The impact of crosstalk in the X-IFU instrument on Athena science …x-ifu.irap.omp.eu/wp-content/uploads/2016/03/The-impact... · 2016. 8. 19. · The impact of crosstalk in the

The impact of crosstalk in the X-IFU instrument on Athena science cases

R. den Hartog1, P. Peille2, T. Dauser3, B. Jackson4, S. Bandler5, D. Barret2, T. Brand3,

J.-W. den Herder1, M. Kiviranta6, J. van der Kuur1, S. Smith5, J. Wilms3

1 SRON Netherlands Institute for Space Research, Sorbonnelaan 2, 3584 CA Utrecht, The Netherlands

2 CNRS, IRAP Institut de Recherche en Astrophysique et Planétologie, Toulouse, France 3 ECAP, Erlangen Centre for Astroparticle Physics, Erlangen, Germany

4 SRON Netherlands Institute for Space Research, Landleven 12, 9700 AV Groningen, The Netherlands

5 NASA GSFC, Goddard Space Flight Center, Greenbelt, Maryland, USA 6 VTT Technical Research Centre of Finland, Espoo, P.O. Box 1000, FI-02044 VTT, Finland

ABSTRACT

In this paper we present a first assessment of the impact of various forms of instrumental crosstalk on the science performance of the X-ray Integral Field Unit (X-IFU) on the Athena X-ray mission. This assessment is made using the SIXTE end-to-end simulator in the context of one of the more technically challenging science cases for the XIFU instrument. Crosstalk considerations may influence or drive various aspects of the design of the array of high-countrate Transition Edge Sensor (TES) detectors and its Frequency Domain Multiplexed (FDM) readout architecture.

Keywords: Athena, X-IFU, SIXTE, TES detectors, FDM, crosstalk, energy resolution

1. INTRODUCTION

The Athena X-ray mission was selected as the second L-class mission in ESA's Cosmic Vision 2015–25 plan, with a launch foreseen in 2028, to address the theme ''Hot and Energetic Universe"1. One of the two instruments on board Athena is the X-ray Integral Field Unit2 (X-IFU) which is based on an array of ~3800 Transition Edge Sensors (TES's) operated at a temperature of ~90 mK. The science cases pose an interesting challenge for this instrument, as they require a combination of high energy resolution (2.5 eV FWHM or better), high spatial resolution (5 arcsec or better) and high count rate capability (several tens of counts per second per detector for point sources as bright as 10 mCrab). The performance at the single sensor level has been demonstrated3, but the operation of such detectors in an array, using multiplexed readout, brings additional challenges, both for the design of the array in which the sensors are placed and for the readout of the sensors. The readout of the detector array will be based on Frequency Domain Multiplexing (FDM)4. In this system of detectors and readout, crosstalk can arise through various mechanisms: on the TES array, neighboring sensors can couple through thermal crosstalk. Detectors adjacent in carrier frequency may suffer from electrical crosstalk due to the finite width of the bandpass filters, and shared sources of impedance in their signal lines. The signals from the individual detectors are summed and then amplified by a pair of SQUID amplifiers before being sent to warm front-end electronics. The transfer function of the SQUID amplifiers is non-linear, which will give rise to higher harmonics of carriers and intermodulation products when multiple signal pulses are simultaneously present in the SQUID. Under high count rate conditions this is another source of crosstalk. The effect of all these crosstalk sources is that parasitic pulses will appear in the record of a signal pulse which will create a stochastic offset of the measured energy and thus a degradation of the energy resolution.

