gianlugi pessina infn milano bicocca università di milano ...pessina.mib.infn.it/wolte2014/tutorial...

Electronics for cryogenic detectors for high energy

particle physics

Gianlugi PessinaINFN Milano BicoccaUniversità di Milano Bicocca

… and also Lorenzo Cassina, Claudio Gotti and Matteo Maino

[email protected], http://pessina.mib.infn.it

mailto:[email protected]

wolte, Grenoble, July 7, 2014 2

Summary

• Preamble;

• Technology;

•

Ionization Detectors;

• High impedance Cryogenic Detectors;

• Low impedance Cryogenic Detectors.

References of other authors will be found at http://pessina.mib.infn.it/WOLTE2014 or starting from http://pessina.mib.infn.it , looking for WOLTE2014. This repository will be maintained for a few months. My references are all found at my web page, and will have direct link to it in the pdf version of this document.

http://pessina.mib.infn.it/WOLTE2014

http://pessina.mib.infn.it/


Preamble: Electronics for cryogenic particle detectors

The subject of this speech is all that I know about the front-end for cryogenic particle detectors.

This does not means that the subject is only “cryogenic” electronics, but, rather, all that is necessary to operate a cryogenic detector at its best.

That implies situations for which the better solution is a good thermal link with a amplifying stage working at room temperature.


• Preamble;

• Technology;

•





Technology (1)

Cryogenic operation of semiconductor devices and Low Frequency, LF, noise are strictly connected. From one side because the noise is an important parameter in detector readout.

From another side because LF noise is a probe of the skillfulness of the device at the given operating temperature.

The mechanism of conduction in semiconductor is LF- dominated also at room temperature, but that is not observed since the roll-off frequency is extremely high. The sense of this sentence is soon known if we consider the way carriers are available for conduction.


Technology (2)

Conduction band

Valence band

Free electrons

Ionized donor atoms

Free holes

Ionized acceptor atoms

Valence band

Conduction band

We need to remember: electron in the conduction band are present if donor atoms are introduced with a small ionizing energy for one electron al least.

Holes are present in the valence band if acceptor atoms are introduced with a small energy gap for accepting at least one electron from the valence band.


Technology (3)

Conduction band

Valence band

Free electrons

Ionized donor atoms

What does it means: any electron can jump from the parent atom to the conduction band if it acquires enough thermal energy, or a force is given.

The Boltzmann statistics applies and we know that at room temperature the thermal energy is about 26 meV per degree of freedom. If the energy gap between the donor atom and the conduction band is within a few tens of meV at most the probability that the electron is most of the time in the conduction band is closed to 1.

Said in an other way: anytime an electron fall in a donor atom it suddenly jump to the conduction band: we can say that the donor atoms are fast trapping centers.


Technology (4)

So far we have considered donor/acceptor atoms with a small energy distance to the conduction band of the semiconductor, we can see Si in the example above.

The important thing is that those atoms are not exhaustive for the process at hand.

Several other atoms behave as donor/acceptor, but with an energy distance which is not optimum: we can see several atoms around Si in the periodic table of elements.


Technology (5)

Here we have the list of the donor/acceptor atoms for Si.

The energies of the donors/acceptors are distributed in the gap. What does it mean?

Why do we consider all of these dopants?

= standard dopants, equal to fast centers at room T


Technology (7)

Why do we consider all of these dopants different from those of interest?

1. During the doping process atoms different from the dopants are around in the environment in small percentage and can mix with the former. The percentage depending of the cleanness of the environment.

2. The bulk material could be loosely contaminated.

This means that a good device is like wine: good years and bad years, good wafer and bad wafer…


Technology (8)

Speaking at room temperature.

The small percentage of spurious atoms does not compromise the working operation of the device and in many standard applications their presence is unimportant.



Technology (9)

But, why we consider and are interested to all these dopants?


