CERN -> ATLAS -> TU Vienna
High-Density Digital Links

3.2 Aggressors & Victims

The TRT faces two kinds of EMC problems: first the system should not interfere with itself, and second it should not interfere with the surrounding subsystems of the ATLAS detector.

3.2.1 Intrasystem Environment

Figure 3.2 shows a principle diagram of the front-end electronics of the TRT. The basic components are the detecting element - the straw, an analogue, bipolar ASIC - the ASDBLR, and a digital CMOS ASIC. Its services consist of digital data and control lines, a high-voltage supply of 1,600 V for the straws, a bipolar low-voltage supply of ±3 V for the analogue part, a unipolar low-voltage supply of 5 V for the digital part, and several pipes for the active-gas supply for the straws and for the cooling of the straws and of the electronics.

Figure 3.2 Principle schematic diagram of the TRT-front-end electronics: the wires of the straws connect directly to the inputs of the ASDBLR which amplifies, shapes, discriminates the signal; the output of the ASDBLR is connected to the DTMROC which gathers the signal information together with its timing data and sends it out when requested; straws which are connected to the same HV group share the same decoupling capacitor CC ; several HV groups are supplied by the same HV line isolated by an isolation resistor RISO .

Each of these components are possible noise sources but also possible victims - see Figure 3.3.

Figure 3.3 EMC-impact matrix for the intrasystem environment of the ATLAS TRT: the dots indicate possible EMC problems.

Starting with the weakest aggressor, the detecting elements produce delta pulses of a r.m.s. of 0.5 ns which contain about 3 to 5% of the total deposited charge from 2 fC to several picocoulombs - see Chapter 2.1.1. Thus the signal has a bandwidth of up to 1 GHz.

Though the straw is a coaxial structure and the wall of the straw shields the sensitive wire from direct coupling, three main coupling paths remain. First, straws which share the same decoupling capacitor suffer conductive coupling on their signal return. Second, the straw wall surrounds its wire all over the length of the straw, but is exposed to the environment at the ends of the straw. Third, the straws of the barrel section of the TRT are read out on both ends. The wire is divided into two equal parts, but these two detecting elements share the same straw cathode.

Figure 3.4 Schematic drawings of the physical positions of straw-to-straw cross talk: (left) at the read-out side of the straws through capacitive, inductive, or conductive coupling; (middle) in read-out channels which share the same straw in the barrel section through capacitive coupling; (right) wires which are opposed to each other at the level of the inner seal of the end-cap section through capacitive coupling.

The coupling paths shown in Figure 3.4 translate into three equivalent circuits representing the conductive coupling through the decoupling capacitor CC , the capacitive coupling from straw to straw, straw to wire, and wire to wire (combined in Cwire_wire ), and inductive coupling from wire to wire ( Lwire_wire ) - see Figure 3.5.

Figure 3.5   Equivalent circuits for the thee possible straw-to-straw coupling paths: (left) conductive coupling through the decoupling capacitor CC ; (middle) capacitive coupling; (right) inductive coupling.

This allows to estimate the conductive coupling over a common high-voltage capacitor to


the capacitive coupling to


and the inductive coupling to


The maximal negative coupling depends on the ratio of the straw capacitance Cstraw (~ 10 pF) and the decoupling capacitor CC (~ 1 nF) and lies thus in the percentage region. The coupling decreases above some 100 MHz depending on the straw capacitance and the frequency-dependent input impedance Zi of the preamplifier.

The capacitive cross talk is the positive derivation of the signal and rises to a maximum which depends on the ratio of the straw capacitance Cstraw (~ 10 pF) and the coupling capacitance from wire to wire Cwire_wire . For read-out channels which share the same straw in the barrel section and at the level of the inner seal of the end-cap section, the coupling capacitance is about 100 fF. Though it can increase to some picofarad at the read-out electronics and thus produces cross talk in the 10% region.

The inductive cross talk depends on the straw capacitance Cstraw and the mutual inductance of the wires Lwire . Though the coupling rises with the square of the frequency, the mutual inductance would have to reach values of some 100 nH to see effects like those from the conductive or capacitive coupling.

The effective cross talk is a superposition of all three coupling paths, but is reduced by the limited input bandwidth of the read-out electronics.

