IEEE TRANSACTIONS ON NUCLEAR SCIENCE - FOR PEER REVIEW - FIRST REVISION 2
Modification of the LM124 Single Event Transients
by Load Resistors
F. J. Franco, I. L
´
opez-Calle, J. G. Izquierdo and J. A. Agapito
Abstract—The influence of a load resistor on the shape of the
single event transients was investigated in the LM124 operational
amplifier by means of laser tests. These experiments indicated
that, as a general rule, load resistors modify the size of the
transients. SPICE simulations helped to understand the reasons
of this behavior and showed that the distortion is related to the
necessity of providing or absorbing current from the load resistor,
which forces the amplifier to modify its operation point. Finally,
load effects were successfully used to explain the distortion
of single event transients in typical feed-back networks and
the results were used to explain experimental data reported
elsewhere.
Index Terms—Laser irradiation, LM124, load effects, opera-
tional amplifier, single event transients, two-photon absorption.
I. INTRODUCTION
O
PERATIONAL amplifiers (op amps) make up a set of
useful active devices for electronic design. In the case
of radiation-tolerant systems, many authors have investigated
the action of heavy ions or pulsed lasers on the generation of
single event transients (SETs) in popular operational amplifiers
[1]–[11]. In particular, one of the most studied devices is the
LM124 [12].
One of the main characteristics of the operational amplifiers
is their versatility. An example of this is the capability of these
devices to bias output loads if a critical current value, so-called
short-circuit current, is not exceeded. Previous papers have
dealt with the influence of the output resistance in comparators
showing that there is a clear dependence of the transient
shape on the resistance value [1], [13]–[15]. In fact, these
works revealed that the cross-section as well as the duration
of the single events strongly depend on the resistive load
connected to the device output. Sharing similar internal blocks,
operational amplifiers are not very different from comparators
so it would be interesting to investigate the way that the single
event transients are modified by the presence of a resistive
load.
This work was supported in part by the EMULASER project (CDTI/PNE-
034/2006) and by the Spanish MCINN through Grant CTQ2008-02578/BQU,
and Consolider SAUUL CSD2007-00013. Finally, I. L
´
opez-Calle gratefully
acknowledges the grant offered by the “Miguel Casado San Jos
´
e’’ private
foundation.
F. J. Franco, I. L
´
opez-Calle, and J. A. Agapito are with the Depar-
tamento de F
´
ısica Aplicada III, Facultad de F
´
ısicas, Universidad Com-
plutense de Madrid (UCM), 28040 Madrid (Spain) (e-mail: isabelcalle, monti,
agapito@fis.ucm.es).
J. G. Izquierdo is with the Servicio de Espectroscop
´
ıa Multifot
´
onica
y de Femtosegundo, CAI de Espectroscop
´
ıa, Facultad de Qu
´
ımicas, Uni-
versidad Complutense de Madrid (UCM), 28040 Madrid (Spain) (e-mail:
jegonzal@quim.ucm.es).
In recent years, the influence of this parameter was also
indirectly investigated by other authors, such as Boulghassoul
et al. [16]. In this paper, the value of an output feedback
resistor of the LM124 & OP-27 placed on an actual satellite
application was changed in order to find the worst case
situation. Thus, it was discovered that the trend was that the
lower the resistor value, the smaller the transients although
no explication of this behavior was provided. In a previous
paper [5], Sternberg et al. simulated an LM124 inside a non-
inverting network with a gain of 11 and changed the resistor
values. Thus, SETs were simulated using several pairs of
resistors ranging from 10 kΩ-1 kΩ to 1 MΩ-100 kΩ observing
a strong dependence on the resistor values even though the
ratio between them was the same. This astonishing result was
attributed by the authors to a coupling between the resistance
values, gain, and bandwidth.
However, some experiments performed at the Universidad
Complutense de Madrid show that the dependence of the SETs
on the feed-back network resistor values could be also related
to simple load effects on the output of the operational ampli-
fier. In fact, the authors believe that, at least in large transients
involving changes of output voltage sign, the distortion of the
SETs is linked to the size of the current provided or accepted
by the output of the operational amplifier, which helps some
of the internal transistors to recover the stable state. Besides,
these results seem to be in agreement with the experiments
reported by Buchner et al. in 2008 [17].