= trapping centers most efficient at room T

Well, the probability that a carrier falls (is trapped) or escapes (is de- trapped) from them is dependent on temperature. The larger is the temperature, the larger is the probability that a deep trap, one with energy close to the middle of the gap, is active in this mechanism.

Expressed in an other way, the time a trapping center stays full/empty depends on temperature.


Technology (10)

EC

EV

Empty trap Filled trap Empty trap again

t-e0=t

2BIAS

TOTL

VeNI 2BIAS

TOTL

V1NeI 2

BIASTOT

LVeNI

TOTNII

I I I

Starting from the -law and expressing the current as a function of the total number of carriers the simplified model of the Generation-Recombination, G-R, is easily drawn:

So far we have a measure of the effect: the effect is noticeable if total number of carrier is small (small volumes or small doping concentration) and if the time constant of the process is very different from that of the signal, much smaller or much larger

Current change:TOTNII


Technology (11)

T = 0e-t TOTNII

Following the Generation Recombination, G-R, theory a noise is associated with this process. The noise being proportional to the square root of the Fourier transform the time dependence of the trapping process (Lorentzian function):

Current change: ionConcentrat_TrapN

IITOT

and time dependance:

A. Van Der Ziel

If several of such trapping centers are present with different concentration, characterized by different time constants, then the convolution of all the spectra lead to a 1/f behavior.

222n

11e


Technology (12)

Current change per single site:TOTNII

The interesting is that the larger is the DC current, the larger is the noise, at least for this mechanism: the biasing is a probe of the LF noise.

The amount of current change is larger if there is a distribution of trapping centers.

When there are a few trap types it is possible to observe small square pulses above the noise. These pulses are known as Random Telegraph Signals, deeply studied by Kandiah.

Kandiah IEEE TNS V 31 p 465 1984Kandiah IEEE TNS V 41 p 2006 1994Kandiah J Appl Phys V 66 p 937 1989

Kandiah NIMA V 288 p 150 1990Kandiah NIMA V 326 p 49 1993Kandiah NIMA V 377 p 197 1996Kandiah Solid State Electronics V 21 p 1079 1978


Technology (13)

Current change per single site:TOTNII

We have observed such signals at 4.2 K in the base current (developed across a resistor) of SiGe bipolar transistors.

IB = 10 A IB = 100 A

SPIE Volume N.5470, p.486-495, 2004NIMA, vol. A517, p.313-336, 2004INFN/TC-03/14. 7 Ottobre 2003

http://pessina.mib.infn.it/Biblio/Biblio_Articoli/SPIE Volume N.5470 p 486 496 2004.pdf

http://pessina.mib.infn.it/Biblio/Biblio_Articoli/NIMA V A517 p 313 336 2004.pdf

http://pessina.mib.infn.it/Biblio/Biblio_Articoli/INFN-TC-03-14.pdf


Technology (14)

Now, let’s come back a bit. A G-R center close to the valence o conduction band, where close means having a distance in energy comparable to the phonon energy KB T, will be empty most of the time (fast G-R), while a G-R center close to the middle of the gap will be filled most of the time.

Both G-R center do not have effect on noise (one is always filled, the other always empty); the latter is useless, while the former is fundamental for the conduction mechanism.

Following this viewpoint we can say that good dopants must be chosen between those that are very, very fast G-R centers at the operating temperature.




Technology (15)

That is the reason why GaAs works well at deep cryogenic temperatures, while Si not:

B Lengeler Cryogenics V Agosto 1974 p 439

T[K] KB T(meV)

300 26

77 6.7

4.2 0.4

Phonon energy:

Actually best performances for Si –

JFET are obtained in the range 100 –

150 K,

where a compromise must be met between LF noise and maximum mobility, III –

V

compounds has no lower limit, practically.


Technology (16)

Again on temperature.

Starting from the middle of gap, G-R always empty, going to the either of the 2 bands, G-R always filled, there will be some location in energy where the trapping/detrapping probability will equal: G-R centers with those energies will be more effective to contribute to LF noise.

