a) low accuracy

(b) high(b)accuracy

Figure 1.17 Data points

A. Static Instrument characteristics


2. Precision
To have distinguishable alternatives
e.g. a measurand of 2.434 V is more precise
than 2.43 V
Related to degree of reproducibility or
repeatability!

Because
precision
has nothing
to do with
true value!

Accuracy and Precision are not the same! Why?

9/16/2015

A. Static Instrument characteristics


2. Precision

Figure 1.16 Data points

(a)
with (a) low
precision

(b)
(b) high precision.

Notes:

1) Think of the standard deviation of the measured values.


2) Best instrument is the one that is both accurate and precise.

A. Static Instrument characteristics


2. Precision
1) Two terms closely related to precision

Repeatability
The precision of a set of measurements taken over a short time interval

Reproducibility
The precision of a set of measurements BUT
taken over a long time interval or
Performed by different operators or
with different instruments or
in different laboratories
2) Think of the standard deviation of the measured values.
3) Best instrument is the one that is both accurate and precise.

9/16/2015

A. Static Instrument characteristics


Question- Accuracy and preciseness? Which one

is what?

High precision does not imply high accuracy not imply


high accuracy!!!

Which is more accurate?


Which is more precise?

9/16/2015

A. Static Instrument characteristics


3. Resolution (or discrimination)
Smallest increment that can be resolved in the
measurement!

It is the degree to which a change can be theoretically detected,


usually expressed as a number of bits. This relates the number of
bits of resolution to the actual voltage measurements. Also, it
related to ADC.

For digital outputs, it is the one LSB (the number of output levels it can
quantize a signal to)

For example8-bit over 2.56 volts range will yield a resolution of


1 LSB = ?
Think of it as how close values can be to
each other but yet the are discriminated

Answer
is

10mv

A. Static Instrument characteristics

4. Sensitivity
This is an absolute quantity. It is the smallest
absolute amount of change that can be detected by
a measurement.
Also,
Gain of a system = change of O/P /change of I/P

10

9/16/2015

A. Static Instrument characteristics

4. Sensitivity
Sensor
signal

Figure 1.7 Data


Sensor
signal

Measurand
(a)
A low-sensitivity sensor
has
low gain.

Measurand
(b) sensor has high
A high sensitivity
gain.

Note- offset does not impact Sensitivity

A. Static Instrument characteristics

4. Sensitivity

Note 1:
Offset does not affect
sensitivity
Note 2:
Instruments saturate in low
and high input regions,
well, almost. It is best to
operate in high sensitivity
region.

11

9/16/2015

A. Static Instrument characteristics


The relationship between the physical measurement
variable
(X) and the signal variable (S)
e
An instrument is calibrated by applying a number of
KNOWN physical inputs and recording the response of
the system
Signal output (Y)

Physical input (X)

A. Static Instrument characteristics

Interfering inputs (Y)


e

Those that the sensor to respond as the linear superposition with the
measurand variable X

Linear superposition assumption: S(aX+bY)=aS(X)+bS(Y)


Modifying
input Z
Physical variable X

Measuran
d

Sensor

Interfering input Y

Modifying inputs (Z)


e

Those that change the behavior of the


sensor and, hence, the calibration curve

Signal
variabl
e
S

Signal output (Y)

Z=Z1
Z=Z2

Temperature is a typical modifying input


Physical input (X)

12

9/16/2015

Systematic errors
Result from a variety of factors

Interfering or modifying variables (i.e., temperature)


Drift (i.e., changes in chemical structure or mechanical
stresses)
The measurement process changes the measurand (i.e.,
loading errors)
The transmission process changes the signal (i.e., attenuation)
Human observers (i.e., parallax errors)

Systematic errors can be corrected with compensation


methods (i.e., feedback, filtering)

Random errors-Also called NOISE: a signal that carries


no information
True random errors (white noise) follow a Gaussian
distribution
Sources of randomness:
Repeatability of the measurand itself (i.e., height of a
rough surface)
Environmental noise (i.e., background noise picked by a
microphone or power line signals)
Transmission noise (i.e., 60Hz hum)
Signal to noise ratio (SNR) should be >>1

With knowledge of the signal characteristics it may be possible to


interpret a signal with a low SNR (i.e., understanding speech in a
loud environment)

13

9/16/2015

Other notes
Linearity
The closeness of the calibration curve to a
specified straight line (i.e., theoretical behavior,
least-squares fit)

Other notes
How about when nonlinear?
Output

Output

Input
(a)

Input
(b)

Figure 1.18 (a) The one-point calibration may miss nonlinearity. (b)
The two-point calibration may also miss nonlinearity.

14

9/16/2015

Other notes
Population standard
deviation

Estimated mean xs
standard deviation sx

xi
n

Frequency

( x i x)
n 1

f (X )

2
2
e X 2
2

Mean

Figure 1.19 For the normal distribution, 68% of the data lies within 1
standard deviation. By measuring samples and averaging, we obtain the
x s sx. is the tail
estimated mean
, which has a smaller standard deviation
probability that xs does not differ from by more than .

0.2

x
x

x
p

0.1
x
x
0
0

Figure 1.20 A typical Poisson distribution for m = 3.