Space Telescopes and Instrumentation 2016: Ultraviolet to Gamma Ray, edited by Jan-Willem A. den Herder, Tadayuki Takahashi, Marshall Bautz, Proc. of SPIE Vol. 9905, 99055T

· © 2016 SPIE · CCC code: 0277-786X/16/$18 · doi: 10.1117/12.2232098

Proc. of SPIE Vol. 9905 99055T-1

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/17/2016 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

Crosstalk requirements are needed as inputs for various design questions which are becoming pertinent, e.g. the necessity of linearization of the SQUID amplifiers, on the basis of the assessed impact of crosstalk on critical science cases. However, the impact of crosstalk on energy resolution is not straightforward. The strength of the crosstalk signal has a (non-linear) dependence on the energy of the interfering photon (and the signals photon in the case of non-linear amplification), the time between the arrival of the photons and pixel or carrier frequency separation are important. As a consequence, the energy spectrum, spatial structure and count rate of the source play a crucial role. This implies the need for detailed simulations. And as varying fractions of events may be affected, it is necessary to assess the impact of crosstalk on the required scientific products, rather than just its impact on the energy resolution. The SIXTE end-to-end (E2E) simulator has been developed exactly for this purpose.5 The detection of the WHIM in absorption in a GRB spectrum is taken as an example of a science case on which crosstalk is likely to have a non-negligible impact. Another example would be the case of a bright extended source, such as a supernova remnant. The WHIM science case has been studied with the SIXTE with a basic set of instrumental effects6, but in absence of crosstalk. In this paper we describe the implementation of the major crosstalk effects in the SIXTE simulator, and derive the first results relevant for the GRB / WHIM science case.

2. CROSSTALK

Crosstalk can be loosely defined as the offset in the inferred energy for a signal pulse on one pixel due to the presence of signals on other pixels. When taking into account the technical details, crosstalk needs to be defined in such a way that: It does credit to the complex reality, while still being implementable in an E2E simulator Is verifiable in laboratory experiments Can be scaled according to a tunable parameter At present 4 of the 5 identified mechanisms have been or are being implemented in SIXTE E2E simulator. A summary of the mechanisms with their dependencies, mitigation methods and implementation status is presented in Table 1.

Crosstalk impact on the inferred energy from a signal pulse may not only occur when it overlaps in time with the perturbing pulse, but may also arise when the perturbing pulse is present well before the signal pulse, or after the signal pulse has decayed, due to the effect that the perturbing pulse has on the determination of the baseline. In this paper we assume a standard read out scheme for X-ray pulses, which is based on event sampling which starts with a return-to-baseline time equal to 10 (with the pulse decay time), and lasts 50 , which is long enough to limit the degradation of the energy resolution due to noise on the baseline determination to 4%.7 Crosstalk from a perturbing pulse will manifest itself as an additional pulse somewhere during the event and create, depending on the timing with respect to the signal pulse, a positive or negative offset. This is illustrated in Figure 1 for the example of a pixel with a = 80 s, typical for a hybrid small pixel array (SPA), which was conceived to achieve the goal of observing 10 mCrab sources.8

Table 1. Overview of crosstalk mechanisms, dependencies, mitigation options and implementation status. Mechanism Dependence Mitigation options Status

1. Thermal leakage (on sensor array)

Inter-pixel distance on array Heatsinking layer

implemented in SIXTE

2. Carrier leakage (in TES bias circuit)

f-2; R2nbr; Lfltr

-2 Rnbr changes during pulse

Increase frequency spacing f in bias circuit: - more wires - more DACs


3. Common impedance (in read-out circuit)

f-1; Lcom; fcarr; L-1fltr Lower Lcom

Increase f provisionally implemented

in SIXTE 4. Coupling between parallel r/o circuits

Mutual L and C Shielding of strategic points in circuit Increase distance between sensitive parts

not yet implemented

5. Non-linear amplification (mainly by SQUIDs)

1 ; 2 ; t; (f) per pulse present on input coil

SQUID linearization Mix SPA and LPA pixels More GBW in BBFB




1.0

0.8

La_' 0.6

a 0.4 -Ñ -N -aa 0.2 -

0.0

-0 2

freq. dompulse shape ............

-10 0 10 20 30 40delta t [tau]

Figure 1. Relative impact of crosstalk from a perturbing pulse as a function of the difference in arrival time in units of decay time with respect to the signal pulse. The normalized criticaly damped pulse shape ( t et/ ) is indicated by the dotted line. Crosstalk pulse that arrives later than ~36 in the event have tails that fall increasingly outside the event, creating the slow return to zero weight towards the end of the event.