Increasing the temperature this energy, that with maximum noise effect, will move toward the meddle of the gap, decreasing the temperature it will move close to the bands.




Technology (17)

R.A.Spaulding, Proc of IEEE, May 1968, 886

Kandiah

Here 2 examples of the behavior with temperature of the noise coming from a G-R center. The maximum on the noise shifts in frequency with temperature.

IMPORTANT: there is no way to know if the trapping center is active for electrons or holes.

Trap ATrap B


Technology (18)

Noise in Solid State Devices and CircuitsAldert Van Der Ziel, J Wiley & Sons, 1986

Another example of G-R in resistors.

Plots on this and the previous page are taken at fixed frequency, while the temperature is changed.

This is due to the fact that at the time of these studies the instrumentation able to calculate the FFT was not widely diffused in the labs.


Technology (19)

INFN/TC-03/14. 7 Ottobre 2003, pp. 53NIMA, vol. A517, p.313-336, 2004.

Using a spectrum analyzer is much more easy to study the behavior of a G-R center, if the device is good and only one center is present. Two or more centers make the analysis much more difficult.

Here an example of a good Si JFET optimized for room T operation and characterized around room temperature. From these spectra the traps energy is close to 0.5 eV from the bands and the candidates are: Mn (0.53), Cd (0.55), Au (0.54 A), Co (0.53A), Cu (0.53), Fe (0.51), O (0.51).

70 pF

http://pessina.mib.infn.it/Biblio/Biblio_Articoli/INFN-TC-03-14.pdf

http://pessina.mib.infn.it/Biblio/Biblio_Articoli/NIMA V A517 p 313 336 2004.pdf


Technology (20)

This is a Si JFET optimized for cryogenic operation and characterized around its optimum T.

From these spectra the trap energy extrapolated were close to 0.15 eV from the bands and the candidates are: O (0.16), Pb (0.17 A), Be (0.17A).

12 pF


Technology (21)

To summarize.

At room T active traps in Si ar at about 0.5 eV from bands.

Lowering the temperature down to 90 K the active traps for noise are at 0.2 eV from bands.

But what does it happen if we lower the temperature further? We expect that the donors or acceptors atoms should become source of noise, either.


Technology (22)Indeed this is the case.

Lowering T below 80 K the noise increases apparently as if it were white in this plot, since the bandwidth of the spectrum analyzer was 52 KHz.(Thermal noise does not justify this behavior unless it is considered hot electron effect at T much greater than room T)

Only at about 14 K the corner frequency from G-R noise from donors is visible.

Here the corner frequency of G-R noise from donors is shown vs T. It enters the spectrum analyzer range only at 14 K.

This rules out the use of JFET below 70 K, unless very special process are developed.

6.5 pF


Technology (23)Working below 80 K is possible with III-V compounds such as GaAs, for their small donor ionization energy (3 meV), or Si-MOS, for the large transversal electric field that is able to free the carriers from dopants.

B Lengeler Cryogenics V Agosto 1974 p 439

n+ n+

p

VD

VGID

n+ n+

p

VD

VGID

n+ n+

p

VD

VGID

High field


Technology (24)Unfortunately, MOS and GaAs and, generally, III-V compounds work well at cryogenic temperatures, but their LF noise is not as small as that of Si JFET, or it must be well.

SG

D

Substrate

Charge tunneling in the oxide or in the substrateT becomes: T = 0e-t , ' e Z '

Z

This is McWhorter model (1956) of LF: tunneling in the oxide

Other models interpret the noise in term of mobility fluctuation or conductance fluctuation, see Hooge.

A single kind of trap can determine several different time constants depending on its distance from the surface of separation between the oxide and the semiconductor.

This model seems to account well of the noise. A proof of this is the very different level of LF noise in PMOS and NMOS: the former shows a lower level that come from the fact that holes can tunnel very little the oxide being slower and heavier of electrons.


