Home / Courses / CSIR NET Physics / Practice / ADCs and DACs

ADCs and DACs

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

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

    An 8-bit flash (parallel) ADC needs how many comparators?
    1. 8
    2. 256
    3. 255
    4. 128
    Answer and solution

    Answer: C

    A flash converter compares the input with every threshold between adjacent codes at once. 28=256 codes have 255 thresholds, so it needs 2n−1=255 comparators (and a 256-resistor ladder). That is why flash ADCs are fast but rarely exceed 8 bits.

    Why people get this wrong: 256 counts codes, not the thresholds between them.

  2. MCQWarm-up

    An n-bit successive-approximation ADC completes one conversion in about
    1. n clock cycles
    2. 2n clock cycles
    3. 2n−1 clock cycles
    4. one clock cycle
    Answer and solution

    Answer: A

    It performs a binary search: one comparison per bit, from the most significant down. So n clock cycles (plus a little overhead), independent of the input voltage.

    Why people get this wrong: 2n−1 cycles is the worst case of a counter-ramp ADC.

  3. NATWarm-up

    An 8-bit DAC produces outputs from 0 to 2.55 V in equal steps. The size of one step is ______ mV.
    Answer and solution

    Answer: 10 mV

    There are 28=256 codes and therefore 255 steps between the lowest and highest output: 2.55 V/255=10 mV per LSB.

    Why people get this wrong: Dividing by 256 instead of 255 when the stated range is from the first to the last output.

  4. MCQWarm-up

    An audio signal contains frequencies up to 20 kHz. To sample it without aliasing, the sampling rate must be
    1. 20 kHz
    2. 10 kHz
    3. 30 kHz
    4. more than 40 kHz
    Answer and solution

    Answer: D

    The sampling theorem requires fs>2fmax=40 kHz. CDs use 44.1 kHz, leaving room for a practical anti-aliasing filter.

    Why people get this wrong: Sampling at the highest signal frequency itself loses everything at that frequency.

  5. MCQWarm-up

    An ideal ADC that rounds to the nearest code has a maximum quantisation error of
    1. 1 LSB
    2. ±12 LSB
    3. zero
    4. ±2 LSB
    Answer and solution

    Answer: B

    Every input inside one code's band is reported as the centre of that band, so the error never exceeds half a step either way. A truncating converter has an error from 0 to −1 LSB instead.

    Why people get this wrong: Zero error would need infinitely many bits.

  6. MSQWarm-up

    Which statements about ADC architectures are correct?
    1. The flash ADC is the fastest type
    2. The dual-slope ADC is slow but rejects mains hum and is used in digital multimeters
    3. The successive-approximation ADC is the common choice for medium speed and resolution
    4. The conversion time of a counter-ramp ADC does not depend on the input voltage
    Answer and solution

    Answer: A, B, C

    Flash converts in one step; SAR in n steps; a counter-ramp ADC counts up until its DAC output passes the input, so a larger input takes longer (up to 2n−1 clocks). Dual-slope integrates the input over a fixed time, averaging out any periodic interference whose period divides that time.

    Why people get this wrong: Only the counter-ramp (and tracking) types have input-dependent conversion times.

  7. MCQWarm-up

    The percentage resolution of an 8-bit DAC (one step as a fraction of the full-scale output) is about
    1. 12.5%
    2. 3.9%
    3. 0.39%
    4. 0.80%
    Answer and solution

    Answer: C

    An 8-bit DAC has 28−1=255 steps from zero to full scale, so one step is 1/255=0.39% of full scale.

    Why people get this wrong: 1/8=12.5% divides by the number of bits instead of the number of steps.

  8. NATStandard

    A 4-bit binary-weighted resistor DAC uses 10 kΩ for the most significant bit. The resistor for the least significant bit is ______ kΩ.
    Answer and solution

    Answer: 80 kΩ

    Each lower bit must contribute half the current of the one above, so its resistor doubles: 10, 20, 40, 80 kΩ. For n bits the spread is 2n−1, which becomes impractical for many bits.

    Why people get this wrong: Answering 160 kΩ (one doubling too many) or 40.

