Kang Min-Seong1
Yoon Kwang Sub1
-
(Inha Univ., Incheon, Korea)
Copyright © The Institute of Electronics and Information Engineers(IEIE)
Index Terms
Reconfigurable, hybrid, SAR, single slope, ADC
I. INTRODUCTION
Nanometer-order Complementary Metal-Oxide Semiconductor (CMOS) technology has driven
the rapid development of mobile and wearable devices, such as smart phones and smart
watches. Concurrent wearable devices have been used to monitor bio-signals for healthcare
applications. Such wearable devices employ low-power analog-to-digital converters
(ADCs) to interface the analog bio-signals from users with the digital signals required
by micro-processing units (MPUs). These devices typically require low-power high-precision
ADCs and multi-signal processing capabilities.
Recently, wearable devices have been actively studied to implement low-power and high-resolution
hybrid ADCs [1-8]. Various architectures have been proposed in literature for low-power high-resolution
hybrid ADCs, including flash-SAR, pipelined SAR, noise-shaping SAR, SAR-SS, and SAR-dual-slope
(DS).
The power consumption and sampling rates of the SAR-SS and noise-shaping SAR architectures
are lower than those of pipelined SAR and flash-SARs. The flash-SAR [1] and pipelined SAR [2] architectures offer high-speed conversion time and high resolution but suffer from
high power consumption owing to the inherent natures of the flash and pipeline structures.
However, noise-shaping SAR [3] and SAR-SS [4-6] architectures offer low-power and high-resolution characteristics with low-speed
conversion time.
The noise-shaping SAR ADC in [3] showed ENOB of 14.2-bit and power consumption of 120 $\mu$W. The SS SAR hybrid ADC
in [4] was shown to have a power consumption of 56 ${\mu}$W because of the ramp signal generator
for the SS ADC to drive the capacitor digital-to-analog converter (C-DAC) inside the
SAR ADC. In [5], the time-interleaved SAR-DS ADC was used to enhance signal processing speed and
resolution, with a large power dissipation of 350 ${\mu}$W. In [6], the SS-SAR hybrid ADC determined MSB and LSB from the SS and SAR, respectively.
This resulted in a poor linearity of +1.65/-1.45 LSBs. In [7], the SS ADC demonstrated better linearity than the SAR ADC. Consequently, the placement
of the SAR structure for MSB and SS structure for LSB is expected to achieve low-power
high-resolution hybrid ADC with good linearity.
This paper presents an architecture comprising a hybrid SAR and SS ADC with reconfigurability
driven by a digital control circuit within the SS block. This allows the resolution
and conversion speed of the proposed hybrid ADC to be reconfigured depending on the
external resolution select signal. The proposed hybrid architecture shares analog
blocks to minimize power consumption between the two segmented structures (SAR and
SS). Section II describes the proposed hybrid architecture and algorithm. Section
III presents the measurement results and comparisons of the measured performances
with those from conventional architectures. Conclusions are drawn in Section IV.
II. PROPOSED ARCHITECTURE
The proposed reconfigurable SAR-SS ADC architecture includes a 14-bit C-DAC with a
split capacitor (256/255 Cu), a latched comparator with a preamplifier, an 8-bit SAR
logic, an SS block with bit select (BS) signal to reconfigure the resolution, a capture
cell to store the y-bit signal from the y-bit binary counter, and an m-bit output
register, as shown in Fig. 1.
In the conventional SAR ADC, the SAR logic generates all the input signals to the
C-DAC. In the proposed design, the 8-bit SAR logic is connected to the 8-bit MSB of
the C-DAC, and the y-bit binary counter is connected to the remaining 8-bit LSB of
the C-DAC. Once the SAR logic is activated, the binary counter increments the output
voltage of the C-DAC by 1 LSB per clock to generate a digital ramp starting from the
residue voltage of the SAR ADC segments. The final SAR logic output bits are not reset
to zero until the binary counter operation ends. The output data of the binary counter
are captured when the digital ramp of the C-DAC crosses the common-mode voltage, and
the output of the comparator is inverted with respect to its previous digital state.
