Capacitor-based Isolation Amplifiers for Harsh Radiation Environments Francisco J. Franco∗, Yi Zong, and Juan A. de Agapito∗ Departamento de F́ısica Aplicada III, Facultad de Ciencias F́ısicas, Universidad Complutense de Madrid, Ciudad Universitaria, 28040 Madrid (Spain) Abstract Commercial-off-the-shelf (COTS) capacitor-based isolation amplifiers were irradiated at the Por- tuguese Research Reactor (PRR) in order to determine its tolerance to the displacement damage and total ionising dose (TID). The set of experimental data shows that some of these devices are suitable for zones inside future nuclear facilities where the expected total radiation damage would be below 2.2·1013 1-MeV neutron/cm2 and 230 Gy (Si). However, some drawbacks must be taken into account by the electronic designers such as the increase of the output offset voltage and the slight modification of the transmission gain. Keywords: COTS, Displacement Damage, Isolation amplifiers, Total Ionising Dose (TID). PACS: 29.90.+r, 28.52.Lf, 28.41.Rc 1. Introduction1 In electronic design, it is often necessary the use of analog subcircuits with separated grounds.2 Thus, the typical low voltage instrumentation systems are protected against high common-mode3 voltages of the measured signal. Also, separated grounds easily break ground loops removing4 interferences or parasitic signals in the measurement circuits. In this framework, isolation amplifiers5 are an important tool to deal with separated grounds. These are a family of devices able to capture6 an analog signal value from a subsystem and accurately transmit it to the other subsystem with7 its own ground, jumping the barrier of the high common-voltage value. Some models are also8 designed to provide a power supply to bias active sensors as well as the signal conditioner placed9 in the system with the isolated ground [1].10 Large systems such as particle accelerators, nuclear facilities, etc. contain instrumentation11 systems that need insulation between different stages and that are also exposed to radiation such12 ∗Corresponding author. Tel.: +34913944434; fax: +34913945196. E-mail address: monti@fis.ucm.es Preprint submitted to Nuclear Physics A February 26, 2015 2 INTERNAL STRUCTURE OF AMPLIFIERS 2 − + osc. − + S/H S/H G=1 G=6 Is ol at io n OUT IN RIN ROUT IR 2IR 2IR IR A B CD E CIN COUT IIN IX Figure 1: Internal structure of a typical ISO12X according to the manufacturer [6]. as particles or very energetic photons (Gamma, X rays). Nowadays, the state-of-the-art offers13 three ways to create a barrier between the input and the output stages of the isolation amplifiers:14 Optocouplers, coupling-transformers or capacitors [2]. Optical based isolation amplifiers are not15 recommended for nuclear facilities given the high sensitivity of the optical devices to the displace-16 ment damage caused by ions or neutrons [3]. Besides, other papers have dealt with the effects of17 the radiation damage on isolation amplifiers with coupling-transformers [4, 5] although the large18 size and cost of these devices can make their use inadvisable. Finally, capacitor-based isolation19 amplifiers are an alternative choice although a study of their behaviour under radiation is necessary20 to advise or discard their use in electronic systems to be exposed to radiation.21 2. Internal Structure of Amplifiers22 This technology was developed by Texas Instruments to build some of its interface devices,23 either digital or analog. Analog isolation amplifiers make up the ISO12X family and the data24 shown in this paper were focused on the ISO122 & ISO124 devices, the datasheets of which can25 be found on the manufacturer’s website [6]. These two devices are quite similar given that the26 internal block shown in Figure 1 is implemented in both of them. Actually, the only difference27 between them is that the elementary devices inside the amplifiers such as resistors, capacitors, etc.28 are more accurately built in the case of the ISO124.29 According to the manufacturer, the principle of working is the following: The amplifier A30 creates a virtual ground at the right side of RIN in such a way that a current IIN = VIN/RIN31 flows into the isolation amplifier. This signal is added to a current IX , the value of which is ±IR32 depending on the state of the comparator D that control