EMC Immunity studies for front-end electronics in high-energy physics experiments

Transcription

1 EMC Immunity studies for front-end electronics in high-energy physics experiments F. Arteche*, C. Rivetta**, *CERN,1211 Geneve 23 Switzerland, **FERMILAB, P.O Box 0 MS341, Batavia IL 510 USA. fernando.arteche@cern.ch, rivetta@fnal.gov Abstract: The Compact Muon Solenoid (CMS) is one of the four high-energy physics experiments under construction at CERN for the Large Hadron Collider (LHC) accelerator. Thousands of channels of read-out electronics are designed to process signals of a few mv and digitise them synchronously at 40 MHz. The characterization of the front-end electronics to EMI disturbances is an important issue before the detector is integrated to guarantee the design goals and the good performance of the system. The characterization of the FEE immunity allows defining the noise level required at the output of the power supplies units, the near/far field interference level, etc. This paper presents results of electromagnetic immunity tests conducted on the front-end electronics of the CMS Hadron Calorimeter (HCAL) to characterize the front-end electronics sensibility to external perturbations. I. INTRODUCTION High-Energy Physics (HEP) detectors are very complex and large systems that identify particle interactions and measure their energy using sensitive detection devices such as silicon and pixel detectors, scintillators, optical devices, wire chambers, etc. The electrical signal measured by the sensitive detection device is amplified and processed by the front-end electronics (FEE) inside the detector. After processing, the data is transferred, via optical links, to the acquisition system located 120 meters away from the detector. The electronic read-out of the calorimeter is designed to process signals of a few mv and digitise them synchronously at 40 MHz. Part of this electronic system is located inside of a harsh environment with particle radiation and a high DC magnetic field of 4 Tesla. The Compact Muon Solenoid (CMS) detector, under construction at CERN for the LHC, is about 25 meter long with a 20-meter diameter, bringing the total weight of the detector to around tones. The CMS detector is divided in five sub-systems located at different layers of the structure. Each sub-system has distinct objectives and has to process different signals and power levels. The CMS experiment powers-up each sub-system individually and requires a total power between 0kW and 7kW to operate. The power is supplied by power supply units placed in two areas located 20 m. and 120 m. away from the detector due to the harsh environment mentioned above. Although all the fast signals are transmitted from the detector to the acquisition system via optical fibres, in the detector a large amount of electromagnetic (EM) interactions exist among the electronic sub-systems. These interactions can be mainly found in the connection between the FEE and the power supply distribution [1]. The most important problems to address before integrating the electronic system into the detector is to define the immunity limits of each sub-system to both RF noise and transient signals and how to test the electronic systems to comply with those limits based on normalized measurement. The CMS experiment is identifying both emission and susceptibility noise limits for the electronic equipment to integrate into the experiment [2]. In addition, CMS is defining normalized tests to analyse the equipment compliance with those limits. In the HEP community, there has not been such a systematic approach to define quantitative limits on both the emission and the susceptibility levels inside the detectors. To define those limits and to quantify the cross effects between emitter-receiver, electromagnetic compatibility (EMC) studies were conducted on FEE and power supply prototypes. The final goal of this approach is to integrate the electronic systems into the detector in a safe way, such that the performance and reliability of the FEE is not degraded. This paper addresses the EMC studies focused on the quantification of the FEE sensitivity to conductive noise coupled through the input/output cables and the definition of the noise level to be specified at the output of the power supply units. II. RF IMMUNITY OF THE FEE The electromagnetic immunity is the ability of a device, equipment or system to operate without performance degradation or failure in presence of electromagnetic disturbances. The RF electromagnetic interference perturbing the FEE is superimposed to the signal processed by the electronic system and its intrinsic noise. The total noise perturbing the signal processed by the front-end electronics is due to the contribution of different sources as: the intrinsic thermal noise of transistors and resistors, the EM noise picked-up by connections between the detector devices and the FEE, the noise generated by power supplies and coupled through power cables, spurious signals, etc. The total noise level defines the minimum signal that the FEE can process. Taking as a common point the output of the FEE analogue stage to analyse

