Data Converter

Posted on Edited on In

Quantization Noise

Quantization Error

image-20250910210909363

image-20250910211207655

image-20250910211034914

Notice \(e_q\in (0, \Delta)\) and its average is \(\Delta/2\). To calculate SNDR, DC component shall be excluded

Don't confuse resolution \(\Delta\) with Bounded Quantization Noise \(-\Delta/2 \sim \Delta/2\)

image-20250909233010702

Quantization Noise Spectrum

image-20240825221754959

Quantization noise is less with higher resolution as the input range is divided into a greater number of smaller ranges

This error can be considered a quantization noise with RMS

image-20240925235213137

image-20260501104900735

image-20260501104930194

ADC Input Noise

Walt Kester, ADC Input Noise: The Good, The Bad, and The Ugly­. Is No Noise Good Noise? [link] [pdf]

—, MT-004: The Good, the Bad, and the Ugly Aspects of ADC Input Noise-Is No Noise Good Noise? [https://www.analog.com/media/en/training-seminars/tutorials/mt-004.pdf]

Understanding ADC Noise for Small and Large Signal Inputs for Receiver Applications [https://www.analog.com/en/resources/technical-articles/understanding-adc-noise-for-small-and-large-signal-inputs-for-receiver-applications.html]

The LSB determines the ADC code resolution, but the minimum reliably detectable input voltage is determined by the total noise floor, including quantization noise

image-20260501103056383

[https://share.google/aimode/aA1V4uj3GofKbBojl]

ADC quantization noise is considered a "deterministic error" rather than random thermal noise because it correlates with the input signal, creating a saw-tooth error waveform rather than random Gaussian noise. While it cannot be reduced by averaging a static signal, it can be reduced through averaging if the signal is oversampled, or if dither (random noise) is added to decorrelate it.

Quantization is NOT Noise

N. Blachman, "The intermodulation and distortion due to quantization of sinusoids," in IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 33, no. 6, pp. 1417-1426, December 1985 [https://sci-hub.st/10.1109/TASSP.1985.1164729]

image-20250902203709226

Carsten Wulff, Oversampling and Sigma-Delta ADCs [https://analogicus.com/aic2026/oversampling_and_sigma-delta_adcs] [video] [slides]

N. Blachman, "The intermodulation and distortion due to quantization of sinusoids," in IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 33, no. 6, pp. 1417-1426, December 1985

The quantization noise is an infinite sum of input signal odd harmonics, where the amplitude of the harmonics is determined by a sum of a Bessel function

"Quantization noise is white", because for a high number of bits, it looks white in the FFT

image-20260602220757703

The quantization noise is odd harmonics of the input signal [Gist link]

l6_quant

img

[Gist link]

n_impact

image-20260501103237761

Sampling Noise in ADC

Kwantae Kim, Integrated Analog Systems D - Lecture 12 (ADC) [https://youtu.be/NkSitVkPNig]

image-20260501170637885

In the power domain, \(\color{red}v_{nS,RMS}/3 \lt \sigma_{q,RMS}\) ensures that sampling noise power is nearly an order of magnitude smaller than the quantization noise

image-20260502084730678

image-20260501171651963

ADC SNR & clock jitter

Akkaya, A. (2021). High-Speed ADC Design and Optimization for Wireline Links (Publication No. 8453) [PhD thesis, EPFL; Supervised by Y. Leblebici]. [https://doi.org/10.5075/epfl-thesis-8453]

CC Chen, Why Absolute Jitter Matters for ADCs & DACs? [https://youtu.be/jBgDDFFDq30]

Thomas Neu, TIPL 4704. Jitter vs SNR for ADCs [https://www.ti.com/content/dam/videos/external-videos/en-us/2/3816841626001/5529003238001.mp4/subassets/TIPL-4704-Jitter-vs-SNR.pdf]

Walt Kester , MT-007: Aperture Time, Aperture Jitter, Aperture Delay Time [https://www.analog.com/media/en/training-seminars/tutorials/MT-007.pdf]

cyclostationary random process

image-20250809170358612

image-20250525134220901

image-20250525135503671

\[\begin{align} \text{SNR}_\text{ADC}[\text{dB}] &= -20\cdot \log \sqrt{\left(10^{-\frac{\text{SNR}_\text{Quantization Noise}}{20}}\right)^2 + \left(10^{-\frac{\text{SNR}_\text{Jitter}}{20}}\right)^2} \\ &= -10\cdot \log \left(\left(10^{-\frac{\text{SNR}_\text{Quantization Noise}}{20}}\right)^2 + \left(10^{-\frac{\text{SNR}_\text{Jitter}}{20}}\right)^2\right) \\ &= -10\cdot \log \left(\left(10^{-\frac{10\log(\frac{3\times2^{2N}}{2})}{20}}\right)^2 + \left(10^{-\frac{-20\log{(2\pi f_\text{in}\sigma_\text{jitter})}}{20}}\right)^2\right) \\ &= -10\cdot \log \left( \frac{2}{3\times 2^{2N}} + (2\pi f_\text{in}\sigma_\text{jitter})^2 \right) \end{align}\]

image-20250525141523199

image-20250525143507747

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
import numpy as np
import matplotlib.pyplot as plt

N = 12  # ADC bit
fin = 100e6 # the frequency of the sinusoidal input signal
jrms = np.linspace(0, 10, 1000)*1e-12 #ps

