This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
IEEE TRANSACTIONS ON VERY LARGE SCALE INTEGRATION (VLSI) SYSTEMS 1
High-Precision PLL Delay Matrix
With Overclocking and Double
Data Rate for Accurate FPGA
Time-to-Digital Converters
Poki Chen , Member, IEEE, Jian-Ting Lan, Ruei-Ting Wang, Nguyen My Qui,
John Carl Joel S. Marquez, Student Member, IEEE, Seiji Kajihara, Member, IEEE,
and Yousuke Miyake, Member, IEEE
Abstract—An extremely high-resolution, 2-D Vernier field-
programmable gate array (FPGA) time-to-digital converter
(TDC) with phase wrapping and averaging has been proposed
recently to get an extremely fine resolution of 2.5 ps. However,
the cell delays in a delay matrix are not fully controlled so
that the TDC performance strongly depends on the stochastic
distribution of cell delays, and the input range is limited to
less than 20 ns. To achieve both high-precision phase division
and wide measurement range, a phase-locked loop (PLL)-based
delay matrix, which is capable of overclocking and double data
rate (DDR), is proposed in this article. All delay cells are under
the precise control of PLLs to generate output phases evenly
divided within the reference clock period. For a concept proof,
the TDC architecture is implemented on an Altera Stratix-IV
FPGA chip to achieve 15.6-ps resolution. The differential nonlin-
earity (DNL), integral nonlinearity (INL), and rms resolution
are measured to be merely -0.157 to 0.137 LSB, -0.176 to
0.184 LSB, and 1.0 LSB, which prove the superiority of the
proposed structure to its stochastic counterparts. The proposed
high-precision phase division technique can be applied to not
only the TDC but also the digital-to-time converter (DTC) to
enrich its future applications.
Index Terms—2-D Vernier, delay matrix, double data rate
(DDR), overclocking, phase division, phase-locked loop (PLL),
time-to-digital converter (TDC).
I. I NTRODUCTION
T
ECHNOLOGY development has increased the need to
measure the start–stop time intervals of physical events
and convert them into digital formats. These measurements are
Manuscript received July 22, 2019; revised October 12, 2019 and
December 1, 2019; accepted December 17, 2019. This work was supported
in part by the Ministry of Science and Technology under Grant MOST 107-
2221-E-011-110, in part by the National Taiwan University of Science and
Technology, and in part by the Kyushu Institute of Technology under Grant
Kyutech-NTUST-106-02. (Corresponding author: Poki Chen.)
Poki Chen, Ruei-Ting Wang, Nguyen My Qui, and
John Carl Joel S. Marquez are with the Department of Electronic
and Computer Engineering, National Taiwan University of Science and
Technology, Taipei 10617, Taiwan (e-mail: poki@mail.ntust.edu.tw).
Jian-Ting Lan is with Himax Technologies Inc., Tainan 74148, Taiwan
(e-mail: m10502207@gapps.ntust.edu.tw).
Seiji Kajihara and Yousuke Miyake are with the Kyushu Institute of
Technology, Fukuoka 804-0015, Japan (e-mail: kajihara@cse.kyutech.ac.jp;
miyake@aries30.cse.kyutech.ac.jp).
Color versions of one or more of the figures in this article are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TVLSI.2019.2962606
performed by the so-called time-to-digital converter (TDC).
In addition to the inherent applications, such as nuclear
physics, biomedical engineering, and time-of-flight measure-
ment [1]–[3], TDCs are also applied in design-for-testability
such as on-chip delay, temperature, and voltage measurements
[4] and [5]. They are even used to test transistor aging by
measuring the increased delay due to aging [6].
To achieve high resolution, fine measurement techniques
or the so-called interpolators are adopted to design TDCs.
Analog interpolators, such as time-stretching [7] and
time-to-amplitude conversion [8], are proposed. However,
they are quite sensitive to process, voltage, and ambient
temperature (PVT) variations and undergo long conversion
time. Consequently, digital techniques such as tapped delay
lines (TDLs) [9], pulse shrinking delay cell [10], gated ring
oscillators [11], and 2-D Vernier [12] are commonly used.
Most of the current TDCs based on digital interpolators are
implemented with full-custom designs to get good linearity,
as well as high resolution down to a picosecond order. A 2-D
Vernier with second-order 16 modulators and a spiral 2-D
arbiter array [13] is adopted to achieve a resolution of 1 ps and
differential nonlinearity (DNL)/integral nonlinearity (INL) of
0.41/0.79 ps at the expense of long conversion time. A TDC
based on pulse shrinking mechanism [14] obtains 1.8-ps
resolution. However, it suffers from a low conversion rate.
The time amplifier, which is based on two switched ring
oscillators [15], is proposed to accomplish 2.6-ps resolution
and increase power efficiency. However, it is limited to a
short dynamic range. Moreover, a 13-bit TDC based on cyclic
time-domain successive approximation [16] is proposed to
achieve 0.61-ps resolution and 5 ns dynamic range. This
structure increases the conversion time and requires highly
accurate matching for offset delay in two loops. Another TDC
using stochastic phase interpolation fabricated on a 14-nm
fin field-effect transistor (FinFET) [17] achieves 1.17-ps
resolution but short measurement range.
Although full-custom TDC designs are better than their
field-programmable gate array (FPGA)-based counterparts in
performance, they require rigorous design and verification
process. In contrast to the application-specified integrated
circuit (ASIC), FPGA with parallel architecture has shorter
1063-8210 © 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.