References for Neutron Irradiation
Tolerance Testing for CSC electronics
Neutron
flux and energy spectrum at the location of Cathode Strip Chamber
|
Neutron energy range
|
Worst radiation (KHz/cm2 )
|
Least radiation (KHz/cm2)
|
|
0~1 KeV
|
100
|
20
|
|
1~10 KeV
|
100
|
10
|
|
10-100 KeV
|
100
|
10
|
|
100 KeV~1 MeV
|
100
|
10
|
|
1~10 MeV
|
20
|
5
|
|
10~100 MeV
|
10
|
1
|
|
100 MeV~1 GeV
|
5
|
1
|
|
Total
|
450
|
50
|
|
Annual flux (1012/cm2)
|
7.0
|
0.75
|
|
1 MeV equivalent neutrons(1012/cm2.yr)
|
1.8
|
0.24
|
|
Annual Total Ionising Dose (Gy)
|
11
|
3.1
|
10 years of neutron fluence at CSC worst radiation location
| 2~20 MeV |
~3x1012 neutron/cm2 |
| > 20 MeV |
~1.5x1012 neutron/cm2 |
* These numbers are crunched from ATLAS detector simulation and are
approximate values. Another approach with formula found in "Review of Particle
Physics" gives somewhat similar numbers.
** The worst radiation occurs in the lowermost area of the CSC; and
the least radiation occurs in the outermost area of the CSC detector.
*** Assuming ATLAS will run for 180 days in a
year.
Figure 1. The green curve shows the neutron energy spectrum for ATLAS.
The red curve shows the energy spectrum of the neutrons coming from 20
MeV deuteron beam stopped by a Beryllium disk.
Electronics to be tested
Any silicon device will be effected in a radiation environment. For the
CSC readout electronics, the following devices will be tested: (These chips
are manufactured by Agilent
Technology)
| G-link data networking components |
HP part no. & link to description |
| transmitter |
HDMP
1022 |
| receiver |
HDMP 1024 |
| transceiver |
HFBR 5305 |
| evaluation kit (transmitter and receiver on separate board) |
HDMP
102K |
Neutron Radiation Mechanism
The de Broglie wavelength of neutrons is comparable to atomic spacing.
So most of the neutrons pass
through materials by elastic scattering or diffraction. But when the
neutron energy is greater than
an energy threshold, nuclear reaction comes in and plays a very
important role in the neutron material interaction. The neutron energy
threshold for n + Si -> p + Al is roughly
2.8 MeV, and for n + Si -> (alpha) + Mg is
around 4.2 MeV. When a nuclear reaction occurs, the charged ion will
create a pulse. If the pulse is big enough, it enables digital flip-flop
to change state. This phenomenon is called "Single
Event Upset (SEU)". Although nuclear reaction
is the main cause for SEU, other mechanism such as inelastic scattering
could also induce SEU. In our CSC neutron irradiation testing, we are more
concerned on the SEU rate of the G-Link connection under ATLAS running
environment.
The SEU cross section vs the neutron
energy has the following characteristic:
-
For thermal neutrons, energy is lower than 2 MeV,
basically no SEU.
-
2 < Eneutron < 20 MeV, the lower
limit corresponds to the threshold of inelastic reactions. In this energy
range, the SEU sensitivity to neutrons varies a lot.
-
Eneutron > 20 MeV, protons and neutrons
have the same probability to induce SEU. With energy greater than 20 MeV,
protons pass the Coulomb barrier, therefore protons and neutrons behave
the same. In this energy range, the SEU sensitivity to protons/neutrons
has a flat distribution as a function of proton/neutron energy.
Figure 2. The SEU cross section for G-Link transmitter (Liquid Argon
result) induced by neutron as a function of deuteron beam energy.
Error Categorization
| SEU |
Single Event Upset |
A change of state or transient induced by an energetic particle
in a device. This may occur in digital, analog and optical components or
may have effects in surrounding interface circuitry. These are "soft" errors
in that a reset or rewriting of the device causes normal behavior thereafter.