# ADC (SNR) with Quantization Noise & Jitter degradation
SNR_ADC = -10 * np.log10(10**(-np.log10(3*2**(2*N)/2)) + (2*np.pi*fin*jrms)**2)
ENOB = (SNR_ADC - 1.76) / 6.02

plt.plot(jrms*1e12, ENOB, label='100 MHZ input')
plt.plot([0, 10], [12, 12], '--', label='12-bit limit')
plt.plot([0, 10], [6, 6], '--', label='6-bit limit')

plt.xscale('linear')
plt.xlim([0, 10])
plt.ylim([5, 15])
plt.xlabel('RMS Jitter (ps)')
plt.ylabel('Effective Number of Bits (ENOB')
plt.grid(which='both')
plt.title('ENOB vs. RMS Clock Jitter (100 MHz)')
plt.legend()
plt.show()

Chun-Hsien Su (蘇純賢). Design of Oversampled Sigma-Delta Data Converters. July, 2006 [pdf]

image-20250809182751263

Chembian Thambidurai, "SNR of an ADC in the presence of clock jitter" [https://www.linkedin.com/posts/chembiyan-t-0b34b910_adcsnrjitter-activity-7171178121021304833-f2Wd/]

Unlike the quantization noise and the thermal noise, the impact of the clock jitter on the ADC performance depends on the input signal properties like its PSD

image-20241123205352661

The error between the ideal sampled signal and the sampling with clock jitter can be treated as noise and it results in the degradation of the SNR of the ADC

image-20241124004634365

For sinusoid input:

image-20241210235817281

image-20241222140258960

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
import numpy as np
import matplotlib.pyplot as plt

ENOB = 8
fin = np.logspace(8, 11, 60)

# quantization noise: SNR = 6.02*ENOB + 1.76 dB
Ps_PnQ = 10**((6.02*ENOB + 1.76)/10)
PnQ = 1/Ps_PnQ

# jitter noise: SNR = 6 - 20log10(2*pi*fin*Jrms) dB @ref. Chembiyan T
Jrms_list = [25e-15, 50e-15, 100e-15, 250e-15, 500e-15, 1000e-15]
for Jrms in Jrms_list:
    # Ps_PnJ_lcl = 10**((6-20*np.log10(2*np.pi*fin*Jrms))/10)     # ref. Chembiyan T
    Ps_PnJ_lcl = 10**((0 - 20 * np.log10(2 * np.pi * fin * Jrms)) / 10)  # ref. Nicola Da Dalt
    PnJ_lcl = 1/Ps_PnJ_lcl
    SNR_lcl = 10*np.log10(1/(PnQ+PnJ_lcl))
    plt.plot(fin, SNR_lcl, label=r'$\sigma_{jitter}$'+'='+str(int(Jrms*1e15))+'fs')

plt.xscale('log')
plt.ylim([0, 55])
plt.xlabel(r'$f_{in}$ [Hz]')
plt.ylabel(r'SNR [dB]')
plt.grid(which='both')
# plt.title(r'ref. Chembiyan T')
plt.title(r'ref. Nicola Da Dalt')
plt.legend()
plt.show()

K. Tyagi and B. Razavi, "Performance Bounds of ADC-Based Receivers Due to Clock Jitter," in IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 70, no. 5, pp. 1749-1753, May 2023 [https://www.seas.ucla.edu/brweb/papers/Journals/KT_TCAS_2023.pdf]

N. Da Dalt, M. Harteneck, C. Sandner and A. Wiesbauer, "On the jitter requirements of the sampling clock for analog-to-digital converters," in IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications, vol. 49, no. 9, pp. 1354-1360, Sept. 2002 [https://sci-hub.se/10.1109/TCSI.2002.802353]

M. Shinagawa, Y. Akazawa and T. Wakimoto, "Jitter analysis of high-speed sampling systems," in IEEE Journal of Solid-State Circuits, vol. 25, no. 1, pp. 220-224, Feb. 1990 [https://sci-hub.se/10.1109/4.50307]

image-20241210232716862

Akkaya, A. (2021). High-Speed ADC Design and Optimization for Wireline Links (Publication No. 8453) [PhD thesis, EPFL; Supervised by Y. Leblebici]. [https://doi.org/10.5075/epfl-thesis-8453]

待学芯. ADC量化结果反推采样时钟抖动（Jitter） [https://mp.weixin.qq.com/s/55xfVQMe_N8zUGpI8ZvmsQ]

—. 关于时钟抖动(Jitter)与ADC的一些讨论 [https://mp.weixin.qq.com/s/GW1keHhfq7zrd036lyG0CQ]

image-20250811210300829

DAC SNR & clock jitter

Boris Murmann ISSCC 2022 SC1: Introduction to ADCs/DACs: Metrics, Topologies, Trade Space, and Applications [pdf]

S. Kim, K. -Y. Lee and M. Lee, "Modeling Random Clock Jitter Effect of High-Speed Current-Steering NRZ and RZ DAC," in IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 65, no. 9, pp. 2832-2841, Sept. 2018 [https://sci-hub.se/10.1109/TCSI.2018.2821198]

Martin Clara. High-Performance D/A-Converters - Application to Digital Transceivers, 2013 [pdf]

Chun-Hsien Su (蘇純賢). Design of Oversampled Sigma-Delta Data Converters. July, 2006 [pdf]

ampling Jitter Effects for ADC/DAC

  • In both DAC or ADC cases, doubling the timing jitter doubles the noise level
  • Also, doubling the frequency or amplitude doubles the jitter induced noise - SNR is not improved

image-20250810213544751

image-20250810213615814

ADC Linearity (DNL/INL)

image-20260426174704406

image-20260425095158692

missing code

image-20260425100334682

TODO 📅

  • Endpoint method
  • BestFit method

image-20241006211529077

image-20241006195931838

INL/DNL Measurements for High-Speed Analog-to Digital Converters (ADCs) [https://picture.iczhiku.com/resource/eetop/sYKTSqLfukeHSmMB.pdf]