2.1 Thermal leakage

The leakage of thermal energy deposited by a photon in one pixel to pixels in the physical neighborhood is a well established mechanism for crosstalk.9,10 The fraction of energy that leaks from one pixel into its nearest neighbor (horizontal or vertical in an array of square pixels) is of the order of a few times 103 of the energy deposited by the photon if no measures are taken,10 but may be reduced to levels below 104 when heatsinking layers are applied underneath the pixels.11

2.2 Electrical crosstalk

Electrical crosstalk is a container term for several mechanisms in Table 1. It includes relatively familiar crosstalk mechanisms from carrier leakage and common impedance,4,12 and mutual inductance between physically nearby circuits, as well as crosstalk from common mode coupling. The latter two mechanisms are not yet considered here, as the assessment of their magnitude requires detailed wiring layout of the sensor array and cold electronics which currently does not exist in sufficient detail.

Crosstalk due to carrier leakage arises when a carrier at frequency s gets modulated by a change in TES resistance in a circuit at frequency p due to the finite bandwidth of the LC filter. The modulated pulse shape is the same, but its amplitude is scaled down by a factor (Rp/2L)2, where =sp.

4,12 When the common impedance is non-zero the situation gets more complicated. A current running in one circuit creates a voltage change in all the other circuits that share this impedance. As a consequence, the TESs in those circuits respond by a change in their set points. A proper treatment of the problem requires to solve a coupled set of equations:

Vb t,ωs =Ip t Rp

t +LdIp

dt+Cp Ip t dt+ Lcom

dIs

dt (1)

Vb t,ωs =Is t Rs Ts,Is +LdIs

dt+Cs Is t dt+ Lcom

dIp

dt

where Vb(t, s) is the sinusoidal voltage bias at frequency s, L and C are filter inductance and capacitance, Lcom is the common impedance, which is in practice an inductance. Shunt resistors and parasitic impedances other than Lcom are omitted here for clarity but are included in f.i. the computations with the TESSIM code.5

If Rp(t) represents the response of the perturbing TES to the absorption of a photon of energy Ep, the change in resistance in the coupled circuit of the signal TES requires solving simultaneously the differential equation for Ts(t) and finding



Rs(Ts,Is) for the signal TES. Here it clearly makes a difference whether the TES is in its equilibrium state or recently received a photon itself when the perturbing pulse arrives. At present, this scheme of coupled readout circuits is a potential extension of the TESSIM code, but not yet implemented. The current, provisional implementation of electrical crosstalk in the E2E simulator assumes that Rs does not change, and solves Eqs. 1 in the frequency domain. TESSIM current pulses Ip(t) are generated for different photon energies Ep, and the crosstalk is assessed by optimal filtering the response Is(t). The electrical crosstalk is thus a function of the energy of the perturbing pulse, Ep, the difference in arrival time t between the signal and the perturbing pulse (for the moment according to the relative impact shown in Figure 1), and depends explicitly on the frequencies of the carriers s and p. Unlike carrier leakage, which depends only on the relative frequency distance , common impedance crosstalk is proportional to s Lcom. The implementation in the E2E simulator is based upon a table of electrical crosstalk values for a grid with dimensions Ep, t, s and p. The E2E simulator generates an event list which contains for each photon that arrives at the focal plane, an energy, a pixel number and an arrival time. By assessing photons that arrive in the same read-out chain and have overlapping event records on a pairwise basis, the resulting electrical crosstalk is then obtained by interpolation of tabulated values on this grid. 2.3 Non-linear amplification

The response of a SQUID amplifier is intrinsically non-linear. Its transfer function from input flux to output voltage V is approximately sinusoidal. Already the second term of a Taylor-like expansion is a non-linear term:

! (2)