two current sources, 2·IR & IR. IX + IIN33 2 INTERNAL STRUCTURE OF AMPLIFIERS 3 + − + − RIN VOS,IN VIN VOUT KVIN+VOS IIN Figure 2: Simplified equivalent macro model of ISO12X, useful for hand analysis. is used to charge and discharge a capacitor, CIN , connected to the output of the amplifier A. This34 node is also connected to a comparator (E ) that evaluates the difference between the output of A35 and a 500-kHz wave generator. Thus, a square signal with a duty cycle depending on the size of36 IIN and, evidently, on VIN is obtained at the output of E.37 This signal as well as its complementary is transmitted through the isolation barrier by means38 of a couple of capacitors so they reach the inputs of another comparator (C ), which acts as a buffer39 to recover the signal. The width-modulated square signal is decoded using several devices in such40 a way that the initial voltage value is regenerated at the output node of the isolation amplifier.41 Unfortunately, it is impossible to reach the internal devices without destroying the isolation42 amplifier. Thus, a simple macro model (Figure 2) containing as much information as possible was43 developed to evaluate the degradation of the device and to allow a later use in simulations or hand44 calculations. In this structure, RIN is the input resistance shown in Figure 1. VOS,IN is the input45 offset voltage of the A operational amplifier. Ideally, the voltage value at the inverting input of46 this operational amplifier should be 0 V but, due to the input offset voltage, the voltage value at47 this node is not 0 but VOS,IN [7] and can be measured as it will be later shown. The output stage48 is modeled by means of a voltage-controlled voltage source the gain of which is ideally K = 1.49 An additional output offset voltage, VOS, is included to take into account non-idealities and other50 defects of the output stage.51 It must be highlighted that the “input offset voltage” given by the manufacturer in the datasheets52 is just the “output offset voltage” defined in this paper. Actually, the offset voltage in the input53 operational amplifier only affects the function associating VIN with IIN . In other words, the input54 characteristic. In fact, even though large values of VOS,IN were measured, the output voltage with55 zero input was very close to 0 V. The input offset voltage could affect the size of the input current56 3 SET-UP FOR THE ON-LINE TESTS 4 in such a way that the modulated-width square signal is distorted. However, the decoding of the57 transmitted information is made using a similar operational amplifier at the output stage. Given58 that this amplifier has been built in the same wafer as the first one, both devices would be carefully59 matched so the error introduced by the first amplifier is removed by the second one. Thus, even60 if the offset voltage of the input operational amplifier is often beyond 100 mV, the offset voltage61 of the complete isolation amplifier never exceeds the typical values provided by the manufacturer62 (50 mV).63 Finally, ideal isolation amplifiers have a transmission coefficient, K, equal to 1. However, in64 actual devices this value is never accomplished being usually above or beneath this value. An65 additional parameter, called “typical output error”, the meaning of which will be explained later,66 was also measured along with the transmission coefficient, K.67 All the parameters depicted in the previous paragraphs were measured on-line but, once the68 devices could be safely handled, more parameters were measured. Some of these parameters were:69 • Power supply rejection ratio (PSRR).70 • Insulation between the stages (IMRR, insulating impedance and electric breakdown field).71 • Quiescent current, parameter related to the power consumption.72 • Frequency behavior73 • Output noise74 3. Set-up for the on-line tests75 3.1. Description of the irradiation facility76 Both kinds of isolation amplifiers were tested at the neutron facility of the Portuguese Research77 Reactor [8] using three samples of each model. These samples, which belonged to the same batch,78 were mounted on different printed circuit boards and distributed along a cylindrical cavity with the79 goal of irradiating each sample with a different total radiation dose. The irradiation took about80 20 h split in three