A 15-bit 125 MS/s Time-Interleaved Pipelined Analog-to-Digital Converter
7.6. EXPERIMENTAL RESULTS 117
Figure 7.19: Measured SNDR and SFDR versus finat normal calibration.
is a differential 1.4Vpp 9.99MHz sinusoidal signal. Without calibration, the offset error is the dominant distortion term. The signal-to-distortion-plus-noise (SNDR) is 54.4 dB and the spurious-free dynamic range (SFDR) is 55.6 dB. After the calibration is activated, the SNDR is improved by 15.5 dB to 69.9 dB and the SFDR is improved by 36.3 dB to 91.9 dB. Fig. 7.19 shows the ADC’s measured SNDR and SFDR versus input frequencies at 125 MS/s sampling rate. The calibration can improve the SNDR by more than 10 dB and the SFDR about 30 dB when the input frequencies up to the Nyquist frequency.
Fig. 7.20 shows the ADC’s measured SNR and SFDR versus input signal levels with the calibration on and off respectively by applying a sinusoid at the frequency of 1.99MHz.
This results reveal that the noise power excluding the distortion is not affected by the input levels, and the injected random term, q × Vr. The measured dynamic range is approxi-mately 73 dB at the supply voltage of 1.8 V.
Fig. 7.21 shows the ADC’s measured signal-to-noise ratio (SNR) versus input fre-quencies at 125 MS/s sampling rate. The input is a differential 1.4 Vpp9.99MHz sinusoidal signal. The SNR is calculated from the ADC’s output FFT spectrum while ignoring the
Figure 7.20: Measured SNR and SFDR versus input levels.
Figure 7.21: Measured SNR versus input frequencies.
7.7. SUMMARY 119
harmonic tones caused by A/D nonlinearity and the spurious tones caused by inter-channel offset and gain mismatches. Also shown in Fig. 7.21 is the calculated SNR of an ADC model which includes the effect of sampling clock jitter ∆t. Its input is A sin(2πfint)+nex
where nexis an external noise source. Its SNR can be expressed as [55]:
SNRn+j = A2/2 2π2fin2A2∆t2+ n2ex
(7.1) By curve-fitting (7.1) against measured data, the root-mean-square (rms) value of nex is estimated to be nex,rms = 137 µV, and the rms value of ∆t is estimated to be ∆trms = 0.82 psec.
7.7 Summary
The comparisons of this prototyping ADC with other works are described in the follow-ing. Table 7.4 shows the measured specifications of this ADC chip and compare it with published works that claim to have a maximum sample rate of more than 100 MS/s and a resolution of more than 14 bits.
Fig. 7.22 shows the performance of pipelined and time-interleaved pipelined ADCs of more than 12-bit resolution reported in the Custom Integrated Circuits Conference (CICC), International Solid-State Circuits Conference (ISSCC), Journal of Solid-State Circuits (JSSC), and commercial product’s data sheets of Analog Device and Texas In-struments Incorporated. The corresponding references are listed in Table 7.5. The 12-bit, 13-bit, 14-bit, 16-bit designs implemented using different technologies are labeled, as well as this work. The ADCs with outstanding performance are highlighted using the corre-sponding item in Table 7.5. The results indicate that BiCMOS implementations can have better effective number of bits (ENOB) and faster sampling rate than CMOS implemen-tations. This is true because BiCMOS is the best technology for the design of an opamp in all aspects except for money cost when it combines the benefits of Bipolar and those of CMOS [58]. The qualitative comparison for an opamp implemented using CMOS, Bipolar and BiCMOS technology is shown in Table 7.6 [59].
For a high resolution ADCs, two dominant noise sources are the thermal noise and the aperture noise. If the thermal noise is considered as the only noise source, the reachable
Table 7.4: Performance Comparison
This Work [3] [56] [57]
Technology 0.18 µm CMOS 0.13 µm CMOS 0.35 µm BiCMOS 0.35 µm BiCMOS
Architecture TI Pipelined Pipelined Pipelined TI Pipelined
Calibration Digital Background Digital Background None Digital Foreground
Supplies 1.8 V 1.5 V 3.3 V/5 V 3.0 V
Input Range (Vpp) 1.4 V 1.5 V 2 V 4 V
Power Consumption 0.909 W 0.224 W 1.95 W 1.4 W
Resolution 15 Bits 14 Bits 14 Bits 14 Bits
Max. Sampling Rate 125 MS/s 100MS/s 125 MS/s 100 MS/s
DNL (LSB) −0.27/+0.25 −1.1/+1.1 −0.2/+0.2 0.97
INL (LSB) −5.7/+5.5 −2.0/+2.0 −0.5/+0.5 6.9
SNR (@fin= 5 MHz) 70.4 dB 70 dB 75 dB
SNR (@fin= 49 MHz) 67.3 dB 65 dB 75 dB
THD (@fin= 5 MHz) −81.7 dB −71.1 dB THD (@fin= 49 MHz) −76.9 dB −68 dB
THD (@fin= 210 MHz) −76.3 dB
SFDR (@fin = 5 MHz) 91.6 dB 100 dB