For the CSC G-Link components, there could be two types of SEU depending
on the location of the bit on the chip, the first kind is that one data
bit is corrupted, in this case, the pulse height on one strip will be changed
thus the position of the track will be shifted due to this data bit corruption;
the second kind is that one control bit of the G-Link is corrupted, in
this case, the G-Link will fail in locking on a frame to transfer, in other
words, the link is down, all the data will be lost during down time. |
| MBU |
Multiple Bit Upset |
An event induced by a single energetic particle such as cosmic ray
or proton that cause multiple upsets during its path through a device.
Energetic neutron is unlikely to cause a MBU. |
| SHE |
Single Hard Error |
an SEU which causes a permanent change to the operation of a device. |
| SEL |
Single Event Latchup |
a condition which causes loss of device functionality due to a single
event induced high current state. An SEL may or may not cause permanent
device damage, it requires power strobing of the device to resume normal
device operations. |
*Note: this single event effect specification is taken from NASA database
and only the category related to CSC readout electronics is listed. The
categorization is still applicable to ATLAS. For a complete description
of the single event effect, please visit http://flick.gsfc.nasa.gov/radhome/papers/seespec.htm
Experimental Setup
The radiation tolerance testing should be done under normal operational
condition. A realistic simulation of the data flow for CSC ATLAS will be
essential to measure the SEU rate. The neutron flux in this testing is
many order of magnitude higher than in the ATLAS environment, therefore
a pile-up effect has to be studied to extrapolate the SEU rate in ATLAS
running environment.
Two possible
test setups (conceptul design).
Radiation Standard
The ATLAS CSC will have 64 chambers in total. Each chamber has 4 layers,
and each layer carries 192 precision strips and 48 transverse strips. The
preamplifier and shaper (P/S) for each strip sends the waveform on the
strip to a Switched Capacitor Array (SCA) continuously with a speed of
40 MHz at each beam crossing. The SCA will sample the waveform (4 or 5
samples per waveform) and store the information in its cells, each SCA
takes in 12 P/S inputs. When a real physical event occurs, a triggering
signal triggers ADC to read in averagely 4 (5?) waveform samples stored
in the SCA for each strip and digitize them to 12 bits data. ATLAS will
have trigger rate of 100 KHz. Then the output of 16 channels of ADCs are
MUXed and transferred to off-chamber electronics through two G-Link connections
for data processing. Another G-link is dedicated for transfering
SCA control data for each ASM board. There are 960 G-Link connections for
the whole ATLAS CSC.
To reach a safe SEU rate for the CSC electronics, we make assumptions
as shown in the following table and then set a standard for G-Link chips.
| Assumption Parameter |
value |
Comment |
| CSC total in-efficiency |
0.1% |
In-efficiency is defined as 100%-efficiency. |
| In-efficiency per G-Link |
1.04E-6 |
960 G-Links for ATLAS CSC. |
| Link-down time |
2 ms |
Up-stream and down stream, each direction takes 1 ms to recover. |
| No. of chips per link |
2 |
transmitter and receiver. |
| Data dropped after re-connection |
8 triggers |
8 triggers worth of data will be dropped to flush the garbled charges
due to the link -down. |
| Link-down probability per SEU |
100% |
Although not every SEU will cause link-down, link-down certainly introduces
more problem. A control bit upset will cause a link-down thus result in
this particular G-Link dropping all the data during down time. A data bit
upset will just result in the change of the pulse height on one strip thus
may shift the position of the hit on this particular strip. To set a safe
SEU rate, we assume the worst for each SEU occurance. |
| Acceptable SEU rate |
one SEU per 32 minutes per chip |
Neutron flux at ATLAS CSC worst radiation area in 32 minutes will be
8.7x108
including
neutrons of all energy. A factor of 5
has
to be included here to account for the uncertainty in the neutron flux
simulation. |
| Acceptable SEL rate |
one SEL per 2.56E6 hours per chip |
Assuming each ATLAS run takes 2 hours and when SEL occurs on a particular
G-Link channel, all the data on this channel will be lost during the run.