2 the total noise, the noise contribution can be written as: n a (t) = nth(t) + nin(t) + n ps(t) + K (1) where n th (t) is the intrinsic thermal noise of the FEE, n in (t) is the noise picked-up by the connection between the sensor device and the FEE, and n ps (t) represents the noise generated by the power supply and coupled through the power cables. These three terms are the most important in the total noise contribution. The criterion generally used to define the total noise contribution in the FEE is to make the thermal noise contribution as the dominant part of the total noise. In order to use this criterion for the design, it is necessary to know and minimize thermal noise and to characterize the EMI contributions to reduce their effect on the total noise. III. RF IMMUNITY TEST OF THE FEE The EMI contribution can be either evaluated at early stage of the system design via modelling and simulation [3] or measured on prototypes. In the first case, corrective actions can be taken at the design stage, whereas in the second case, it is possible to identify in prototypes critical elements and inappropriate layouts that are responsible for the performance degradation of the FEE. As the FEE is linked to the acquisition system via optical fibres, the conductive noise is mainly coupled into the FEE through the input power cables and slow control network. To define the immunity level of the FEE to conductive disturbances, four tests are conducted by injecting currents through the FEE input power cables. These tests are: Injection of current through the shield. Injection of common mode (CM) currents into the central conductors. Injection of differential mode (DM) currents into the central conductors. Injection of CM and DM currents. They allow quantifying the FEE sensitivity to conductive noise to define the output noise level of power supplies, the magnitude of external EM fields and ground currents, etc. III.1 Experimental Set-up The RF immunity tests are carried out using a FEE prototype of one of the CMS sub-systems, the Hadron Calorimeter (HCAL). In this prototype, the input amplifier is a custom chip design based on a charge integrator encoder (QIE) [4]. This device amplifies and digitises the signal generated by a hybrid photomultiplier located a few centimetres from the amplifier. The target value for the thermal noise level of this device is 2.16 counts RMS at the output of the ADC or an Equivalent Noise Charge (ENC) at the input of 0.72 fc = 40 electrons. In the HCAL front-end electronics, each board contains 6 QIE chips and digital electronics to control the chips, serialize the data collected by the QIE s and send it to the acquisition system via optical links. In this prototype, the output noise level for all channels is between 2.64 to 2.94 counts RMS when no RF perturbation is injected. The prototype used to perform the immunity tests consisted of 12 channels distributed in two identical boards, each of them connected to the back-plane. The front-end boards, the back-plane and the input power filter are placed into a metallic read-out box (RBX) that not only gives mechanical support to the electronic system but also EM shielding and thermal management. Further information about the HCAL FEE can be found in [5]. The experimental set up to study the immunity of the FEE is shown in Fig. 1. The basic idea of this test setup is to keep its topology as close as possible to the final one. The FEE and the auxiliary equipment are placed on a copper plane as suggested by IEC [6]. This copper sheet (2x2 meters) is the reference ground plane. The perturbing signal is injected to the FEE input power cables using a bulk injection current probe, a RF amplifier and a RF signal generator. The level of the injected signal is monitored using an inductive current clamp and a spectrum analyser. To represent the effect of very long cables, normalized common impedances (CI) (Common Mode and Differential Mode impedance) based on lumped components are inserted between the power supplies and the FEE to standardize the measurements. The output signal of the FEE is measured by its own acquisition system. PS RF Generator I p CI Copper plane Spectrum analyzer Fig.1 Test set-up. FEE Acquisition System V out RBX The test procedure consists of injecting a sine-wave perturbing current at different frequencies and amplitudes to the FEE through the input cables (mainly the input power) and evaluating the performance of the FEE measuring the output noise. The frequency range of the sine-wave signal is between 1 khz and MHz.