We must remember…

Passed 13 November 2013, Born 1945, Buenos Aires

Before addressing the following results I must remember Daniel Victor Camin who was a pioneer in cryogenic front-end with GaAs MESFET.

He passed the last November 2013 after a long illness.

I was his student and collaborator since the middle ’80 of the last century till the beginning of the present century.


Technology (25)

Cryogenic V 29 p 857 862 1989

< 10 pF

GaAs MESFET work well at deep cryogenic T. Unfortunately LF noise, that improves with T lowering, remains quite high and is a limit for several applications.

LF in GaAs comes from the crystal lattice imperfections (el-x traps), defects created between the semiconductor and metal surfaces, etc.

http://pessina.mib.infn.it/Biblio/Biblio_Articoli/Cryogenic V 29 p 857 862 1989.pdf


Technology (26)

Cryogenic V 29 p 857 862 1989

< 10 pF

The very important consideration is that LF noise from G-R is not an intrinsic process and depends mainly from the purity of the starting material. In this sense an analogy exist: only from a good “grape harvest”

it is possible to hope in a good wine.

Si is a semiconductor that can be grown “easily “

with a good level of purity. This is much more difficult with GaAs.

The uncertainty principle applies here in stating that nothing is perfect: GaAs works well at cryogenic T, but LF noise could be its performance limit.

As a confirmation of the above consideration the noise we measured with the first 3SK166 GaAs MESFET from Sony was never obtain again in all the following batches we had.

Of course, an attempt to buy all the MESFET remained available from that “golden”

batch

failed as already distributed around the world…

http://pessina.mib.infn.it/Biblio/Biblio_Articoli/Cryogenic V 29 p 857 862 1989.pdf


Technology (27)

One way to try to optimize LF noise in III-V compounds is to grow by himself the starting wafers, or to select the wafers on which to diffuse the own transistors.

This is what is done for instance at the CNRS (Laboratoire de Photonique et de Nanostructures, LPN) where they develop HEMT designed for cryogenic operation with LF noise.

100 pF

That is a very promising component.


Technology (28)

What about white noise with T?

Thermal noise improves with lowering T, being proportional to T, but, generally, its improves less than expected.

That is due to the so called hot-electron effect or, said in an alternative way, to the fact that electron gas has a T larger than that of the lattice.

IEEE TNS, Vol. 51, pp. 2975-2982, 2004.

Here one example with a Si-JFET from room T down to 80 K. Transconductance does not change so much and input thermal noise has an improvement that is lower at large channel current, proving the hot-electron effect.

http://pessina.mib.infn.it/Biblio/Biblio_Articoli/IEEE TNS V 51 p 2975 2982 2004.pdf


Technology (29)

In this second example the series white noise of the HEMT from CNRS can be seen from 80 K down to about 2 K.

The square root of noise improves a factor of about 2 instead of a factor larger than 5.


Technology (30)

Parallel noise stops to exist for practically all the devices, but bipolar SiGe.

IEEE TNS V 50, pp. 921, 2003.

Here an example of SiGe characterized for parallel and series noise from room temperature down to 4.2 K.



• Preamble;

• Technology;

•





Ionization Detectors (1)

So far we had a brief view of the operation and noise performances of devices at low T.

LF frequency noise and white noise need to be considered and the application will drive the more suitable technology. As it will be seen.



Before entering into the discussion it is interesting a short historical break about the origin of the very front-end for particle detectors. Apparently this has not much to do with the subject of this tutorial, but …

What triggered me was a first paper I found that dated 1956: it was the first voltage preamplifier for particle detectors, I suppose. It employed valves!

The author of the circuit was Emilio Gatti et Al.

The circuit and …

…the configuration

Look at the supply voltage:300 V!

Il nuovo cimento V 3 p 473 1956

Cascode


Ionization Detectors (3)At this point I asked to Pier Franco Manfredi about the first charge sensitive preamplifier. He told me that Veljko Radeka was the first to develop a genuine charge sensitive preamplifier.

