Professional Documents
Culture Documents
A 12bit 250MSPS Pipeline ADC With 4 Gbps Serial Output Interface
A 12bit 250MSPS Pipeline ADC With 4 Gbps Serial Output Interface
https://doi.org/10.1007/s10470-019-01389-y (0123456789().,-volV)
(0123456789().,-volV)
Abstract
A 12bit 250MSPS pipeline analog-to-digital converter (ADC) with serial output interface is presented. The pseudo random
digital calibration dithered sub-ADC in first stage is used to lower non-ideal errors and improve the dynamic performance
in the high speed ADC. An integrated serial output interface is implemented to convert 12bit parallel data into a differential
serial data stream. The pipeline ADC was fabricated with CMOS 180 nm 1.8 V 1P5 M process. The active ADC with the
serial output interface consumes a power consumption of 395 mW and occupies an area of 8.0 mm2, where the active area
of the interface is 0.75 mm2. The measurement results show that the differential non-linearity and integral non-linearity of
the proposed ADC are - 0.22/? 0.16LSB and - 0.4/? 0.6LSB, respectively. The spurious free dynamic range and signal-
to-noise ratio can get 81.17 dB and 69.92 dB with 20 MHz input signal at full sampling speed. The serial output interface
provides an eye height greater than 800 mV for data rates of 4 GHz bits per second with a power of 75 mW.
123
Analog Integrated Circuits and Signal Processing
16 6 6 6 7
Voltage
Reference Delay Array and Digital Correction Logic
Generator
D[0:11]
250MHz
CLKP 2GHz 2GHz 4Gbps Serial Output Interface
Receiver PLL
ADC. Besides, differential clock receiver was used to gain and high bandwidth. The Fig. 2(b) gives the structure
minimize the clock jitter noise. The pipelined stage circuits of the 4.5-bit dithered sub-ADC. There are 16 dynamic
are implemented in switched capacitor scheme. The 41bits comparators used in this sub-ADC. The dynamic com-
thermometer code from the five pipelined sub-stages is parator is implemented by commonly used common mode
aligned by Delay Array and encoded by the Digital Cor- insensitive switched capacitor multi-stage latched com-
rection Logic and gets the 12bit parallel digital conversion parator. The left reference voltage of Refn1 to Refn16 and
code. The parallel 12bit digital output codes are transported right reference voltage of Refp1 to Refp16 are generated by
to the 4 Gbps serial output interface in order to minimize resistor array from band-gap signal. The PN1 to PN16 signal
the pin numbers and improve the digital signal speed. is generated by PN circuit. The sequence of control
As a non-overlapping multi-phase clock signal with switches is random and not repetitive. The left control
250 MHz frequency is used for the ADC core, a clock signal PN1–16 is same with right control signal. The PN
input receiver is used to receive the differential input clock, noise is injected in the voltage references by randomly
and the distributed clock generation and buffer system is varying the comparator threshold level, which is realized
selected to transmit the on-chip non-overlapping clocks. A by randomly selecting the reference control switches PN1
fully integrated 2 GHz PLL is designed for generating the to PN16. Thus, the PN signal is injected to input end of sub-
serializing clock for the 4 Gbps serial interface, without ADC by the dithering reference voltage.
any external component. In order to provide precision The block diagram of the PN calibration algorithm in
voltage reference with fast settling, an on-chip precision sub-ADC is illustrated in Fig. 3. The PN signal is injected
bandgap reference voltage generator and buffer amplifiers into the input end of the sub ADC, which is actually
are also integrated on the chip. equivalent to adjusting the threshold voltage of the com-
