Testing Design Problems

Exam 2021-01-25
Problem 1
A data acquisition system has to be developed in order to monitor the solar irradiance Ee along a Photo Voltaic (PV) plant.
The plant includes 4 strings that cover a 20 m × 20 m square area; each string is made up of 10 series-connected PV
modules. The solar irradiance on each PV module has to be measured and is expected to be in the range
2
Ee ∈ [50, 1300] W/m .

Silicon reference cells are available and are characterized by a voltage output
2
Vcell = 60 mV, ±0.1% at Ee = 1 kW/m .

The main requirements and constraints are:

2
• Irradiance resolution: 1 W/m ;
2
• Irradiance uncertainty: 5 W/m ;
• 100 multiple readings acquired at each measuring node and then sent to a central PC, which is located at a corner of
the plant;
• Data from all measurement nodes updated with a rate of 2 s.

Design the data acquisition system and provide details about:

1. Conditioning circuitry;
2. Analogue-to-Digital Converter (ADC) specifications (range, number of bits, sampling frequency);
3. Data-communication bus.

Solution
2) Professor’s solution
2.1 Sensor sensitivity S
Define the sensitivity:
Vcell
S=
Ee
2
At Vcell = 60 mV and Ee = 1000 W/m :

60 × 10−3 V V
S= 2 = 60 × 10−6 2
1000 W/m W/m

Hence:
2
S = 60 µV/(W/m )
Sensor relative uncertainty:
εVcell = 0.001 ⇒ εS = 0.001

1
2.2 Irradiance range → sensor voltage range Vs
The sensor output is:
Vs = S · Ee
At the limits:
Vs,min = S · 50 = 60 µV · 50 = 3 mV
Vs,max = S · 1300 = 60 µV · 1300 = 78 mV
Thus:
Vs ∈ [3, 78] mV

2
2.3 Resolution requirement 1 W/m ⇒ ∆Vs
2
For ∆Ee = 1 W/m :
dVs
∆Vs = ∆Ee = S∆Ee = 60 µV
dEe
So:
∆Vs = 60 µV

2.4 Preliminary check for ADC bits (#levels)

Sensor span:
Vs,max − Vs,min = 78 − 3 = 75 mV
Number of resolution steps across the span:

75 mV 75 × 10−3
#levels ≈ = = 1250
60 µV 60 × 10−6
Thus:
2Nb ≥ 1250 ⇒ Nb ≈ 11
Therefore:
Nb = 11

3) Conditioning circuitry + ADC reference (explaining the sketch)

3.1 Text explanation of the sketch (no drawing)
The block chain is:
1. Input irradiance Ee enters the sensor block of sensitivity S and produces Vs .
2. Vs is amplified by a non-inverting amplifier with resistors R1 and R2 , giving gain
R2
G=1+
R1
and output VADC = G Vs .
3. VADC is sampled by an ADC defined by reference/full-scale voltage VF R , resolution Nb , and sampling frequency fs .
4. The ADC output code Dout is transmitted over the bus to the central PC.

3.2 Meaning of the VF R / Arduino UNO note

The professor lists possible VF R values and notes (Arduino UNO):

VCC = 5 V, Vref,int ≈ 1.1 V

This indicates VF R is a design choice; a smaller VF R improves voltage resolution provided the amplified signal does not
saturate. In the solution, the working assumption is:

VF R ≈ 1 V

2
3.3 Gain constraint and chosen gain G
Non-inverting amplifier:  
R2
VADC = Vs 1+ = G Vs
R1
Impose no saturation:
1 1
VADC ≤ VF R ≈ 1 V ⇒ G ≤ = ≈ 12.8
Vs,max 0.078
Hence:
G ≤ 12.8
Choose:
G = 11
Select resistors:
R1 = 1 kΩ, R2 = 10 kΩ
since:
10k
G=1+ = 11
1k

3.4 ADC input range after gain

VADC = G Vs
VADC,min = 11 · 3 mV = 33 mV = 0.033 V
VADC,max = 11 · 78 mV = 858 mV = 0.858 V
Therefore:
VADC = (0.033 to 0.858) V