The baseband feedback (BBFB)4 suppresses all the carriers on the input coil, hence the flux on the input coil only differs from zero when a pulse signal is modulated on a carrier by a TES. In the case of a non-zero at a frequency f, the non-linear terms give rise to higher harmonics at 3f, 5f, etc… which is a first source of crosstalk. This crosstalk is in principle avoidable by simply avoiding carrier frequencies at the location of these higher harmonics. This requires a non-trivial control over the resonance frequencies of the LC filters. For the moment we assume that this is feasible. In order to see how this non-linearity gives rise to other crosstalk, consider an input signal consisting of two pulses separated by a time interval t:

(3)

The output signal V contains a contribution of the 2nd pulse at the carrier frequency of 1st, and vice versa. This can be seen in the first few terms of the output:

! (4)

+ terms proportional to cos(ω2t), and higher-order intermodulation terms

The second term in brackets in Eq. 4 is responsible for crosstalk between overlapping pulses irrespective of their distance in frequency space!

The application of gain G in the baseband feedback suppresses the flux on the input coil as /(1+G), which has already a linearizing effect, and which is further amplified by an additional 1/1+G term applied to the non-linear terms, since they can be regarded as gain errors in the forward path of the feedback loop:

sorder termhigher )(1

)(

!3)(1

)(

!2)(1

1

)(1

)(3

1

3

2

1

21 fG

f

k

k

fG

f

k

k

fGfG

fkV ininin

out (5)

In order to fully assess the crosstalk between two pulses which overlap in time, while parallel baseband feedback loops are active on both carriers, a BBFB simulator was developed, with the following features: The ability to simulate the action of 2 or 3 parallel BBFB loops The possibility to include the SQUID non-linearity in the BBFB in a self-consistent way The flexibility to include additional effects such as a SQUID setpoint offset or non-sinusoidal (measured or

simulated) V() transfer functions.



1.0

0.8

V.O

+

0.4

0.2

0.0

sinusoidal -.

+flux offsetmeasured _....__._

- - -- linear

10 100 1000 10000

GBW [Hz]

1000.0 -

100.0

10.0 -

1.0

01

6 keV pe. nurber.2keVpetlurbew _

1 2 3 4 5

Frequency [MHz]

.0 r

óverlapPn9 P

-2 0 2 4

Time difference [tau]

0sr ig náÌampÌiÌutle.:. -

__-permd SrartMimde-.;-

0.0 0.1 0.2 0.3Amplitude [PhiO]

0.4 0.5

It mimics the functionality of the Digital Readout Electronics (DRE): It performs simulations in the domain simulation with a 3.3 ms event length and a 20 MHz sampling rate, equivalent

to 65k samples of 50 ns per event. The input is a sum of signal pulses modulated on carriers with frequencies of interest A digital gain factor g and a time delay d of 0.5 s are introduced, consistent with current FPGA implementations. It allows parallel I and Q loops per carrier The integrator is similar to the firmware implementation on the FPGA A factor 128 decimation is achieved by subsequent frequency and time domain low-pass filtering Frequency domain optimal filtering is applied to the decimated events, to measure the energy of the signals Two calibrations have to be made: For a given pulse shape, the peak amplitude of 0.25 0 has to be associated with a photon energy. We assume here

that the cold electronics are dimensioned such that an amplitude of 0.25 0 occurs for a 7 keV pulse. Next the digital gain has to be tuned to produce a known BBFB loopgain G. For the simulations in this work we tune

it to G = 1, i.e. at SQUID input, an 0.25 0 peak is reduced by a factor 2. The relation between the gain factor and the gain bandwidth product, defined as GBW = g x 0.18/d,

14 is shown in Figure 2a for various choices for the SQUID transfer function.

a b

c d

Figure 2 a. Suppression factor for the maximum flux excursion during a pulse as a function of the gain bandwidth setting, for different options for the V- curve. The ideal sinusoidal V- and the V- measured for a VTT F5 SQUID almost fully coincide. A GBW = 550 Hz is applied to achieve the G = 1 setting for assessing the crosstalk. b. The simulated crosstalk on a 0.7 keV signal pulse due to non-linear amplification as a function of its carrier frequency, from a 2 keV perturbing signal at 5 MHz or a 6 keV perturber at 1 MHz, both with a time offset t = 0.1 c. Crosstalk from a 2 keV perturbing signal with a frequency offset f of 1.6 MHz with respect to the 0.7 keV signal pulse as a arrival time difference t. d. Given a f of 1.6 MHz and a t of 0.1