  9. NATStandard

    An 8-bit R-2R ladder DAC with reference 10 V and output amplifier gain such that |Vo|=VrefD/256 receives the code 1100 1000. The magnitude of its output is ______ V (to two decimal places).
    Answer and solution

    Answer: 7.81 V (±0.01)

    1100 1000 = 128 + 64 + 8 = 200. |Vo|=10×200/256=7.8125 V. Each bit contributes Vref/2,Vref/4,… in turn.

    Why people get this wrong: Reading the code backwards (LSB first) gives 19.

  10. NATStandard

    The ideal signal-to-quantisation-noise ratio of a 12-bit ADC for a full-scale sine wave is ______ dB (to one decimal place).
    Answer and solution

    Answer: 74 dB (±0.1)

    Quantisation noise has rms q/12; a full-scale sine has rms 2nq/(22). Their ratio is 1.52n, or 6.02n+1.76 dB =72.24+1.76=74.0 dB.

    Why people get this wrong: Using only 6.02n (72.2 dB) or 20log⁡2n without the sine correction.

  11. MCQStandard

    A 7 kHz sine wave is sampled at 10 kHz with no anti-aliasing filter. The reconstructed signal appears at
    1. 7 kHz
    2. 3 kHz
    3. 17 kHz
    4. 10 kHz
    Answer and solution

    Answer: B

    The Nyquist frequency is 5 kHz. A 7 kHz tone folds back to |7−10|=3 kHz: its samples are identical to those of a 3 kHz tone.

    Why people get this wrong: 17 kHz is also consistent with the samples, but reconstruction keeps only the band below fs/2.

  12. NATStandard

    A dual-slope ADC integrates the input for a fixed T1=20 ms, then integrates a −2.0 V reference until the integrator returns to zero. For an input of 0.75 V, the second phase lasts ______ ms.
    Answer and solution

    Answer: 7.5 ms (±0.01)

    The charge in equals the charge out: VinT1/RC=VrefT2/RC. So T2=T1Vin/Vref=20×0.75/2=7.5 ms. R, C and the clock frequency all cancel, which is why the method is accurate.

    Why people get this wrong: Thinking the integrator's RC must be known.

  13. MCQStandard

    A dual-slope ADC rejects 50 Hz mains interference completely when
    1. its integration time T1 is a whole multiple of 20 ms
    2. its clock frequency is exactly 50 Hz
    3. its reference voltage is 50 V
    4. its integrator time constant is 20 ms
    Answer and solution

    Answer: A

    Integrating a sine over a whole number of periods gives zero. With T1=20 ms (or 100 ms, which also covers 60 Hz cycles of 16.7 ms six times), the hum integrates away.

    Why people get this wrong: The time constant RC cancels; only the integration window matters.

  14. NATStandard

    An 8-bit counter-ramp ADC is clocked at 1 MHz. Its worst-case conversion time is ______ μs.
    Answer and solution

    Answer: 255 μs

    The counter starts at 0 and steps the DAC up one LSB per clock until it passes the input. A full-scale input needs 28−1=255 clocks: 255 μs. An 8-bit SAR at the same clock needs only 8 μs.

    Why people get this wrong: Using n=8 clocks (that is the SAR).

  15. MCQStandard

    A 10-bit successive-approximation ADC takes one clock per bit with a 1 MHz clock (ignore sampling time). Its maximum sampling rate is
    1. 1 MS/s
    2. 10 kS/s
    3. 1 kS/s
    4. 100 kS/s
    Answer and solution

    Answer: D

    One conversion takes 10 clocks =10 μs, so at most 105 conversions per second: 100 kS/s.

    Why people get this wrong: Taking one clock per conversion (1 MS/s).

  16. MSQStandard

    Which statements about an R-2R ladder DAC are correct?
    1. It needs only two resistor values
    2. The resistance looking into the ladder from any node towards the LSB end is R (with the proper termination)
    3. Its output is proportional to the binary input code
    4. It needs a resistor ratio of 2n−1 between its largest and smallest resistors
    Answer and solution

    Answer: A, B, C

    Every node sees 2R∥2R=R further down, so the current halves at each node, giving binary weights with only R and 2R: easy to match on a chip. A spread of 2n−1 is the drawback of the weighted-resistor DAC.

