Loading study handbook…

Instrumentation and Controls

4% of exam

Sensors and transducers, data acquisition, sampling and A/D resolution, signal conditioning, and logic diagrams.

3 concepts

A. Sensors

Sensors and Transducers

How temperature, pressure, motion, and chemical sensors turn a physical quantity into a usable signal, and how sensitivity, range, linearity, and calibration govern what you can trust.

Unlocks with an access pass — one-time payment, no auto-renew. View passes

Pass holders

B. Data acquisition

Data Acquisition, Sampling, and A/D Conversion

The signal chain from sensor to digital code: how the Nyquist rate prevents aliasing, how n-bit quantization sets resolution, and how conditioning, filtering, and gain make a small signal measurable.

Data acquisition is where a continuous physical signal becomes numbers a computer can use, and it is one of the most reliably tested corners of the FE Other Disciplines exam because the math is clean and the traps are sharp. The handbook's Instrumentation, Measurement, and Control chapter gives you the four equations that matter — sampling rate, the Nyquist criterion, A/D resolution, and the digital-reading-to-voltage relation — but the points come from using them together: choosing a sample rate that avoids aliasing, picking enough bits to resolve the signal, and conditioning a tiny sensor output so it fills the converter's range. This concept walks the whole signal chain in order, so the formulas snap onto a picture instead of floating free.

The acquisition signal chain

A data-acquisition system is a pipeline: the transducer produces a small analog signal, signal conditioning amplifies and filters it, an anti-aliasing filter removes frequencies above half the sample rate, a sample-and-hold freezes each sample, and the A/D converter assigns a digital code. Understanding the order matters — the anti-alias filter must come before the sampler, and amplification should happen before digitizing so the signal spans the converter's input range. Each stage adds or limits error, and exam questions usually probe one stage while assuming the rest are ideal.

Sampling rate and the Nyquist criterion

Sampling converts a continuous signal into a sequence taken every

\Delta t

seconds, so the sample rate is

f_s = 1/\Delta t

. Shannon's sampling theorem sets the rule: to reconstruct a signal faithfully, the sample rate must exceed twice the highest frequency present,

f_s > 2 f_N

. Watch the terminology, because conventions clash: the FE Reference Handbook names that highest signal frequency

f_N

the 'Nyquist frequency,' so match the handbook on exam day — but most DSP texts instead call

f_s/2

the Nyquist frequency and call

2 f_N

the 'Nyquist rate.' The physics is identical either way:

f_s/2

is the highest frequency the system can represent (the folding limit), and you must sample strictly faster than

2 f_N

. Real systems sample several times faster than the strict minimum to give the anti-alias filter room to roll off.

f_s = \dfrac{1}{\Delta t},\qquad f_s > 2 f_N

Aliasing

When the sample rate is too low, frequencies above

f_s/2

do not vanish — they fold back and masquerade as lower frequencies, an irreversible corruption called aliasing. A component at frequency

f

appears at an alias

f_a = |\,f - k f_s\,|

, where

k

is the integer that brings the result into the band

[0, f_s/2]

. The cure is an anti-aliasing low-pass filter applied before sampling, removing any energy above the Nyquist limit. Once a signal is aliased, no amount of post-processing can recover the true frequency, which is why this is the most consequential mistake in data acquisition.

f_a = \left|\,f - k\,f_s\,\right|,\qquad 0 \le f_a \le \tfrac{f_s}{2}

A/D resolution and the LSB

An

n

-bit converter divides its input span

[V_L, V_H]

into

2^n

equal steps, so the voltage resolution — the value of one least-significant bit (LSB) — is the span divided by

2^n

. This is the smallest voltage change the converter can distinguish; the quantization error of any single reading is at most half an LSB. More bits means finer resolution but more data and cost. Note the asymmetry the handbook flags: the digital codes run from

0

to

2^n - 1

, so the highest representable voltage is one resolution step below

V_H

, namely

V_H - Q

.

Q = \dfrac{V_H - V_L}{2^n}\qquad(\text{one LSB})

From digital code back to voltage

The converter outputs an integer

N

in the range

0

to

2^n - 1

. To turn that code back into a voltage you scale by the resolution and add the low end of the range. This is the equation you use to interpret a reading, and combined with the sensor's calibration it carries a raw count all the way to engineering units. The quantization uncertainty

\pm Q/2

rides along with every such conversion.

V = N\,Q + V_L,\qquad Q = \dfrac{V_H - V_L}{2^n}

Quantization error and dynamic range

Rounding a continuous voltage to the nearest code introduces quantization noise bounded by

\pm Q/2

. For a full-scale sinusoid the resulting signal-to-noise ratio is approximately

\text{SNR} \approx 6.02\,n + 1.76

decibels — the famous 'about 6 dB per bit' rule, a fast way to judge how many bits a measurement needs. Each added bit halves the step size, doubles the number of levels, and buys roughly 6 dB of dynamic range. When a problem asks how many bits are required to resolve a given voltage, set

2^n \ge (V_H - V_L)/\Delta V

and round up.