Code Density Test

Apply a linear ramp to ADC input

image-20241214100849243

Jungwirth, Patrick. Sampling Theory and Analog-to-Digital Conversion. Independently published, 2018.

image-20260907235127782

image-20260907235159600

DAC Linearity (DNL/INL)

image-20241215101400962

The worst INL of three DAC Architecture is same

image-20260913165357043

DAC DNL

One difference between ADC and DAC is that DAC DNL can be less than -1 LSB

In a DAC, DNL < -1LSB implies non-monotonicity

image-20241006215420568

image-20260913124553643

DAC INL

image-20260913165242029

  • \(A = \sum_{j=1}^k I_j\), \(B=\sum_{j=k+1}^N I_j\)
  • A and B are independent with \(\sigma_A^2 = k\sigma_u^2\) and \(\sigma_B^2=(N-k)\sigma_u^2\)

Therefore \[ \mathrm{Var}\left(\frac{X}{Y}\right)\approx \frac{k^2}{N^2}\left(\frac{\sigma_i^2}{kI_u^2} + \frac{\sigma_i^2}{NI_u^2} -2\frac{\mathrm{cov}(X,Y)}{kNI_u^2}\right) \] and \[\begin{align} \mathrm{cov}(X,Y) &= E[XY] - E[X]E[Y] = E[A(A+B)] - kNI_u^2 \\ &= E[A^2]+E[A]E[B] - kNI_u^2= \sigma_A^2+E[A]^2 + k(N-k)I_u^2 - kNI_u^2\\ &= k\sigma_i^2 + k^2I_u^2+ k(N-k)I_u^2 - kNI_u^2 \\ &= k\sigma_i^2 \end{align}\]

Finally, \[ \mathrm{Var}\left(\frac{X}{Y}\right)\approx \frac{k^2}{N^2}\left(\frac{\sigma_i^2}{kI_u^2} + \frac{\sigma_i^2}{NI_u^2} -2\frac{k\sigma_i^2}{kNI_u^2}\right) = \frac{k^2}{N^2}\left(\frac{1}{k}- \frac{1}{N}\right)\sigma_u^2 \] i.e. \[ \boxed{\mathrm{Var(INL(k))} = k^2\left(\frac{1}{k}- \frac{1}{N}\right)\sigma_u^2 = k\left(1- \frac{k}{N}\right)\sigma_u^2} \]

Standard deviation of INL is maximum at mid-scale (k=N/2)

image-20241215114755896

image-20241215101727644

INL/DNL analysis of current steering DAC

Spectral Metrics

image-20260425164053212

SNR, SNDR (SINAD)

Understanding Key Parameters for RF-Sampling Data Converters White Paper (WP509) [https://docs.amd.com/v/u/en-US/wp509-rfsampling-data-converters]

image-20260425164740278

image-20260503100924412

image-20260501160344805

ENOB

Qasim Chaudhari, On Analog-to-Digital Converter (ADC), 6 dB SNR Gain per Bit, Oversampling and Undersampling [https://wirelesspi.com/on-analog-to-digital-converter-adc-6-db-snr-gain-per-bit-oversampling-and-undersampling/]

The quantization noise power \(P_Q\) for a uniform quantizer with step size \(\Delta\) is given by \[ P_Q = \frac{\Delta ^2}{12} \] For a full-scale sinusoidal input signal with an amplitude equal to \(V_{FS}/2\), the input signal is given by \(x(t) = \frac{V_{FS}}{2}\sin(\omega t)\)

Then input signal power \(P_s\) is \[ P_s = \frac{V_{FS}^2}{8} \] Therefore, the signal-to-quantization noise ratio (SQNR) is given by \[ \text{SQNR} = \frac{P_s}{P_Q} = \frac{V_{FS}^2/8}{\Delta^2/12}=\frac{V_{FS}^2/8}{V_{FS}^2/(12\times 2^{2N})} = \frac{3\times 2^{2N}}{2} \] where \(N\) is the number of quantization bits

When represented in dBs \[ \text{SQNR(dB)} = 10\log(\frac{P_s}{P_Q}) = 10\log(\frac{3\times 2^{2N}}{2})= 20N\log(2) + 10\log(\frac{3}{2})= 6.02N + 1.76 \]

the maximum achievable SNR of N-bit ADC — theoretical SNR limit

image-20250705100706289

image-20250705101619687

image-20250705101635533

Dan Boschen, GRCon25: Quantifying Signal Quality: Practical Tools for High-Fidelity Waveform Analysis

[linkedin GRCon25]

img

SDR, THD

image-20260425165031673

Understanding Key Parameters for RF-Sampling Data Converters White Paper (WP509) [https://docs.amd.com/v/u/en-US/wp509-rfsampling-data-converters]

image-20260503101042354

Walt Kester. Evaluating High Speed DAC Performance [https://www.analog.com/media/en/training-seminars/tutorials/mt-013.pdf]

via other definition

THD: signal to distortion

SINAD: noise and distortion to signal

image-20260503102806229

image-20260425165827715

SFDR & INL

image-20260425164947635

image-20250524172307785

Beware, this is of course only true under the same conditions at which the INL was taken, i.e. typically low input signal frequency

Dynamic Range (DR)

image-20260503113228361

image-20260503113130108

image-20250825220134900

image-20250825220536821

Noise Spectral Density (NSD)