rounds followed by technical reactor shutdown periods. Thus, the samples81 received the total radiation dose shown in Table 1. The neutron fluence was obtained with 58Ni82 foil detectors and multiplied by a factor of 1.27 to express the neutron fluence in standard 1-MeV83 n/cm2 units [5, 8]. The total ionising dose was measured by an ionisation chamber. From now on,84 3 SET-UP FOR THE ON-LINE TESTS 5 Table 1: Total radiation dose and dose rate received by the samples. Sample Neutron Fluence TID Dose Rate TID/N.F. A 2.20 236 11.8 107.3 B 0.95 148 7.4 155.8 C 0.34 104 5.2 305.9 ·1013 1-MeV n/cm2 Gy(Si) Gy(Si)/h ·Gy/1013 1-MeV n/cm2 the total radiation dose will be expressed in units of 1-MeV n/cm2, the TID value being calculated85 using the ratios of TID vs. neutron fluence found on Table 1.86 The temperature was measured with PT-100 resistive temperature detectors distributed along87 the facility cavity, which has an injecting-air cooling system so the temperature kept stable around88 26-27 oC during the whole radiation.89 3.2. Acquisition system set-up90 All the printed circuit boards had separated ground for the input & output stage, and a couple91 of ±15 V power supplies to bias the devices. These power supplies were not switched off until92 the end of the test. During the irradiation, the devices were characterised every ten minutes by93 an acquisition system consisting in a personal computer, an accurate digitally controlled voltage94 source, two precision multimeters, and a matrix switching system, all of them controlled by a95 general purpose interface bus (GPIB). The distance between the samples at the reactor cavity and96 the instrumentation system was on the order of 3-4 m so low-resistance shielded pipes were used97 to connect both parts. It is necessary to say that all the voltages were measured on the boards.98 This fact is especially important in the case of the input voltage, which was not measured at the99 input voltage source but directly on the board. Also, the isolation amplifiers were disconnected100 from the input source and voltmeters using mechanical relays and connected again only during the101 interval needed to characterise the devices.102 This system performed a DC sweep at the input voltage from –1 V to +1 V with a step of 0.2103 V to obtain the transmission coefficient, K, and the output offset voltage, VOS, with a linear fit104 after the data coming from the multimeters.105 These linear fits also allowed the calculation of the Typical Output Error (∆VOUT ), defined as106 follows. Supposing that there are N pairs of input and output values (VIN,k, VOUT,k) that were107 4 EXPERIMENTAL RESULTS AND DISCUSSION 6 + − RIN VOS,IN VOUT RS VA VIN SA Figure 3: Test set-up to measure RIN & VOS,IN . Using RS makes VA different from VIN and the values of RIN & VOS,IN are easily extracted. linearly fitted to obtain the values of K and VOS, ∆VOUT is:108 ∆V 2 OUT = 1 N − 2 N∑ k=1 [VOUT,k − (VOS +K · VIN,k)] 2 (1) In order to measure the input offset voltage, VOS,IN , and the input resistance, RIN , the pro-109 cedure was as follows: A mechanical relay connects the input source to the input of the isolation110 amplifier with RS, a 10-kΩ precision resistor (Figure 3). A voltage on the order of +1 V is set at111 VIN and a pair of voltages, VIN,1 & VA,1 are measured and stored. Immediately, the voltage source112 changes to –1V to measure a new couple of values, VIN,2 & VA,2. Using Kirchoff’s current law it is113 easy to demonstrate that the values of the unknown parameters are:114 VOS = VA,1 − α · VA,2 1− α (2) RIN RS = α 1 + α · VA,1 + VA,2 VIN,1 − VIN,2 (3) where115 α = VIN,1 − VA,1 VIN,2 − VA,2 (4) Initial values of the input resistance are shown in Table 2.116 4. Experimental results and discussion117 4.1. Transmission coefficient, K118 In an ideal isolation amplifier, the transmission coefficient is 1 in order to accurately regenerate119 the input signal at the output stage. However, actual devices do not accomplish this theoretical120 4 EXPERIMENTAL RESULTS AND DISCUSSION 7 Table 2: Initial values of the resistance, RIN . Sample ISO122 ISO124 A 196.0 ± 0.1 199.7 ± 0.1 B 169.9 ± 0.1 198.3 ± 0.1 C 178.0 ± 0.1 199.2 ± 0.1 kΩ kΩ 0,0 0,3 0,6 0,9 1,2 1,5 1,8 2,1 2,4 0,988 0,992 0,996 1,000 1,004 1,008 1,012 A B ISO124 C Tr an sm is si on C oe ffi ci en t, K