II. EXPERIMENTAL SET-UP
A. Laser configuration
The experiments were performed at the UCM Multiphotonic
Spectroscopy and Femtosecond Facility using a Ti:Sapphire
laser followed by a regenerative amplifier. The laser wave-
length is tunable between 300-3000 nm. For two-photon ab-
sorption processes in silicon, 60-fs laser pulses at a frequency
of 1 kHz and a wavelength of 1300 nm was fixed. The energy
was measured with a typical commercial powermeter and set
to 1.2 nJ.
The device was mounted on a motorized xyz stage with 0.1
µm resolution and it could be observed with an infrared CCD
camera to allow the correct placement of the laser. Laser beam
was focused with a 50x long working-distance microscope
objective, appropriate for infrared light and making the spot
diameter on the order of 1.5 µm. Afterwards, a sweep along
the z-axis was performed in order to store a large set of output
transients and, this way, to statistically validate the results.
This kind of test could be done since the laser wavelength
IEEE TRANSACTIONS ON NUCLEAR SCIENCE - FOR PEER REVIEW - FIRST REVISION 3
XYZ
x50
CCD
60 fs
1.3 µm
1 kHz
Filters
White
light
probe
Trigger
GPIB
Fig. 1. Laser configuration and test set-up.
J
1
R
L1
V
IN
J
N
R
LN
o
o
o
o
o
o
Fig. 2. Electric configuration of the LM124 operational amplifier during
the laser irradiation. Jumpers were used instead of other devices such as
analog switches due to the absence of parasitic capacitances. All of them
were removed for unloaded configuration.
was chosen to induce two-photon absorption processes, able
to induce transients in buried layers of the device [3]. Fig. 1
shows a graphical description of the laser system.
B. Electronic set-up
The LM124 operational amplifier, in quad CERDIP pack-
age, was mechanically decapsulated and tested as follows. The
amplifier to be tested was configured as a buffer the input of
which, V
IN
, was connected either to ground or to an external
source of –1 V. These values were selected in order to modify
the bias point of the output stage as it will be explained later
(Section IV-A). A set of jumpers allowed the selection of
the 1%-tolerance resistors with different values to load the
operation amplifier output (Fig. 2). Finally, the device was
biased with ±15 V power supplies. The other three operational
amplifiers were configured as voltage followers and the input
connected to ground.
The output was connected to a digital oscilloscope with 8-
pF probes that was triggered by means of an external signal
coming from the laser (Fig. 1). Data preceding the laser impact
no longer than 10 µs were also saved to determine the DC
output value before the event. Given that the step was 10 ns,
every set of data amounted to 10.000 points that were saved
by a specific LabView application, which also controlled all
the devices by means of the GPIB protocol.
Q04Q03
Q18 Q20
Q17 Q21
Q05
Q06
QR1
Q09
C
Q12
I
BO
R
SC
Q13
Q14
Q11
I
BIN
I
BG1
I
BG2
V- V+
+V
CC
OUT
-V
EE
Fig. 3. Simplified LM124 diagram. The values or I
BIN
, I
BG1
, I
BG2
and
I
BO
are 6, 4, 100 & 50 µA. On the other hand, R
SC
= 20 Ω. Finally, a
resistor, connected to the base of Q09 and called R1 at the manufacturer’s
datasheet, turned out to be a open-base NPN transistor [20].
Q03
Q02
Q04
Q05
Q06
Q07
Q08
Q09
Q10
Q11
Q12
Q13
Q14
Q15a
Q15b
Q16a
Q16b
Q19a
Q19b
Q19c
Q19d
Q19e
I
REF
Q17
Q18 Q20
Q21
QR1
+V
CC
-V
EE
OUT
V
-
V
+
25
40k
C
C
Fig. 4. Actual structure of the LM124 according to Savage et al. [20]. More
information about the current source, I
REF
, can be found in the original
paper. The transistors hit by the laser are highlighted.
III. RESULTS
Four points were chosen to investigate the effects of the
resistive load on the shape on the laser induced transients.
These points were the bases of Q09, Q18, Q20, and QR1,
the first being a transistor of the gain stage, the other two
transistors at the differential pair of the input stage and the
last an open-base transistor working as a resistor in the gain
stage. Fig. 3 shows the simplified structure provided by the
manufacturer in the device datasheet. We have preferred not
to follow the nomenclature used by the manufacturer but that
used in most of the papers [18]. For instance, Q09 in Fig. 3 is
called Q12 at the datasheet by National Semiconductors [12].