3 III.2 Susceptibility to currents injected through the power cable shield High frequency currents flowing through the power cable shield can be induced by far and near electromagnetic fields coupled to the shield or by high frequency ground currents flowing in the system. To emulate its effect on the FEE, a sine-wave current is injected to the power cable shield. This shield current, in addition, couples CM currents to the internal power conductors through the surface transfer impedance of the cable. All these currents affect the performance of the FEE and the interference depends on the amount of noise current coupled into the sensitive areas of the FEE. Due to slight differences in the connection between the QIE amplifier and the photo-detector the perturbing current does not affect equally all the channels. Fig. 2 depicts the digitised RMS value of the amplifier output voltages for all channels when perturbing currents of 6 ma RMS at 5 MHz and 10 MHz are injected. These values are compared with the output voltage noise of each channel when no perturbation is injected (Reference). RMS [Counts] Reference 5 Mhz 10 Mhz The transfer function of channel 5 measured between 0 khz and MHz is depicted in Fig. 3. Lower frequency values are not shown because the FEE output voltage is dominated by the thermal noise and the transfer function is poorly measured. These measured values can be fitted to a mathematical model of the FEE transfer function in the frequency range between 1 khz and 100 MHz. This model is proposed by combining the transfer function of the QIE with an additional term that models the external noise coupled into the FEE. The mathematical model of FEE transfer function is: τ sin ω T ˆ F 2 1 ( ω) = j ω L (4) m τ ω j ω 1+ 2 ωb where τ represents the sampling period of the sampled integrator (1/34MHz during the tests, 1/40MHz for the final version), ω b the bandwidth of the QIE input stage (~ MHz) and Lm a parameter that quantifies the coupled noise. This parameter depends on both the QIE channel and the grounding layout and it is used to parameterise the proposed model. The model is fitted to the measured values adjusting the parameter Lm by a least square method. Fig. 3 shows the transfer function measured for channel 5 and the function described by Eq. 4 fitted to the data. Measured data E stim ate d values Channel number Fig.2 Noise perturbation distribution per channel. TF( ω ) - [fc/ma] The transfer function defined as the ratio between the AC output voltage and the injected current is used to quantify the sensitivity of the FEE and analyse the noise contribution in the system for any perturbing signal. The transfer function is defined as: V1out ( ω ) TF ( ω ) = (2) I1shield ( ω ) where I 1shield (ω) is the magnitude of the perturbing sine wave signal and V 1out (ω) is the magnitude of the AC output voltage. To compare the results with the input signal level, it is more convenient to represent the output signal as equivalent signal at the input of the FEE by dividing V 1out (ω) by the low frequency charge gain G o of the FEE. In this case, the transfer function can be expressed by: V1 out( ω) TF( ω ) = [ fc / ma] (3) G I ( ω) o 1shield Fig.3 Measured and fitted transfer function of channel 5 for shield currents. The FEE immunity to currents flowing through the power cable shield basically depends on the connection between the input filter box and metallic box holding the electronics (RBX). This connection can be made with a cooper strap of about 5-15 cm. To evaluate the impact of this connection on the FEE noise sensitivity different straps were tested. Modifying the length and the routing of the strap connection to the RBX changes the strap s inductance. Fig. 4 depicts the transfer function to current shields corresponding to channel 5 for different strap configurations. The best configuration is labelled as Ground 3 and corresponds to the lowest inductance connection between the filter box and the RBX. It is achieved by a short strap and located close to the RBX.

4 TF ( ω ) - [fc / ma] GND 1 GND 2 GND 3 The FEE configurations used during these studies correspond to: FEE-A: a FEE with good grounding and good EMI input power filter (TFcm-A); FEE-B: a FEE without EMI input filter (TFcm-B), and FEE-C: a FEE with a poor grounding design at detector level and without EMI input filter (TFcm-C). III.4 Susceptibility to DM currents Fig.4 Fitted transfer functions corresponding to channel 5 for different strap configurations. III.3 Susceptibility to CM currents To study the effect of the common mode noise currents flowing through the internal power conductors, the perturbation current is injected to both the active and return power cables. In practice, this CM noise is generated by power supplies and coupled to the FEE through long cables. The test procedure followed is similar to the previous one. A sine-wave current is injected as CM perturbation and the FEE output signal is measured by its acquisition system. Similarly to the previous results, the noise does not distribute equally in all channels. The FEE immunity to CM currents is quantified by the transfer function, defined as the ratio between the AC output voltage, referred to the FEE input, and the perturbing CM current. Mathematically it can be expressed as: V1 out ( ω) TF( ω) = (5) Go (2 I1 CM ( ω)) where 2.I 1CM (ω) is the injected current and I 1CM (ω) is the common mode current. The TF(ω) magnitude is mainly defined by the input power filter attenuation to CM components and the grounding layout around the detector. These power filters cannot use magnetic material due to the fact that the FEE operates inside a magnetic field of 4 Tesla. The transfer function is measured for three FEE configurations involving different input EMI power filters and detector grounding layout. Fig 5 depicts the measured results fitted to eqn. (4). TF ( ω ) - [fc/ma] 10 1 TF cm - C TF cm - B TF cm - A Fig.5 Fitted CM transfer functions for different FEE configurations. To