Present:
B. Allongue
CERN
A. Bigga
CERN
N. Bingefors
Uppsala
G. Blanchot
Barcelona
I. Crotty
CERN/RS California
M. Dentan
CERN-Saclay
P. Farthouat
CERN
P. Fontaine
CERN
B. Hallgren
CERN
B. Hajduk
Cracow
M. Huhtinen
CERN
P. Jarron
CERN
M. Letheren
CERN
N. Lupu
Technion
R. Richter
MPI Munich
C. Rivetta
CERN
G. Stefanini
CERN
H. Takai
BNL
Date of next PROSPERO irradiation: october, 20th, 1999 (wednesday),
see below.
1. B. Allongue: Future test program for power supplies
Bruno briefly summarized the results of the PROSPERO run of september,
30th, 98, where seven different power supply types had been tested, one
unit of each up to about 2 * 10^11 n/cm^2 (nominal ATLAS dose in 10
years, including a safety factor of 4) and another one up to 2.3*10^12
n/cm^2.
All units survived the lower dose and operated inside specs. Units
started to fail beyond 6 * 10^11 n/cm^2. Only one unit, made by CEA, survived
up to the end (see prev. minutes for details). It is now clear that all
failures are caused by the failure of optocouplers used in the circuits
(Siemens and Toshiba). The HP optocoupler 6N138/139, however, is now proven
to be more rad.tolerant by a factor of > 5 (see presentation by B.Hallgren,
below).
For the irrad. test at PROSPERO, foreseen for october 99, a number
of modified commercial PS units has been prepared. The modification is
that the previously used optocouplers are replaced by the HP coupler,
mentioned above. 10 PSs from DELTA, 2 from LAMBDA and 4 from Melcher will
be included in the test. In addition, 6 unmodified VICOR, 2nd generation
PSs will also be tested.
The test setup is similar to the one used in sept., 1998, allowing
online supervision of all relevant parameters. Some of the PSs are in series
with ACTIL ripple filters, to see, whether these filters could be used
in the case that the ripple of the PSs was too high for our applications.
A PS from WIENER, containing some custom defined controls, arrived
at CERN, and will be included in the test, if the company agrees to this
potentially destructive test and if there is enough time for an extensive
test of the unit at CERN beforehand. The same is true for a number of units
from CAEN, which are expected to come to CERN in the next 2 weeks.
Discussion:
Remark (G. Stefanini): Power MOSFET in the PSs may fail at higher n-energies
than available at PROSPERO (< 3 MeV). Preliminary results from measurements
done at ETH show that certain MOSFETs stand much higher ionising events
than others. Derating from the nominal operating voltage may strongly decrease
the failure rate, because gate oxyd thickness is the crucial parameter
for destructive SEGR (single event gate rupture). These results will be
published in the next weeks and will be distributed (RR gets a copy).
Remark: Neutrons up to 35 MeV are available at the CERI facility
(Orleans, France). An experimental study of SEE should follow the PROSPERO
studies and could be done end-of-99/beginning-of-2000.
2. Helio Takai: Test of VICOR bricks (rad. & magn. field)
Helio reported about tests, which were mainly done by Mauro Citterio
at BNL.
The VICOR 2nd generation PS units are improved relative to the 1st
generation insofar as optocoupling has been replaced by transformer coupling
in the feedback loop. One of the 2nd gen. units was tested and survived
10^13 n/cm^2 and 100 krad. It operated at 100 kHz switching frequency,
300 V input, 12 V output voltage and was loaded with 2 A (about 10% of
maximum). It is clear that tests of a larger number of units ("statistics")
would be very useful. Also tests with a higher load, corresponding to the
situation in the experiment, will be necessary.
The n-irradiation was done at a synchrotron. The energy spectrum of
the neutrons will be communicated to our working group.
The use of active ripple filters (R.A.M.s) improved the noise from
the LAr FE amplifiers only by an negligible amount (0.5%). This, however,
may be a consequence of the many local voltage regulators used in the system,
which may have a similar ripple rejection like the R.A.M. filters.
Concerning magn. field, this unit destructively failed at 400 G with
the field perpendicular to the block and at 180 G parallel to the block.
Inspite of the fact that it will be operated in an environment of about
50 G, this seems not to be a sufficient safety margin and passive magn.
shielding with soft iron will have to be applied.
The VICOR blocks, if selected, would have to be packaged in a suitable
form together with control electronics etc. The problem to find a company
for this work was also addressed.
3. B. Hallgren: LMB radiation test 1998 and 1999
With the exception of the opto-couplers the Local Monitor Boards (LMBs)
had well survived 2.3 * 10^12 n/cm^2, accumulated during the PROSPERO run
in sept. 98. Tests immediately after the run showed that the CAN protocol
was still operating correctly.
Looking for a replacement of the failing optocoupler, Bjoern irradiated
the following optocouplers at the TCC2 test area at CERN:
MOCD223 Motorola
HCPL-0731 HP (SMD package
of the 6N139, mentioned above).
The irradiation was extended up to about 20 krad and 3.3*10^12 n/cm^2.
While the CTR (current transfer ratio) of the Motorola coupler was
at 10% after 10^12, the HP was still at about 80% after 3.3*10^12 n/cm^2.
Though the neutron and gamma spectra at TCC2 are not known in detail (in
fact they vary from day to day depending on the HE beam parameters) this
measurement shows the superior performance of the HP coupler.
Bjoern pointed out that lowering the threshold for the opto-coupler
output signal (used in a digital mode) can give another important safety
factor, because the system could still reliably operate at CTRs as low
as 4 %.
