Abstract—A 0.25µm CMOS serializer ASIC, designed using
radiation tolerant layout practice, was exposed to proton beam at
various flux levels and accumulated fluence over 1.9×1015
protons/cm2 (100 Mrad (Si)). The ASIC survived this total
ionizing dose (TID) with no degradation in function. Single event
effect (SEE) cross-sections are also calculated.
HE Gigabit Optical Link (GOL) is a serial transmitter
developed by CERN Microelectronics Group with data
rate up to 1.6Gbps . The ASIC chip is implemented in
0.25µm CMOS technology with radiation tolerant layout
practices . It is developed for the LHC experiments front-
end electronics, and therefore subjected to high doses of
ionizing irradiation during the lifetime of the experiments.
System performance of this transmitter ASIC has been
tested . Radiation resistance for current LHC applications
has been evaluated using X-rays (10 keV peak) in a single step
to accumulate total dose of 10 Mrad (SiO2) at dose rate of 160
rad (SiO2)/s. Single event upset (SEU) crosstalk was also
measured using 60 MeV proton beam with fluence up to
3.14×1012 protons/cm2. No SEU events were observed.
Today, accelerator based high energy physics experiments
are conducted at increasingly higher collision rates. The
proposed LHC luminosity upgrade will reach 1035/cm2/sec, 10
times more than current design luminosity. The fluences of
secondary particles will also increase by a factor of 10,
respectively. Previous studies , indicated that the use of
radiation tolerant layout practices on deep-submicron CMOS
technology, such as enclosed layout, majority voting and up-
sizing analog components, extends the tolerable total dose well
beyond the inherent technology limit. In the following, we
present the results of GOL radiation resistance experiment at
the highest reported proton dose and fluence. Section II
outlines the design of the GOL chip as well as data analysis
Manuscript received July 20, 2007. This work was supported by the NSF-
US ATLAS program.
Chu Xiang, Tiankuan Liu, Cheng-An Yang, Jingbo Ye and Ryszard
Stroynowski are with Department of Physics, Southern Methodist University,
Dallas, TX 75275 USA (telephone: 214-768-1472, e-mail: firstname.lastname@example.org).
Ping Gui, Wickham Chen, Junheng Zhang and Peiqing Zhu are with
Department of Electrical Engineering, Southern Methodist Univeristy, Dallas,
TX 75275 USA (telephone: 214-768-1733, e-mail: pgui@engr. smu.edu).
system. Experiment setup and irradiation results are described
in section III, and section IV concludes the paper.
II. ASIC ARCHITECHURE
The GOL chip deploys basic principles and common
architecture of a serializer. It also supports dual-protocol and
dual-speed transmission. The configuration registers are
accessed via I2C bus and JTAG interface. Serial data output is
fed to a 50Ω line driver or a laser driver. Block diagram of the
ASIC is shown in Fig. 1. Operation mode stayed the same
before, through and after irradiation, as listed in Table I.
Fig. 1. ASIC architecture of configurable dual-protocol, dual-speed GOL
ASIC OPERATION MODE CONFIGURATION
To evaluate system performance and to carry out online
testing during irradiation, we developed a testbed that consists
of a complete optical data link and a build-in error rate tester.
Schematic drawing of the testbed is shown in Fig. 2. An FPGA
mother board generates test patterns and controls, feeds clock
and signal to the GOL serializer, detects received data,
Total Ionizing Dose and Single Event Effect
Studies of a 0.25µm CMOS Serializer ASIC
Chu Xiang, Member, IEEE, Tiankuan Liu, Member, IEEE, Cheng-An Yang, Ping Gui, Member, IEEE,
Wickham Chen, Junheng Zhang, Peiqing Zhu, Jingbo Ye, Member, IEEE and Ryszard Stroynowski
compares with transmitted data, as well as counts the number
of individual bit errors and detected frame errors. Parallel data
and control bits are converted into LVDS format to maintain
signal integrity between board interfaces. Payload data
contains valid 8b/10b encoded frames and intermittent IDLE
frames. A Texas Instrument TLK2501 board is used as the
receiver. The optical transceivers are a Truelight TTR-1F43-
107 VCSEL and a Stratos SLC-25-C-1E SFP.
Fig. 2. Block diagram of the test system with complete optical data-link
and build-in error rate tester.
Fig. 3. Picture of the fabricated boards fit into separate VME chaises for
in-lab and irradiation test.
The testbed is developed into separate boards to
accommodate both system evaluation and irradiation
experiment. Only the GOL carrier board is exposed to proton
beam while the rest of the system are well shielded from
irradiation. These boards fit into two chaises of 3U and 6U
VME respectively, as pictured in Fig. 3.
III. RESULTS AND ANALYSIS
Two GOL carrier boards, connected with independent data
links and bit error testers, were irradiated at the Massachusetts
General Hospital Proton Facility (NPTC) with 230 MeV
proton beam. We studied total ionizing dose effect and single
event effect in the same experiment. Flux stepped up from
1×107 to 5×1011 protons/cm2/s, and fluence reached 1.9×1015
protons/cm2, corresponding to the total ionization dose of 100
Mrad(Si). The two systems rendered similar results. In the
following, we present data from the system that exhibited more
A. Total ionizing dose results
During beam pauses at scheduled flux steps, and after
irradiation reached the planned fluence, data transmissions ran
error-free. Both GOL chips withstood a total ionizing dose of
100 Mrad(Si). Currents supplied to GOL chips decreased by
less than 4% during the process. Currents consumed by the
VCSEL lasers, which were also exposed to proton beam but at
reduced density due to their off-center positions, increased 1%.