Figure 2b-d give an impression of the complexity of the crosstalk as a function of frequency difference f between the carriers on which the pulses are modulated, the flux excursion amplitudes s and p for the signal and perturbing pulses and the time difference t between the pulses. Figure 2b indicates that the dependence of crosstalk on the frequency difference between the carriers is modest, and that harmonic effects relatively minor. Figure 2c shows a dependence on the time difference which is more complex than the one in Figure 1. The zero crosstalk near +4 is retrieved, but near 4 it appears absent, and there are additional dips near + and . The global dependence on the amplitudes of the signal and perturbing pulse was expected from Eq. 4, but the level of detail shown in Figure 2d is quite unexpected.

To convey this complexity to the E2E simulator, the non-linear crosstalk is encoded via a lookup table with dimensions s, p, t and f, and the level of crosstalk for any pair of pulses occurring on the SQUID input is obtained by interpolation in these 4 dimensions.

An even more complex situation arises when 3 signals are simultaneously present on the SQUID input coil. During high count rate conditions, such as the observation of a 10 mCrab point source, up to 10% of the events might be involved in this type of crosstalk situations. As the size of the lookup table and the time to generate it became prohibitively large, we decided to treat these situations in a pairwise fashion, and accept the associated uncertainty in the derived crosstalk levels until it becomes clear that a more accurate assessment is indispensable.

3. IMPLEMENTATION AND FIRST RESULTS

One of the driving science cases for the X-IFU instrument is to measure the density of the missing baryons to 10% accuracy. Roughly half of the baryonic content of the Universe is suspected to be locked up in the WHIM, and detectable by its OVII and OVIII absorption lines in the 0.2 - 1 keV X-ray part of spectra from distant (z > 3), bright GRBs. This case requires the detection of OVII / OVIII absorption lines with expected equivalent widths down to 0.15 eV. For the X-IFU instrument this is a technically challenging science case as it requires a combination of: high energy resolution (better than 2.5 eV), high count rate (several 10s of cps per pixel for ~10 mCrab GRBs) limit to the observation time of 50 – 100 ks due to the source flux that decays as t1.2. The combination of narrow absorption lines and a high count rate point source make these observations in principle sensitive to all the crosstalk mechanisms discussed in the previous Section. And the decay of the source prevents mundane solutions for crosstalk issues such as a reduction of the count rate by a neutral density filter. This makes this science case a very interesting test case for the study of the various types of crosstalk. For the first tests of the implementation of the four crosstalk mechanisms discussed above assumptions were made that would bring out the maximum impact of the crosstalk. This pessimistic scenario was based on the following choices: Non-decaying point sources. As we were at first interested in the statistics of the effects at event level, the source flux

levels mentioned below refer to a constant source strength and not the start flux of a decaying GRB afterglow. Small Pixel Array with high multiplexing density. The electrical crosstalk is stronger with smaller carrier spacing and

higher bandwidth pixels. The hybrid SPA pixels have a bandwidth of ~1.7 kHz, so if they can be operated close to critical damping this translates in a crit 80 s.8

Figure 3 shows the assumed implementation of the SPA in the sensor array. The assumed multiplexing factor is 27, which is the maximum number of SPA pixels that can be logically assumed to be read out by one chain, not exceeding the baselined multiplexing factor of 40 uniform LPA pixels and consistent with dividing up the array over a multiple of 6 read out chains. Placing this many pixels in one read out chain promotes the probability for electrical and non-linear crosstalk.