4) ADC bits confirmation using quantization

2
4.1 Voltage step at ADC input for 1 W/m
∆VADC = G∆Vs = 11 · 60 µV = 660 µV = 0.66 mV
∆VADC = 0.66 mV

4.2 Quantization step and bits condition

Quantization step:
VF R
Vq =
2Nb
Require:
VF R
Vq ≤ ∆VADC ⇒ 2Nb ≥
∆VADC
With VF R = 1 V and ∆VADC = 0.66 mV:
1
2Nb ≥ ≈ 1515 ⇒ Nb = 11
0.66 × 10−3
Thus:
Nb = 11

5) Sampling frequency fs
Minimum sampling rate per node:
#readings 100
fs,min = = = 50 Sa/s
UR 2
Professor’s selected value (with margin):
fs = 500 Sa/s

3
6) Calibration function (measurement model) and uncertainty
6.1 Digital output relationship and calibration equation
Approximate relation:
VADC
Dout ≈
Vq
Also:
VADC = G Vs = G S Ee
and:
R2 R1 + R 2 1 R1
G=1+ = ⇒ =
R1 R1 G R1 + R 2
Solving for Ee gives the calibration (measurement) model:

Dout Vq R1
Ee = ·
S R1 + R 2

6.2 Sensitivity uncertainty δS

δS = S · εS
With εS = 0.001:
V
δS = 60 × 10−6 · 0.001 = 60 × 10−9 2
W/m
2
δS = 60 nV/(W/m )

6.3 Worst-case uncertainty propagation

∂Ee ∂Ee ∂Ee ∂Ee 2
δEe = δDout + δS + δR1 + δR2 ≤ 5 W/m
∂Dout ∂S ∂R1 ∂R2

6.4 Partial derivatives (professor’s numeric coefficients)

From:
Dout Vq R1
Ee = ·
S R1 + R 2

(a) With respect to Dout

2
∂Ee Vq R1 Vq W/m
= · = ≈ 0.74
∂Dout S R1 + R 2 GS LSB

∂Ee 2
≈ 0.74 W/m /LSB
∂Dout

(b) With respect to S

∂Ee Dout Vq R1 Dout Vq
= 2
· =
∂S S R1 + R 2 S2G
In the notes the resulting contribution is:
∂Ee 2
δS ≈ 1.3 W/m
∂S

(c) With respect to R1 and R2 Professor’s coefficients:

2
∂Ee W/m
≈ 1.18
∂R1 Ω

2
∂Ee W/m
≈ 0.118
∂R2 Ω

4
6.5 Final inequality and budget allocation
Assembled inequality:
0.74 δDout + 1.3 + 1.18 δR1 + 0.118 δR2 ≤ 5
Remaining budget after the sensor term:
2
5 − 1.3 = 3.7 W/m
Allocate:
0.74 δDout ≤ 1.85, 1.18 δR1 ≤ 0.925, 0.118 δR2 ≤ 0.925
Thus:
1.85
δDout ≤ ≈ 2.5 LSB
0.74
0.925
δR1 ≤ ≈ 0.78 Ω
1.18
0.925
δR2 ≤ ≈ 7.8 Ω
0.118
Relative tolerances:
0.78 7.8
εR1 ≈ · 100% ≈ 0.078%, εR2 ≈ · 100% ≈ 0.078%
1000 10000
Conclusion:
Use approximately 0.1% resistors (or better)

7) Communication bus sizing and choice

Required throughput:
Nnodes · Nreadings · Nb
TR =
UR
Substitute:
40 · 100 · 11
TR = = 22000 bit/s ≈ 22 kbit/s
2
With overhead/margin, the professor indicates:

T R ≈ 31.25 kbit/s

Suggested solution:
Fieldbus (Profibus PA)
Topology note:
• 2 sections (1 repeater)
• ∼ 20 nodes per section
• section length on the order of ∼ 50 m

8) Final design statement

2
• S = 60 µV/(W/m ), εS = 0.001
2
• Vs ∈ [3, 78] mV, ∆Vs = 60 µV for 1 W/m
• Choose VF R ≈ 1 V, non-inverting amplifier with G ≤ 12.8, select G = 11
• R1 = 1 kΩ, R2 = 10 kΩ
• VADC ∈ [0.033, 0.858] V
• ADC: Nb = 11 bits, unipolar
• Sampling: fs,min = 50 Sa/s, choose fs = 500 Sa/s
• Calibration model:
Dout Vq R1
Ee = ·
S R1 + R2