Neutron Fluence (·1-MeV 1013 n/cm2) Figure 4: Transmission coefficient of the ISO124. requirement. In fact, the pristine samples of the ISO122 showed a scattering of this parameter121 between 1.000 & 1.004 (0.4%), this error being smaller in the ISO124 where the transmission122 coefficient values were between 1.000 & 1.001 (0.1%). Let us remember that the ISO124 is similar123 to the ISO122 with more accurately trimmed internal components.124 This can be the reason of the different behaviour of the transmission coefficient in both devices.125 Figure 4 shows the evolution of K at the ISO124. The value of this parameter keeps quite stable126 even at the most irradiated sample and only deviations up to 1.003 were registered in some of the127 devices. On the contrary, the evolution of K in the ISO122 (Figure 5) is much more problematic128 given that, a priori, it is impossible to know if this parameter will increase or decrease and that,129 at any rate, the shift in this parameter makes the value of K be placed between 0.990 & 1.009. In130 other words, the possible error goes beyond 1 %.131 In the authors’ opinion, this different behaviour is a consequence of the worse trimming of132 4 EXPERIMENTAL RESULTS AND DISCUSSION 8 0,0 0,3 0,6 0,9 1,2 1,5 1,8 2,1 2,4 0,988 0,992 0,996 1,000 1,004 1,008 1,012 A B ISO122 C Tr an sm is si on C oe ffi ci en t, K Neutron Fluence (·1-MeV 1013 n/cm2) Figure 5: Transmission coefficient of the ISO122. the internal devices of the ISO122. Probably, the radiation damage accentuates the mismatch133 between supposed similar devices making the transmission coefficient move away from the ideal134 value. Given that the internal devices of the ISO124 are better trimmed, the deviation is smaller.135 4.2. Offset Voltage, VOS136 Pristine samples of both devices have a typical output offset voltage between ±20 mV. Unfor-137 tunately, these limits are quickly exceeded in most of the samples. Figures 6 & 7 show the exact138 evolution of this parameter. Some important conclusions can be drawn from these figures. First139 of all, it is impossible to forecast the exact evolution of the offset voltage since it grows in some140 devices, decreases in other of them and keeps quite constant in one of the ISO122. In any case,141 it seems evident that the shift is larger in the ISO122 samples than in the ISO124: The ISO122142 offset voltage keeps between -40 & 140 mV whereas in the ISO124 the limits are wider (-130 &143 220 mV).144 The origin of this difference of behaviour would come from the same fact as the transmission145 gain. Offset voltages are caused by mismatches among the internal components of a specific device.146 Radiation damage makes these differences more significant changing the value of the offset voltage.147 These mismatches are unknown so, provided the great number of parameters involved in the value148 of the offset voltage, the final evolution is impossible to predict. This evolution of the offset149 voltage is very similar to that observed in irradiated operational amplifiers, especially those with150 JFET input stage where the matching is worse than those completely built with bipolar junction151 4 EXPERIMENTAL RESULTS AND DISCUSSION 9 0,0 0,3 0,6 0,9 1,2 1,5 1,8 2,1 2,4 -150 -100 -50 0 50 100 150 200 250 A B ISO122 C O ffs et V ol ta ge , V O S (m V ) Neutron Fluence (·1-MeV 1013 n/cm2) Figure 6: Offset voltage of the ISO122. 0,0 0,3 0,6 0,9 1,2 1,5 1,8 2,1 2,4 -150 -100 -50 0 50 100 150 200 250 A B ISO124 CO ffs et V ol ta ge , V O S (m V ) Neutron Fluence (·1-MeV 1013 n/cm2) Figure 7: Offset voltage of the ISO124. The Y-axis scale is similar to that of the ISO122 (Figure 6) to make the comparison between both of them easier. 