\text{SNR}_{\text{dB}} \approx 6.02\,n + 1.76

Signal conditioning: gain, offset, filtering

Conditioning matches the sensor to the converter. Amplification with gain

G

scales a small signal so it spans the A/D input range — a sensor delivering tens of millivolts into a

0

–

5\,\text{V}

converter wastes nearly all the resolution unless amplified. Choose the gain so the largest expected signal just reaches full scale:

G = (V_H - V_L)/V_{\text{sensor,max}}

. Offset (level shifting) centers a bipolar or biased signal, and filtering removes noise and band-limits the signal ahead of sampling. Crucially, amplifying before digitizing improves the effective resolution referred to the input: the input-referred LSB becomes

Q/G

.

V_{\text{out}} = G\,V_{\text{in}},\qquad Q_{\text{input-referred}} = \dfrac{Q}{G}

Exam strategy

Sort every question into one of four molds: (1) sampling — compare

f_s

with

2f_N

, or compute an alias with

f_a = |f - k f_s|

; (2) resolution —

Q = (V_H - V_L)/2^n

, watching that the span is the full range and the count is

2^n

, not

2^n - 1

; (3) code-to-voltage —

V = NQ + V_L

; (4) bits-needed — solve

2^n \ge \text{span}/\Delta V

and round up. Always confirm whether a stated frequency is the Nyquist frequency (highest signal component) or the sample rate, and remember the smallest code is

0

V while the largest is

V_H - Q

. For conditioning, set the gain so the maximum signal fills — but does not exceed — the converter range.

Key equations

Sampling rate

f_s = \dfrac{1}{\Delta t}

Sample frequency from the sampling interval $\Delta t$ (s). Units: Hz (samples/s).

Nyquist (Shannon) criterion

f_s > 2 f_N

Sample faster than twice the highest signal frequency to reconstruct it. The FE handbook labels that highest frequency $f_N$

Alias frequency

f_a = \left|\,f - k\,f_s\,\right|

Apparent frequency of an undersampled component; $k$

A/D voltage resolution (LSB)

Q = \dfrac{V_H - V_L}{2^n}

Number of quantization levels

L = 2^n,\qquad N \in [\,0,\ 2^n - 1\,]

An $n$

Code to voltage

V = N\,Q + V_L

Reconstruct the analog voltage from integer code $N$ using resolution $Q$ and range offset $V_L$

Maximum representable voltage

V_{\max} = V_H - Q

Top code is $2^n - 1$

Quantization error

e_q = \pm\dfrac{Q}{2}

Worst-case rounding error of a single conversion, half a least-significant bit.

Quantization SNR

\text{SNR}_{\text{dB}} \approx 6.02\,n + 1.76

Ideal SNR for a full-scale sinusoid; roughly 6 dB per bit of dynamic range.

Bits required for a resolution

2^n \ge \dfrac{V_H - V_L}{\Delta V}

Amplifier gain

V_{\text{out}} = G\,V_{\text{in}},\qquad G = \dfrac{V_H - V_L}{V_{\text{sensor,max}}}

Input-referred resolution

Q_{\text{in}} = \dfrac{Q}{G}

Resolution referred back to the sensor input after a gain stage $G$ ; amplifying before the A/D improves effective resolution.

Worked examples

Resolution and code-to-voltage of a 12-bit converter

Problem. A 12-bit A/D converter has an input range of $0$ to $10\,\text{V}$ . Find the resolution (one LSB), the voltage for a code of $N = 1500$ , and the highest voltage the converter can represent.

Solution. Resolution:

Q = \dfrac{V_H - V_L}{2^n} = \dfrac{10 - 0}{2^{12}} = \dfrac{10}{4096} = 2.44\times10^{-3}\,\text{V} = 2.44\,\text{mV}

. Voltage at

N = 1500

:

V = N\,Q + V_L = 1500(2.4414\times10^{-3}) + 0 = 3.66\,\text{V}

. Maximum: the top code is

4095

, so

V_{\max} = V_H - Q = 10 - 0.00244 = 9.998\,\text{V}

. Sanity check:

1500/4096 \approx 0.366

of full scale, and

0.366\times10\,\text{V} \approx 3.66\,\text{V}

— consistent. Answers:

Q = 2.44\,\text{mV}

,

V = 3.66\,\text{V}

,

V_{\max} = 9.998\,\text{V}

.

Q = \dfrac{10}{4096} = 2.44\,\text{mV},\quad V = 1500\,Q = 3.66\,\text{V}

Aliasing of an undersampled tone

Problem. A $90\,\text{Hz}$ sinusoid is sampled at $f_s = 100\,\text{Hz}$ . Is the Nyquist criterion satisfied, and at what frequency does the tone appear in the sampled data?