5
 • Uncertainty:
0.74 δDout + 1.3 + 1.18 δR1 + 0.118 δR2 ≤ 5
leading to δDout ≤ 2.5 LSB and εR1 ≈ εR2 ≈ 0.078%.
• Bus throughput: T R ≈ 22 kbit/s raw; with overhead select ∼ 31.25 kbit/s fieldbus (Profibus PA), optionally split into
2 sections with repeater.

Problem 2
A data acquisition system has to be developed in order to monitor the temperature θ of the modules of a Photo Voltaic (PV)
plant. The plant includes 4 strings that cover a 20 m × 20 m square area and each string is made up of 10 series-connected PV
modules. The temperature of each PV module, which is expected in the range (0–80) ◦ C, is measured by means of sensors
characterized by the following input/output relationship:

Iout = S θ + Ioff ,

where S = 100 µA/◦ C with ±0.1% relative uncertainty and Ioff = 10 mA ± 5 µA. The compliance voltage of the temperature
sensors is 1 V.
The main requirements and constraints are:
• temperature resolution: 0.1 ◦ C;

• temperature uncertainty: 0.5 ◦ C;

• 100 multiple readings acquired at each measuring node and then sent to a central PC, located at a corner of the plant;
• data from all measurement nodes updated with a rate of 1 s.

Design the data acquisition system and provide details about: conditioning circuitry, Analogue-to-Digital Converter
specifications (range, number of bit, sampling frequency), and data-communication bus.

Solution
Number of nodes
There are 4 strings and 10 modules per string, hence:

Nnodes = 4 × 10 = 40.

Sensor output range

The sensor model is:
Iout = S θ + Ioff , S = 100 µA/◦ C, Ioff = 10 mA.
At θmin = 0◦ C:
Iout,min = 10 mA.
◦
At θmax = 80 C:
Iout,max = 100 µA/◦ C · 80◦ C + 10 mA = 8 mA + 10 mA = 18 mA.
Therefore:
Iout ∈ [10, 18] mA.

Resolution requirement mapped to current

With ∆θ = 0.1◦ C:
dIout
∆Iout = ∆θ = S ∆θ = 100 µA/◦ C · 0.1◦ C = 10 µA.
dθ

6
Conditioning circuitry: shunt resistor (I/V conversion)
Use a shunt resistor RS to convert current to voltage:

VADC = RS Iout .

The sensor compliance voltage is CV = 1 V, hence:

CV 1
RS,max = = ≈ 55.6 Ω.
Iout,max 18 × 10−3

Choose:
RS = 50 Ω.
Then:
VADC,min = 50 Ω · 10 mA = 0.5 V, VADC,max = 50 Ω · 18 mA = 0.9 V,
so:
VADC ∈ [0.5, 0.9] V.

ADC range and number of bits

Adopt a unipolar ADC with full-scale reference:
VF R = 1 V.
The voltage step corresponding to ∆θ = 0.1◦ C is:

∆VADC = RS ∆Iout = 50 Ω · 10 µA = 500 µV = 0.5 mV.

The ADC quantization step is:

VF R
Vq = .
2Nb
Impose Vq ≤ ∆VADC :
1 1
≤ 0.5 × 10−3 ⇒ 2Nb ≥ = 2000 ⇒ Nb = 11.
2Nb 0.5 × 10−3

Sampling frequency
Each node must acquire 100 readings every U R = 1 s:
100
fs,min = = 100 Sa/s.
1
Choose:
100
fs = 1 kSa/s, ∆t100 = = 0.1 s.
1000

ADC code relation and maximum code

Approximate the ADC output code by:
VADC VF R
Dout ≈ , Vq = .
Vq 2Nb
Since VADC = RS (Sθ + Ioff ):
RS (Sθ + Ioff )
Dout = .
Vq
With VF R = 1 V and Nb = 11:
1 1
Vq = = V.
211 2048
At VADC,max = 0.9 V:
0.9
Dout,max ≈ = 0.9 · 2048 ≈ 1843 LSB.
1/2048