Starting a new run will reset the G-link to resume its normal function. |
Statistics discussion: In the situation
of not observing any SEU/SEL in a period of time T,
the probability function to describe it using Poisson
statistics is P(0)=exp(-T/T0)
where T0 is
the average time for one SEU/SEL event (unknown
here). The probability of not observing any SEU/SEL will
drop down to 10% in timeT=2.3T0
then
we may inteprete it as the average time for one SEU/SEL to occur is greater
than T/2.3 with 90%
confidence level.
Probability for observing number
of events in time 2.3T0
|
n
|
0
|
1
|
2
|
3
|
4
|
5
|
6
|
|
P(n) (%)
|
10
|
23
|
26.4
|
20.3
|
11.6
|
5.3
|
2
|
The average number of events expected is 2.3.
To measure T0 to
a significance of 10 sigma, the number of occurances has to be greater
than 100.
Note: The electronics have to sustain three types of irradiation tests
to be qualified
-
Single Event Effect on the Digital electronics
with 10 years of neutron fluence for A) En
> 20 MeV and B) 2 < En <
20 MeV.
-
Displacement Demage with the total fluence
of 10 years of 1 MeV equilavent neutrons. The main effect will be on the
analog microelectronics for gain loss, leakage current increase and threshold
shift etc.
-
10 years of Total Ionising Dose with 1 MeV
photons.
Neutron Radiation Stations
-
CERI: a cyclotron provides 20 MeV deuteron beam. The deuterons are stopped
on a Be disk producing neutron flux by stripping reactions.
The maximum energy a neutron can get is 20 MeV;
the mean energy is 9 MeV. This source
could provide a flux of 3x109 neutron/cm2/sec. One
hour of irradiation will accumulate more than 10 years of ATLAS CSC neutron
fluence. Centre
d'Etudes et de Recherchesé par Irradiation, Orleans France .
Related link: Liquid
Argon-ATLAS.
-
88" cyclotron at Lawrence Berkeley Laboratory.
This
source can provide 5-55 MeV proton beam with a profile of up to 4" in diameter
and the beam is uniform to about 5-10%. The cost of beam time for physics
use could be free if the test schedule is flexible OR
a charge of $637/hr could buy high priority beam time. A chunk
of minimum 8 hours has to be scheduled. Contact: Peggy
McMahan, Research Coordinator <p_mcmahan@lbl.gov>. A
more detailed description of this source.
-
Crocker Proton Cyclotron at University
of California, Davis. This source can provide 63 MeV proton beam
and deliver a uniform flux over a diameter of 7 cm. The flux can be in
the range 3x105--2x1010 protons/cm2/sec.
The cost of using this facility is $500/hour. Contact:
Carlos Castaneda, <castaneda@crocker.ucdavis.edu> (530) 752-4228.
-
Indiana University
Cyclotron Facility . This source can provide 30-200 MeV proton beam
with a flux in the range 1E6 to 1E11 p/cm2/sec. The beam size could be
less than 2 cm to 7 cm in diameter. The uniformity variation is less than
30%. The cost of using this facility is $1000/hr and no less than
16 hours will be scheduled. Contact: Charles
Foster <foster@iucf.indiana.edu>.
-
Cyclotron in Svedberg Laboratory
at Uppsala University, Sweden. The cyclotron provides 25-180
MeV proton beams and with a maximum flux of 108 protons/cm2/sec
in a solid angle of 0.5 msr measured for a beam diameter of 7.5 cm. The
flatness of flux over the beam profile is about 10%. Mono-energetic
neutrons could also be obtained from nuclear reaction7Li(p,n)7Be.
The
neutron flux from this source is about 104 neutrons/cm2/sec.
The cost of using this facility is free for pure scientific research.
Contact:
Per-Ulf Renberg <renberg@tsl.uu.se> A
more detailed description of this neutron source .
-
Tandem Van de Graaf Accelerator
at Brookhaven National Lab provides proton and deuteron beams with
a maximum energy of 29 MeV. Using the same
method as CERI, TVDG could provide neutron source of E(max)=29
MeV and E(mean)=13 MeV. This source
has a tight schedule at this moment.
-
A list of neutron sources http://neutrons.ornl.gov/NSatHFIR/othern.html
Related links: Radiation
Hard Electronics page at CERN by Martin Dentan.
Created by Yong Li of
UCI on January 26, 2000.