SFDR (@fin = 70 MHz) 82.9 dB 95 dB
SFDR (@fin = 210 MHz) 79.9 dB
Active Area 18.5 mm2 1.02 mm2 70 mm2 10.2 mm2
7.7. SUMMARY 121
Table 7.5: Nyquist-rate ADC Survey
Item Publication/ Sampling ENOB
Author Title
Year Rate (MS/s) (bits)
CMOS
1 ISSCC / 2006 100 10.5 P. Bogner, etc. A 14b 100MS/s Digitally Self-Calibrated Pipelined ADC in 0.13µm CMOS
2 ADI / 2006 125 11.8 Analog Device AD9246 Data Sheet 3 ADI / 2006 150 11.9 Analog Device AD9254 Data Sheet
4 JSSC / 2005 110 10.7 T. N. Andersen, etc. A Cost-Efficient High-Speed 12-bit Pipeline ADC in 0.18-µm Digital CMOS
5 CICC / 2005 180 10.6 k. Gulati, etc. A Highly-Integrated CMOS Analog Baseband Transceiver with 180MSPS 13b Pipelined CMOS ADC and Dual 12b DACs 6 Texas In. / 2005 125 11.3 Texas Instrument ADS5500 Data Sheet
7 JSSC / 2005 80 11.8 C. R. Grace, etc. A 12b 80MS/s Pipelined ADC with Bootstrapped Digital Calibration
8 ISSCC / 2004 50 12.5 K. Nair, etc. A 96 dB SFDR 50MS/s Digitally Enhanced CMOS Pipeline A/D Converter
9 JSSC / 2003 75 11 B. Murmann, etc. A 12-bit 75-MS/s Pipelined ADC Using Open-Loop Residue Amplification
10 JSSC / 2001 54 10.5 H. v. d. Ploeg, etc. A 2.5-V 12-b 54-MSample/s 0.25-µm CMOS ADC in 1-mm2 with Mixed-Signal Chopping and Calibration
11 JSSC / 2001 75 12 W. Yang, etc. A 3-V 340-mW 14-b 75-MSample/s CMOS ADC with 85-dB SFDR at Nyquist Input
12 ISSCC / 2000 65 11.6 L. Singer, etc. A 12b 65MSample/s CMOS ADC with 82dB SFDR at 120MHz 13 JSSC / 2000 50 10.5 H. Pan, etc. A 3.3-V 12-b 50-MS/s A/D Converter in 0.6-µm CMOS with
over 80-dB SFDR
14 ISSCC / 1997 128 10.1 R. Jewett, etc. A 12b 128MSample/s ADC with 0.5LSB DNL BiCMOS
15 JSSC / 2006 125 12.5 A. M. Ali, etc. A 14-bit 125-MS/s IF/RF Sampling Pipelined ADC with 100dB SFDR and 50fs Jitter
16 Texas In. / 2006 170 11.4 Texas Instrument ADS5545 Data Sheet 17 Texas In. / 2006 210 12.1 Texas Instrument ADS5547 Data Sheet
18 JSSC / 2005 65 12.9 A. Zanchi, etc. A 16-bit 65-MS/s 3.3-V Pipeline ADC Core in SiGe BiCMOS with 78-dB SNR and 180-fs Jitter
19 ADI / 2005 100 13.1 Analog Device AD9446 Data Sheet
20 Norchip / 2004 100 12.1 V. Hakkarainen, etc. A 14b 200MHz IF-Sampling A/D Converter with 79.9dB SFDR Bipolar
21 JSSC / 2000 100 12.5 A 14-bit 100-MSample/s Subranging ADC
Figure 7.22: High-resolution (≥ 12 bits) high-speed (≥ 50 MS/s) pipeline and time-interleaved ADCs.
Table 7.6: Qualitative Comparison for An Opamp Implemented Using Bipolar, CMOS and BiCMOS Technologies
Qualitative Comparison CMOS Bipolar BiCMOS
High Gain √ √
High Gain-Bandwidth Product √ √
Ease of Compensation √ √
Low Voltage Noise √ √
Low Current Noise √ √
Low Power Supply Limit √ √
Common Rejection √ √ √
Low Voltage Offset √ √
High Input Impedance √ √
7.7. SUMMARY 123
Figure 7.23: Figure of merit.
resolution is given by
Nthm = 1
2log2 VF S2 6kT Ref ffs
!
−1 (7.2)
where k is the Boltzmann’s constant, T is the absolute temperature, fs is the sampling rate, and Ref f is an effective thermal resistance, which includes the effects of all noise sources [60]. If the aperture noise is only considered, the attainable resolution is limited as
Napt = log2
2
√3πfsσa
−1 (7.3)
where σa is the rms aperture jitter [60]. Both Nthm and Napt boundaries are plotted in Fig. 7.22 with Ref f = 300 Ω and σa = 0.1 psec. The thermal noise truly limits the resolution above 14 bits. The aperture jitter degrades the available resolution more as the sampling rate increases. However, the designs labeled in Fig. 7.22 do not close to the limitation due to physical uncertainty, so does this work.
The figure of merit, the energy per conversion step, defined as Econv = Pd
2ENOB× fs (7.4)
is widely used to evaluate the efficiency of ADCs, where Pd is ADC’s power dissipation, and fs is ADC’s sampling rate. Fig. 7.23 shows the efficiency comparison of the ADCs surveyed in Table 7.5 which have the resolution more than 14 bits and the sampling rate more than 100 MS/s. The three boundaries of Econv = 1 pJ, Econv = 2 pJ and Econv = 3 pJ are also plotted in Fig. 7.23 to clearly exhibit this comparison. Most designs locate between Econv = 1 pJ and Econv = 2 pJ. The CMOS design of reference 3 has the best conversion efficiency among these references. BiCMOS designs reveal better 2ENOB× fs, but larger power consumption. The design of reference 17 has outstanding efficiency among the BiCMOS designs. Although it has little worse conversion efficiency than that of CMOS designs of reference2 and 3, it has much better 2ENOB× fs.