Home / Courses / CSIR NET Physics / Practice / Noise, filtering and lock-in detection

Noise, filtering and lock-in detection

Experimental methods · 77 questions and templates for CSIR NET Physics, in order from warm-up to top rank. Check each answer, then read the full solution.

New to this topic? Read the lesson first: Noise, filtering and lock-in detection.

Loading questions… (they need JavaScript; nothing you type is sent anywhere)

Written questions with solutions

The same questions as a list, so you can print them or read without the practice box. Templates (fresh numbers) appear only in the practice box above.

  1. MCQWarm-up

    The thermal (Johnson-Nyquist) noise voltage across a resistor depends on
    1. the current through it
    2. its temperature, its resistance and the measurement bandwidth
    3. the material it is made of
    4. only the bandwidth
    Answer and solution

    Answer: B

    v2=4kBTRB: thermal agitation of the carriers in any resistance at temperature T, measured over bandwidth B. It is there with no current flowing and is the same for a carbon or a metal resistor of the same value.

    Why people get this wrong: Noise that grows with current is shot noise or excess (1/f) noise, not Johnson noise.

  2. MCQWarm-up

    Shot noise in a photodiode current arises because
    1. the diode is warm
    2. the junction capacitance charges and discharges
    3. charge arrives as discrete, independent electrons
    4. the light intensity drifts slowly
    Answer and solution

    Answer: C

    A current made of independent charges e arriving at random has Poisson fluctuations: i2=2eIB. It needs a current crossing a barrier (a junction or vacuum gap) and is white up to very high frequencies.

    Why people get this wrong: Shot noise is absent in a resistor carrying current in equilibrium; there only Johnson noise appears.

  3. MCQWarm-up

    Noise whose power spectral density falls as 1/f has
    1. equal power in every 1 Hz band
    2. no power below 1 Hz
    3. power only at multiples of the mains frequency
    4. equal power in every decade of frequency
    Answer and solution

    Answer: D

    Power in a band f1 to f2 is ∫Adf/f=Aln⁡(f2/f1): the same for 1 to 10 Hz as for 1 to 10 kHz. White noise has equal power per hertz instead. 1/f noise dominates at low frequencies, which is why weak signals are moved to higher frequencies by chopping.

    Why people get this wrong: Equal power per hertz describes white noise.

  4. MCQWarm-up

    A repetitive signal buried in random noise is averaged over 100 sweeps. The signal-to-noise ratio (in amplitude) improves by a factor of
    1. 100
    2. 10
    3. 50
    4. 2
    Answer and solution

    Answer: B

    The signal adds coherently (∝N) while uncorrelated noise adds in quadrature (∝N), so SNR ∝N=10.

    Why people get this wrong: Averaging gains only as the square root of the number of repetitions.

  5. MCQWarm-up

    An amplifier has a voltage gain of 20 dB. Its output voltage is larger than the input by a factor of
    1. 20
    2. 100
    3. 10
    4. 20
    Answer and solution

    Answer: C

    For voltages, dB =20log10⁡(V2/V1), so 20 dB is a factor of 10 in voltage (and 100 in power).

    Why people get this wrong: 100 is the power ratio; 10 log is for power, 20 log for voltage.

  6. MCQWarm-up

    An audio signal contains frequencies up to 20 kHz. To digitise it without aliasing, the sampling rate must exceed
    1. 40 kHz
    2. 20 kHz
    3. 10 kHz
    4. 80 kHz
    Answer and solution

    Answer: A

    The Nyquist criterion: sample faster than twice the highest frequency present, here 40 kHz (CD audio uses 44.1 kHz, with an anti-aliasing filter in front).

    Why people get this wrong: Sampling at exactly the signal frequency sees a sine as a constant.

  7. MCQWarm-up

    White noise is measured through a filter whose bandwidth is then made 100 times wider. The rms noise voltage becomes larger by a factor of
    1. 100
    2. 10
    3. 1 (it does not change)
    4. 50
    Answer and solution

    Answer: B

    White noise has constant power per hertz, so the noise power is proportional to bandwidth and the rms voltage to its square root: 100=10.

    Why people get this wrong: The power grows 100 times; the rms voltage only 10 times.

  8. NATStandard

    The rms Johnson noise voltage of a 1.0 MΩ resistor at 300 K measured in a 10 kHz bandwidth is ______ μV (to one decimal place). Use kB=1.381×10−23 J/K.
    Answer and solution

    Answer: 12.9 μV (±0.1)

    v=4kBTRB=4×1.381×10−23×300×106×104=1.657×10−10=1.29×10−5 V =12.9 μV.