Then I asked to Veljko, which sent me a very nicely short report.

Veljko mail:

“The first concept of charge to voltage conversion on a capacitor in feedback appeared in 1948 in a Swiss patent, slide 1; Only one vacuum tube was needed! “

This configuration is equivalent to a “common source topology”.


Ionization Detectors (4)According to Veljko, charge sensitive preamplifier for nuclear detector having valves as active components have been used and known since World War II.

Veljko mail:

“Vacuum tube charge amplifiers were used until ~1965; an example is shown in slide 2; Cascode configuration with tubes was known from World War II“



Then, it arrived the transistors….

Veljko mail:“I think, I was the first to use a JFET at the input (BJTs the rest) in 1963 (none were available before then); published in an obscure place; I am on travel now and don't have access to that paper;“


Ionization Detectors (6)Here is the first charge sensitive preamplifier based on the use of transistors. It was the 1968.

This preamplifier is, at the same time, the first example of cryogenic electronic

Veljko mail:“In 1968 (paper you requested) I described in some detail, a charge amplifier for germanium detectors, with a JFET at low temperature, slide 3;;“

Cold stage


Ionization Detectors (7)This preamplifier was designed to readout a Ge ionization detector that works at 78 K.

The input JFET is a Si transistor operated closed to the detector. The second stage of this dominant pole amplifier is at room temperature.

At that time both JFET and bipolar transistors were available.

The JFET here is the 2N4416, one of the most popular JFET ever used.


Ionization Detectors (8)The input transistor T1 is in common source configuration and is loaded by T2, in common base configuration: T1 and T2 form a cascode that has a twofold effect. Maximizing the bandwidth and minimizing the miller effect of T1.


Ionization Detectors (9)T2 is loaded with a current source, a large impedance RL in this case. Buffer T3 load RL with a large input impedance.

Capacitance Cs introduces the main pole and allows to compensate the network.


Ionization Detectors (10)Rf and Cf are the feedback elements that close the loop of this feed-backed amplifier.

If Qi is the input signal charge, then the output signal voltage has an amplitude of approximately -Qi/Cf, followed by an exponential roll-off with time constant RfCf.

T4 is the line driver.


Ionization Detectors (11)This preamplifier deserve this attention since it is the more widely used configuration since then.

Commercial Ge detectors (working T at 78 K) usually have the input JFET transistor and the feedback element very close to the detector pin, at cold, the rest of the preamplifier is placed at a few cm outside the Dewar, or cold finger.

The second stage is located at room temperature for various reasons, one of which is for minimizing dissipation in the liquid, so minimizing the boil off.

Noise performance is important for the selection of the cryogenic device and to understand while Si JFETs have been the only used transistor for this cryogenic detector.


Ionization Detectors (12)The typical acquisition chain with a charge sensitive preamplifier includes a shaper at the output. The shaper has 2 aims: filtering of the noise and shaping of the signal.

VoA-A

CD CA

CF

RF

Qi (t)

F()Vo

1np1p)p(F

2

Di2Ai2

Rfi

2Ae

Shaper CR-RCn

The noise sources are the parallel noise of the preamplifier, the feedback resistor and the detector and the series noise of the preamplifier.

Parallel noise follows the same path as that of the signal, and is integrated on the feedback. Its contribution at the output increases as the filtering time constant increases.

Parallel noise is generally shot noise, and is a white noise source, unless a bipolar transistor is at the input of the preamplifier.



Series preamplifier noise, eA , is amplified by the ratio of the input to the feedback capacitance (preamplifier input is a virtual ground), and it is unfiltered at the preamplifier output.

eA is the quadratic sum of white and LF contributions.

White series noise has therefore greater effects at small shaping time, or large filtering bandwidth.

LF series noise is almost unaffected from the filtering bandwidth as long as is 1/f. This is mainly due to the fact that 1/x is a function that can not be integrated at large x.