    Why people get this wrong: Mixing the two DAC types.

  17. NATStandard

    An 8-bit ADC has an LSB of 20 mV (range 0 to 5.12 V) and truncates (reports the code of the band the input lies in, counting from 0). For an input of 3.17 V its output code, in decimal, is ______.
    Answer and solution

    Answer: 158

    3.17/0.020=158.5, so the input lies in band 158 (from 3.16 to 3.18 V). A rounding converter (thresholds shifted by half an LSB) would give 159 for this input.

    Why people get this wrong: Rounding when the converter is said to truncate.

  18. MCQStandard

    An 8-bit two-step (half-flash) ADC converts the upper 4 bits with a 4-bit flash ADC, subtracts the DAC value of that result, and converts the remainder with a second 4-bit flash ADC. Its total number of comparators is
    1. 30
    2. 255
    3. 16
    4. 510
    Answer and solution

    Answer: A

    A 4-bit flash needs 24−1=15 comparators, so two of them need 30, against 28−1=255 for a full 8-bit flash. The price is a second conversion step and an accurate DAC and subtractor between the steps.

    Why people get this wrong: Counting 24=16 per flash, or using 255 as if it were a single flash.

  19. NATStandard

    A single-slope ADC starts a counter clocked at 1.0 MHz when an analogue ramp of 1.0 V/ms starts from 0 V, and stops it when the ramp reaches the input. For an input of 2.5 V the final count is ______.
    Answer and solution

    Answer: 2500

    The ramp reaches 2.5 V after 2.5/1.0=2.5 ms. At 106 counts per second the counter reaches 2.5×10−3×106=2500. The result depends directly on the ramp slope and clock, which is why the dual-slope design (which cancels them) is preferred for accuracy.

    Why people get this wrong: Using the clock period in ms and the time in s (or the reverse) gives an answer off by 103.

  20. NATExam level

    A 4-bit SAR ADC has an LSB of 1 V (code D means D volts) and keeps a trial bit when the DAC output does not exceed the input. For an input of 10.3 V, the number of trial bits that are tried and then rejected is ______.
    Answer and solution

    Answer: 2

    Trials: 1000 (8 V ≤ 10.3, keep), 1100 (12 V > 10.3, reject), 1010 (10 V ≤ 10.3, keep), 1011 (11 V > 10.3, reject). Final code 1010 = 10, with 2 bits rejected.

    Why people get this wrong: Clearing the earlier kept bits when a later trial fails.

  21. MCQExam level

    The same 4-bit SAR ADC (1 V per LSB) converts 10.3 V. The sequence of trial words placed on its internal DAC is
    1. 1000, 1100, 1010, 1011
    2. 1000, 0100, 0010, 0001
    3. 0001, 0010, 0100, 1000
    4. 1000, 1100, 1110, 1111
    Answer and solution

    Answer: A

    Each trial adds the next bit to the bits already kept. 1000 kept; 1100 rejected so the next trial is 1010 (not 1110); 1010 kept; then 1011, rejected. Result 1010.

    Why people get this wrong: Testing each bit alone forgets that the search builds on the kept bits.

  22. NATExam level

    A 12-bit ADC digitises a full-scale 1 kHz sine wave (peak Vp, so 1 LSB =2Vp/212). For the sampling error caused by aperture uncertainty to stay below half an LSB, the aperture uncertainty must be less than ______ ns (to one decimal place).
    Answer and solution

    Answer: 38.9 ns (±0.1)

    The steepest slope of Vpsin⁡ωt is ωVp. Half an LSB is Vp/212. So ta<Vp/2122πfVp=12π×1000×4096=38.9 ns. A sample-and-hold makes this easy.

    Why people get this wrong: Using a full LSB or the whole peak-to-peak range changes the answer by factors of 2.

  23. NATExam level

    A converter samples 16 times faster than the Nyquist rate and averages (decimates) the result. Assuming white quantisation noise, the resolution improves by ______ bits.
    Answer and solution

    Answer: 2

    Averaging N samples lowers the in-band noise power by N: 10log⁡16=12 dB. Each bit is worth 6.02 dB, so 2 bits. Every factor of 4 in oversampling gives one bit (more with noise shaping).