Solution. Nyquist limit is

f_s/2 = 50\,\text{Hz}

Choosing bit depth and conditioning gain

Problem. A sensor outputs $0$ to $50\,\text{mV}$ . It feeds an A/D with a $0$ to $5\,\text{V}$ input range. (a) What amplifier gain fills the range? (b) How many bits are needed to resolve $1\,\text{mV}$ at the converter input (referred to the sensor)?

Common pitfalls

•Dividing the span by $2^n - 1$ instead of $2^n$ for resolution. The handbook defines $Q = (V_H - V_L)/2^n$ ; there are $2^n$ steps, even though the largest code is $2^n - 1$ .
•Assuming the top code maps to $V_H$ . The maximum reading is $V_H - Q$ , one LSB short of full scale — a frequent off-by-one error.
•Confusing the Nyquist frequency with the Nyquist rate. Conventions conflict: the FE handbook names the highest signal component $f_N$ the 'Nyquist frequency,' while most DSP texts call $f_s/2$ the Nyquist frequency and $2f_N$
•Sampling exactly at $2f_N$ and assuming it works. The theorem requires $f_s$ strictly greater than $2f_N$
•Believing aliasing can be filtered out after sampling. The anti-alias filter must precede the sampler; once folded, the alias is indistinguishable from a real low-frequency signal.
•Forgetting to amplify before digitizing. A small sensor signal that uses only a sliver of the converter range throws away most of the available bits — set the gain to fill the span.
•Mixing units in the bit-count inequality. Keep span and target step in the same units before solving $2^n \ge \text{span}/\Delta V$ , then round $n$ up.

References

NCEES FE Reference Handbook — Instrumentation, Measurement, and Control: Sampling and Analog-to-Digital Conversion
NCEES FE Reference Handbook — Instrumentation, Measurement, and Control: Signal Conditioning
Oppenheim & Willsky, Signals and Systems — Sampling theorem, aliasing, and reconstruction.

C. Logic diagrams

Logic Diagrams and Control Logic

Read and build digital logic and control diagrams: gates and truth tables, De Morgan, ladder logic and seal-in circuits, interlocks, block diagrams, and how a PLC scan executes them.

Unlocks with an access pass — one-time payment, no auto-renew. View passes

Pass holders

the Nyquist frequency;

f_s/2

is the folding frequency, which most DSP texts instead call the Nyquist frequency (with

2 f_N

the Nyquist rate).

is the integer placing

f_a

in

[0, f_s/2]

. Irreversible once it occurs.

Smallest distinguishable voltage step. $V_H, V_L$ = input range limits, $n$ = number of bits (typ. 4, 8, 10, 12, 16).

-bit converter has

2^n

levels and codes 0 through

2^n - 1

.

, so the highest reading is one LSB below

V_H

— a common off-by-one trap.

Smallest $n$ that resolves a target step $\Delta V$ over the span; round $n$ up to the next integer.

Scale the sensor signal to fill the converter range without exceeding it.

. The signal at

90\,\text{Hz}

exceeds

50\,\text{Hz}

, so

f_s > 2f_N

is violated — the tone will alias. Alias frequency with

k = 1

:

f_a = |\,f - k f_s\,| = |\,90 - 100\,| = 10\,\text{Hz}

, which lies in

[0, 50]\,\text{Hz}

, so it is the alias. Sanity check: a true

90\,\text{Hz}

component masquerades as

10\,\text{Hz}

— and no filtering after sampling can undo it; an anti-alias filter below

50\,\text{Hz}

was needed. Answer: criterion violated; the tone appears at

10.0\,\text{Hz}

.

f_a = |\,90 - 100\,| = 10\,\text{Hz}

Solution.

(a) Gain to fill the range:

G = \dfrac{V_H - V_L}{V_{\text{sensor,max}}} = \dfrac{5\,\text{V}}{50\,\text{mV}} = 100

. (b) After the gain, a

1\,\text{mV}

change at the sensor becomes

100\times1\,\text{mV} = 100\,\text{mV}

at the converter. Bits needed:

2^n \ge \dfrac{5\,\text{V}}{0.1\,\text{V}} = 50

. Since

2^5 = 32 < 50 \le 64 = 2^6

, take

n = 6

. Check: a 6-bit,

0

–

5\,\text{V}

converter has

Q = 5/64 = 78\,\text{mV} < 100\,\text{mV}

, so it resolves the amplified step. Sanity check: input-referred LSB

= Q/G = 78\,\text{mV}/100 = 0.78\,\text{mV} < 1\,\text{mV}

, confirming the requirement. Answers:

G = 100

,

n = 6

bits.

G = \dfrac{5\,\text{V}}{50\,\text{mV}} = 100,\qquad 2^n \ge \dfrac{5}{0.1} = 50 \Rightarrow n = 6

the 'Nyquist rate.' Anchor on the unambiguous facts — you must sample faster than

2f_N

and

f_s/2

is the folding limit — rather than on the label.

; sampling at exactly twice can lose the signal entirely.