    Why people get this wrong: Remember the factor 4, and take the square root at the end.

  9. NATStandard

    A DC current of 1.0 mA flows through a photodiode. Its shot-noise current density is ______ pA/Hz (to one decimal place).
    Answer and solution

    Answer: 17.9 pA/√Hz (±0.1)

    in=2eI=2×1.602×10−19×10−3=1.79×10−11 A/Hz =17.9 pA/Hz.

    Why people get this wrong: Shot noise grows as the square root of the current, so the relative noise falls as the current rises.

  10. MCQStandard

    An RC low-pass filter has R=10 kΩ and C=10 nF. Its −3 dB frequency is
    1. 10 kHz
    2. 159 Hz
    3. 6.28 kHz
    4. 1.59 kHz
    Answer and solution

    Answer: D

    RC=104×10−8=10−4 s. fc=1/(2πRC)=1/(6.283×10−4)=1.59 kHz.

    Why people get this wrong: 1/RC=104 s−1 is an angular frequency; divide by 2π for hertz.

  11. MCQStandard

    A dual-phase lock-in shows X=0 while R=X2+Y2 is clearly non-zero. This means
    1. there is no signal at the reference frequency
    2. the signal is 90° out of phase with the reference
    3. the time constant is too short
    4. the signal is at twice the reference frequency
    Answer and solution

    Answer: B

    X=Vcos⁡ϕ and Y=Vsin⁡ϕ (in the lock-in's rms convention). X=0 with R≠0 means cos⁡ϕ=0: the signal is in quadrature with the reference. Shifting the reference phase by 90° moves it all into X.

    Why people get this wrong: A single-phase lock-in would show zero here and could fool you into thinking there is no signal.

  12. MCQStandard

    In an optical experiment the light is chopped at about 1 kHz and the detector signal is fed to a lock-in referenced to the chopper. The main reason is to
    1. move the signal away from 1/f noise and slow drifts near zero frequency
    2. increase the average light power
    3. make the detector faster
    4. remove shot noise completely
    Answer and solution

    Answer: A

    A DC measurement collects all the low-frequency noise (1/f, drifts, room light). Chopping puts the signal at 1 kHz, where the noise is mostly white and lower, and the lock-in measures only a narrow band around it.

    Why people get this wrong: Shot noise is white, so chopping does not remove it; the narrow bandwidth only reduces how much of it is accepted.

  13. MSQStandard

    Which of these reduce the rms Johnson noise voltage of a resistive sensor at the amplifier input?
    1. Cooling the sensor
    2. Narrowing the measurement bandwidth
    3. Using a smaller resistance (if the signal allows)
    4. Increasing the bias current through it
    Answer and solution

    Answer: A, B, C

    vn=4kBTRB: lower T, R or B all reduce it. The bias current does not enter Johnson noise (it can increase the signal and add excess noise).

    Why people get this wrong: More current may raise the signal-to-noise ratio, but it does not reduce the Johnson noise itself.

  14. MCQStandard

    A first-order RC low-pass filter attenuates a sine wave at ten times its corner frequency by about
    1. 3 dB
    2. 10 dB
    3. 40 dB
    4. 20 dB
    Answer and solution

    Answer: D

    |H|=1/1+(f/fc)2=1/101≈0.0995, which is 20log10⁡(0.0995)=−20.0 dB. Far above the corner a first-order filter falls 20 dB per decade (6 dB per octave).

    Why people get this wrong: 3 dB is the attenuation at the corner itself; 40 dB per decade needs a second-order filter.

  15. MCQStandard

    The ideal signal-to-quantisation-noise ratio of a 12-bit ADC for a full-scale sine wave is about
    1. 48 dB
    2. 74 dB
    3. 86 dB
    4. 72 dB
    Answer and solution

    Answer: B

    Quantisation noise has rms Δ/12; a full-scale sine has rms 2NΔ/(22). The ratio gives SNR =6.02N+1.76 dB =72.2+1.8=74.0 dB.

    Why people get this wrong: 72 dB is only the 6.02N part.

  16. MCQStandard

    A 7 kHz sine wave is sampled at 10 kHz with no anti-aliasing filter. The samples look like a sine of frequency
    1. 7 kHz
    2. 17 kHz
    3. 3 kHz
    4. 1.5 kHz
    Answer and solution

    Answer: C

    Frequencies above the Nyquist frequency (5 kHz) fold back: |7−10|=3 kHz. The samples of 7 kHz and 3 kHz sines are identical (up to sign).