VoA-A

CD CA

CF

RF

Qi (t)

F()Vo

1np1p)p(F

2

Di2Ai2

Rfi

2Ae

Shaper CR-RCn



Performing the calculation the final expression for the so called ENC, or Equivalent Noise Charge, namely the charge to be injected at the input for obtaining a S/N of 1 has the magical and simple expression of:

2A

2Rf

2Df

2sw2

FAD2 iiiAeCCCENC

A good tutorial on this is: La Riv del Nuo Cim V 9 pp 1 71 1986

,

and

are coefficients that depend on the kind of filter, while

is the shaping time constant.

esw is the series white noise and Af /f is the LF series noise of the preamplifier.

VoA-A

CD CA

CF

RF

Qi (t)

F()Vo

1np1p)p(F

2

Di2Ai2

Rfi

2Ae

Shaper CR-RCn



At small shaping time series white noise dominates, while at large shaping time dominates the parallel noise. LF noise adds a plateau.

If the rate of the incoming signals is not a constraint, the shaping time can be optimized on the basis of the input transistor used.

Here an example of ENC as a function of shaping time constant.

2A

2Rf

2Df

2sw2

FAD2 iiiAeCCCENC

0 10 20 30 40 500

1000

2000

3000

4000

tau ( sec)

EN

C (e

l)

Solo bianco serieSolo 1/f serieSolo paralleloLa somma di tutto

10-1 100 101 1020

1000

2000

3000

4000

tau ( sec)

EN

C (e

l)

Solo bianco serieSolo 1/f serieSolo paralleloLa somma di tutto



Let’s estimates the maximum LF noise in a good working condition in which the detector has a capacitance, CD , of 30 pF. The maximum noise from electronic would be about 500 eV FWHM, that correspond to about 70 el ENC in Ge.

Contribution from LF noise should be 1/10, or 7 el ENC.

In matching condition the input JFET of preamplifier has a capacitance similar to the detector capacitance, or CA =30 pF.

is almost independent on shaping and can be considered equal to .

Putting the numbers together we expect for Af a value close to 10-16 V2, namely a noise contribution of the order of 10 nV/Hz @ 1 Hz.

With

about 0.5 from semi-Gaussian shaping white noise should be close to 1 nV/Hz.

2A

2Rf

2Df

2sw2

FAD2 iiiAeCCCENC



Experiment with accelerators at LHC, CERN, foresees very high rates, 25 ns or 40 MHz, and large energies, 15 TeV maximum.

Both the detectors and the electronics must be very fast and able to sustain high radiation levels.

At such high rate thermal detectors, that are limited by the speed of sound roughly, are not used.

At LHC the only cryogenic detector is used at ATLAS. This detector is the Liquid Argon Calorimeter, LAR, with an Accordion geometry.

The detector is a drift chamber in which ionization is created in the liquid. A high voltage is applied to create a large electric field in the liquid so as the drifting charge is collected by the electrodes where the electronics can amplify the signal.



Accordion is the shape of the electrodes in the argon.

With this geometry the particle tracks can be reconstructed saving in number of readout channels. We are speaking here of thousands of readout channels, anyway.

Between the electrodes there is Argon and also some material, lead, able to generate secondary particles.

Electrodes extend in length the whole detector and their shape allow to reconstruct the track



The drifting speed in argon is quite slow and for maintaining the high rate the shaping of signals is fast, and bipolar, so that only the first part of the signal is taken.

The peaking time is a few tens of ns. At such high speed LF noise is not required to be small, while the important is the minimization of white series noise: remember this latter is proportional to 1/.

At that time there were developed 2 Charge Sensitive Preamplifiers, CSP, competitors: the first cryogenic CSP in Si-JFET and the first cryogenic CSP in GaAs-MESFET.



The first preamplifier was in Si-JFET. The process was from INTERFET that developed its first (and last) monolithic circuit.

NIMA V 360 p 158 1995

To match the accordion electrodes the capacitance of the input transistor is a few hundred pF.