The 8-bit SAR logic and y-bit digital counter codes are produced simultaneously once
the conversion process ends.
The block diagram presented in Fig. 2 represents the SS block of Fig. 1.
The SS block consists of the y-bit reconfigurable binary counter, Done generator,
SS control circuit, and 10-clock counter with start-up circuit. After the 10-clock
counter generates a pulse signal, the STATE signal from the SS control circuit enables
the binary counter to activate the counter signal. The Done generator comprises a
simple combination logic with a single D-flip flop that produces the Done signal at
the end of the binary counter. The Boolean expression of the Done generator can be
written as (1).
, where $B_{i}$ is the $i^{th}$ binary code of the binary counter.
The block diagram of the reconfigurable binary counter presented in Fig. 3 includes the binary counter and a decoder circuit. The decoder is capable of generating
three RES signals ($RES_{0}$, $RES_{1}$, $RES_{2}$) to control six output signals
($B_{0}$, $B_{1}$, $B_{2}$, $B_{3}$, $B_{4}$, $B_{5}$). $RES\left[2\colon 0\right]$
can be described as
The remaining signals ($RES_{3}$, $RES_{4}$, $RES_{5}$) are unchanged at 1 since the
upper three bits of the binary counter are not reconfigured. The decoder circuit is
implemented as per the truth table given in Table 1.
The resolution of the reconfigurable binary counter can be determined by the state
of the two-bit BS signal, as given in Table 2. The states 00, 01, 10, and 11 of the BS signals configure the binary counter resolution
to 3-bit, 4-bit, 5-bit, or 6-bit.
As the binary counter activation ends, the bit-detection circuit produces the Done
signal to reset the overall circuit, including the 10-clock counter with start-up
circuit. The preamplifier is implemented to prevent kick-back noise from the latched
comparator along with enhancement of the dynamic performance and linearity. The timing
diagram of the proposed ADC is presented in Fig. 4.
As the Done signal becomes high, the reset generator becomes zero, C-DAC is fully
discharged, and SAR logic commences operation. The BS signal determines the resolution
(3–6 bit) of the binary counter in the SS ADC. As the 10-clock counter produces the
SAR_Done signal, the 8-bit MSB of the C-DAC is determined and the STATE signal becomes
high. Once the STATE signal goes high, the reconfigurable binary counter is changed
according to the digital state of the BS signal. The binary counter in the SS ADC
is connected to the LSB of the C-DAC, which increases the output voltage of the C-DAC
during each clock to generate the digital ramp signal. If the binary counter becomes
full, the STATE signal becomes low and Done signal becomes high to enable the 10-clock
counter.
Table 1. Truth table of the decoder circuit
$BS_{1}$
|
$BS_{0}$
|
$RES_{5}$
|
$RES_{4}$
|
$RES_{3}$
|
$RES_{2}$
|
$RES_{1}$
|
$RES_{0}$
|
0
|
0
|
1
|
1
|
1
|
0
|
0
|
0
|
0
|
1
|
1
|
1
|
1
|
1
|
0
|
0
|
1
|
0
|
1
|
1
|
1
|
1
|
1
|
0
|
1
|
1
|
1
|
1
|
1
|
1
|
1
|
1
|
Table 2. Relationship between the BS signal state and resolution of the reconfigurable binary counter
State of the BS signal
|
Resolution of binary counter
|
00
|
3-bit
|
01
|
4-bit
|
10
|
5-bit
|
11
|
6-bit
|
Fig. 1. Block diagram of the proposed m-bit reconfigurable SAR-SS ADC.
Fig. 2. Block diagram of the SS block.
Fig. 3. Block diagram of the reconfigurable binary counter.
Fig. 4. Timing diagram of the proposed ADC.
