Indian Institute of Technology Bombay

Department of Electrical Engineering

Deadline: 10th August 2017 EE 617: Sensors in Instrumentation Homework 1 Solution Key

Note: Homeworks are to be solved individually. There is zero tolerance for copying.
1. Consider an accelerometer that we are using to measure gravity (defined as 1g). With a precise
measurement instrument, we have recorded 1000 samples for the accelerometer’s output and observed
the mean (µ) value as 1.02g and standard deviation (σ) as 10mg. The measurements follow a normal
(Gaussian) distribution. The accelerometer has bandwidth of 100Hz and full scale range of ±2g.
(a) (2 marks) Based on the definition taught in class, what is the typical value of the measurement?

Answer:
Typical value = µ ± 3σ = 0.99g to 1.05g.

(b) (2 marks) Your marketing manager would like to spec the noise in terms of the noise spectral
density instead
√ of RMS noise. What is the noise power spectral density of the accelerometer (in
units of µg/ Hz) based on the measurement mentioned above? Assume that the bandwidth
in the accelerometer is set by 1 pole filter.

Answer:
π
Equivalent noise bandwidth for 1 pole filter system, EN BW = 2 f3dB = 157Hz.
10,000 õg
∴ Noise spectral density = noise
√ RM S
EN BW
= √
157 Hz
= 798.1 õg
Hz
.

(c) (2 marks) In our application, the accelerometer experiences random vibration signals that have
a mean of 0g and root mean square (rms) vibration amplitude of Xg. If we use the same precise
instrument and record 1000 samples for the output in presence of vibration, what are the mean
(µ0) and standard deviation (σ 0) that we will measure? Assume that the vibration spectrum
lies inside the sensor’s bandwidth. Assume that the vibration amplitude is not large enough to
cause clipping of the accelerometer output.

Answer:
p
µ0 = 1.02g, σ 0 = Xg. For standard deviation, both Xg and X 2 + (0.01)2 g are acceptable
answers. Since the question asks what will we measure, the details involve variation of
the instrument (repeatability) and operator variation (reproducibility). However, if X is
assumed to be much larger compared to these variations and well above the noise floor of
the sensor, the standard deviation, σ 0 ≈ Xg

(d) (2 marks) Now assume that the vibration amplitude is large enough to cause clipping of the
accelerometer output at positive rail (+2g). What are the mean (µclip 0) and standard deviation
(σclip 0) that we will measure?

Answer:
The new distribution will be 1-sided truncated normal distribution. Fore more details refer
“https://en.wikipedia.org/wiki/Truncated normal distribution”.
s
 2 
0 0 0 φ(β) 0 0 φ(β) φ(β)
∴ µclip = µ − σ Φ(β) , σclip = σ 1 − β Φ(β) − Φ(β)
0
where β = 2g−µ , φ(x) is the probability density function of standard normal distribution
σ0
and Φ(x) is its cumulative distribution function.

(e) (2 marks) Now assume that the vibration amplitude is large enough to cause clipping of the
accelerometer output at both positive rail (+2g) and negative rail (−2g). What are the mean
(µrail 0) and standard deviation (σrail 0) that we will measure?
EE617 Quiz 0

Answer:
The new distribution will be 2-sided truncated normal distribution. Fore more details refer
“https://en.wikipedia.org/wiki/Truncated normal distribution”.
s
 2 
φ(α)−φ(β)
∴ µrail 0 = µ0 + σ 0 Φ(β)−Φ(α) , σrail 0 = σ 0 1 + αφ(α)−βφ(β)
Φ(β)−Φ(α) − φ(α)−φ(β)
Φ(β)−Φ(α)

0 0
where α = −2g−µ , β = 2g−µ , φ(x) is the probability density function of standard normal
σ0 σ0
distribution and Φ(x) is its cumulative distribution function.