15

9/16/2015

Other notes
Threshold

Normal
population

True
negative

False
positive, p =

Diseased
population
False
negative, p =

True
positive

Figure 1.21 The test result threshold is set to minimize false positives and false
negatives.

Other notes
Amplitude
5 mV

(a)
Time

Dynamic
Range

-5 mV

Amplitude
1V

(b)

Time
-1 V

Figure 1.9 (a) An input signal which exceeds the dynamic range. (b) The resulting
amplified signal is saturated at 1 V.

16

9/16/2015

The reason for dynamic characteristics is the presence of energystoring elements


- Inertial: masses, inductances
- Capacitances: electrical, thermal

1/A

time

time

Amplitude

Amplitude

Amplitude

Amplitude

Amplitude

Dynamic characteristics are determined by analyzing the response of the


instrument to a family of variable input waveforms:
- Impulse, step, ramp, sinusoidal, white noise

time

time

time

The dynamic response of the sensor is (typically) assumed to be


linear, therefore, it can be modeled by a constant-coefficient linear
differential equation:
d2 y(t)

ak

d y(t)
a
dt k

dt

dy(t)

a
1

dt

a y(t) x(t)
0

In practice, these models are confined to zero, first and


second order. Higher order models are rarely applied

These dynamic models are typically analyzed with the


Laplace transform which converts the differential
equation into a polynomial expression
Think of the Laplace domain as an extension of the
Fourier transform.
Fourier analysis is perfect for sinusoidal signals

17

9/16/2015

The Laplace transform of a time signal y(t) is denoted by

L[y(t)] = Y(s)

The s variable is a complex number s=+j


e

The real component defines the real exponential behavior


The imaginary component defines the frequency of oscillatory behavior

The fundamental relationship is the one that concerns the


transformation of differentiation
d
L y(t) sY(s) - f 0

dt

Other useful relationships are


Impulse : L (t) 1
1
Lu(t)
Step :
s
1

r(t)

L
Ramp :
s2

Decay :

Lexpat s - a -1

Sine :

Lsin t

Cosine :

Lcos

s2
s
t 2
s

Applying the Laplace transform to the sensor model yields


d2 y

dk y
dy
a1
L ak k a 2
a o y(t) x(t)
2
dt
dt
dt

a s

a s 2 a s1 a o Y(s) X(s)

Y(s)
1

G(s)
X(s)
a sk a s2 a s a
k

G(s) is called the transfer function of the sensor

The position of the poles of G(s) -zeros of the denominator- in


the s-plane determines the dynamic behavior of the sensor
such as
e

Oscillating components
Exponential decays
Instability

18

9/16/2015

- j

2t

1t

e 1t sin 1t

-1 j

+j1

-2
1

-j1

+j1

+j1

sin

+j2

t
t

+j2

si
t

1t

+1

-1

e 1t sin 1t

j
+j1

-j1

Input and output are related by an equation of the type


y(t) k x(t)
e

Y(s)
k
X(s)

Zero-order is the desirable response of a sensor

No delays
Infinite bandwidth
The sensor only changes the amplitude of the input signal

Example- zero-order
sensor
A potentiometer used to
measure linear and rotary
displacements
This model would not work for
fast-varying displacements

VCC

X
Y

Note: Zeroorder systems


do not include
energy-storing
elements

19

9/16/2015

Inputs and outputs related by a first-order differential equation

dy

a
1

a y(t) x(t)

dt

Y(s)

X(s)

1
k

a1s a0 s 1

First-order sensors have one element that stores energy and one that dissipates it
Step response

y(t) = Ak(1-e-t/)
A is the amplitude of the step
k (=1/a0) is the static gain, which determines the static response
(=a1/a0) is the time constant, which determines the dynamic response
Ramp response
y(t) = Akt - Aku(t) + Ake-t/
Frequency response
Better described by the amplitude and phase shift plots

Step response

Frequency response
e

Corner frequency c=1/


Bandwidth

Ramp response

20

9/16/2015

Inputs and outputs are related by a second-order differential


equation

a2
e

d2 y
1
a1 dy a0 y(t) x(t) Y(s)
2
dt 2
dt
X(s) a 2s a 1s a 0
We can express this second-order transfer function as
Y(s)
k

X(s) s2 2

2
n

s n
1
with k ,
a0

Where

a1
,
2 a 0 a1

a0
a2

k is the static gain


is known as the damping coefficient
n is known as the natural frequency

Response types
Underdamped (<1)
Critically damped
(=1)
Overdamped (>1)
e

Response parameters
e

Rise time (tr)


Peak overshoot (Mp)
Time to peak (tp)
Settling time (ts)

From [PAW91]

21

9/16/2015

Ramp response

Frequency response

From [PAW91]

A thermometer covered for protection


e
Adding the heat capacity and thermal resistance of the
protection yields a second-order system with two real poles
(overdamped)
Spring-mass-dampen accelerometer
e
The armature suffers an
acceleration
We will assume that this
acceleration is orthogonal
to the direction of gravity
e
x0 is the displacement of the
mass M with respect to the
armature
Mx i x 0 Kx 0 Bx 0
e
The equilibrium equation is:

Ms 2 X i (s) X 0 (s) K Bs Ms 2

x0

M
M

xi

K/M
X 0 (s) M

s 2 X i (s) K s 2 s(B/M) K/M

22

9/16/2015

End chapter 1

23