Input JFET J1 is cascoded by J2 and Rext provides additional current necessary to the large input J1.



J3 is configured as a current generator: it works at Vgs=0 V providing its maximum current (the channel is closed to current if Vgs is lower than its threshold Vt, about -1 V for the JFETs of this preamplifier. At intermediate currents we expect 0< Vgs < Vt.).

J7 buffers the output (source) of J3, J4 buffers the output of J7 to the drain of J3. As a consequence the drain and the source voltage of J3 are dynamically at the same potential and the current in the source to drain resistor of J3 is negligible: this enhance greatly the impedance of current generator J3.



J8 buffers J7 for driving the 50 transmission line.

DC feedback is closed with Rf to R2 the current Ir that flows in R2 flows also in R1. The consequence is that a DC shift is added at the J8 source that rises the output, hence J3 source, at a suitable positive value.

AC feedback is closed through Cf to the J8 source allowing to obtain a small output impedance.

IrVG -0.5 V



The preamplifier showed remarkable performances.

A very interesting characterization of the noise with temperature is shown at 2 different shaping time constant of 50 ns and 20 ns, the targets of the experiment.

As it can be seen the optimum for the noise was got between 100 and 150 K, as expected.

Power dissipation of this preamplifier was around 80 mW.This is an important parameter considering the large number of channel foreseen and the liquid boil off.



A CSP in monolithic GaAs technology was developed.

The introduced technique for the temperature compensation of the working condition of every transistor was patented. The designed network set the working condition of |Vgs|=Vds, for minimizing the applied voltage, hence power consumption.



After the patent there was an evolution in the circuit design: auto bias.

Q1 is cascode by Q2. Q3 buffers the source of Q2 to its drain, bootstrapping it. This way the output impedance from Q1, Q2 and Q3 is very high although the small value of Rds of Q1.

The same consideration is valid for the current generator Q4, bootstrapped by Q5. This time double cascode is unnecessary since Q4 is a small area transistor.

IEEE TNS V 40 p 759 763 1993

|Vgs|

2x|Vgs|




Small output impedance from short gate length devices such as GaAs, HEMT, MOS, etc deserve attention in the design of closed loop amplifiers (where large gain is needed). That is due to the short channel effect at saturation.

This effect is less important for wide band amplifiers that works on small loading impedances, typically 50 .

IEEE TNS V 41 p 1260 1267 1994




IEEE TNS V 43 p 1649 1996

Noise: 100 nV/Hz @ 1HzNoise from the monolithic GaAs process was very good at high frequency, but LF noise was large. Nevertheless at LHC the demand was for high speed and that was not a limitation.

ENC performance of the GaAs CSP and Si-JFET CSP were comparable.

Power dissipation of the 2 CSP were also comparable. GaAs CSP allowed 2 working modes at 22 mW and 60 mW.



Ionization Detectors (28)Both Si-JFET and GaAs-MESFET CSPs were validated against the level of radiation expected at LHC with success (see ref below).

But… at the time that the selection of one of the two alternatives was supposed to be taken a third alternative arose…

IEEE TNS V 42 p 2266 1995

NIMA V 330 p 228 1993

Actually the alternative was under study since a while, but the results arrived just then. This was an example for which a cryogenic detector was readout with room temperature operated front-end…

In essence it was proved experimentally that when the shaping time of the signal is smaller than the delay time of the connecting transmission line the capacitance of the line has marginal effect on the noise performance



At the end the current preamplifier for the LAr at ATLAS was located at room temperature. Its feedback is set for having a “cold”

input matching impedance to the

transmission line.

Noise was worst with respect to that of the CSP, but not so worse to convince the adoption of the cold electronics.

This solution offers 2 advantages: the electronics is not facing directly the particle beam and the radiation level on it is small, it does not work in the liquid.


• Preamble;

• Technology;

•





High impedance Cryogenic Detectors (1)

In this section of high impedance detectors I consider mainly Cryogenic detectors for low background physics.