4 EXPERIMENTAL RESULTS AND DISCUSSION 10 0,0 0,3 0,6 0,9 1,2 1,5 1,8 2,1 2,4 0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 ISO122 Ty pi ca l O ut pu t E rr or , V O U T, ( m V ) Neutron Fluence (·1-MeV 1013 n/cm2) Sample A Sample B Sample C Figure 8: Typical output error of the ISO122, ∆VOUT . transistors [9, 10, 11, 12, 13, 14].152 From the system designer’s point of view, the large values of the offset voltage are a serious153 concern. Fortunately, there are some state-of-the-art techniques that minimise this drawback, such154 as that depicted in [15].155 4.3. Typical Output Error, ∆VOUT156 In ideal isolation amplifiers, the value of this parameter is 0 V, situation that is never achieved157 in actual devices due to the output noise and the nonlinearity of the device.158 Before the tests, samples of the ISO122 showed a value of ∆VOUT about 0.25 mV while the159 other device showed a lower value, 0.1 mV. The evolution of this parameter is shown in Figures 8160 & 9 from which we can see that the value of ∆VOUT increases as the irradiation is carried out. In161 the case of the ISO122, the highest value is on the order of 0.65 mV, being lower in the case of the162 ISO124, where the value of ∆VOUT never went beyond 0.45 mV.163 The reason of this behaviour is not completely understood. It is well-known that all the164 irradiated electronic devices show a higher noise level due to the creation of defects inside the165 silicon lattice. However, the complexity of the device does not allow accepting that this is the166 only cause. E. g., ionising radiation can create finite impedance paths below the epitaxial oxide167 allowing the interferences of any of the 500-kHz signals at the output node.168 4 EXPERIMENTAL RESULTS AND DISCUSSION 11 0,0 0,3 0,6 0,9 1,2 1,5 1,8 2,1 2,4 0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 ISO124 Ty pi ca l O ut pu t E rr or , V O U T, ( m V ) Neutron Fluence (·1-MeV 1013 n/cm2) Sample A Sample B Sample C Figure 9: Typical output error of the ISO124, ∆VOUT . 4.4. Input resistance, RIN169 According to the manufacturer, the value of this parameter is 200 kΩ. However, actual devices170 do not accomplish this requirement (Table 2). In fact, the value of RIN in the ISO122 varies from171 170 to 200 kΩ although, in the case of the ISO124, the variation range is much smaller (198-200172 kΩ). This fact is clearly related to the more careful process used by the manufacturer to build this173 model.174 Concerning the effects of the radiation, no change was observed during the tests since their175 values kept constant until the end. Figure 10 shows the behaviour of the ISO124 input resistances176 as the irradiation was performed. The fact that the input resistance are implemented in metallic177 thin-film resistor technology explains the great tolerance of this part of the isolation amplifiers since178 it is commonly accepted that metals are insensitive to either displacement or ionisation damage179 [16].180 4.5. Offset voltage of the input operational amplifier, VOS,IN181 The values of these parameters were initially distributed between ±150 mV in all the tested182 devices. Unlike the offset voltage, the change was steady and monotonic without a constant shift183 rate that could strongly vary from one sample to another (Figure 11). For instance, the most184 irradiated sample of the ISO122 showed a shift rate of 1.02 mV/1013 n/cm2 while, in the second185 sample of the same device, the ratio was -9.44 mV/1013 n/cm2, more than nine times larger.186 4 EXPERIMENTAL RESULTS AND DISCUSSION 12 0 10 20 30 40 50 60 70 197 198 199 200 201 C B A Irradiation Sessions IS O 12 4 In pu t R es is ta nc e, R IN (k ) Time (h) Figure 10: Input resistances of the ISO124 during the irradiation. Because of unknown reasons, the noise level is greater during some periods. 0,0 0,3 0,6 0,9 1,2 1,5 1,8 2,1 2,4 -150 -120 -90 -60 -30 0 30 60 90 120 150 ISO122 A ISO124 C ISO124 B ISO122 B ISO122 C ISO124 A O ffs et V ol ta ge o f t he in pu t o pe ra tio na l a m pl ifi er , V O S ,IN (m V ) Neutron Fluence (·1-MeV 1013 n/cm2) Figure 11: Offset voltage of the input operational amplifier during the irradiation. 4 EXPERIMENTAL RESULTS AND DISCUSSION 13 4.6. Off-line parameters187 Other parameters could not be measured during the test. They were taken about one month188 later once that the radioactive isotopes generated during the irradiation had vanished and a safe189 handling could be done.190 4.6.1. Power supply rejection ratio (PSRR)191 This parameter measures the dependence of the offset voltages with the power supply values.192 In the case of VOS,IN , no significant change was found. In the case of the output offset voltage,193 the high noise level typical of these devices masked the little variations of this parameter needed194 to estimate the experimental PSRR value.195 4.6.2. Isolation barrier196 The main characteristic of these devices is the ability to insulate input and output