Understanding Key Parameters for RF-Sampling Data Converters White Paper (WP509) [https://docs.amd.com/v/u/en-US/wp509-rfsampling-data-converters]

image-20260503102953803

image-20260503103307603

image-20260503103531776

image-20250902010512726

Spectral Leakage

Two ways to deal with spectral leakage: Ensure integer number of periods or Windowing

image-20260426075613908

image-20260501161349428

Coherent Sampling

Choosing M/N non-prime repeats the signal quantization periodically and fewer quantization steps are measured. The quantization repeats periodically and creates a line spectrum that can obscure real frequency lines (e.g. the red lines in the images below, created by non-linearities of the ADC).[https://www.dsprelated.com/thread/469/coherent-sampling-very-brief-and-simple]

image-20250705085139758

\[ \frac{f_{\text{in}}}{f_{\text{s}}}=\frac{M_C}{N_R} \]

  • \(f_\text{in}\) and \(f_s\) must be incommensurate (\(f_s/f_\text{in}\) is irrational number. btw, co-prime is sufficient but not necessary)

  • \(M_C\) and \(N_R\) must be co-prime

  • Samples must include integer # of cycles of input signal

An irreducible ratio ensures identical code sequences not to be repeated multiple times.

Given that \(\frac{M_C}{N_R}\) is irreducible, and \(N_R\) is a power of 2, an odd number for \(M_C\) will always produce an irreducible ratio

Assuming there is a common factor \(k\) between \(M_C\) and \(N_R\), i.e. \(\frac{M_C}{N_R}=\frac{k M_C'}{k N_R'}\)

The samples (\(n\in[1, N_R]\))

\[ y[n] = \sin\left( \omega_{\text{in}} \cdot t_n \right) = \sin\left( \omega_{\text{in}} \cdot n\frac{1}{f_s} \right) = \sin\left( \omega_{\text{in}} \cdot n\frac{1}{f_{\text{in}}}\frac{M_C}{N_R} \right) = \sin\left( 2\pi n\frac{M_C}{N_R} \right) \]

Then

\[ y[n+N_R'] = \sin\left( 2\pi (n+N_R')\frac{M_C}{N_R} \right) = \sin\left( 2\pi n \frac{M_C}{N_R} + 2\pi N_R'\frac{M_C}{N_R}\right) = \sin\left( 2\pi n \frac{M_C}{N_R} + 2\pi N_R'\frac{kM_C'}{kN_R'} \right) = \sin\left( 2\pi n \frac{M_C}{N_R}\right) \]

So, the samples is repeated \(\color{red}y[n] = y[n+N_R']\)

\(N_R\) & \(M_C\) irreducible ratio (mutually prime)

  • Periodic sampling points result in periodic quantization errors
  • Periodic quantization errors result in harmonic distortion

image-20250705091742434

image-20260426080431593

GCD(2048, 67)=1

image-20260426083812415

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
N = 2048;
cycles = 67;
fs = 1000;
fx = fs*cycles/N;
LSB = 2/2^10;
%generate signal, quantize (mid-tread) and take FFT
xc = cos(2*pi*fx/fs*[0:N-1]);
x = round(xc/LSB)*LSB;
s = abs(fft(x));
s = s(1:end/2)/N*2;
% calculate SNR
sigbin = 1 + cycles;
noise = [s(1:sigbin-1), s(sigbin+1:end)];
snr = 10*log10( s(sigbin)^2/sum(noise.^2) );

% frequency vector
f = [0:N/2-1]/N;
subplot(1,3,1)
stem(f, s);
 
% some FFT bins for "noise" to be exactly zero, empty plot
subplot(1,3,2)
plot(f, 20*log10(s))

% Use a small offset to avoid -Inf
subplot(1,3,3)
plot(f, 20*log10(s + 1e-6))
ylim([-120, 0])

image-20260426101717290

Periodic Quantization Noise if N and cycles are not mutually prime, i.e. cycles=64 then GCD(2048, 64)=64, then \(N_R' = 2048/64=32\), so quantization noise manifests as odd harmonics \(n/32\mid n = 2k + 1, k \in \mathbb{Z}\) because of Half Wave Symmetry

image-20260501155434170

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
N = 2048;
cycles = 64;
fs = 1000;
fx = fs*cycles/N;
LSB = 2/2^10;
%generate signal, quantize (mid-tread) and take FFT
xc = cos(2*pi*fx/fs*[0:N-1]);
x = round(xc/LSB)*LSB;

% frequency vector
f = [0:N/2-1]/N;
nq = x - xc;
NTn = N / gcd(N, cycles);

subplot(2,1,1)
plot(nq(1:NTn/2), '-s')
hold on
plot(nq(1+NTn/2:NTn), '-o')
xticks(1:1:16); grid on

sn  = abs(fft(nq));
sn = sn(1:end/2)/N*2;
subplot(2,1,2)
plot(f, 20*log10(sn))

image-20260501162130482

image-20250705092503925

image-20250705103213974

Using FFT in Cadence Spectre [https://www.eecis.udel.edu/~vsaxena/courses/ece614/f14/Homeworks/fft_calculation.pdf]

image-20260504082308071

Hideo Okawara's Mixed Signal Lecture Series, DSP-Based Testing - Fundamentals 6 - Spectrum Analysis – FFT [https://www3.advantest.com/documents/11348/8c5e06b8-85b8-407d-b253-b671ca9ac85c]

—, DSP-Based Testing - Fundamentals 7 - Coherent Condition [https://www3.advantest.com/documents/11348/7f5f00bb-f5f0-41da-b154-fc1d974ad201]

image-20260523205409666

Kwantae Kim, ELEC-E3530 [https://github.com/KwantaeKim/ELEC-E3530/blob/main/CAD8.ipynb]

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
import numpy as np
from matplotlib.ticker import EngFormatter

def coherent_sampling(fin, Fs, N):
    """
    Returns the adjusted input frequency for coherent sampling.