The amount of change in GOL currents did not affect normal
chip operation. And the amount of change in VCSEL currents
would not cause data transmission error given the margin of
optical signal power before irradiation. After two weeks of
unbiased, room temperature annealing, the currents returned to
their values before irradiation.
Electrical waveforms of serial data were shown in Fig. 4. No
changes in rise/fall time were observed, while the amplitude
slightly decreased from 350 mV to 300 mV. Both waveforms
before and after irradiation comply with modified Gigabit
Ethernet standard --- system data rate of 1.6Gbps is determined
by the LHC master clock of 40 MHz.
Fig. 4. Serial data waveform (upper: before irradiation; lower: after
irradiation with total dose of 100 Mrad(Si)).
Jitter measurements were also performed before and after
irradiation to facilitate margin budgeting for application
design. GOL jitter generation, which is essentially the
integrated phase-noise, was measured as serial output intrinsic
jitter in the absence of applied jitter to parallel input. As
shown in Table II, change in GOL jitter generation was
minimal after irradiation.
TRANSMITTER INTRINSIC JITTER COMPONENTS
Jitter transfer function of a component is the ratio of the
amplitude of output jitter to an applied input jitter . This
measurement specifies that no parts of a system will cause an
unacceptable increase in jitter and that the recovered signal
will follow a tolerance template. We measured transfer
function of GOL and of the complete data link by injecting
sinusoidal jitter of 200ps at various modulation frequencies up
to 1.5MHz to the parallel input of GOL. Results are shown in
Fig. 5. Jitter transfer function (upper: GOL jitter transfer before and after
irradiation; lower: system jitter transfer before and after irradiation).
Flattening of the GOL jitter transfer at high modulation
frequency after irradiation indicates that the transmitter’s
multiplexer bandwidth might have increased. This requires
strict control on input signal jitter spectrum. However, there is
little change in the overall system jitter transfer after
irradiation, showing that the receiver is able to tolerate the
above specified injected jitter with GOL transfer function
degraded after irradiation.
Jitter tolerance of the complete data link was also tested
before and after irradiation. We measured the jitter penalty of
equivalence to 1dB power degradation while maintaining
better than 1×10-12 bit error ratio. As shown in Fig. 6, the
applied sinusoidal jitter magnitude and frequency that caused
the specified degradation follows SONET/SDH OC48
template well. Irradiation caused no effect on the overall
Fig. 6. Jitter tolerance before and after irradiation.
B. Single event effect results
In Table III we summarize the step raised flux levels,
accumulated fluences and corresponding SEE results. No hard
or soft reset was needed for the data link to restore normal
operation; therefore the SEE events observed were transient
effects. At fluence lower than 1×1011 protons/cm2 no SEE
events were observed. As fluence increased, a small number of
transmission errors were recorded.
PROTON BEAM RUNS AND SEE RESULTS
The absolute majority of events are frame errors, caused by
loss of synchronization of the link (LoL). Only few events
were shown as data corruption (Bit error). This indicates that
the analog circuitry in GOL chip is more susceptible to proton
irradiation than the digital circuitry . All loss-of-
synchronization errors, characterized by detected frame control
bits changing from valid data to error propagation, endured for Download full-text
a number of consecutive frames. For the purpose of analysis,
each block of consecutive frame errors was treated as one
event. SEE cross-section is derived from the number of SEE
events registered during a period of time divided by fluence.
Cross-section derived from the highest flux run is smaller than
cross-sections derived from the lower flux runs, but is within
3σ variance determined by the number of error samples. We
evaluate the proton irradiation transient SEE cross-section to
be (2.5±0.6) ×10-13 cm2 for loss of synchronization events and
to be (5.3±2.6) ×10-14 cm2 for bit corruption events.
A test system to characterize the radiation hardness of a
Gigabit per second serializer (GOL) has been developed. Two
GOL chips were exposed to total ionizing dose of 100
Mrad(Si) with no functional degradation. Single event effect
cross-sections are calculated to be (2.5±0.6) ×10-13 cm2 for loss
of synchronization events and (5.3±2.6) ×10-14 cm2 for bit
The authors would like to thank Dr. Paulo Moreira at CERN
Microelectronics Group and Mr. Ethan Cascio at Massa-
chusetts General Hospital Proton Facility for their help.
 P. Moreira, T. Toifl, A. Kluge, G. Cervelli, F. Faccio, A. Marchioro and
J. Christiansen, "G-Link and Gigabit Ethernet compliant serializer for
LHC data transmission, " IEEE Nuclear Science Symposium., vol. 2, pp.
96-99, Oct. 2000.
 G. Anelli et al, "Radiation tolerant VLSI circuits in standard deep
submicron CMOS technologies for the LHC experiments: practical
design aspect, " IEEE Nuclear Trans. Nucl. Sci., vol. 46, no. 6, pp. 1690-
1696, Dec. 1999.
 P. Moreira et al., "A radiation tolerant gigabit serializer for LHC data
transmission," Workshop on Electronics for LHC Experiments, 2001.
 G. Anelli et al., "Total dose behavior of submicron and deep submicron
CMOS technologies," Workshop on Electronics for LHC Experiments,
 F. Faccio et al., "Total dose and single event effects in a 0.25um CMOS
technology," Workshop on Electronics for LHC Experiments, 1998.
 C. Claeys and E. Simoen, Radiation Effects in Advanced Semiconductor
Materials and Devices, Springer-Verlag, Berlin Heidelberg, 2002
 IEEE standard 802.3ae-2002: Gigabit Ethernet; Institute of Electrical and
Electronics Engineers, NY, 2002.
 D. Derickson et al., Fiber Optic Test and Measurement, Prentice Hall
PTR, NJ, 1998.