Measurements of the barrel section [45] to [47] identified the conductive coupling due to a common HV capacitor as main coupling path with a cross talk of about 1 to 2%. A bad design of the connecting traces from the straw to the front-end electronics in one area was blamed for detected straw-to-straw crosstalk of up to 8%. The crosstalk between the two readouts of a single straw was less than 0.5%. Measurements with a prototype of the end-cap section of the TRT showed the same behavior [48].

Although calculations and measurements show results in the percentage region, the cross talk can reach a manifold of these results. As eight straws are connected to a common high-voltage capacitor, one straw will see the cross talk of the other seven straws - mainly conductive - when they are fired. 

The internal channel-to-channel crosstalk of the analogue-read-out chip - the ASDBLR - is, by design, less than 0.5%. Its peaking time is set to 8 ns and its input resistance behaves like a low-pass filter - see Chapter 2.1.3, in order to maximize the signal-to-noise ratio by limiting the bandwidth of the chip to about 100 MHz.

A possible problem arrives from the output currents - see Figure 3.6 - of the ASDBLR and the supply currents of its digital part. The chip is supplied by ± 3 V, but only the input stage and the shaper are truly bipolar. The discriminator and the output driver work unipolar. Thus, the switching of the output driver could harm the symmetry of the supply for the analogue part of the chip.

Figure 3.6 Equivalent circuits for a possible cross talk from the outputs of the ASDBLR to its inputs: (left) conductive coupling of a shared ground impedance Zground of the signal returns of the in- and outputs. coupling paths: (right) capacitive coupling over the parasitic capacitance Cout_in from the output to the input traces - note: the parasitic capacitance of the output traces Ctrace limits this kind of crosstalk for high frequencies.

The output currents are very fast current signals with a rise time of 1 ns and amplitudes of some 100 uA whereas the input currents with levels of a few micro ampere are a factor 100 smaller. A possible conductive coupling descends from the impedance Zgroundof a shared ground plane when the path of the current return of the output signal overlaps the path of the current return of the input signal. If this impedance (> mOhm ) is not infinitely small, a part of the signal return of the output signal will flow over the input impedance Zi of the preamplifier. A capacitive coupling derives from the parasitic capacitance Cout_in from the output traces to the input traces and leads to noise in the percentage region even with a few femtocoulombs.

The intrinsic noise of the ASDBLR can be described as function of the threshold of the low-level discriminator [49]. Setting the low threshold to 240 eV, the noise counting rate - the rate of the detection of a "noise" particle - of an unconnected channel was 1 Hz. Connecting 5 cm of cable with a LEMO connector increased this rate to 28 Hz. Adding a short straw which was placed into a metallic box increased the counting rate further to 500 Hz. It becomes 12 kHz for 200 eV, and reaches 100 kHz for 150 eV - see Figure 3.7.

Figure 3.7 Measured noise counting rate [49] of the ASDBLR as a function of the threshold: one straw is connected to the input of the ASDBLR; the noise counting rate is the rate of detected "noise particles" for a certain threshold level of the low discriminator of the ASDBLR.

The DTMROC is a digital CMOS chip. It is supplied by 5 V and clocked at 40 MHz. It uses a delay-locked loop to measure the time difference of the incoming signal from the ASDBLR with respect to the internal clock. This value is used to calculate the drift time of the electrons in the straw. Its input and output signals from and to the back-end electronics are low-voltage differential signals with signal rates of 40 Mbit/s. The differential output drivers are designed to draw a constant current from the supply. However the fall time of the signal is faster than its rise time. This skew introduces a spike-like common-mode signal which can interfere with the system - see Figure 3.8. In addition, the internal activity of the DTMROC can affect the power supply lines.

The possible victims of the DTMROC are not only the input signals of the ASDBLR but also its output and control signals. Especially the threshold lines over which the DTMROC sets the thresholds of the discriminators of the ASDBLR and the test-pulse lines can easily inject noise into the ASDBLR.

Figure 3.8 Spice simulation of the output driver of the DTMROC in the worst case situation where the chip is irradiated and supplied only by 4.5 V: (top) the differential signal and (bottom) the common mode signal.

The services for the front-end electronics of the TRT consist of data and control lines, low-voltage lines for the supply of the chips, high-voltage lines for the supply of the straws, pipes with active gas for the straws, and pipes for the cooling of the front-end electronics and the straws.