Marginally optimized wiring. Ideally, pixels which are physically close on the array should be far apart in frequency of assigned carriers and vice versa. In the implementation of this principle illustrated in

Figure 3b. No nearest neighbors in frequency appear in the nearest 8 physical neighbors of a pixel. This principle breaks down in the center of the array as it was optimized for one quadrant, which was then replicated by rotation. This is of no consequence as coupling between different readout chains is presently not part of the implemented crosstalk mechanisms.



MHO EIMMMEI MUDD= IIIMIMI %MN MOM= DICHM= DDC MMEHLdiudIMI

CHUM 'Me MM IMMIMBrirlaDDDD 311 EOM !SWIM

lanai IBM IIE MM DM=MERL_ MM MM =MUMPRIM"' EN EN IMECHICHOM

IBMINEMECOMIDEMNIMMUUME IHMMEMMUMMM

IMIKOMMMMMMMMINIM COMMENOMMMMMMMINIM COMMMEREIMMMMMMM !Err MEMBOMEIMMMMCBMEt] 181 EIIMMICHOMMMMIMBM

iiDI MIMMEMIMMEMMMMM-OCOMMEIMEMEMMIEMMMMMMCUMMIIIHMEMEIMMMMMIEMM

No heatsinking layers. To assess the effect of thermal crosstalk we have assumed that no heatsinking layers are applied, and the crosstalk to nearest neighbors in horizontal directions is 103, 410 4 for diagonal neighbors and 810 5 for second nearest neighbors in horizontal directions.9

Low baseband feedback loopgain. In order to bring out the impact of crosstalk due to non-linear amplification of the SQUID, a low loopgain of 1 is assumed for the baseband feedback.

Figure 3 a. Illustration of a hybrid array, which is one of the possible alternatives for the hexagonal uniform baseline array, with the allocation of pixels to carrier frequencies and readout chains in a Small Pixel Array. Shading indicate s different read-out chains. Figure 4 summarizes the first results obtained with the E2E simulator. In the implementation in SIXTE, each incoming photon generates a series of crosstalk events that can then influence subsequent or previous events. At present all crosstalk mechanisms are simulated independently, i.e. possible non-linear effect of several mechanisms piling on one another are not yet taken into account. All crosstalk effects are (necessarily) computed as energy exchanges between pairs of events, even if more than two events may mutually influence each other. Under the current assumptions, crosstalk has almost no impact on the throughput of high-resolution events, as can be derived from Figure 4a. The 'grading' of events is performed on the basis of timing differences between subsequent triggers,13 and crosstalk events are not very likely to be so large as to set off a trigger. By far the most dominant type of crosstalk is thermal, as is clear from Figure 4b. Together with Figure 4c, which shows the average energy offset due to different types of crosstalk, the picture that emerges is that thermal crosstalk give numerous small nudges to the energy measured for an event, while electrical and non-linear crosstalk cause fewer but larger offsets in the inferred energy. Finally, Figure 4d shows the lowest crosstalk level in the 1%, 5% and 10% top percentile of events, when sorted on crosstalk impact. The 5% curve is also decomposed in thermal, electrical and non-linear crosstalk contributions. The dotted lines indicate that at a count rate of 1 mCrab (~4.5 counts per second per pixel) and under the above assumptions 95% of the events have an offset of their inferred energy due to crosstalk below 0.1 eV. Electrical and non-linear crosstalk only start to affect more than 5% of the events above the 0.1 eV level for count rates above ~5 mCrab. As the SPA is likely to see up to 10 mCrab some crosstalk mitigation is required.



o

Mea

n of

fset

due

to c

ross

talk

[eV

]o it

1+

Thr

ough

put H

ighr

es E

vent

so

óó

óó

óó

oF

otJ

Ca

IAM

O!