The actual structure has been determined by several authors
[7], [8], [10], [19], [20], although there are minor changes
among the schematics provided by the different works. In
particular, Fig. 4 shows the structure provided by Savage [20].
Finally, Fig. 5 shows the exact placement of the transistors at
which the laser was aimed.
Once the resistive load was set using the jumpers, a z-
scan was performed with a step of 5 µm penetrating in the
device. At each depth value, the data after the laser impact
were saved in the hard disk of the computer to be analyzed
IEEE TRANSACTIONS ON NUCLEAR SCIENCE - FOR PEER REVIEW - FIRST REVISION 4
Fig. 5. Microphotograph of the LM124 lay-out. Spots where the laser was
focused are marked with black dots. P1, P2, P3, and P4 are the bases of Q09,
Q18, Q20, and QR1 [9].
by a specially developed SCILAB program in order to obtain
typical parameters such as the peak voltage, full width-half
maximum (FWHM), etc.
According to the results, there was not a significant differ-
ence among the output transients observed after hitting Q18 or
Q20. Thus, similar transients to those depicted in the literature
[9] were registered and the duration and size of the transients
seemed to be independent of the load connected to the output.
On the contrary, the behaviors of Q09 & QR1 are by far much
more interesting.
0 10 20 30 40 50 60
-16
-14
-12
-10
-8
-6
-4
-2
0
3.3 k
4.7k
10 k
47 k
OUTPUT VOLTAGE (V)
TIME ( s)
No load
100 k
Q09
INPUT 0V
Fig. 6. Output transients observed after hitting Q09 at a depth of 30 µm.
Every line is associated with a resistive load. The input was 0 V.
A. Transients at Q09
Transients related to this transistor are fast drops down
to a value close to −V
EE
followed by a slower recovery
0 10 20 30 40 50 60
-16
-14
-12
-10
-8
-6
-4
-2
0
3.3 k
47k
100 k
4.7 k
OUTPUT VOLTAGE (V)
TIME ( s)
No load
10 k
Q09
INPUT -1V
Fig. 7. Output transients registered after hitting the base of Q09 at a depth
of 30 µm with an input voltage of –1 V.
until reaching the DC output voltage, the speed of which
is determined by the operational amplifier slew rate (∼ 0.3-
0.5 V/µs). Fig. 6 shows the modification of the single event
transients as the load resistance decreases. In general, the
effects of the resistive load are mainly two:
1) A fast recovery during the first steps of the transients
before the signal changing according to the slew rate
value.
2) A hump at the end of the transient, making the output
value almost constant when V
OU T
/R
L
≈ −60 µA.
Later, the input voltage switched to V
IN
= −1 V, the
results of this experiment being shown in Fig. 7. The main
conclusions that we can derive from this figure are:
1) The transients of the unloaded amplifier seems to be
independent of the input value with the only fact that
transients with V
IN
= −1 V are a little shorter given that
the DC output voltage is closer to the negative saturation
voltage.
2) Loads of 47 kΩ & 100 kΩ induce shorter and smaller
transients than lower load resistor values.
3) Transients associated with loads of 3.3, 4.7 & 10 kΩ do
not have the characteristic hump observed in Fig. 6.
In general, transients in unloaded amplifiers are bigger than
those observed in loaded ones. Fig. 8 shows that the points
related to the transients approach to the y-axis as the load
resistance decreases. Besides, very few transients reach the
negative power supply value if the resistance value is low.
B. Transients at QR1
Those transients induced on the base of QR1 are usually
short spikes that lead the output voltage up to the positive
saturation voltage [8], [9]. They are followed by a swift
decrease down to 0 V, sometimes followed by a negative
transient, smaller but longer. The reason of this transient is
that the activation of QR1 takes the whole of the Q09 base
current, momentarily cutting this transistor off so the output of
the gain stage goes to a high positive value. These transients
were recreated at the laser facility and the results of the
IEEE TRANSACTIONS ON NUCLEAR SCIENCE - FOR PEER REVIEW - FIRST REVISION 5
0 2 4 6 8 10 12 14 16
-15
-12
-9
-6
-3
0
Q09
INPUT 0 V
No Load
100 k
47 k
10 k
4.7 k
3.3 k
AMPLITUDE (V)
DURATION
(
s
)
Fig. 8. Amplitude vs. FWHM of all the transients induced on the base of
Q09. Once set the load resistor value and placed the laser over Q09, the laser
performed a z-sweep to obtain a large set of transients with different shapes.