III. MEASUREMENT RESULTS
The chip photograph of the fabricated chip for the proposed ADC is shown in Fig. 5. The ADC is implemented in a standard CMOS 28-nm n-well 1-poly 8-metal process. The
active chip area occupied $700\mu m\times 500\mu m$, excluding the bonding pads. The
14-bit C-DAC is placed atop the chip with a double guard for protection from the digital
signals, and the latched comparator with preamplifier is placed adjacent to the C-DAC.
The SS block, C-DAC control SAR logic circuit, y-bit binary counter, 8-bit SAR logic,
and output register are placed at the bottom of the layout.
The photograph presented in Fig. 6 is the printed circuit board (PCB) with the proposed ADC. The PCB encloses two SMA
connectors for the input signal (VIN) and clock signal (CLK), with analog power supply
(AVDD, AVSS) and digital power supply (DVDD, DVSS). Two pins to apply the 2-bit BS
signal and two test pins to measure the output of the comparator and SS block are
placed at the center and on the edge of the right-hand side of the PCB, respectively.
The 14-bit digital output pins are available at the bottom of the PCB.
The measured waveforms of the 2-bit BS (11, 10, 01, 00) and Done signals are presented
in Fig. 7. The period time $t_{sample}$ of the Done signal associated with the 2-bit BS signal
is designed to decrease approximately 1.5 times per bit.
The measured waveforms of the Done and comparator signals from the two test pins in
Fig. 6 are presented in Fig. 8, as shown in Fig. 1. The rising edge of the Capture signal enables the capture cell to save the output
data of the binary counter. These results verify the basic functionality and reconfigurability
of the SS block and overall circuit.
The linearity (differential nonlinearity (DNL) and integral nonlinearity (INL)) of
the reconfigurable ADC associated with the 14-bit mode (8-bit of SAR and 6-bit of
SS) are measured as -0.96/+0.90 LSB and -0.73/+0.92 LSB, respectively. The analog
and digital power consumption values of the ADC (14-bit mode) shown are measured to
be 4.23 ${\mu}$W and 10.27 ${\mu}$W, respectively. As the proposed ADC is reconfigured
by the external 2-bit BS signal to change a single bit, the digital power consumption
of a single J/K flip flop associated with the bit is only increased by a negligible
amount (of the order of a few nanowatt) compared to the total power consumption of
14.5 ${\mu}$W.
The measured Fast Fourier Transform (FFT) results are presented in Fig. 9 in order of the resolution (11–14 bit). The input and clock frequency are 1 kHz and
2 MHz, respectively. The 14-bit mode shows a signal to noise and distortion ratio
(SNDR) of 76.5 dB and ENOB of 12.4 bit.
Fig. 10 illustrates the measured ENOB of 12.4 to 6.2-bit. The clock frequency varies from
100 kHz to 20 MHz with the input frequency fixed to 1 kHz. The measured ENOB stays
close to 12.4-bit up to a clock frequency of 2 MHz and degrades thereafter. As the
input frequency changes from 0.1 to 20 kHz with the clock frequency fixed to 26.7
kHz, the measured ENOB remains at almost 12.1-bit up to an input frequency of 2 kHz
and decreases thereafter, as shown in Fig. 11.
A comparison of the static/dynamic performance parameters of the proposed hybrid ADC
with those of the conventional ADC is made in Table 3. The proposed hybrid ADC shows a relatively low power consumption and comparable
ENOBs with respect to the other hybrid devices [3-5]. The FoM of the proposed ADC is higher than those of the others owing to the low
sampling rate. As observed, the conventional devices do not provide reconfigurability.
On the other hand, the proposed ADC is capable of offering reconfigurable resolution
(11–14 bit), so that it is applicable to several bio-signal processing cases.
Fig. 5. Chip photograph of the proposed reconfigurable ADC.
Fig. 6. Photograph of the proposed ADC on a PCB.
Fig. 7. Waveforms of the 2-bit BS (11, 10, 01, 00) and Done signals.
Fig. 8. Waveforms of Done and Capture signals (x=8, y=3).