-1

F, E A m ç,r o

d

Low

est o

ffes

t in

top

perc

entil

e of

aff

ecte

d ev

ents

[eV

]Fr

actio

n of

Eve

nts

infl

uenc

ed b

y C

ross

talk

o

óo

F,o

0o-

i+

ó o

0 ir,,,

,1

1,,,

,111

,,

1,11

,1

1~ C x E C

oI

(I

I 4iii

liIi

1iii

iiI1

i, 1

I,i1

1

Figure 4 a.: Throughput of high-resolution events, including and excluding crosstalk. b.: Fraction of total number of events affected by crosstalk. c.: Mean offset in energy due to crosstalk in affected events for different mechanisms. b and c together show that thermal crosstalk has a smaller impact but on far more events. d.: Minimum offset for various top percentiles of events after sorting on crosstalk impact. For the assumed detector, a flux of 10 mCrab corresponds to a (maximum) count rate per detector of ~45 cps.

4. DISCUSSION AND CONCLUSIONS

The work presented in this paper is in full progress. The conclusions are based on a first implementation of crosstalk mechanisms in the SIXTE end-to-end simulator. Under conservative assumptions we have assessed the potential impact of four different crosstalk mechanisms. Clearly, the inferred impact of each of the crosstalk mechanisms depends heavily on these assumptions. Yet, already on the basis of Figure 4 some preliminary conclusions are possible: Thermal crosstalk dominates through the number of events it affects. This is a consequence of the fact that we are

looking at a strong point source, which results in a deposition of most of the photons in a small number of adjacent pixels.

Crosstalk appears to have a small impact on high-res throughput, in the sense that only a small fraction of the events are rejected after event grading. This is because crosstalk seldomly creates pulses that set off the trigger, and would thus influence the event grading.

Even under these conservative assumptions we find that for a 1 mCrab source, at ~4.5 cps per detector, >95% of the events have an energy offset due to crosstalk <0.1 eV

The conservative assumptions underlying these conclusions also point towards relatively easy mitigation measures that can be taken to reduce the impact of crosstalk:

elec

nlin

all

therm

all 5%

nlin elec

therm

all 1%

all 10%



The conservative assumption behind thermal crosstalk was that no heatsinking layers were applied. The heatsinking layers so far have been demonstrated on solid substrates,11 and they bring down the thermal crosstalk by at least an order of magnitude. For membrane-based pixels similar reductions could be achieved by backside coating with a heatsinking layer. This would render the impact of thermal crosstalk below 0.1 eV for over 95% of the events for count rates as high as 10 mCrab, or ~44 counts per second per pixel.

The impact of electrical crosstalk is based on a maximum number of high bandwidth pixels in one readout channel. In a hybrid array consisting of LPA and SPA pixels8 the two types of pixels could also be mixed per readout chain in a ratio of 1:11, in a way that would increase the frequency separation f between the SPA pixels by a factor ~6, and thus the energy offset due to electrical crosstalk by a factor ~36. The LPA pixels adjacent in frequency to the SPA pixels would receive electrical crosstalk signals, but no science signals, as the point source is (approximately) centered on the SPA.

The impact of crosstalk due to non-linear amplification by the SQUID was computed on the basis of a minimal baseband feedback gain of 1. In the baselined FDM architecture, gain bandwidth products of ~15 kHz are possible4, implying that even for the SPA pixels (with a bandwidth of ~1.7 kHz) gain factors of up to 8 might be realized. It seems therefore that also this type of crosstalk could be further reduced by at least an order of magnitude.

Observing strong point sources with a defocused telescopes has the effect of spreading the photons over a larger number of pixels, and read out chains.

The next steps to be taken are: Use the results obtained for an assessment of the impact of the crosstalk on critical science cases. For the WHIM

detection this raises questions such as: What is the effect on equivalent width of the relevant absorption lines? And which targets suffer most from crosstalk: the high count rate GRBs on which the crosstalk has the targets impact, or the relatively low count rate GRBs, which suffer less from crosstalk but for which the number of high-resolution events is already marginal? How much do the answers to these questions change when we assume that the above mitigation measures are implemented?

Assessment of other science cases on which crosstalk may have a critical impact, in particular with extended sources. A comparison with the case of a defocused high count rate point source observed with a uniform array of relatively slow ('hybrid') LPA pixels, is in particular interesting.

Based on the results of these investigations, we should be able to formulate an advice on the need for SQUID linearization.