7
Measurement model (calibration equation)
From Dout Vq = RS (Sθ + Ioff ):
Dout Vq
Sθ = − Ioff ,
RS
hence:
Dout Vq Ioff
θ= − .
RS S S

Worst-case uncertainty propagation

Let xi ∈ {Dout , RS , S, Ioff }. A worst-case estimate is:
X ∂θ
δθ = δxi .
i
∂xi

Dout Vq Ioff
Partial derivatives Using θ = − :
RS S S
∂θ Vq VF R
= = N ≈ 0.098 ◦ C/LSB,
∂Dout RS S 2 b RS S
∂θ Dout Vq
= ≈ 3.6 ◦ C/Ω (evaluated at Dout,max ),
∂RS RS2 S
◦
∂θ 1 Dout Vq C
= 2 Ioff − ≈ 8 × 105 ,
∂S S RS A/◦ C
∂θ 1
= = 104 ◦ C/A.
∂Ioff S

Uncertainty data
εS = 0.1% = 10−3 ⇒ δS = SεS = (10−4 ) · 10−3 = 10−7 A/◦ C,
δIoff = 5 µA.
Impose:
δθ ≤ 0.5◦ C.

Communication bus throughput

Each update (every U R = 1 s) carries 100 readings per node, Nb = 11 bits per reading, and Nnodes = 40 nodes:
Nnodes · Nreadings · Nb 40 · 100 · 11
T Rraw = = = 44,000 bit/s = 44 kbit/s.
UR 1
A bus rate must be selected with margin to account for protocol overhead.

Problem 3
The system shown measures the illuminance Ev by means of a photodiode, an operational amplifier with feedback resistor
R, and a voltmeter. The photodiode has sensitivity

S = (50 ± 0.5) nA/lx (deterministic model).

The resistance R is obtained as the average of n = 10 repeated readings provided by a digital ohmmeter with range 100 kΩ
and expanded uncertainty

U (R) = 0.20% reading + 0.10% range Ω, k = 3,
obtaining
R̄ = 19.76 kΩ, uA (R) = 0.15 kΩ
(experimental standard deviation). The voltmeter is characterized by an expanded uncertainty U (V ) = 5 mV with confidence
level p = 95%. It measures the offset voltage of the whole measuring chain as

Voff = −10 mV.

8
When the same voltmeter provides the measurement

Vout = 120 mV,

evaluate the value and the standard uncertainty of the illuminance Ev .

Solution
1) Measurement model
The transimpedance stage converts the photodiode short-circuit current into a voltage:

Vout = Isc R + Voff .

With photodiode current proportional to illuminance, Isc = S Ev , the model becomes

Vout = S Ev R + Voff ,

hence
Vout − Voff
Ev = .
SR

2) Numerical estimate of Ev
Using
Vout,0 = 0.12 V, Voff,0 = −0.01 V, S0 = 50 × 10−9 A/lx, R0 = 19.76 × 103 Ω,
0.12 − (−0.01) 0.13
Ev,0 = −9
= ≈ 1.3158 × 102 lx.
3
(50 × 10 )(19.76 × 10 ) 9.88 × 10−4
Therefore,
Ev,0 ≈ 131.6 lx.

3) Standard uncertainties of inputs

Sensitivity S (deterministic model). Treat ±0.5 nA/lx as a rectangular distribution with half-width ∆S = 0.5 nA/lx:

∆S 0.5
u(S) = √ = √ nA/lx ≈ 0.289 nA/lx = 0.289 × 10−9 A/lx.
3 3

Resistance R̄ (Type A + Type B). Type B (ohmmeter): first compute the expanded uncertainty at the indicated value
and range:
U (R) = 0.002 (19760) + 0.001 (100000) ≈ 39.5 + 100 = 139.5 Ω.
With k = 3:
U (R) 139.5
uB (R̄) =≈ = 46.5 Ω.
3 3
Type A (mean of n = 10 readings), with experimental standard deviation uA (R) = 0.15 kΩ = 150 Ω:
150
uA (R̄) = √ Ω ≈ 47.4 Ω.
10
Combine: q p
u(R̄) = u2B (R̄) + u2A (R̄) ≈ (46.5)2 + (47.4)2 Ω ≈ 66 Ω.