Fig. 9. Measured FFT result for an input frequency of 1 kHz: (a) 11-bit, fs = 105.3 kS/s; (b) 12-bit, fs = 74.1 kS/s; (c) 13-bit, fs = 46.5 kS/s; (d)14-bit, fs = 26.7 kS/s.
Fig. 10. Plot of ENOB as a function of clock frequency ($f_{in}=1kHz$).
Fig. 11. Plot of ENOB as a function of the input frequency ($f_{sample}=26.7kS/s$).
Table 3. Comparison of performances of the proposed and conventional ADCs
Parameter
|
[3]
|
[4]
|
[5]
|
[6]
|
[7]
|
[8]
|
This work
|
Process
|
28 nm
|
90 nm
|
28 nm
|
180 nm
|
90 nm
|
90 nm
|
28 nm
|
Architecture
|
NS SAR
|
SAR – SS
|
SAR – SS
|
SS – SAR
|
SAR
|
LC
|
SAR – SS
|
Supply voltage [V]
|
1.0
|
1.2
|
0.9
|
3.3
|
1.0
|
0.8
|
0.8
|
Resolution [bit]
|
16
|
12
|
12
|
11
|
10
|
10
|
11
|
12
|
13
|
14
|
ENOB [bit]
|
14.26
|
N/A
|
10.41
|
9.39
|
9.0
|
6.8
|
10.24
|
11.17
|
11.85
|
12.42
|
Sampling rate [S/s]
|
$2M$
|
$370k$
|
$100M$
|
83k
|
500
|
1k
|
105.3k
|
74.1k
|
46.5k
|
26.7k
|
DNL [LSB]
|
N/A
|
-0.45
/0.84
|
-0.53
/0.53
|
-1.45
/1.65
|
N/A
|
N/A
|
-0.87
/0.83
|
-0.86
/0.87
|
-0.90
/0.88
|
-0.96
/0.90
|
INL [LSB]
|
N/A
|
-1.5
/0.74
|
-0.82
/0.79
|
N/A
|
N/A
|
N/A
|
-0.69
/0.88
|
-0.73
/0.86
|
-0.77
/0.91
|
-0.73
/0.92
|
Power [W]
|
$120\mu $
|
$56\mu $**
|
$350\mu $
|
$7\mu $
|
$5.8\mu $
|
0.18$\mu $
|
14.1$\mu $
|
14.2$\mu $
|
14.4$\mu $
|
14.5$\mu $
|
FoM* [fJ/step]
|
29.6
|
36.9
|
2.2
|
37.8
|
34
|
656
|
110.7
|
83.2
|
83.9
|
99.1
|
*FoM = [Power] /( 2^[ENOB]${\times}$ [Sampling rate])
**Power dissipation of [4] is divided by the number of channels
IV. CONCLUSION
This paper presents a CMOS hybrid ADC with high resolution and reconfigurability for
bio-signal processing. The proposed hybrid ADC consists of two segmented architectures,
namely the SAR and SS architectures that are associated with the MSB and LSB, respectively.
The SS block includes a reset generator, control circuitry, reconfigurable binary
counter, and Done signal generator to enable reconfigurable resolution (11–14 bit)
for the proposed ADC. The reconfigurable binary counter driven by an external two-bit
signal can change the resolution of the proposed ADC, which is implemented on a standard
CMOS 28-nm 1-poly 8-metal process. The active layout area occupied 700 ${\mathrm{\mu}}$m
${\times}$ 500 ${\mathrm{\mu}}$m, excluding the bonding pads. The measurement results
demonstrated a power consumption of 14.5 ${\mu}$W (analog and digital power of 4.2~${\mathrm{\mu}}$W
and 10.3 ${\mathrm{\mu}}$W, respectively), ENOB of 12.42 bit, DNL/INL of $\pm $0.96
LSB and $\pm $0.92 LSB, and FoM of 99.1 fJ/step.