study the effect of differential mode noise currents flowing through the power cables, the perturbation current is injected to the active power cable. In addition, to increase the CM impedance of the system, the connection to ground of both the power supply and normalized common impedance box is removed. In this case, the injected current in the active power cable is forced to return through the return power cable and the CM noise injected to the FEE is negligible. Similarly to previous tests, a sine-wave current is injected as DM perturbation and the FEE output signal is measured. The FEE immunity to DM currents is quantified by the transfer function, defined as the ratio between the AC output voltage, referred to the FEE input, and the perturbing DM current. Fig. 6 depicts the transfer function to DM perturbing currents. From this figure, it is possible to observe a resonance at 20MHz. It is due to the interaction between the parasitic capacitance between the ground plane and the RBX and the inductance of the straps connecting both structures. The resonance appears in this test set-up because the normalized common impedance box is disconnected from the ground, eliminating the damping introduced by that impedance. TF ( ω ) - [fc / ma] Measured values Estimated values Fig.6 Transfer function of channel 5 for DM currents. III.5 Susceptibility to CM+DM currents In general, normalized specifications for the output noise of power supplies are based on the terminal current. This current can be decomposed in two components, the differential and common mode currents. The transfer function relating the output voltage noise to the terminal current should be equivalent to the transfer function obtained by decomposing the terminal current into common mode

5 and differential mode currents. This transfer function is obtained by multiplying the I CM and I DM components of the terminal current by its respective TF, and adding the resulting output voltages vectors. This test is performed to verify the assumption defined above and to quantify the FEE sensitivity to noise currents at the input power conductor. Using a set-up similar to the one used previously for shield current and CM current tests, a sine wave current was injected into the active power cable. The transfer functions obtained for the three FEE configurations defined in III.3 are depicted in Fig. 7. TF ( ω) - [fc/ma] 10 1 TF - C TF - B TF - A Fig.7 Fitted (CM+DM) Transfer function of channel 5 for different FEE configurations. Let us use as example the magnitude obtained at 10MHz and compare it with the TF obtained by decomposing the injected current into I CM and I DM. At 10MHz, a perturbing signal Ip = 19.5 ma is injected into the active power conductor and the RMS output noise measured at channel 5 was equal to 3.81 counts. The injected current Ip is decomposed in the two orthogonal modes I CM and I DM, whose magnitude is mainly defined by the input impedance of the RBX. Based on preliminary measurements, the magnitudes for these modes are I DM = 0.63 Ip and I CM = 0.37 Ip. Multiplying these components by the transfer functions depicted in Fig. 5 (TFcm-A curve) and 6, the estimated RMS output noise is 4.1 counts (error 7.5%), which is quite close to the measured value. For these calculations, the I CM and I DM currents are in phase because the CM and DM input impedance of the RBX have similar characteristics. From all these measurements, it is possible to observe that the FEE is more sensitive to CM mode current perturbations than the DM mode currents. It is worth noting from the last calculations that 71% of the output noise contribution is due to CM currents and only 29 % is induced by DM currents, although I DM = 1.7 I CM. required at the output of switching power supplies used to power-up the FEE electronics. Although several committees, like IEC and IEEE [7][8], have suggested noise emission levels for the power supply output, the EU and FCC standards, that regulate the conductive noise emissions of commercial power supplies, do not enforce any emission limit for the outputs. The current noise level at the power supply outputs can be defined based on the inverse curve of the FEE sensitivity functions described above. From Figs. 5 and 7, the common mode noise level and the noise level for the active terminal required at the output of the power supply can be defined as a function of the frequency by a level increase of 20dB/dec up to 100KHz and a constant curve above 20MHz. To fix the level of the proposed limit curve, the noise at the output of the FEE is calculated based on a testing power supply switching at 100KHz with a conducted noise emission similar to the described above. A criterion to define the noise level is to set the noise contribution at the FEE output due common mode currents equal to 20% of the total thermal noise of the FEE and due to currents in the positive terminal equal to 25% of the thermal noise. Additionally, the noise spectrum is considered between 100KHz and MHz to include components above 40MHz that are not rejected by the FEE electronics. Following this procedure, Figs. 8 and 9 show the noise level proposed for the three different FEE designs as previously presented in section III-3. In those figures, the current levels are converted to voltage, assuming a normalized impedance of ohms, and compared with the levels imposed by EN 522 classes A-B to the input terminal of power supplies. V [dbµv] EN A HCAL - A HCAL - B HCAL - C EN B Fig.8 European (EN 522 ) and HCAL standard levels for positive input terminal noise. 100 CM HCAL - A IV. POWER SUPPLY EMISSION LEVEL I [dbµa] Based on the FEE sensitivity to common and differential mode noise coupled through the input power cables, the conductive noise limits to be imposed to the power supply outputs can be derived. It is particularly important to specify the noise 40 CM EN A CM HCAL - B 30 CM EN B 20 CM HCAL - C Fig.9 European (EN 522 ) and HCAL standard levels for common mode noise.