The data and control lines are shielded-twisted-pair cables which transport low-voltage differential signals. Though, the differential signals can translate into common-mode signals due to inhomogeneities in the geometry of the cables, the termination, and the signal form. In addition, common-mode currents can couple into the cable from the front-end electronics and its environment.

The LV lines can introduce noise picked up from the environment into the system, but also pollute the system with the noise of the digital part of the front-end electronics. The base line of the design of the detector is to reduce the material budget due to confined space and to the influence on the physical characteristics of the detector. Thus, it was not foreseen to use separate return lines for the analogue and the digital power supply.

Figure 3.9 Effect of a common current return for the analogue and the digital supply: each change of the current through the cable (e.g. digital noise from the DTMROC) translates over the finite cable impedance Zcable into a change of the reference voltage; however the ASDBLR will not see this change but the relative change to this on its supply lines.

However, this would introduce massive coupling from the digital part of the front-end electronics to the analogue part. The confined space dictates the use of very small cables and thus high cable impedances of some tens of milliohms. Each change of the digital current would change the voltage drop over a common impedance. The analogue chip would not see the change of its reference point but would translate it to a change of its voltage supply - see Figure 3.9.

The HV lines are also very delicate, as the high voltage connects to the most sensitive part of the detector - the straw.

The pipes of the active gas and the cooling can effect the detector in several ways: a change of the gas pressure or of the temperature influences the gas gain of the straw, a temperature change could affect the analogue electronics, and noise picked up from the environment can be carried into the system.

3.2.2 Intersystem Environment

Figure 3.10 shows the environment of the TRT with the LAr calorimeter outside the TRT, the SCT and PIXEL detectors and the beam pipe inside the TRT, and the services penetrating the whole system.

Figure 3.10 Axial view of the environment of the TRT: the LAr calorimeter surrounds the TRT; the SCT, the PIXEL detector, and the beam pipe are inside the TRT; the services of the TRT barrel run from its electronics at its head face over a patch panel at PPB1 along the inner wall of the calorimeter to the repeater patch panel at PPB/F2 (close to the front-end electronics of the LAr); the services from the TRT end cap run from its electronics at its circumference along the calorimeter wall over patch panel at PPF1 to the repeater patch panel at PPB/F2; the SCT and PIXEL services penetrate the TRT between the barrel and the end cap and between the section B and C.

The detailed EMC-impact matrix of Figure 3.11 illustrates the complex aggressor-victim relationships at the system level. Basically, they concentrate in five zones:

Figure 3.11 EMC-impact matrix for the intersystem environment of the ATLAS TRT: the dots indicate possible EMC problems.

A quantitative estimation is not possible at the system level. ATLAS is still in a fast-changing design phase where some problems are not yet solved. The space envelopes are not yet frozen. 

The LAr calorimeter is the most likely victim of the noise generated by the TRT detector as it measures the electromagnetic energy of the particles produced in ATLAS. Thus the noise level of the TRT must stay below the threshold of the LAr detector to avoid a degradation of the energy measurements. Additional crosstalk could appear at PPB/F2 where the active patch panels of the TRT are located and all services have to pass close to the LAr-front-end electronics. 
SCT electronics

The electronics of the SCT is located close to the open end of the TRT-end-cap straws where the anode wire is not protected by the straw.

Figure 3.12 Screen shots [51] of threshold scans without (left) and with (right) the powered SCT power tape running in parallel with the straws: the plots show the numbers of fired channels over the threshold.
SCT & PIXEL services

The services of the SCT and the PIXEL detector are running either close to the open end of the TRT-end-cap straws, in parallel to the straws, or in parallel to the services of the TRT. Although the data is transmitted over optical fibres, the power lines and supply pipes stay substantial aggressors. Due to the confined space, SCT and PIXEL use a special tape structure [50] to transmit their power. Its high capacitance allows an effective filtering of the noise produced by the digital front-end electronics, but produces high electric fields which couple to the TRT. Preliminary qualitative measurements [51] showed massive coupling from the SCT power tapes to the TRT - see Figure 3.12
beam pipe

The beam pipe which carries the particle beam is also predicted to be a possible noise source [52].

Only near field coupling is possible inside ATLAS due to the short distances among the sub detectors. Possible conductive coupling is eliminated by definition. The ATLAS Policy on Grounding and Power Distribution [53] enjoins the electrical isolation of all detector subsystems.

February 9, 2000 - Martin Mandl
Copyright © CERN 2000