Finally, further refinement of the crosstalk models that underlie the look-up tables used in the SIXTE simulator and a verification of the derived crosstalk levels against laboratory measurements are much needed to ensure that this complex set of problems related to crosstalk is sufficiently understood.

REFERENCES

[1] Nandra, K., Barret, D., Barcons, X., Fabian, A., den Herder, J.-W., Piro, L., et al., "The Hot and Energetic Universe: A White Paper presenting the theme motivating the Athena+ mission," ArXiv e-prints (June 2013).

[2] Barret, D., Trong, L.T., den Herder L., Piro, L., et al., "The Athena X-ray Integral Field Unit," Proc. SPIE 9905, (2016), these proceedings

[3] Akamatsu, H., Gottardi, L., et al., "Development of frequency domain multiplexing for the x-ray integral field unit (XIFU) on the Athena," Proc. SPIE 9905, (2016), these proceedings

[4] van der Kuur, J., Gottardi, L., Akamatsu, H., van Leeuwen, B.J., et al., "Optimising the multiplex factor of the frequency domain multiplexed readout of the TES-based microcalorimeter imaging array for the X-IFU instrument on the Athena X-ray observatory," Proc. SPIE 9905, (2016), these proceedings

[5] Wilms, J., Brand, T., Barret, D., Beuchert, T., et al., "ATHENA end-to-end simulations," Proc. SPIE 9144, 91445 (2014)

[6] Brand, T., Wilms, J., Dauser, T., Peille, P., et al., "Observing the WHIM with Athena," Proc. SPIE 9905, (2016), these proceedings

[7] Doriese, W.B., Adams, J.S., Hilton, G.C., Irwin, K.D., et al., "Optimal filtering, record length, and count rate in transition-edge-sensor microcalorimeter", CP 1185, Low Temperature Detectors LTD13, Proceedings of the 13th International Workshop, eds. B. Cabrera, A. Miller, B. Young, pp 450-453 (2009)



[8] Smith, S.J., Adams, J.S., Bandler, S.R., Betancourt-Martinez, G.L., et al. " Transition-edge sensor pixel parameter design of the microcalorimeter array for the X-ray Integral Field Unit on Athena," Proc. SPIE 9905, (2016), these proceedings

[9] Kilbourne, C.A., Doriese, W.B., Bandler S.R. et al., “Multiplexed read-out of uniform arrays of TES x-ray microcalorimters suitable for Constellation-X,” in Proc. SPIE 7011, 701104 (2008)

[10] Iyomoto, N., Bandler, S.R., Brekosky, R.P., Brown, A.D., et al., “Heat sinking, crosstalk, temperature stability for large close-packed array of microcalorimeters,” IEEE Trans. Applied Superconductivity, vol. 19, no. 3, pp. 557–560 (2009)

[11] Finkbeiner, F.M., Bailey, C.N, Bandler, S.R., Brekosky, et al., "Development of Embedded Heatsinking Layers for Compact Arrays of X-Ray TES Microcalorimeters," IEEE Trans. Applied Superconductivity, vol. 21, no. 3, pp. 223-226 (2011)

[12] Dobbs, M.A., Lueker, M., Aird, K., Bender, A., et al. “Frequency multiplexed superconducting quantum interference device readout of large bolometer arrays for cosmic microwave background measurements,” Review of Scientific Instruments 83, 073113, (2012)

[13] Peille, P., Ceballos, M. T., Cobo, B., Wilms, J., et al. "Performance assessment of different pulse reconstruction algorithms for the Athena X-ray Integral Field Unit," Proc. SPIE 9905 (2016), these proceedings

[14] D. Drung, and M. Mück, "SQUID electronics" in The SQUID Handbook, edited by J. Clarke and A.I. Braginski, Weinheim: WILEY-VCH, 127 – 170 (2004)



Documents

The impact of crosstalk in the X-IFU instrument on Athena science …x-ifu.irap.omp.eu/wp-content/uploads/2016/03/The-impact... · 2016. 8. 19. · The impact of crosstalk in the