0 10 20 30 40 50
-15
-12
-9
-6
-3
0
3
6
9
12
15
10k
3.3k
4.7 k
47 k
OUTPUT VOLTAGE (V)
TIME ( s)
No load
100 k
R1
INPUT 0V
0 10 20 30 40 50
-15
-12
-9
-6
-3
0
3
6
9
12
15
10k
3.3k
4.7 k
47 k
OUTPUT VOLTAGE (V)
TIME ( s)
No load
100 k
R1
INPUT -1V
Fig. 9. Induced single event transients at a depth of 20 µm below the base
of QR1. The input voltage was 0 V & –1 V.
experiment are shown in Fig. 9. According to these results, it
seems clear that the second stage of the transients, where the
output voltage is negative, is attenuated as the load resistance
decreases. However, in the case of using an input voltage of
–1 V, the characteristics corresponding to loads of 100 & 47
kΩ show humps resembling those depicted in the previous
sections and making the transient longer than usual.
0 10 20 30 40 50 60
0
2
4
6
8
10
12
14
16
3.3k
4.7k
10 k
47 k
OUTPUT VOLTAGE (V)
TIME ( s)
No load
100 k
R1
INPUT 0V
0 10 20 30 40 50 60
-2
0
2
4
6
8
10
12
14
16
4.7k
10 k
47 k
OUTPUT VOLTAGE (V)
TIME ( s)
No load
100 k
QR1
INPUT -1V
Fig. 10. Transients originated below QR1 at a depth of 30 µm with an input
of 0 V & –1V and different values of resistive loads.
Other interesting transients are those originated at a depth
of 30 µm (Fig. 10), which are unusually large and strictly
positive. Provided that these transients appears in a narrow
interval of z-values, it is likely that these events are related to
a charge generation near a buried layer.
If there is no load, the transient is just a quick increase
followed by an interval at the positive saturation voltage
finally, and ending with a slow decrease to a stable value ac-
complishing the slew rate limitation. Besides, the dependence
of the transients on the input voltage is not significant. In fact,
the transient associated with V
IN
= −1 V is longer since the
DC value is a bit farther from the positive saturation voltage
than the ground. However, as the load resistance decreases,
the original transient seems to split up into two parts: The
original spike associated with QR1, quite insensitive to the
load, and a slower signal the size and duration of which is
strongly affected.
The transients show some interesting facts: First of all, the
line corresponding to a 3.3 kΩ resistor is just a short and small
spike. In fact, most of the transients registered during the z-
sweep with other loads vanished when this resistor value was
used instead. Besides, lines related to the 4.7 & 10 kΩ in Fig.
10 with V
IN
= −1 V also show a little hump prior to recover
the original value. As in the case of Q09, this phenomenon
occurred when V
OU T
/R
L
≈ −60 µA.
IEEE TRANSACTIONS ON NUCLEAR SCIENCE - FOR PEER REVIEW - FIRST REVISION 6
Q09
Q12
I
BO
R
SC
Q13
Q14
Q11
I
BG2
+V
CC
OUT
-V
EE
R
L
I
O
V
O,G
Fig. 11. Output stage of the LM124 with a load resistor.
IV. DISCUSSION
Experimental data show that the load resistance affects the
shape of the output transients. With the help of the analysis
of the output stage and simulations, a theory to explain the
distortions and attenuation of the single event transients, and
based on the switch of some transistors to anomalous states,
will be developed in the following sections.
A. DC behavior of the output stage
Unlike most of the typical general-purpose operational am-
plifiers that make use of a class AB output stage, the LM124
output stage is a combination of two emitter followers that
cannot simultaneously work. The way that the stage operates
depends on the value of the output current, I
O
, (Fig. 11). In
fact, if the output current is positive or, at least, negative but
lower than I
BO
, Q13 & Q14 must be in forward active zone
to accomplish the Kirchoff’s current law at the OUT node.