Voltmeter readings. Given expanded uncertainty U (V ) = 5 mV at p = 95%, take k0.95 ≈ 2:

U (V ) 5 mV
u(Vout ) = u(Voff ) = = = 2.5 mV.
2 2

9
4) Uncertainty propagation with correlation
Let
Vout − Voff
Ev = f (S, R, Vout , Voff ) = .
SR
The sensitivity coefficients are
∂Ev Voff − Vout ∂Ev Voff − Vout
= , = ,
∂S S2R ∂R SR2
∂Ev 1 ∂Ev 1
= , =− .
∂Vout SR ∂Voff SR
The combined variance is
X  ∂f 2 ∂f ∂f
u2 (Ev ) = u2 (xi ) + 2 ρ(Vout , Voff ) u(Vout ) u(Voff ),
i
∂xi ∂Vout ∂Voff

with xi ∈ {S, R, Vout , Voff } and all other correlations taken as zero.
Since the same voltmeter is used for both Vout and Voff , take

ρ(Vout , Voff ) = 1.
∂f 1 ∂f 1
With ∂Vout = + SR and ∂Voff = − SR and u(Vout ) = u(Voff ), the voltage terms cancel:
 2  2  2
1 1 1
u2 (Vout ) + u2 (Voff ) − 2 u(Vout )u(Voff ) = 0.
SR SR SR
Therefore,
 2  2
2 ∂Ev 2 ∂Ev
u (Ev ) = u (S) + u2 (R̄).
∂S ∂R

5) Numerical evaluation of u(Ev )

Evaluate the coefficients at nominal values:
∂Ev −0.13 lx
= ≈ −2.63 × 109 ,
∂S (50 × 10−9 )2 (19.76 × 103 ) A/lx
∂Ev −0.13 lx
= −9 3 2
≈ −6.67 × 10−3 .
∂R (50 × 10 )(19.76 × 10 ) Ω
Compute contributions:
∂Ev
uS (Ev ) = u(S) ≈ (2.63 × 109 )(0.289 × 10−9 ) lx ≈ 0.760 lx,
∂S
∂Ev
uR (Ev ) = u(R̄) ≈ (6.67 × 10−3 )(66) lx ≈ 0.439 lx.
∂R
Thus, q p
u(Ev ) = u2S (Ev ) + u2R (Ev ) ≈ (0.760)2 + (0.439)2 lx ≈ 0.878 lx.

6) Final result
Ev ≈ 131.6 lx, u(Ev ) ≈ 0.88 lx (≈ 0.9 lx).

Problem 4
The system shown measures the illuminance Ev by means of a photodiode, a trans-resistance amplifier and a voltmeter.
The photodiode has been selected from a batch of 1000 devices calibrated against a reference instrument with a relative
standard uncertainty of 0.3%. The batch is characterized by an average sensitivity

S = 30 nA/lx

and an experimental standard deviation

uA (S) = 0.12 nA/lx.

10
 The conditioning circuitry is characterized with the switch in position 2 using a known reference current

Iref = 100 µA, U (Iref ) = 0.5 µA (p = 95%),

and measuring the voltage output Vref by means of a voltmeter with range 20 V and expanded uncertainty


U (V ) = 1.0% reading + 0.5% range V, k = 3,

obtaining
Vref = 9.95 V.
Evaluate the value and the standard uncertainty of the illuminance Ev when the same voltmeter provides the measurement

Vout = 12.0 V.

Solution
1) Measurement model
With the switch in position 2, the reference current is applied to the trans-resistance:
Vref
Vref = R Iref ⇒ R= .
Iref
With the switch in position 1, the photodiode current is converted into voltage:

Vout = R Isc , Isc = S Ev ,

hence
Vout = R S Ev .
Substituting R = Vref /Iref gives
Vref Vout Iref
Vout = S Ev ⇒ Ev = .
Iref Vref S

2) Numerical estimate of Ev
Using Vout = 12.0 V, Vref = 9.95 V, Iref = 100 µA = 10−4 A, S = 30 nA/lx = 30 × 10−9 A/lx:

12.0 10−4
Ev,0 = ≈ 4.020 × 103 lx.
9.95 30 × 10−9
Therefore,
Ev,0 ≈ 4020 lx.