6 To show results of the effect induced on the FEE by the conducted noise emitted by power supplies, Figs. 10 and 11 depict the conducted output noise of a switching DC-DC converter. In Fig. 10, the conducted noise emission is measured at the positive output terminal of a DC-DC converter composed by a Vicor unit and an additional EMI output filter operating at 7.5V/7.5A. The noise current is measured by a current transformer and a spectrum analyser and converted to voltage considering a normalized impedance of ohms. The noise measured complies with the emission level called HCAL-A and the output noise for the FEE-A is equal to 2.91 counts RMS, which is similar to the FEE thermal noise. When the same converter is connected to the FEE-B and FEE-C, the output noise measured is equal to 3.05 counts and 5.28 counts. In both cases, it is possible to observe, from Figs 10 and 11 that the noise emission does not comply with the levels defined by HCAL-B and HCAL-C. V [dbµv] Ch 5 - A = 2.91 counts 110 Ch 5 - B = 3.05 counts Ch 5 - C = 5.28 counts (CM+DM) Output noise EN Class B EN Class A HCAL - A HCAL - B HCAL - C Fig.10 Conductive noise emission measured at the output positive terminal of a DC-DC converter. I [dbµa] CM output noise CM EN Class B CM EN Class A CM HCAL - A CM HCAL - B CM HCAL - C Fig.11 Conductive common mode noise emission measured at the output of a DC-DC converter. Similarly, Fig. 11 depicts the common mode emission measured at the output of the DC-DC converter unit. The noise measured complies with the level defined by HCAL-A. V. CONCLUSIONS The noise immunity level of front-end electronics used in high-energy physics experiments has been evaluated. The immunity level is quantified by transfer functions relating the FEE output voltage to different perturbing currents flowing through the input power cables. Based on these transfer functions, the FEE overall noise can be estimated based on the spectral content of perturbing currents flowing through the power cable shields (near/far field effects), the central conductors (power supplies effect), etc. The application of these transfer functions to define the conductive emission level at the output of switching converters has been presented. Further research in this topic is being conducted to define a standardized measuring procedure for the conducted emissions at the output of power supplies to get results reliable, repetitive and also representative of the environment where the device is going to operate. REFERENCES [1] C. R. Paul, "Introduction to Electromagnetic Compatibility", John Wiley & Sons, Inc., New York, [2] F. Arteche, C. Rivetta and F. Szonsco Electromagnetic compatibility plan for the CMS detector at CERN, Proc. of 15th Int. Zurich Symposium on EMC, February 18-20, 2003, Zurich, Switzerland, pp [3] F. Arteche and C. Rivetta, "Noise Susceptibility Analysis of the HF Front-End Electronics for the CMS High -Energy Experiment", Proc. of IEEE Int. Symposium on EMC, August 2003, Boston, USA, pp [4] T. Zimmerman, J. R. Hoff, The Design of a Charge-Integrating Modified Floating-Point ADC Chip, IEEE JSSC, Vol. 39, No. 6, June [5] T. Shaw et. al. Front End Readout Electronics for the CMS Hadron Calorimeter, Proc. of IEEE Nuclear Science Conference, November 2002, Virginia, USA, pp [6] IEC "Electromagnetic compatibility (EMC) - Testing and measurement techniques Immunity to conducted disturbances, induced by radio-frequency fields Basic EMC publication, Ed [7] CISPR 22 - "Information technology equipment - Radio disturbance characteristics - Limits and methods of measurement, Basic EMC publication, Ed [8] IEEE1515 Recommended practice for electronic power sub-systems: Parameter definitions, test conditions and test methods, Ed