    Why people get this wrong: Aliases appear at |f−kfs|, folded into the range 0 to fs/2.

  17. NATStandard

    Averaging 64 sweeps of a repetitive signal in white noise improves the amplitude signal-to-noise ratio by ______ dB (to one decimal place).
    Answer and solution

    Answer: 18.1 dB (±0.1)

    Amplitude SNR improves by 64=8; in decibels 20log10⁡8=18.06 dB. (The power SNR improves by 64, which is the same 18.06 dB.)

    Why people get this wrong: Using 10log10⁡8=9.0 dB mixes a voltage ratio with the power formula.

  18. MCQStandard

    The Johnson noise of a resistor can be described as a current source in parallel with it. For a 1.0 MΩ resistor at 300 K, this noise current density is about
    1. 1.3 pA/Hz
    2. 0.13 nA/Hz
    3. 129 nV/Hz
    4. 0.13 pA/Hz
    Answer and solution

    Answer: D

    in=4kBT/R=4×1.38×10−23×300/106=1.66×10−26=1.29×10−13 A/Hz =0.13 pA/Hz. A larger resistor has more noise voltage but less noise current, which is why transimpedance amplifiers use large feedback resistors.

    Why people get this wrong: 129 nV/Hz is the noise voltage 4kBTR, a different (equivalent) description.

  19. NATStandard

    A resistor is cooled from 300 K to 4.0 K (its resistance unchanged). Its Johnson noise voltage in a fixed bandwidth falls by a factor of ______ (to two decimal places).
    Answer and solution

    Answer: 8.66 (±0.01)

    vn=4kBTRB∝T, so the factor is 300/4.0=75=8.66. That is why sensitive preamplifier inputs and bolometers are cooled.

    Why people get this wrong: The noise power falls 75 times, the rms voltage only 75 times.

  20. MCQStandard

    A sensitive measurement shows strong 50 Hz hum and its harmonics only when the instrument and its preamplifier are both plugged into the mains and joined by a coaxial cable. The likely cause and cure are
    1. a ground loop; ground the system at one point only, or use a differential or isolated input
    2. Johnson noise in the cable; cool the cable
    3. shot noise; increase the signal current
    4. aliasing; add an anti-aliasing filter
    Answer and solution

    Answer: A

    The two mains earths and the cable shield form a closed loop. Mains magnetic fields induce a current round it, and earth currents give a potential difference between the two grounds; either way a 50 Hz voltage appears along the shield, in series with the signal. Breaking the loop (single-point grounding, isolation) or rejecting the common-mode voltage (differential input) removes it. Random noise would not be locked to the mains frequency.

    Why people get this wrong: Interference at exactly the mains frequency and its harmonics is pickup, not intrinsic noise.

  21. NATExam level

    White noise passes through a first-order RC low-pass filter with RC=1.0 ms. The noise-equivalent bandwidth (the width of the ideal rectangular filter passing the same noise power) is ______ Hz.
    Answer and solution

    Answer: 250 Hz (±1)

    Beq=∫0∞df1+(2πfRC)2=12πRC⋅π2=14RC=250 Hz. It is π/2 times the −3 dB frequency (159 Hz), because the response has a tail.

    Why people get this wrong: Using the −3 dB frequency as the noise bandwidth underestimates the noise.

  22. MCQExam level

    A lock-in has a single-pole (6 dB/octave) output filter with time constant 300 ms. The input noise is white with density 10 nV/Hz around the reference frequency. The rms noise on the output is about
    1. 18 nV
    2. 5.8 nV
    3. 3.0 nV
    4. 9.1 nV
    Answer and solution

    Answer: D

    ENBW of a single RC section =1/(4τ)=1/1.2=0.83 Hz. Noise =10 nV/Hz×0.83 Hz=9.1 nV.

    Why people get this wrong: The time constant sets the bandwidth through 1/4τ, not 1/τ or 1/2πτ.

  23. NATExam level

    A signal 2.0cos⁡(ωt+60∘) mV is multiplied by a reference cos⁡ωt (unit amplitude) and the product is averaged over many cycles. The result is ______ mV (to two decimal places).
    Answer and solution

    Answer: 0.5 mV (±0.01)

    cos⁡(ωt+ϕ)cos⁡ωt=12[cos⁡ϕ+cos⁡(2ωt+ϕ)]. The 2ω term averages to zero, leaving 12Vscos⁡ϕ=12×2.0×0.5=0.50 mV.