    Parameters:
        fin  : input frequency (Hz)
        Fs   : sample rate (Hz)
        N    : number of samples

    Returns:
        Adjusted fin (Hz) such that M/N = fin/Fs, where M is the nearest prime to N*fin/Fs.
    """
    x = int(np.ceil(N * fin / Fs))
    primes = [n for n in range(2, x*2) if all(n % i for i in range(2, n))]
    M = primes[np.argmin(np.abs(np.array(primes) - x))]
    return Fs * M / N

fmt = EngFormatter(places=6)
print(f"Adjusted f_in: {fmt(coherent_sampling(10, 100, 2**6))}Hz")

# adjusted f_in: 10.937500Hz

Midrise vs Midtread Quantizers

\(\Gamma_x\) is no-overload range

image-20250809170637486

image-20250907071636266

Top-Plate vs Bottom-Plate Sampling

[https://class.ece.iastate.edu/ee435/lectures/EE%20435%20Lect%2044%20Spring%202008.pdf]

Bottom-Plate Sampling

Sample signal at the "grounded" side of the capacitor to achieve signal independent sampling

image-20240825231816582

The capacitor voltage is defined as

\[ V_C = V_{\text{out}}-V_X \]

Before \(M_2\) turns off, \(M_1\) and \(M_2\) are both ON, so approximately

\[ V_{\text{out}}=V_{\text{in}},\qquad V_X=0 \]

and therefore

\[ V_C=V_{\text{in}} \]

When \(M_2\) turns off first, the slide writes

\[ V_C=V_{\text{in}}+\frac{\Delta Q_2}{C} \]

At this moment \(M_1\) is still ON, so the top plate is still held at

\[ V_{\text{out}}\approx V_{\text{in}} \]

Hence

\[\begin{aligned} V_X &=V_{\text{out}}-V_C\\ &=V_{\text{in}} -\left(V_{\text{in}}+\frac{\Delta Q_2}{C}\right)\\ &=\boxed{-\frac{\Delta Q_2}{C}}. \end{aligned}\]

image-20240825232007848

image-20240825232717342

image-20240825233801855

image-20240825233821389

image-20240825233859540

[https://indico.cern.ch/event/1064521/contributions/4475393/attachments/2355793/4078773/esi_sampling_and_converters2022.pdf]

EE 435 Spring 2024 Analog VLSI Circuit Design - Switched-Capacitor Amplifiers Other Integrated Filters, https://class.ece.iastate.edu/ee435/lectures/EE%20435%20Lect%2044%20Spring%202008.pdf

Top-Plate Sampling

TODO 📅

image-20250622235355760

Maintain constant common-mode during conversion

D. Pfaff et al., "7.3 A 224Gb/s 3pJ/b 40dB Insertion Loss Transceiver in 3nm FinFET CMOS," 2024 IEEE International Solid-State Circuits Conference (ISSCC), San Francisco, CA, USA, 2024 [https://iccircle.com/static/upload/img20240529101747.pdf]

—, "A 224Gb/s 3pJ/bit 42dB Insertion Loss Post-FEC Error Free Transceiver in 3-nm FinFET CMOS (Invited)," 2025 IEEE Custom Integrated Circuits Conference (CICC), Boston, MA, USA, 2025, pp. 1-8, doi: 10.1109/CICC63670.2025.10983461.

E. Swindlehurst et al., "An 8-bit 10-GHz 21-mW Time-Interleaved SAR ADC With Grouped DAC Capacitors and Dual-Path Bootstrapped Switch," IEEE Journal of Solid-State Circuits, vol. 56, no. 8, pp. 2347-2359, 2021, [https://sci-hub.se/10.1109/JSSC.2021.3057372]

Tracking nonidealities

Tracking Settling Accuracy

image-20260913185637548

image-20260913185723050

image-20250729005852740

Finite Acquisition Time - Consider a sinusoidal input

\[\begin{align} V_\text{in}(t)=\cos(\omega t+\theta) & \overset{\mathcal{L}}{\Rightarrow} \frac{s\cos \theta-\omega \sin \theta}{s^2+\omega^2} \\ h(t) & \overset{\mathcal{L}}{\Rightarrow} \frac{\frac{1}{\tau}}{s+\frac{1}{\tau}} \end{align}\]

Then,

\[\begin{align} V_\text{out}(s) &= V_\text{in}(s)\cdot H(s) \\ &= \frac{s\cos \theta-\omega \sin \theta}{s^2+\omega^2} \cdot \frac{\frac{1}{\tau}}{s+\frac{1}{\tau}} \\ &= \frac{A}{s+\frac{1}{\tau}} + \frac{Bs+C}{s^2+\omega^2} \end{align}\]

Obtain,

\[ A = -\frac{\cos(\theta - \phi)}{\sqrt{ \omega^2\tau ^2 +1}} \qquad\qquad B = -A \qquad\qquad C = -\frac{\omega \sin(\theta - \phi)}{\sqrt{\omega^2\tau ^2 +1}} \]

That is \[ \boxed{V_\text{out}(s) = -\frac{\cos(\theta - \phi)}{\sqrt{\omega^2\tau ^2 +1}} \frac{1}{s+\frac{1}{\tau}} + \frac{1}{\sqrt{\omega^2\tau ^2 +1}}\frac{s\cos(\theta - \phi) - \omega \sin(\theta - \phi)}{s^2+\omega^2}} \]

where \(\phi = \arctan(\omega \tau)\)

The relation \(V_\text{out}(s) = H(s)\,V_\text{in}(s)\) is the zero-state response only. The transfer function is defined under the assumption \(V_\text{out}(0^-) = 0\), so you can't recover the natural response from it. You have to go back one step, to the differential equation.