parator according to the PN sequence to change the output
2.1 PN calibration technique in sub-ADC encoding of the sub-ADC.
In the sub ADC, the pseudo random sequence signal can
The first PN calibrated 4.5-bit stage circuit is showed in be illustrated as PN Vcal . The PN modulation is actually
Fig. 2(a); the detailed structure of 4.5-bit dithered sub- changed the comparator threshold voltage in the sub-ADC,
ADC is illustrated in Fig. 2(b). The stage circuit is con- which makes the output signal of the corresponding stage
sisted of a 4.5-bit dithered sub-ADC, a traditional 4.5-bit change with random margin in the threshold of the com-
sub-DAC, a 4.5-bit thermometer to binary code encoder, a parator. The Output Dsub of sub-ADC is illustrated as:
MDAC with the gain of 16 and the PN generator. The 4.5- Dsub ¼ Vin =Vref þ PN Vcal =Vref þ QS ð1Þ
bit dithered sub-ADC compares the analog input signal Vi
with PN modulated voltage references and gets the 16-bit where QS is the quantization error of sub-ADC, the output
thermometer code. The 16-bit thermometer code is enco- of the back-end ADC is:
ded by the 4.5-bit encoder and gets the 5-bit binary code. Dout ¼ GðPN Vcal þ QS Vref ÞVref þ Qb ð2Þ
The MDAC used in this 4.5-bit sub-stage circuit is
implemented by traditional switched capacitor MDAC where Qb is the quantization error of the back-end ADC,
scheme, which consumes major current to achieve accurate the expectation value of D0out is:
123
Analog Integrated Circuits and Signal Processing
(b)
Average
123
Analog Integrated Circuits and Signal Processing
to the input of the sub-DAC, injecting PN at the input of telescopic amplifier and a common-source amplifier, which
the sub-ADC is much easier to achieve high speed with is shown in Fig. 4(a). The voltage gain of the designed op-
simplified system structure. First, the injected pseudo-ran- amp is:
dom noise can be compensated by the digital redundancy Av ffi gm1 f½gm3 ro3 ro1 jjðgm5 ro5 ro7 Þg gm9 ðro9 jjro11 Þ ð6Þ
of the pipeline ADC without any performance degradation.
Therefore it is unnecessary to reduce the input signal ro is the resistor of corresponding transistor, gm is the trans-
magnitude when injecting the pseudo-random noise. Sec- conductance of corresponding transistor.
ond, we do not need to face the complication of an accurate The transistor of MN2,3 and MP1–4 is the differential
scaling (about 1/4 value) in the analog domain. Third, in input and load transistor. The first and second common-
the circuit implementation, the injection of pseudo-random mode feed-back (CMFB) is used to stabilize the common-
noise can be done by randomly varying the comparator mode signal. A improved frequency compensation with the
threshold level (comparator dithering), as long as this capacitor of Cc2,3 is adopted in this opamp to improve the
variation/dithering does not exceed the bounds of digital sufficient phase margin, which is shown in Fig. 4 [9].
redundancy. Finally, MDAC do not need to design dither Because of the frequency compensation theory of Cc2
circuits, which is the most critical and sensitive block in the and Cc3, the small signal model of the op-amp with only
pipeline ADC design. Thus, we can obtain faster operation with Cc3 is shown in Fig. 5(a) for quantitative analysis.
and less noise coupling from this radix detection The gmn2,3 and RI is the input trans-conductance and output
scheme with low power. Because of these advantages, the impedance of first stage; The gmn7,8 and RII is the input
PN is dithered at the input of the sub-ADC for the 4.5-bit trans-conductance and output impedance of first stage. CL
first stage circuit for the 12-bit pipeline ADC. is the load capacitor. The gmp3,4 is the MP3 and MP4 trans-
conductance of first stage; The gmp7,8 is the MP7 and MP8
2.2 High speed operational amplifier trans-conductance of first stage. Vin and Vout is small input
and output voltage, relatively. Then, the node equation of
As a critical module used for SH and MDAC in each stage, A point is:
the operational amplifier requires careful design to meet the V1
1
g
requirements of the system, so as to minimize the gain and gmn2;3 Vin þ gmp3;4 1 mp3;4 1 Vout ¼ 0 ð6Þ
gmp3;4 þ sCc3
RI
settling error as possible.
The setting accuracy should be smaller than 1/4 least the node equation of B point is:
significant bit (LSB). The voltage gain needs to satisfy the !
following formula: 1 1
gmn7;8 V1 þ 1 1
þ þ s CL Vout ¼ 0 ð7Þ
gmn3;4 þ sCc3
A b 2Nþ2 ð4Þ RII
A is the gain of the op-amp, b is feedback factor, N is The transfer function is:
the resolution of corresponding stage in ADC. Consider gmn2;3 gmn7;8 RI RII 1 þ gCmp3;4
c3
s
Vout
some non-ideal factors, such as input parasitic capacitor, ¼
Vin RII CL Cc3 2 Cc3
amplifier gain should be greater than ideal gain. In SH, b is gmp3;4 s þ gmp3;4 þ RII CL þ RII Cc3 þ gmp7;8 RI RII Cc3 s þ 1
123
Analog Integrated Circuits and Signal Processing
CMFB1 A
Vo1 Vo2
MN9 MN10
MN3 VBP2 MN4
Cc2
Cc2
VFo1 VFo2
VBIAS1
MN0
Voutp
FB
Voutn CMFB2 MN13
+
V1 gmp3,4*Vs3,4
gmn2,3*Vin RI RII CL
_ gmn7,8*V1
+
gmn2,3*Vin gmp3,4*Vs3,4 gmn7,8*V1
V1
RI RII CL
1/gmp3,4
_
gmp3,4R1 after the compensation. Then, The compensation second stage, further lower the whole op-amp’s current.