Cost of the HP device is about 6.2 CHF, compared to 1.7 CHF for Motorola
and Siemens.
Bjoern reported on the rad. behaviour of the "High performance analog
MUX (ADG438F)" which failed at about 7 krad. This is in agreement with
measurements done in Saclay (see minutes of M. Dentan of radiation meeting
on sept. 3rd).
The AT90S2313 Microcontroller failed at about 12.5 krad because of
excessive power consumption. As it had safely survived 10^13 n/cm^2
at PROSPERO (sept. 98, see minutes), the failure is attributed to the
gamma dose, which was a factor 10 higher than the one received at PROSPERO.-
Early destructive failures under Co60 irradiation of this device were confirmed
by G. Blanchot.
Finally a number of SEEs have been observed in the CAN chip SAE81C91.
It could be rebooted by a power-off/power-on cycle and then continued to
work correctly.
In summary, the LMBs with the new optocoupler can be considered qualified
only for the Muon system, due to the low gamma doses in this area and because
of the proven tolerance against neutrons. However, potential damage from
neutrons with E > 10 MeV will still have to be studied.
Discussion:
Are there alternative producers of the Microcontroller, which give better
tolerance?- Bjoern will look at it.
Bjoern was also encouraged to have a new version of the LMB produced
with the HP opto-coupler incorporated.
4. N. Lupu: Irradiation of linear temp. sensors
The devices TMP35/TMP37/LM35D (Analog Devices) were tested at PROSPERO
(april 99) up to a dose of 2.3*10^12 n/cm^2.
Pre-rad tests had shown individual, constant offsets of +- 1 deg C
and a spread of the individual reading of about +- 0.1 deg C.
During the irrad. the sensors were constantly monitored. No particular
problems were observed, apart from the fact that the spread of individual
readings was 5 times higher than in the pre-rad test. After the irrad.
(several weeks) the spread was back to the pre-rad conditions. 31 sensors
out of 31 had survived the test in good conditions. There was a permanent
shift of about 2 deg C observed in the average temperature readings.
There were some isolated SEU errors observed during the irradiation
(e.g. output = 0).
The devices seem to be well suited for the application in the TGC temperature
control system.
5. M. Huhtinen: Simulation of SEU rates at LHC
Mika described the computational method, based on FLUKA. Simulated data
are compared to published SEU rates observed in experiments with heavy
ions. The energy distribution of ionising nuclear fragments (mainly recoils)
deposited in a predefined "sensitive volume" of typically 1*1*1 mycrometer^3
is observed for different projectiles like protons, neutrons and pions
as a function of projectile energy.
The main uncertainty of the method is the interpretation of the heavy
ion data w.r.t. a realistic definition of the energy threshold for SEU.
Energy loss of ions in Silicon:
Recoiling Si28 and Mg25, the latter from the (n, alpha)-pickup reaction,
give high energy deposits (up to 4*10^4 MeV/cm). The highest values of
deposited energy density come from projectiles with energies around 40
MeV.
Fragment spectra:
The distribution of fragment energies is computed as a function of
projectile energy. E.g.: while fragments produced by 5 MeV neutrons have
a sharp cut-off above 1 MeV fragment energy, those produced by 19 MeV neutrons
extend up to 5 MeV. Maximum fragment energies are produced by neutrons
around 200 MeV.
Fragment range and max. energy deposition are also given (see transparents).
The use of heavy ion data for comparision with computational results
is discussed in detail. The computed rates are compared to those of SEU
rates induced by 200 MeV neutrons in 18 different memory chips. The average
agreement is of the order of 15%.
The relative SEU rate is computed for different projectiles and projectile
energies. This allows the prediction of the SEU rate from one projectile
type to another. E.g.: the SEU rates of protons of 20 MeV/200MeV/10GeV
energy go like 0.45/1.0/1.2. Neutrons and protons are identical at 20 MeV
as well as at 200 MeV, while 200 MeV pions give a 2 times higher SEU rate
than protons. (Is there a qualitative explanation for this? One would expect
less recoil energy of a nucleus if hit by a pion compared to a proton (?).
Remark added by RR.) Previous publications had given a factor 5 for
this number, which can not be confirmed.
Concerning SEU rates, there seems to be no difference between mono-directional
flux (beam) and isotropic flux.
For comparision, the particle spectra in the CMS cavern were given,
which, for neutrons, show a fall-off above about 1 MeV and a second
peak in the n-spectrum around 100 MeV (similar to the situation in
the ATLAS hall).
A practical conclusion is, that SEU rate tests with e.g. protons of
60 MeV would be representative for all hadrons above 20 MeV.- No
SEU is expected from neutrons below 3 MeV.
The work presented in this talk will soon be published.
6. The next n-irradiation (R.R.):
The next PROSPERO run is foreseen for wednesday, oct., 20th. (October 13th, is not available at the Valduc center.)
The following projects have announced their interest in this run (active tests):
project institute contact
item(s) to be tested
1) ATLAS Clermont Ph. Rosnet Tile calorimeter
2) ESS CERN B. Allongue Power supplies
Please send us a mail a.s.a.p. if other projects want to participate (active of passive test). Please remember that the personal coordinates of "new" participants must be given to the Valduc center 4 weeks before the run.
Typical schedule:
Arriving at the PROSPERO site (close to Dijon) in the morning of oct,
19st (3 hours by car from CERN). Install and test the equipment.
oct., 20th: irradiation, whole day (9:00h to 16:00h).
Robert Richter