3) Standard uncertainties of inputs

Reference current. Given U (Iref ) = 0.5 µA at p = 95%, take k0.95 ≈ 2 (normal distribution):

U (Iref ) 0.5 µA
u(Iref ) = = = 0.25 µA.
k0.95 2

Voltmeter readings. For the voltmeter, k = 3 and

U (V ) = 0.01 Vreading + 0.005 Vrange , Vrange = 20 V.

Thus
0.22
U (Vout ) = 0.01(12.0) + 0.005(20) = 0.22 V ⇒ u(Vout ) = = 0.0733 V,
3
0.1995
U (Vref ) = 0.01(9.95) + 0.005(20) = 0.1995 V ⇒ u(Vref ) = = 0.0665 V.
3

11
Sensitivity. Type B (calibration against reference instrument, relative standard uncertainty 0.3%):
uB (S) = 0.003 S = 0.003 (30 × 10−9 ) = 9 × 10−11 A/lx.
Type A (batch experimental standard deviation):
uA (S) = 0.12 nA/lx = 0.12 × 10−9 A/lx.
Combine: q p
u(S) = u2A (S) + u2B (S) = (0.12 × 10−9 )2 + (9 × 10−11 )2 = 1.5 × 10−10 A/lx.

4) Uncertainty propagation (with correlation)

Let
Vout Iref
Ev = f (Iref , S, Vout , Vref ) = .
Vref S
Sensitivity coefficients:
∂Ev Vout ∂Ev Vout Iref
= , =− ,
∂Iref Vref S ∂S Vref S 2
∂Ev Iref ∂Ev Vout Iref
= , =− 2 .
∂Vout Vref S ∂Vref Vref S
Assume all input quantities are uncorrelated except the two voltmeter readings, measured with the same voltmeter:
ρ(Vout , Vref ) = 1, ρ(xi , xj ) = 0 for all other pairs.
Then
 2  2  2  2
∂Ev ∂Ev ∂Ev ∂Ev ∂Ev ∂Ev
u2 (Ev ) = u2 (Iref )+ u2 (S)+ u2 (Vout )+ u2 (Vref )+2 ρ(Vout , Vref ) u(Vout ) u(Vref ).
∂Iref ∂S ∂Vout ∂Vref ∂Vout ∂Vref

5) Numerical evaluation of u(Ev )

At the nominal values:
∂Ev 12.0 lx
= = 4.0201 × 107 ,
∂Iref 9.95 · 30 × 10−9 A
−4
∂Ev 12.0 · 10 lx
=− = −1.3400 × 1011 ,
∂S 9.95 · (30 × 10−9 )2 A/lx
∂Ev 10−4 lx
= = 3.3501 × 102 ,
∂Vout 9.95 · 30 × 10−9 V
−4
∂Ev 12.0 · 10 lx
=− = −4.0403 × 102 .
∂Vref (9.95)2 · 30 × 10−9 V
Contributions: 2
∂Ev
2
u(Iref ) = 4.0201 × 107 · 0.25 × 10−6 ≈ 101.0 lx2 ,
∂Iref
 2
∂Ev
2
u(S) = 1.3400 × 1011 · 1.5 × 10−10 ≈ 404.0 lx2 ,
∂S
 2
∂Ev
2
u(Vout ) = 335.01 · 0.0733 ≈ 603.6 lx2 ,
∂Vout
 2
∂Ev
2
u(Vref ) = 404.03 · 0.0665 ≈ 721.9 lx2 ,
∂Vref
and the covariance term (with ρ = 1):
∂Ev ∂Ev
2 u(Vout )u(Vref ) = 2(335.01)(−404.03)(0.0733)(0.0665) ≈ −1320.1 lx2 .
∂Vout ∂Vref
Thus
u2 (Ev ) ≈ 101.0 + 404.0 + 603.6 + 721.9 − 1320.1 = 510.3 lx2 ,
√
u(Ev ) = 510.3 ≈ 22.6 lx ≈ 23 lx.

6) Final result
Ev ≈ 4020 lx, u(Ev ) ≈ 23 lx.

12