Hint: Variables in a Gaussian distribution can have very large values, but with low probabilities
associated to such measurements. Can the accelerometer’s output exceed the rails ±2g? Beyond
this value, the probability density function for the distribution
R +∞ will be 0. The expected value
(mean) of a random variable X is given as E[X] = −∞ xf (x)dx, where f (x) represents the
probability density function. Thus you expect a change in the mean and standard deviation
due to railing and clipping. Now it’s up to you to find an elegant expression for this change!
2. Consider the circuit shown below implemented with an ideal op-amp. The circuit designer in my
project team implemented this block for signal conditioning in our sensor application. She has asked
me to ensure that it is compatible with my sensor specifications.

Vout
(a) (2 marks) Assuming an ideal op-amp, calculate the transfer function Vin of the circuit.

Answer:
This circuit is a Sallen-Key low-pass filter topology. For more details, refer to
“https://en.wikipedia.org/wiki/Sallen-Key topology”.
Vout Z3 Z4
Vin = Z1 Z2 +Z3 (Z1 +Z2 )+Z3 Z4

(b) (2 marks) The circuit designer chose resistors for Z1 and Z2 (Z1 = R1 , Z2 = R2 ) and capacitors
for Z3 and Z4 (Z3 = sC1 1 , Z4 = sC1 2 ). Simplify the transfer function obtained in part (a) and
identify how many poles and zeros this circuit has and where they are located.

Answer:
This circuit is a Sallen-Key low-pass filter topology.

Vout ω02 1 1 R1 +R2

Vin = s2 +2αs+ω02
, where ω0 = 2πf0 = √
R1 R2 C1 C2
and 2α = C1 R1 R2 .
Based onp
the transfer function above, there are 2 poles and no zeros. The poles are located
at −α ± α2 − ω02 .

(c) (2 marks) In part (b), you must have obtained some sort of filter response for this circuit.
Identify the type of filter, and its order.

Answer:
This circuit is a second order low pass filter.

Page 2
EE617 Quiz 0

(d) (2 marks) In a practical implementation in an integrated circuit (IC), we typically care about
variations in the actual values of resistors and capacitors from chip to chip, as well as between
different elements on the same chip. To address this issue, it is a common practice to place all
elements of the same type close to each other physically in the IC and adopt design practices
where key performance specifications appear as ratios of such quantities, so that sensitivity to
variations in absolute values are reduced. Assume that R1 = mR2 and C1 = nC2 , obtain an
expression for the quality factor of the circuit in terms of m and n.

Answer:
√ √
ω0 R1 R2 C1 C2 mn
Q= 2α = C2 (R1 +R2 ) = m+1 .

(e) (2 marks) My marketing manager mentioned that the customer requires that the system should
not have any form of ringing or peaking to any input profile. This requires us to ensure critical
damping in the circuit (i.e. damping ratio ζ = 1). Assuming m = 1, what value of n does this
require us to design for?

Answer:
√
1 n
For critical damping, ζ = 1, ∴ Q = 2ζ = 0.5 = 2 (substituting m = 1 in above expression
obtained in answer for part d. ∴ n = 1.

Bonus points: If you answer the following sub-section correctly, you will get bonus points that
can make up for any lost marks in this homework. In system design, the system either works,
or it does not work, so no partial marks will be awarded!
(f) (4 marks) Bonus: The input impedance of this circuit is crucial to ensure proper interface to
the sensor output. Obtain an expression for the input impedance of this circuit.

Answer:
The expression for input impedance of second order unity gain Sallen-Key low pass filter is
also presented on the same wikipedia page. Follow this reference for a detailed derivation:
Cartwright, K. V., E. J. Kaminsky “Finding the minimum input impedance of a second-
order unity-gain Sallen-Key low-pass filter without calculus”. Lat. Am. J. Phys. Educ. 7
(4): 525535 (2013).
0
s0 + sQ +1
2
s R1 m
Z(s) = R1 02 0 , where s0 = ω0 and k = R1 +R2 = m+1 .
s + sQk

Page 3