    Why people get this wrong: Expecting one bit per doubling.

  24. NATExam level

    An ADC measures a signal-to-noise-and-distortion ratio of 62.0 dB. Its effective number of bits is ______ (to two decimal places).
    Answer and solution

    Answer: 10 (±0.02)

    ENOB =(SINAD−1.76)/6.02=60.24/6.02=10.0. A 12-bit converter with this SINAD performs like an ideal 10-bit one.

    Why people get this wrong: Dividing 62 by 6.02 without removing the 1.76 dB.

  25. MCQExam level

    A sample-and-hold circuit uses a 1 nF hold capacitor, and 10 pA leaks from it in the hold mode. The droop rate of the held voltage is
    1. 10 V/s
    2. 0.1 mV/s
    3. 10 mV/s
    4. 1 V/s
    Answer and solution

    Answer: C

    dV/dt=I/C=10−11/10−9=0.01 V/s =10 mV/s. Over a 10 μs conversion that is only 0.1 μV.

    Why people get this wrong: Slipping a factor of 1000 between pF, nF and pA.

  26. MSQExam level

    Which statements about DAC errors are correct?
    1. A DAC is monotonic if its differential non-linearity is greater than −1 LSB at every code
    2. Integral non-linearity is the deviation of the transfer curve from a straight line
    3. An offset error shifts every output by the same amount
    4. A gain error can be removed by adding the same constant to every output
    Answer and solution

    Answer: A, B, C

    DNL compares each step with 1 LSB; a step smaller than 0 (DNL below −1) makes the output go down as the code goes up. INL accumulates the DNL. Offset is a constant shift; gain error scales the slope, so it needs a multiplication, not an added constant.

    Why people get this wrong: Offset is additive, gain is multiplicative.

  27. NATExam level

    A 4-bit binary-weighted DAC has input resistors 10, 20, 40 and 80 kΩ (MSB to LSB) switched to +5 V for a 1 and to ground for a 0, all feeding the virtual-ground input of an inverting amplifier with Rf=5 kΩ. For the input 1011, the output is ______ V (to four decimal places).
    Answer and solution

    Answer: -3.4375 V (±0.002)

    Bits 3, 1 and 0 are 1: currents 5/10+5/40+5/80=0.5+0.125+0.0625=0.6875 mA. Vo=−IRf=−0.6875×5=−3.4375 V =−(11/16)×5 V.

    Why people get this wrong: Forgetting the sign of the inverting amplifier, or giving bit 1 the 20 kΩ resistor.

  28. MCQExam level

    A 48 kHz tone reaches an ADC sampling at 44.1 kHz with no anti-aliasing filter. It appears in the digitised audio at
    1. 48 kHz
    2. 3.9 kHz
    3. 22.05 kHz
    4. 18.2 kHz
    Answer and solution

    Answer: B

    Alias frequency =|f−kfs| folded into 0 to fs/2: |48−44.1|=3.9 kHz, which is below 22.05 kHz, so it lands right in the audible band.

    Why people get this wrong: Thinking ultrasound is harmless: without a filter it folds down.

  29. NATExam level

    A 10-bit ADC spans 0 to 10 V (LSB =10/1024 V). The rms quantisation noise, assuming the error is uniform over one LSB, is ______ mV (to two decimal places).
    Answer and solution

    Answer: 2.82 mV (±0.01)

    A uniform error between −q/2 and q/2 has variance q2/12. q=9.766 mV, so rms =9.766/12=2.82 mV.

    Why people get this wrong: Taking the peak error q/2=4.88 mV as the rms.

  30. MCQExam level

    A 6-bit flash ADC uses a resistor string to set its thresholds and a priority encoder for the output. How many comparators does it need?
    1. 64
    2. 6
    3. 63
    4. 32
    Answer and solution

    Answer: C

    26=64 codes need 63 thresholds and so 63 comparators; their outputs form a thermometer code that the encoder turns into 6 bits. Each extra bit doubles the comparator count.

    Why people get this wrong: 64 is the number of output codes.

← All topics · Timed mock test →