    Why people get this wrong: The factor 1/2 from the product of two cosines is easy to drop.

  24. MCQExam level

    A dual-phase lock-in reads X=3.0 μV and Y=4.0 μV. The signal magnitude and its phase relative to the reference are
    1. 7.0 μV, 45°
    2. 5.0 μV, 36.9°
    3. 5.0 μV, 53.1°
    4. 1.0 μV, 53.1°
    Answer and solution

    Answer: C

    R=X2+Y2=5.0 μV and θ=tan−1⁡(Y/X)=tan−1⁡(4/3)=53.1°.

    Why people get this wrong: 36.9° is tan−1⁡(X/Y), measured from the wrong axis.

  25. MCQExam level

    A lock-in uses a square-wave reference at frequency f (switching the signal's sign each half cycle). Compared with a sinusoidal signal of the same amplitude at f, a signal at 3f in the right phase gives an output
    1. equal to it
    2. zero
    3. three times larger
    4. one third as large
    Answer and solution

    Answer: D

    A square wave contains 4π[sin⁡ωt+13sin⁡3ωt+15sin⁡5ωt+…]. Multiplying by the signal and averaging picks out the matching harmonic, so a 3f signal is weighted by 1/3 (and even harmonics give zero). Modern digital lock-ins multiply by a pure sine to avoid this.

    Why people get this wrong: Square-wave (switching) demodulators respond to odd harmonics too.

  26. NATExam level

    A 10 pF capacitor connected to a resistor at 300 K has a total rms noise voltage, integrated over all frequencies, of ______ μV (to one decimal place). Use kB=1.381×10−23 J/K.
    Answer and solution

    Answer: 20.4 μV (±0.1)

    The resistor noise 4kBTR filtered by RC (noise bandwidth 1/4RC) gives v2=4kBTR/4RC=kBT/C: independent of R, as equipartition demands (12Cv2=12kBT). 1.381×10−23×300/10−11=2.04×10−5 V.

    Why people get this wrong: A bigger resistor makes more noise density but a narrower bandwidth; the two cancel.

  27. MCQExam level

    An amplifier has a noise figure of 3.0 dB. Its equivalent input noise temperature (standard T0=290 K) is about
    1. 580 K
    2. 145 K
    3. 289 K
    4. 870 K
    Answer and solution

    Answer: C

    F=100.3=2.00. Te=(F−1)T0=0.995×290=289 K: the amplifier adds as much noise as a matched source resistor at 290 K.

    Why people get this wrong: Convert the noise figure from dB to a ratio before using Te=(F−1)T0.

  28. NATExam level

    A low-noise preamplifier (power gain 20 dB, noise factor 1.5) drives a second stage of noise factor 10 (both as power ratios). The noise factor of the chain is ______ (to two decimal places).
    Answer and solution

    Answer: 1.59 (±0.01)

    Friis: F=F1+(F2−1)/G1=1.5+9/100=1.59. The first stage's gain divides the noise of everything after it, which is why the best amplifier goes first.

    Why people get this wrong: Use the gain as a ratio (100), not in dB (20).

  29. MSQExam level

    Which statements about lock-in detection are correct?
    1. The DC output is proportional to the component of the signal in phase with the reference
    2. Noise at frequencies far from the reference produces output fluctuations that the low-pass filter removes
    3. A longer time constant lowers the output noise but makes the reading respond more slowly
    4. A lock-in can recover a signal whose frequency is unrelated to the reference
    Answer and solution

    Answer: A, B, C

    Mixing shifts components at fr+Δf to Δf; the low-pass filter keeps only |Δf|≲1/4τ. So the output is the in-phase amplitude at the reference frequency, with noise bandwidth ∝1/τ, settling in a few τ. A signal not locked to the reference averages to zero.

    Why people get this wrong: The reference must be coherent with the signal; that is the whole trick.

  30. MCQExam level

    A photodetector is shot-noise limited. If the light power on it is quadrupled, the signal-to-noise ratio
    1. stays the same
    2. doubles
    3. quadruples
    4. halves
    Answer and solution

    Answer: B

    Signal ∝I, shot noise ∝I, so SNR ∝I: four times the light gives twice the SNR.

    Why people get this wrong: In the Johnson-noise-limited regime (noise independent of light) the SNR would quadruple.

← All topics · Timed mock test →