For the RC (track) network with \(\tau = RC\):

\[ \tau \dot{V}_\text{out} + V_\text{out} = V_\text{in} \]

Transform with the full derivative rule \(\mathcal{L}\{\dot{V}_\text{out}\} = s V_\text{out}(s) - V_0\), where \(V_0 \equiv V_\text{out}(0^-)\) (the capacitor voltage is continuous, so \(0^-\) and \(0^+\) agree): \[ \tau\big(s V_\text{out}(s) - V_0\big) + V_\text{out}(s) = V_\text{in}(s) \]

\[ \boxed{\;V_\text{out}(s) = \underbrace{\frac{V_\text{in}(s)}{\tau s + 1}}_{\text{zero-state}} + \underbrace{\frac{\tau V_0}{\tau s + 1}}_{\text{zero-input}} = H(s)\,V_\text{in}(s) + \frac{V_0}{s + \frac{1}{\tau}}\;} \]

Then the Complete Laplace Expression is \[ \boxed{V_\text{out}(s) = -\frac{\cos(\theta - \phi)}{\sqrt{\omega^2\tau ^2 +1}} \frac{1}{s+\frac{1}{\tau}} + \frac{1}{\sqrt{\omega^2\tau ^2 +1}}\frac{s\cos(\theta - \phi) - \omega \sin(\theta - \phi)}{s^2+\omega^2} + \frac{V_\text{out}(0)}{s + \frac{1}{\tau}} } \] The Total Time-Domain Solution \[ \boxed{V_{\text{out}}(t) = \underbrace{V_{\text{out}}(0)e^{-\frac{t}{\tau}}}_{\text{Pure Natural Response}} + \underbrace{\frac{1}{\sqrt{\omega^2\tau^2 + 1}} \left[ \cos(\omega t + \theta - \phi) - \cos(\theta - \phi) e^{-\frac{t}{\tau}} \right]}_{\text{Forced Response (Transient + Steady-State)}} } \] And grouped by transient vs. steady-state components: \[ \boxed{V_{\text{out}}(t) = \underbrace{\left[ V_{\text{out}}(0) - \frac{\cos(\theta - \phi)}{\sqrt{\omega^2\tau^2 + 1}} \right] e^{-\frac{t}{\tau}}}_{\text{Total Transient Response}} + \underbrace{\frac{\cos(\omega t + \theta - \phi)}{\sqrt{\omega^2\tau^2 + 1}}}_{\text{Steady-State Response}}} \]

image-20260913181152511

image-20260913185936010

image-20260913190230077

Tracking Nonlinearity

Wei Yu, Subhajit Sen and B. H. Leung, "Distortion analysis of MOS track-and-hold sampling mixers using time-varying Volterra series," in IEEE Transactions on Circuits and Systems II: Analog and Digital Signal Processing, vol. 46, no. 2, pp. 101-113, Feb. 1999 [https://sci-hub.ru/10.1109/82.752910]

B. Murmann, EE315B VLSI-Data-Conversion-Circuits [https://dl.sabzdanesh.com/Electronic/VLSI-Data-Conversion-Circuits_(WWW.SabzElco.IR).pdf]

\[\begin{align} HD_2 &\approx \frac{1}{2}\omega C V_m \frac{\mathcal{d} R_{on}}{\mathcal{d}v_{in}}\bigg\rvert_{v_{in}=0}=\frac{1}{2}\omega C V_m\frac{R_{ON}}{V_{OV}} = \frac{1}{2}\frac{\omega}{\omega_c}\frac{V_m}{V_{OV}} \\ HD_3 &\approx \frac{1}{4}\omega C V_m^2 \frac{\mathcal{d}^2 R_{on}}{\mathcal{d}v_{in}^2}\bigg\rvert_{v_{in}=0}=\frac{1}{4}\omega C V_m^2\frac{R_{ON}}{V_{OV}^2} = \frac{1}{2}\frac{\omega}{\omega_c}\left(\frac{V_m}{V_{OV}}\right)^2 \end{align}\]

where \(V_{OV}=V_{DD}-V_{IN}-V_t\), \(R_{ON}=\frac{1}{\mu C_{ox}W/LV_{OV}}\) and \(\omega_c=1/R_{ON}C\)

Even-order distortion products cancel in perfectly symmetrical circuits; But phase and amplitude imbalances lead to finite \(HD_2\) in practice

Pipeline ADC

Vishal Saxena, "Pipelined ADC Design - A Tutorial"[https://www.eecis.udel.edu/~vsaxena/courses/ece517/s17/Lecture%20Notes/Pipelined%20ADC%20NonIdealities%20Slides%20v1_0.pdf] [https://www.eecis.udel.edu/~vsaxena/courses/ece517/s17/Lecture%20Notes/Pipelined%20ADC%20Slides%20v1_2.pdf]

Bibhu Datta Sahoo, Analog-to-Digital Converter Design From System Architecture to Transistor-level [http://smdpc2sd.gov.in/downloads/IGF/IGF%201/Analog%20to%20Digital%20Converter%20Design.pdf]

Bibhu Datta Sahoo, Associate Professor, IIT, Kharagpur, [https://youtu.be/HiIWEBAYRJY]

image-20241006174924686

CMP reference voltage is 0.5vref, DAC output is 0.5vref or 0

pipelineADC.drawio

residual error \[ V_{r,n} = (V_{r,n-1}-\frac{1}{2}b_{n})\cdot 2 \] and \(V_{r,-1}=V_i\) \[ V_{r,n-1} = 2^{n}V_i -\sum_{k=0}^{n-1}2^{n-k-1}b_k = 2^{n}\left(V_i - \sum_{k=0}^{n-1}\frac{b_k}{2^{k+1}}\right) \]

here, \(b_0\) is first stage and MSB

It divides the process into several comparison stages, the number of which is proportional to the number of bits