The proposed hybrid ADC has the advantage of not only a reconfigurable architecture
but also low power consumption owing to the segmented architecture compared to those
of conventional hybrid ADCs. This ADC is expected to be employed in wearable devices
for bio-signal processing and various IoT applications because of the reconfigurable
resolution capability and low power dissipation.
ACKNOWLEDGMENTS
This work was supported by Inha Research Grant. MPW and Cadence design tools were
supported by IDEC.
References
D. R. Oh, et al, “An 8-Bit 1-GS/s Asynchronous Loop-Unrolled SAR-Flash ADC With Com-plementary
Dynamic Amplifiers in 28-nm CMOS,” Solid-State Circuits, IEEE Journal of, Vol. 56,
No. 4, pp. 1216-1226, April 2021.
T. Ramkai, et al, “A 5-GS/s 158.6-mW 9.4-ENOB Passive-Sampling Time-Interleaved Three-Stage
Pipelined-SAR ADC With Analog–Digital Corrections in 28-nm CMOS,” Solid-State Circuits,
IEEE Journal of, Vol. 55, No. 6, pp. 1553-1564, April 2020.
L. Jie, et al, “A Cascaded Noise-Shaping SAR Architecture for Robust Order Extension,”
Solid-State Circuits, IEEE Journal of, Vol. 55, No. 12, pp. 3236-3247, Dec. 2020.
M. Kim, et al, “An Area-Efficient and Low-Power 12-b SAR/Single-Slope ADC Without
Calibration Method for CMOS Image Sensors,” IEEE Trans. Electron Devices, Vol. 49,
No. 5, pp. 3599-3604, Sept. 2014.
C. Liu, et al, “A 12 pJ/Pixel Analog-to-Information Converter Based 816 × 640 Pixel
CMOS Image Sensor,” Solid-State Circuits, IEEE Journal of, Vol. 56, No. 4, pp. 2941-2950,
April 2014.
F. Tang, et al, “Low-Power CMOS Image Sensor Based on Column-Parallel Single-Slope/SAR
Quantization Scheme,” IEEE Trans. Electron Devices, Vol. 60, No. 8, pp. 3599-3604,
August 2013.
R. Inanlou, et al, “Arithmetic Tracking Adaptive SAR ADC for Signals with Low-Activity
Periods,” IEEE Access, Vol. 8, pp. 211621-211629, 2020.
B. Yazdani, et al, “A Low Power Fully Differential Level-Crossing ADC with Low Power
Charge Redistribution Input for Biomedical Applications,” IEEE Trans. Circuits and
Systems II: Express Briefs, Vol. 69, No. 3, pp. 864-868, March 2022.
Author
Min-Seong Kang received the Bachelor and Master degree in department of electronics
engi-neering, Inha University, Korea, in 2020 and 2022, respectively. As he joined
Pixel Plus Inc. in 2023, he has involved CIS design project. His research interest
includes design of a low power high resolution SAR ADC.
Kwang S. Yoon received BS degree in EE from Inha University in 1981, MS, and Ph.D.
degrees in EE from Georgia Institute of Technology in 1983 and 1990, respectively.
He worked at Silicon Systems Inc. (Tustin, Calif.) as a senior design engineer for
1988-1992. Since 1992, he joined the department of Electronic Engineering at Inha
University, Korea. He established IEEE ED/SSC joint chapter in 1998 and served as
chairman until 2016. He served as TPC chairman and general chairman of ISOCC (International
SoC Conference) for 2012 and 2013, respectively. He also served as associate editor
of JSTS (Journal of Semiconductor Technology and Science) and associate editor of
JICAS, IDEC. His research interests include mixed-signal CMOS circuit design such
as high performance data converters (high speed (flash-pipeline, time-interleaving),
high resolutions (sigma-delta), low power (SAR) ADCs, hybrid ADCs, and DACs), PLLs,
PMICs (Buck (PFM-PWM)/Boost) for IoTs, Smart sensor systems, and Bio-signal processing.