stages.197 Therefore, it is essential to verify the integrity of the isolation barrier.198 The isolation mode rejection ratio (IMRR) is a parameter that evaluates the influence of the199 common-mode voltage, VISO, on the output and is defined as:200 IMRR = ∂VOUT ∂VISO (5) In order to measure it, a common-mode voltage between ±1000 V was applied between the201 stages with the purpose of measuring the slight output voltage variations. However, and as it202 occurred with the PSRR, the noise was so significant that we could only estimate that the IMRR203 was always larger than 120 dB in the pristine samples of both kinds of amplifiers. After the test,204 even the most irradiated samples did not show a lower value.205 To account for these results, we proceeded to calculate the value of the impedance between206 both stages. This impedance is modeled just as a resistor and a capacitor in parallel. The resistor207 was measured with a Kyoritsu high voltage insulation tester with the final conclusion that its value208 was, at least, higher than 100 TΩ in all the samples (This value is the upper measure limit of the209 instrument).210 The capacitor between the isolated grounds was measured by means of a Hewlett Packard 4192211 impedance analyser in the 10-100 kHz frequency range, the results being shown in Table 3. The212 first fact to bear in mind is that the actual value of the capacitors is higher than that specified213 by the manufacturer (2 pF). However, in any case they are very close so the discrepancy can be214 4 EXPERIMENTAL RESULTS AND DISCUSSION 14 Table 3: Total radiation dose and dose rate received by the samples. Sample Neutron Fluence ISO122 ISO124 – 0.00 3.3 ± 0.2 3.4 ± 0.2 A 2.20 3.2 ± 0.2 3.2 ± 0.2 B 0.95 3.5 ± 0.2 3.3 ± 0.2 C 0.34 3.2 ± 0.2 3.2 ± 0.2 ·10131−MeV n/cm2 pF pF attributed either to little errors during the manufacture or to the appearance of parasitic capacitors215 not included in the simplified ISO12X schematic. The second conclusion from Table 3 is that there216 is no evidence of modification of the capacitance since differences among the values in this table217 are always within the experimental error.218 Finally, the tolerance of the irradiated isolation barrier to very strong electric fields was tested219 applying a high voltage between the ground pins of each stage. The manufacturer guarantees the220 tolerance of these devices to common-mode voltages below 1500 V, fact that was confirmed by221 our measures. Indeed, we applied a common-mode voltage between ±5000 V for a few seconds222 without destroying even the most irradiated samples.223 4.6.3. Quiescent currents224 Before the irradiation, the typical quiescent current was on the order of 4.8-4.9 mA for each225 stage using ±15 V power supplies. Later, we observed a little decrease in this parameter. Thus,226 the most irradiated sample of the ISO124 only required 4.15 mA to work.227 4.6.4. Frequency behavior228 Figure 12 shows the dependence on the frequency value of the transmission coefficient obtained229 from the second sample of the ISO124. The gain was measured applying an input sinusoidal signal230 of an r.m.s. value of 100 mV and measuring the r.m.s output voltage. This sample was selected231 to be represented since it was that one where the degradation was greater. Figure 13 shows the232 evolution of the characteristic frequencies of the ISO124 samples. Similar results were found on233 the ISO122.234 Radiation damage always causes degradation in the frequency response of operational amplifiers235 and derived devices [17]. Therefore, it is possible that the degradation observed in the devices is236 4 EXPERIMENTAL RESULTS AND DISCUSSION 15 10k 100k 0,0 0,2 0,4 0,6 0,8 1,0 -3 dB Not Irradiated ISO124 Sample B A C T ra ns m is si on G ai n Frequency (Hz) -6 dB Figure 12: Bode diagram with linear Y-axis of the second sample of the ISO124. Pre and post irradiation lines can be found in the plot. The sample received 0.95·1013 1-MeV n/cm2. 0,0 0,4 0,8 1,2 1,6 2,0 2,4 55 60 65 70 75 80 85 90 K ey fr eq ue nc ie s in th e IS O 12 4 (k H z) Neutron fluence (·1-MeV 1013 n/cm2) ISO124 -3dB -6dB Figure 13: Frequency values where decreases of 3 & 6 dB were observed in the irradiated ISO124. 