Due to the pipeline structure of both analog and digital signal path, inter-stage residue amplification is needed which consumes considerable power and limits high speed operation

image-20241214164740706

image-20260915220254339

Non-Flip-Around & Flip-Around Amplifier

Non-Flip-Around Amplifier

image-20260929075323994

The amplifier's differential input stays approximately zero after settling; its individual input voltages need not be constant

With matched capacitors, their common-mode contributions cancel when we subtract the two charge-conservation equations

Flip-Around Amplifier

image-20260929075641833

Multiplying DACs (MDAC)

image-20260915230732911

Connection Sampling phase Amplification phase
(V_{}) to left plate of (C_S) Closed Open
Left plate of (C_S) to ground Open Closed
Node (x) to ground Closed Open
Output to ground Closed Open

Phase 1: sample the input

  • \(C_S\) has \(V_{\text{in}}\) on its left plate and 0 V on its right plate, so it stores the input as charge.
  • Both ends of \(C_F\) are grounded, resetting its voltage to zero.
  • The small ADC samples the same input and determines a digital code.

At the end of sampling:

\[ Q_{x,\text{sample}}=-C_SV_{\text{in}} \]

Phase 2: subtract and amplify

  • Switching (C_S)’s left plate from (V_{}) to zero disturbs node (x).
  • Switching the DAC bottom plate from zero to (V_D) produces an opposing disturbance.
  • The amplifier changes (V_{}), through (C_F), to balance the remaining charge.

After amplification settles, both ends of \(C_S\) are approximately zero:

\[ Q_{x,\text{amp}}=-C_FV_{\text{out}}-C_{\text{DAC}}V_D \]

Equating them:

\[ \boxed{ V_{\text{out}} = \frac{C_S}{C_F}V_{\text{in}} - \frac{C_{\text{DAC}}}{C_F}V_D } \]

If \(C_S=C_{\text{DAC}}\), this becomes:

\[ \boxed{V_{\text{out}}=G(V_{\text{in}}-V_D)}, \qquad G=\frac{C_S}{C_F} \]

Thermometer to Binary encoder

image-20241214152349217

R-2R & C-2C

B. Razavi, "The R-2R and C-2C Ladders [A Circuit for All Seasons]," in IEEE Solid-State Circuits Magazine, vol. 11, no. 3, pp. 10-15, Summer 2019 [https://www.seas.ucla.edu/brweb/papers/Journals/BR_SSCM_3_2019.pdf]

\(N_b\) bit binary + \(N_t\) bit thermometer DAC

R-2R.drawio

\(N_b\) bit binary can be simplified with Thevenin Equivalent \[ V_B = \sum_{n=0}^{N_b-1} \frac{B_n}{2^{N_b-n}} \] with thermometer code

\[ V_o = V_B\frac{\frac{2R}{2^{N_t}-1}}{\frac{2R}{2^{N_t}-1}+ 2R}+\sum_{n=0}^{2^{N_t}-2}T_n\frac{\frac{2R}{2^{N_t}-1}}{\frac{2R}{2^{N_t}-1}+ 2R} = \frac{V_B}{2^{N_t}} + \frac{\sum_{n=0}^{2^{N_t}-2}T_n}{2^{N_t}} = \sum_{n=0}^{N_b-1} \frac{B_n}{2^{N_t+N_b-n}} + \frac{\sum_{n=0}^{2^{N_t}-2}T_n}{2^{N_t}} \]

4bit binary R2R DAC with Ru=1kOhm; RVB equivalent R

image-20241214190045688

reference

Maloberti, F. Data Converters. Dordrecht, Netherlands: Springer, 2007.

Ali, Ahmed M. A. High Speed Data Converters. The Institution of Engineering and Technology, 2016.

Razavi B. Analysis and Design of Data Converters. Cambridge University Press; 2025.

Aaron Buchwald, ISSCC2010 T1: "Specifying & Testing ADCs"

Ahmed M. A. Ali. CICC 2018: High Speed Pipelined ADCs: Fundamentals and Variants

John P. Keane, ISSCC2020 T5: "Fundamentals of Time-Interleaved ADCs"

Yun Chiu, ISSCC2023 T3: "Fundamentals of Data Converters" [https://personal.utdallas.edu/~yxc101000/courses/7327/handout/isscc2023_tutorial.pdf]

—， "Design and Calibration Techniques for SAR and Pipeline ADCs" [http://formation-old.in2p3.fr/microelectronique15/IN2P3_ADC.pdf]

—， Radiation-Tolerant SAR ADC Architecture and Digital Calibration Techniques [https://indico.cern.ch/event/385097/attachments/768706/1054353/CERN_May15.pdf]

—， Recent Advances in Multistep Nyquist ADC's [https://www.eecis.udel.edu/~vsaxena/courses/ece614/Handouts/Recent%20Advances%20in%20Nyquist%20rate%20ADCs.pdf]

Aaron Buchwald, ISSCC 2008 T2 Pipelined A/D Converters: The Basics

Yohan Frans, CICC2019 ES3-3- "ADC-based Wireline Transceivers" [pdf]

Samuel Palermo, ISSCC 2018 T10: ADC-Based Serial Links: Design and Analysis

Jan Mulder Broadcom. ISSCC2015 T5: High-Speed Current-Steering DACs

Zhang, Milin, Zhihua Wang, Jan van der Spiegel and Franco Maloberti. "Advanced Tutorial on Analog Circuit Design." (2023)

V. Chen, "Tutorial: High-Speed Analog-to-Digital Converters," 2025 IEEE International Solid-State Circuits Conference (ISSCC), San Francisco, CA, USA, 2025, pp. 1-1, doi: 10.1109/ISSCC49661.2025.11076112.