Thus:
V
OU T
≈ V
O,G
− 2V
BE,(ON)
(1)
V
O,G
being the output of the gain stage (Q09 collector) and
V
BE,(ON)
the voltage drop at a forward biased base-emitter
junction (on the order or 0.6-0.7 V). A consequence of it is
that Q11 is in cutoff state since V
EB,Q11
≈ −1.2 V (The
voltage drop along R
SC
is negligible). However, if the output
current is negative and higher than I
BO
, the only way to drain
the excess of current towards −V
EE
is through Q11, a PNP
transistor. This fact changes the operation point of the output
stage since, in this situation, the mathematical relation between
V
O,G
and V
OU T
becomes:
V
OU T
≈ V
O,G
+ V
BE,(ON)
. (2)
In this case, Q13 & Q14 switch to cutoff state since the
voltage difference between the Q14 base and the Q13 emitter
is on the order of -0.6 V. Finally, the purpose of Q12 is to
protect the device from overcurrent. If something forces the
operational amplifier to provide too much output current, a
voltage difference appears between the base and the emitter
of Q11 due to the presence of R
SC
in such a way that the
Q12 collector grabs some current from the base of Q14 making
the output current decrease. Besides, when the amplifiers does
TABLE I
RELATIVE TRANSISTOR AREA FOR SPICE SIMULATIONS. TH E REST OF
TRANSISTORS H AVE A RELATIVE AREA OF 1.0.
Transistor Area Transistor Area Transistor Area
Q11 2.0 Q16B 1.3 Q19C 0.11
Q15B 3.0 Q19A 0.3 Q19D 0.12
Q16A 1.3 Q19B 0.11 Q19E 0.12
not provide a lot of positive output current, the voltage drop
through R
SC
is very low so the transistor remains in cutoff
state. Finally, if Q11 is in forward-active zone instead of
Q13 & Q14, Q12 switches to reverse-active zone since the
collector-base voltage is on the order of 0.6-0.7 V.
B. SPICE simulations
SPICE simulations were performed in order to find out the
reasons of the change of the SETs. The internal topology of the
LM124 (Fig. 4) has been depicted in several papers [7], [8],
[10], [19], [20] although scarce information is provided about
the characteristic of the internal transistors. In fact, the typical
procedure is separating the individual transistors by means of
laser or ion beams and extract the SPICE parameters using a
microprobe and specific instrumentation [18]. Unfortunately,
this technology is not at the authors’ disposal. However, given
that the purpose of the simulations is to broadly understand
the behavior of the device, we proceeded to do the following
approaches:
1) All the NPN or PNP transistors were identical except
area ratios.
2) Spice parameters were the typical values of a 5Ω-cm, 17
µ-epi 44-V technology [21]. Junction capacitances were
reduced in order to stabilize the amplifier.
3) Mirror transistor areas were trimmed so that the bias
currents were those of Fig. 3.
The transients were simulated using an 1-ps rise & fall time
triangular current source between the involved nodes. The base
resistance was removed of the transistor model and placed
outside [22]. The by-pass capacitor was calculated by means
of the slew rate value (18 pF) [23], in agreement with the
value suggested by other authors [20].
Even after all these simplifications, the results are meaning-
ful. Fig. 12 shows the simulation of a SET on the bases of Q09
& QR1, which are similar to the experimental results (Fig. 6).
In general, the simulations recreate most of the experimental
results except those related to parasitic elements, not included
in the SPICE netlist. This way, most of the typical QR1 SETs
were emulated except those large transients depicted in Fig.
10. Probably, the reason of this failure is that a unique current
source cannot modelate a multi-junction charge collection
process [24].
C. Transients at Q09
Usually, the distortion of the transients related to the change
of the load resistance, either pure load or resistive network,
are explained supposing a coupling effect. However, a careful
study of the behavior of the transistors during the transients
![Fig. 4. Actual structure of the LM124 according to Savage et al. [20]. More information about the current source, IREF , can be found in the original paper. The transistors hit by the laser are highlighted.](/figures/fig-4-actual-structure-of-the-lm124-according-to-savage-et-1xlm9233.webp)
![Fig. 3. Simplified LM124 diagram. The values or IBIN , IBG1, IBG2 and IBO are 6, 4, 100 & 50 µA. On the other hand, RSC = 20 Ω. Finally, a resistor, connected to the base of Q09 and called R1 at the manufacturer’s datasheet, turned out to be a open-base NPN transistor [20].](/figures/fig-3-simplified-lm124-diagram-the-values-or-ibin-ibg1-ibg2-227oefeh.webp)