4 EXPERIMENTAL RESULTS AND DISCUSSION 16 0,0 0,4 0,8 1,2 1,6 2,0 2,4 0 50 100 150 200 250 300 ISO122 S ta nd ar d de vi at io n of th e ou tp ut n oi se ( V ) Neutron fluence (·1-MeV 1013 n/cm2) ISO124 Figure 14: Standard deviation of the output voltage, σ, related to the intrinsic output noise. Supposing that the theoretical value of the output voltage is VMEAN , most of the times the actual output voltage would be a random value between VMEAN ± 2 · σ. a manifestation of the slower response of the internal devices such as operational amplifiers or237 comparators. Nevertheless, this could not be the only reason: Isolation amplifiers give up being238 linear devices at high frequency. Provided that the modulator-demodulator works at 500 kHz,239 the Nyquist frequency is 500/2 = 250 kHz. This is the threshold that limits the correct work240 of sampling amplifiers, usually being even lower than the theoretical value [18]. Therefore, the241 degradation of the frequency response could also be related to a decrease on the oscillator frequency242 in such a way that the Nyquist frequency effects appear at lower frequency values.243 4.6.5. Output noise244 This parameter is closely related to the typical output error and was measured following this245 method. The device was biased with noise-free power supplies and the input connected to ground.246 After waiting for a few minutes in order to stabilise the output, a high-accuracy multimeter mea-247 sured the output voltage 1000 times in less than one minute. The output voltage usually shows a248 random drift that was calculated using a 9th-order Savitzky-Golay filter and later removed from the249 output signal. Thus, the output voltage was centered on about 0 mV and the values distributed250 following a typical Gauss bell. In such a situation, the output noise is easily described by means251 of the standard deviation.252 Figure 14 shows the evolution of this parameter in both kinds of devices. According to these253 5 CONCLUSION 17 results, the output noise voltage soars from 20 µV up to 200-250 µV in the most irradiated samples.254 These results are on the order of the typical output error but always lower. There are several factors255 that explain this behavior. First of all, output noise were measured one month later so the devices256 partially recovered by natural annealing. Second, during the on-line tests the devices were several257 meters far from the instrumentation system so the measurements were not as accurate as those258 performed at the laboratory. Finally, the intrinsic non-linearities of the isolation amplifiers do not259 affect the output noise since it was measured with an only input value (0 V) whilst the typical260 output error was extracted from a DC sweep.261 5. Conclusion262 Capacitor-based isolation amplifiers undergo degradation if they are exposed to radiation. How-263 ever, they can keep operative even when the total radiation dose reaches a value of 2.4·1013 1-MeV264 n/cm2 & 235 Gy(Si). In this situation, the most affected parameters are the offset voltages, the265 transmission gain and the typical output error. Therefore, in case of wishing to use capacitor-based266 isolation amplifiers in instrumentation systems under radiation, some rules must be followed in267 order to guarantee the accuracy of the whole system. Techniques to remove the offset voltage must268 be integrated in the design along with low-pass filters to attenuate the output noise. Besides, the269 system must be able to correct the isolation amplifier gain drift. Finally, precision versions seem270 to be more radiation-tolerant than the devices with lower quality.271 6. Acknowledgements272 This work was supported by the cooperation agreement K476/LHC between CERN & UCM,273 by the Spanish Research Agency CICYT (FPA2002-00912), by the Spanish Agency for the Inter-274 national Cooperation (AECI) and by the Miguel Casado San José Foundation.275 7. References276 [1] M. Rowe, Isolation boosts safety and integrity, Test & Measurement World.277 URL http://www.reed-electronics.com/tmworld/article/CA235286278 [2] A. J. Peyton, V. Walsh, Analog Electronic with Op Amps: A Source Book of Practical279 Circuits, Cambridge University Press, Cambridge (UK), 1993, iSBN: 0-521-3305-X.280 7 REFERENCES 18 [3] G. Messenger, M. Ash, The Effects of Radiation on Electronic Systems, 2nd Edition, Van281 Nostrand Reinhold, New York City (USA), 1992, iSBN: 0-442-23952-1.282 [4] V. Remondino, Radiation damage in amplifiers used for quench detection in a su-283 perconducting accelerator, in: Proceedings of the 3rd European Conference on Radia-284 tion and its Effects on Components and Systems (RADECS95), RADECS Association,285 French Space Agency (CNES), Arcachon (France), 1995, pp. 89–93, iSBN: 0-7803-3093-5.286 doi:10.1109/RADECS.1995.509757.287 [5] Y. Zong, F. J. Franco, J. A. de Agapito, A. C. Fernandes, J. G. Marques, Radiation toler-288 ant isolation amplifiers for temperature measurement, Nuclear Instruments and Methods in289 Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equip-290 ment 568 (2) (2006) 869–876. doi:10.1016/j.nima.2006.09.007.291 [6] Texas instruments website, On-line at http://www.ti.com.292 [7] B. Carter, R. Mancini, Op Amps for Everyone, 3rd Edition, Texas Instruments, USA, 2009,293 Ch. 13: Understanding Op Amp Parameters, pp. 220–221, iSBN: 978-1-85617-505-0.294 [8] J. G. Marques, A. C. Fernandes, I. G. Gonçalves, A. J. G. Ramalho, Test facility at the295 portuguese research reactor for irradiations with fast neutrons, in: E. Ragel, R. Tamayo,296 C. Sánchez (Eds.), Proceedings of the 5th Workshop on Radiation Effects on Components and297 Systems (RADECS04), RADECS Association, Instituto Nacional de Tecnica Aeroespacial298 (INTA), Madrid (Spain), 2004, pp. 335–338, iSBN 84-930056-1-4.299 [9] C. Lee, B. Rax, A. Johnston, Hardness assurance and testing techniques for high resolution (12300 to 16-bit) analog-to-digital converters, IEEE Transactions on Nuclear Science 42 (6) (1995)301 1681–1688, iSSN: 0018-9499. doi:10.1109/23.488766.302 [10] L. Bonora, J. David, An attempt to define conservative conditions for total dose evaluation of303 bipolar ics, IEEE Transactions on Nuclear Science 44 (6) (1997) 1974–1980, iSSN: 0018-9499.304 doi:10.1109/23.658972.305 [11] C. Lee, A. H. Johnston, Comparison of total dose effects on micropower op-amps: bipolar306 and cmos, in: Proceedings of the IEEE Radiation Effects Data Workshop, IEEE Nuclear and307 7 REFERENCES 19 Plasma Sciences Society, The Institute of Electrical and Electronics Engineers (IEEE) , Inc.,308 Newport Beach, CA, USA, 1998, pp. 132–136. doi:10.1109/REDW.1998.731492.309 [12] R. Pease, M. Gehlhausen, J. Krieg, J. Titus, T. Turflinger, D. Emily, L. Cohn, Evaluation310 of proposed hardness assurance method for bipolar linear circuits with enhanced low dose311 rate sensitivity (eldrs), IEEE Transactions on Nuclear Science 45 (6) (1998) 2665–2672, iSSN:312 0018-9499. doi:10.1109/23.736512.313 [13] R. Pease, J. Krieg, M. Gehlhausen, D. Platteter, J. Black, Total dose induced increase in314 input offset voltage in jfet input operational amplifiers, in: Proceedings of the 5th European315 Conference on Radiation and its Effects on Components and Systems (RADECS99), RADECS316 Association, French Atomic Energy Commission (CEA), Fontevraud (France), 1999, pp. 569–317 572, iSBN: 0-7803-5726-4. doi:10.1109/RADECS.1999.858649.318 [14] D. M. Hiemstra, High total dose performance of various commercial off the shelf operational319 amplifiers during irradiation, in: Proceedings of the IEEE Radiation Effects Data Work-320 shop, IEEE Nuclear and Plasma Sciences Society, The Institute of Electrical and Electron-321 ics Engineers (IEEE) , Inc., Reno, Nevada (USA), 1999, pp. 32–38, iSBN: 0-7803-6474-0.322 doi:10.1109/REDW.2000.896266.323 [15] J. Agapito, J. Casas-Cubillos, F. Franco, B. Palan, M. Rodriguez Ruiz, Rad-tol field electronics324 for the lhc cryogenic system, in: Proceedings of the 6th European Conference on Radiation325 and its Effects on Components and Systems (RADECS03), RADECS Association, European326 Space Agency (ESA), Noordwijk aan Zee (The Netherlands), 2003, pp. 653–657, iSSN: 0379-327 6566, ISBN: 92-9092-846-8.328 [16] J. Srour, J. McGarrity, Radiation effects on microelectronics in space, Proceedings of the329 IEEE 76 (10) (1988) 1443–1469, iSSN: 0018-9219. doi:10.1109/5.90114.330 [17] F. Faccio, Cots for the lhc radiation environment: The rules of the game, in: Proceedings331 of the 6th Workshop on Electronics for the LHC Experiments (LEB2000), Krakow (Poland),332 2000, pp. 50–63.333 [18] P. Horowitz, W. Hill, The Art of Electronics, 2nd Edition, Cambridge University Press, United334 States of America, 1990, Ch. 13, pp. 900–902, iSBN:0-521-37095-7.335