S. Su, "Principles and Practices of High-Speed DAC Design: From Conversion Fundamentals to Layout-Aware Implementation," in IEEE Solid-State Circuits Magazine, vol. 18, no. 3, pp. 26-43, Summer 2026, doi: 10.1109/MSSC.2026.3704105

M. Gu, Y. Tao, Y. Zhong, L. Jie and N. Sun, "Timing-Skew Calibration Techniques in Time-Interleaved ADCs," in IEEE Open Journal of the Solid-State Circuits Society [https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10804623]

everynanocounts. Memos on FFT With Windowing. URL: https://a2d2ic.wordpress.com/2018/02/01/memos-on-fft-with-windowing/

How to choose FFT depth for ADC performance analysis (SINAD, ENOB). URL:https://dsp.stackexchange.com/a/38201

Computation of Effective Number of Bits, Signal to Noise Ratio, & Signal to Noise & Distortion Ratio Using FFT. URL:https://cdn.teledynelecroy.com/files/appnotes/computation_of_effective_no_bits.pdf

Kester, Walt. (2009). Understand SINAD, ENOB, SNR, THD, THD + N, and SFDR so You Don't Get Lost in the Noise Floor. URL:https://www.analog.com/media/en/training-seminars/tutorials/MT-003.pdf

T. C. Hofner: Dynamic ADC testing part I. Defining and testing dynamic ADC parameters, Microwaves & RF, 2000, vol. 39, no. 11, pp. 75-84,162

—: Dynamic ADC testing part 2. Measuring and evaluating dynamic line parameters, Microwaves & RF, 2000, vol. 39, no. 13, pp. 78-94

AN9675: A Tutorial in Coherent and Windowed Sampling with A/D Converters https://www.renesas.com/us/en/document/apn/an9675-tutorial-coherent-and-windowed-sampling-ad-converters

APPLICATION NOTE 3190: Coherent Sampling Calculator (CSC) https://www.stg-maximintegrated.com/en/design/technical-documents/app-notes/3/3190.html

Coherent Sampling (Very Brief and Simple) https://www.dsprelated.com/thread/469/coherent-sampling-very-brief-and-simple

Signal Chain Basics #160: Making sense of coherent and noncoherent sampling in data-converter testing https://www.planetanalog.com/signal-chain-basics-160-making-sense-of-coherent-and-noncoherent-sampling-in-data-converter-testing/

Signal Chain Basics #104: Understanding noise in ADCs https://www.planetanalog.com/signal-chain-basics-part-104-understanding-noise-in-adcs/

Signal Chain Basics #101: ENOB Degradation Analysis Over Frequency Due to Jitter https://www.planetanalog.com/signal-chain-basics-part-101-enob-degradation-analysis-over-frequency-due-to-jitter/

Clock jitter analyzed in the time domain, Part 1, Texas Instruments Analog Applications Journal (slyt379), Aug 2010 https://www.ti.com/lit/an/slyt379/slyt379.pdf

Clock jitter analyzed in the time domain, Part 2 https://www.ti.com/lit/slyt389

Measurement of Total Harmonic Distortion and Its Related Parameters using Multi-Instrument [pdf]

Application Note AN-4: Understanding Data Converters' Frequency Domain Specifications [pdf]

Belleman, J. (2008). From analog to digital. 10.5170/CERN-2008-003.131. [pdf]

HandWiki. Coherent sampling [link]

Luis Chioye, TI. Leverage coherent sampling and FFT windows when evaluating SAR ADCs (Part 1) [link]

Coherent Sampling vs. Window Sampling | Analog Devices https://www.analog.com/en/technical-articles/coherent-sampling-vs-window-sampling.html

Understanding Effective Number of Bits https://robustcircuitdesign.com/signal-chain-explorer/understanding-effective-number-of-bits/

ADC Input Noise: The Good, The Bad, and The Ugly. Is No Noise Good Noise? [https://www.analog.com/en/resources/analog-dialogue/articles/adc-input-noise.html]

Walt Kester, Taking the Mystery out of the Infamous Formula, "SNR = 6.02N + 1.76dB," and Why You Should Care [https://www.analog.com/media/en/training-seminars/tutorials/MT-001.pdf]

Dan Boschen, "How to choose FFT depth for ADC performance analysis (SINAD, ENOB)", [https://dsp.stackexchange.com/a/38201]

B. Razavi, "A Tale of Two ADCs - Pipelined Versus SAR" IEEE Solid-State Circuits Magazine, Volume. 7, Issue. 30, pp. 38-46, Summer 2015 [https://www.seas.ucla.edu/brweb/papers/Journals/BRSummer15ADC.pdf)]

Dr. Tai-Haur Kuo (郭泰豪 教授) Analog IC Design (類比積體電路設計) [http://msic.ee.ncku.edu.tw/course/aic/aic.html]

Converter Passion for data-converter professionals sharing thoughts on ADCs and DACs [https://converterpassion.wordpress.com/]

Boris Murmann, EE315B VLSI Data Conversion Circuits, Autumn 2013

MPScholar Analog-to-Digital Converters (ADCs) [https://www.monolithicpower.com/en/learning/mpscholar/analog-to-digital-converters]

tomverbeure. List of Analog Devices Tutorials [https://tomverbeure.github.io/2021/02/15/Analog-Devices-Tutorials.html]