Professional Documents
Culture Documents
A 1.2ps-Jitter Fully-Synthesizable Fully-Calibrated Fractional-N Injection-Locked PLL Using True Arbitrary Nonlinearity Calibration Technique
A 1.2ps-Jitter Fully-Synthesizable Fully-Calibrated Fractional-N Injection-Locked PLL Using True Arbitrary Nonlinearity Calibration Technique
Delay
Delay
fractional-N injection-locked PLL in 65 nm CMOS. A true REF 2xREF
arbitrary non-linearity calibration scheme is specifically pro- Coarse
Code
Fine
Code
Inj. Edge &
Window Gen
Fine Coarse
and -246.7 dB. PD Coarse
Tuning Gain Cal. & Medium
9781538624838/18/$31.00©2018IEEE
Authorized licensed use limited to: MIT-World Peace University. Downloaded on January 20,2024 at 11:56:29 UTC from IEEE Xplore. Restrictions apply.
ELW&RDUVH'7&
IJFRDUVHk IJFRDUVHk [5()
IJILQHnk
ELW)LQH
D )&:)5$&
'7&
ǻ
&XVWRPGHVLJQHG'7&ZLWK )XOO\V\QWKHVL]DEOH'7&ZLWK >@ '7&&75/ >@ >@ >@
3:/, FRQYHQWLRQDO 3:/, FRQYHQWLRQDO
&RDUVH )LQH
SV
SV
/06 /06
*DLQ&DO *DLQ&DO
7RWDO,1/>SV@
7RWDO,1/>SV@
'7&287
3'>@
Nominal Delay[ps] Nominal Delay[ps]
Fig. 3. Block diagram of the proposed TANC implementation.
)XOO\V\QWKHVL]DEOH'7&ZLWK )XOO\V\QWKHVL]DEOH'7&DIWHU
7$1& SURSRVHG 7$1& SURSRVHG
SV
7RWDO,1/>SV@
To solve this issue, this paper proposes a true arbitrary
nonlinearity calibration (TANC) for the coarse DTC, which
Nominal Delay[ps] Nominal Delay[ps]
works for DTC with arbitrary INL profile. First, the INL
E of each coarse code is estimated from the PD output, then
)XOO\V\QWKHVL]DEOH'7&ZLWK &XVWRPGHVLJQHG'7&ZLWK
7$1& SURSRVHG 3:/, FRQYHQWLRQDO
a corresponding compensation code is applied for each
∙∆IJ
∙∆IJ coarse code. The residue INL after the calibration is only
∆IJ ,GHDOGHOD\ ,QWHUSRODWLRQ
&DOLEUDWHG
&RDUVH'7&
WDEOH &DOLEUDWHG
'7&
from the fine stage, which is negligible to the total jitter.
IJRIIVHWk
'7&GHOD\
'7&GHOD\
FRDUVH
ILQH ,GHDOGHOD\ 8QFDOLEUDWHG
'7&
Compared to the conventional DTC calibration techniques,
8QFDOLEUDWHG
&RDUVH'7&
the proposed calibration scheme can work for arbitrary
IJFRDUVHk
INL profile, thus it is suitable for the synthesizable DTC
t
75() 75() 75()
t which has large random INL.
75() 75() 75()
Authorized licensed use limited to: MIT-World Peace University. Downloaded on January 20,2024 at 11:56:29 UTC from IEEE Xplore. Restrictions apply.
,QMHFWLRQ(GJH 5()'RXEOHU
9:,1 9,1- '&(&DO
:LQGRZ*HQ
'7&287 9:,1
5()'RXEOHU )LQH
6\P08; &RDUVH *DLQ&DO
7XQLQJ
>@ 3'>@
3'
9RXW
)) 9,1- >@
[5() '7&287
5() ELW ELW 6\QWKHVL]DEOH
3' 5() 'RXEOHU &RDUVH'7& )LQH'7&
)) '&2 '&2Z,QM 3'
926&
6\P3'
3'>@
'/) >@ 75()
75() 5()
75() 7 ≠7
9,1-
[5()
9:,1
HGJH
HGJH
1 7'&2>@ FRPSDUH 1 7'&2>N@
926& UHSODFH
,QW1PRGH
7’ 7
'7&287
9287
IJULVH IJIDOO IJULVH IJIDOO IJULVH
3'>@
)UDF1PRGH
7’ 7∆ʏ
'7&287
Fig. 4. Conceptual operation of symmetric phase detector and IJULVH IJIDOO∆ʏ IJULVH∆ʏ IJIDOO∆ʏ IJULVH∆ʏ
symmetric multiplexer for DCO calibration.
Fig. 5. Concept of proposed reference double calibration, the
DCO and replica elements, and the gated injection tech- duty cycle error (DCE) in the reference doubler is compensated
nique reduces the injection efficiency and degrades the by the fine DTC.
jitter performance. Thus, the DCO calibration used in this
paper extracts the frequency error by performing phase
comparison and edge replacement simultaneously. The os-
cillator edge is compared against the injection edge in the
symmetric PD, and replaced by the same injection edge in P
the symmetric MUX at the same time. Thanks to the latch-
P
based implementation, the proposed symmetric PD has a
larger gain than the conventional flip-flop implementation, )XOO\V\Q
63,
which gives larger loop gain and helps to minimize the )UDF13//
frequency error.