can obtain relatively larger non-dominant pole with low Besides, The Cc3 draw into a left zero point of - gmp3,4/Cc3
gmp7,8 in the second stage. The opamp can obtain enough in the plane. The simulated gain and unit-gain bandwidth is
phase margins and obviously minimize the current of the
123
Analog Integrated Circuits and Signal Processing
shown in Fig. 6. The gain and unit-gain bandwidth can get Vin Vout Vout Vout
gmp Vin gmn Vin Vout Cs
85 dB and 3.28 GHz with the phase margin of 57.3. R ron rop
¼0
2.3 Serial output interface ð10Þ
The topology of the 4 Gbps serial output interface for the Then, the Av of output stage can be obtained as follows:
high speed ADC is shown in Fig. 7. The block is composed Vout 1 ðgmp þ gmn ÞR
of an encoder which converts the parallel 12-bit code to Av ¼ ¼ ð11Þ
Vin 1 þ R r1on þ r1op
64-bit code, a serializer which changes the parallel 64-bit
code to the serial packets under the clock of 2 GHz and a Thus, the R can reduce the voltage gain and improve the
current model (CM) output interface that drives the seri- unity-gain bandwidth in order to get bandwidth of higher
alized data stream Din to differential output of Dout ? and than 4 GHz.
Dout -.
To fulfill the design target,the data from the 12-bit ADC
is sent serially in packets of 64 bits, which consist of an 3 Experimental results
8-bit header, 48-bit data, and 8-bit error correction code
(ECC). The output data streaming of 16 9 the encode The 12-bit 250-MSPS ADC with 4 Gbps serial interface
clock is designed for the transmitter to drive the digital has been implemented in a 0.18-lm 1.8 V 1P5 M mix-
output code of the 12-bit 250MSPS ADC at the data signal CMOS process. The die photograph of the proposed
transmission rate of 4 Gbps. A 2 GHz PLL is used for ADC is displayed in Fig. 9. The central part is the SHA
generating the serializing clock for the 4 Gbps interface. circuit and the five pipeline sub-stages, and the left side
The differential output driver is the important part in the shows reference voltage generator and buffers, the digital
interface, as it is the bottleneck of the bandwidth and error correction logic block follows the pipelined stages,
usually consumes a large part of power. The current mode the PLL and clock buffer is in the right down of the chip
(CM) output driver is designed to obtain data rate of and the scramble and ECC block is in the right up of the
4 GHz, which is shown in Fig. 8. The Vbiasp and Vbiasn are prototype ADC. The total die area is about 8.9 mm2, where
the bias voltage to ensure the 4 mA driving current. the active area of the pipeline ADC core is 2.52 mm2, the
The R is used to reduce the voltage gain and improve the die area of the PLL is 0.6 mm2, the active area of scramble
unity-gain bandwidth. According to Kirchhoff’s law of and ECC block is 0.55 mm2, and the active area of the
electric current, the following formula can obtained: driver is 0.25 mm2.
123
Analog Integrated Circuits and Signal Processing
12bit ADC
.
Serializer
d11 .
.
d0 D3 12b
Din Driver Dout -
d1
d10 .
d11 .
.
d0 D4 12b
d1
d10 .
d11 .
.
2GHz Clk
d10
d11
Vbiasp
R Dout +
Dout -
Din
Vbiasn
123
Analog Integrated Circuits and Signal Processing
123
Analog Integrated Circuits and Signal Processing
4 Conclusion
123
Analog Integrated Circuits and Signal Processing
ADC can provide 11.3 effective bits of resolution to satisfy 11. Liechti, T., Tajalli, A., Akgun, O. C., Toprak, Z., & Leblebici, Y.
the high speed industrial applications in low power design. (2008). A 1.8 V 12-bit 230-MS/s pipeline ADC in 0.18 lm
CMOS technology. In IEEE International Circuits and Systems
(pp. 21–24).
Acknowledgements This work was supported by Natural Science 12. Shin, S. K., Rudell, J. C., Daily, D. C., et al. (2014). A 12-bit,
Foundation of China (No. 61704161), Higher Education Important 200MS/s zero-crossing based pipelined ADC with early sub-ADC
Science Foundation of Anhui Province (No. KJ2017A396) and decision and output residue background calibration. IEEE Jour-
Guangzhou Industry-Academia-Research Program (No. nal of Solid-State Circuits, 49(6), 1366.
201604016122). The authors would like to thank analog group for the 13. Sahoo, B. D., & Razavi, Behzad. (2009). A 12-bit, 200 MHz
technical discussions, layout and test group for their layout and test CMOS ADC. IEEE Journal of Solid-State Circuits, 44(9), 2366.
contributions.
123
Analog Integrated Circuits and Signal Processing
123