The reference doubler calibration is shown in Fig.5.
After the duty cycle error is roughly calibrated, the residue
error is precisely calibrated by the fine DTC. In the
integer-N mode, the fine DTC will introduce two different
delays τrise and τf all to rising edge and falling edge Fig. 6. Die micro-graph of fully synthesizable fractional-N IL-
respectively to eliminate the residue. In the fractional-N PLL.
mode, τrise and τf all will added with the desired delay
n ∗ Δτ thus residue DCE will not degrade the jitter and
spur performance. power consumption is 2.2 mW and 2.5 mW for integer-
N and fractional-N modes, respectively. The measured
IV. M EASUREMENT R ESULTS fractional spur is -58.5 dBc at 1.0725 MHz offset.
The entire PLL is synthesized using commercial digital Table.I summarizes the measured performance and com-
design tools with a non-modified standard cell library. The pares them with the state-of-the-art DTC-based fractional-
core area of the PLL is 0.12 mm2 , which is fabricated N digital PLLs. The proposed PLL achieves an FoM of
in a 65 nm CMOS process. Fig.6 shows a die photo. -246.7 dB in integer-N mode, which is 5 dB better than
The phase noise is measured by a signal source analyzer the state-of-the-art [1]. The PLL also achieves an FoM of
(Keysight E5052B), and the spectrum is measured by a -234.4 dB in fractional-N mode, which is also 10 dB better
spectrum analyzer (Keysight E4407B). Fig.7 shows the than [3]. The measured worst fractional spur is -58.5 dBc
measured phase noise with a 100 MHz reference clock. after the proposed TANC calibration. This represents the
The PLL achieves 0.3 ps jitter at 1 GHz output in integer-N best results among the fully-synthesizable PLLs, and the
mode, and 1.2 ps jitter at 971.24 MHz output in fractional- results are comparable to that of custom-designed digital
N mode, which is integrated from 10 kHz to 10 MHz. The PLLs.
Authorized licensed use limited to: MIT-World Peace University. Downloaded on January 20,2024 at 11:56:29 UTC from IEEE Xplore. Restrictions apply.
TABLE I
P ERFORMANCE SUMMARY AND COMPARISON WITH STATE - OF - THE - ART
FRACTIONAL -N DIGITAL PLL S .
0
ACKNOWLEDGMENT
Output Power [dBm]
-80
Interpolative Phase Coupled Oscillator, Current-Output DAC,
and Fine-Resolution Digital Varactor Using Gated Edge
-100 Injection Technique,” IEEE J. Solid-State Circuits, vol. 50,
-120 pp. 68–80, Jan. 2015.
[3] ——, “A 0.048 mm2 3mW Synthesizable Fractional-N PLL
-140
Int-N mode Frac-N mode DCO free-run with a Soft Injection-Locking Technique,” in IEEE ISSCC
-160 Dig. Tech. Papers, Feb. 2015, pp. 252–253.
10k 100k 1M 10M
Offset Frequency [Hz]
[4] S. Levantino, et al., “An Adaptive Pre-Distortion Technique
to Mitigate the DTC Nonlinearity in Digital PLLs,” IEEE J.
Solid-State Circuits, vol. 49, pp. 1762–1772, Aug. 2014.
Fig. 7. Measured DCO and PLL phase noise and spectrum.
[5] N. Markulic, et al., “A Self-Calibrated 10Mb/s Phase Mod-
ulator with -37.4dB EVM based on a 10.1-to-12.4GHz, -
V. C ONCLUSION 246.6dB-FOM, Fractional-N Sub-sampling PLL,” in IEEE
ISSCC Dig. Tech. Papers, Feb. 2016, pp. 176–177.
[6] S. Choi, et al., “A PVT-Robust and Low-Jitter Ring-VCO-
Based Injection-Locked Clock Multiplier with a Continuous
A fully-synthesizable fully-calibrated fractional-N IL- Frequency-Tracking Loop Using a Replica-Delay Cell and
PLL is presented in this paper. Extensive digital calibration a Dual-Edge Phase Detector,” IEEE J. Solid-State Circuits,
of DTC, DCO and reference doubler is used to mitigate vol. 51, pp. 1878–1889, Aug. 2016.
the analog impairments caused by layout randomness. The [7] A. Elkholy, et al., “A 6.75-to-8.25GHz, 250fsrms-Integrated-
fully-synthesizable fractional-N IL-PLL achieves 1.2ps Jitter 3.25mW Rapid on/off PVT-insensitive Fractional-N
Injection-Locked Clock Multiplier in 65nm CMOS,” in IEEE
RMS jitter with 2.5mW power consumption. An FoM of ISSCC Dig. Tech. Papers, Feb. 2016, pp. 192–193.
-234.4dB is achieved, which is 10dB better than the best [8] ——, “A 3.7 mW Low-Noise Wide-Bandwidth 4.5 GHz Dig-
fractional-N PLL [3]. ital Fractional-N PLL Using Time Amplifier-Based TDC,”
IEEE J. Solid-State Circuits, pp. 867–881, Apr. 2015.
Authorized licensed use limited to: MIT-World Peace University. Downloaded on January 20,2024 at 11:56:29 UTC from IEEE Xplore. Restrictions apply.