In this class I include thermal detectors to which we have to consider associated also light detectors, that are becoming important in several experiments.

In considering this class of detectors I would address also Ge-BEGE detectors to be used in the experiment GERDA, located at Gran Sasso underground laboratory. This experiment introduces to the low background applications, a field totally complementary to accelerator physics.

Experiment dedicated to the low background physics for the study of the neutrino mass and dark matter are CUORE, EDELWEISS, GERDA, LUCIFER …



What characterizes these experiment is a sort of LF noise that does not come form electronics or detectors but from the environment.

)2Z,A()Z,A(

2

0

Double

decay with neutrinos, 2

(standard model)

Neutrinoless double

decay, 0

The detector is sensitive to the electron only.

Electron spectrum of energy



Both decays are rare process, the first is foreseen and measured, the possible existence of second is under study.

The present limits depend on the candidate isotope to decay. Anyway the probability that an atom undergoes either decay varies between 1023 to 1025

years: a very, very slow trapping center!

These numbers are of the same order of magnitude of the Avogadro number and the hope of measuring some events requires a large number of detectors channels containing atoms candidate to the decay.

As an instance CUORE foresees 1000 detector crystals with a few hundred Kg of Te. In this experiment 0

events, if exist at the quoted rate, are expected of a

few counts per detector per year.

)2Z,A()Z,A(

2

0

Double

decay with neutrinos, 2

(standard model)

Neutrinoless double

decay, 0



Decays from any other source with similar energy are source of noise, background noise.

Material to be located close to the detectors must be selected for not having such sources. Here the background level for several materials typically used in electronics:

Jinst, Vol.4, P09003, p. 1-17, 2009.

As it can be seen Fiberglass is strongly prohibited. Other elements that are to be avoided are resins, etc. Tins,..

http://pessina.mib.infn.it/Biblio/Biblio_Articoli/JINST V 4 P09003 pp 1 17 2009.pdf



In general electronics devices have to be used in bare die.

This is the case for instance of GERDA experiment

First stage: JFET and feedback only close to the detector pin

Second stage at 1 mt for minimizing background.



The structure of the preamplifier is similar to Radeka scheme.

Only, in this case, the second stage is based on the use of a commercial CMOS OA: AD8651.

This is the very new and interesting feature of this circuit: the proof that CMOS devices works well at cryogenic temperature.

rsTe

RT

RT

rsTe

Tank

VOUT1rsTe 2rsTe

BufferZF

Charge Sensitive Preamplifier

II st

In this approach the degradation on the phase margin coming from the delay and the stray capacitance induced by the connecting lines must be accounted to avoid instabilities.


High impedance Cryogenic Detectors (7)In GERDA the fast part of the signal is exploited for the rejection of double hits, while the shaping time is at 10 s.

To take into account both the slow and fast component of the signal having only the input transistor and the feedback at cold and alternative approach has been studied, although the classical scheme is being used: GEFRO.

In GEFRO the 10 mt of the connecting link is accounted in the compensation network and the signal bandwidth is limited to 1 s of rise time, suitable to the 10 s shaping time. The fast signal is read at the drain, in open loop condition and shows a rise time of about 10 ns being read at the terminated transmission line.

rsTe

RT

RT

rsTe

Tank

VOUT_fast

ZF

GEFRO

VOUT_CSC

IEEE TNS Vol 61, p 1259, 2014 Jinst, V 6, P05006, pp 1 19, 2011.

http://pessina.mib.infn.it/Biblio/Biblio_Articoli/IEEE TNS V 61 p 1259 2014.pdf

http://pessina.mib.infn.it/Biblio/Biblio_Articoli/JINST V 6 P05006 pp 1 15 2011.pdf


High impedance Cryogenic Detectors (8)A further parenthesis on CMOS a cryogenic temperatures is CUBE a CMOS preamplifier in monolithic (0,35 m?) technology, spin-off with the Politecnico di Milano.

It has a PMOS at the input optimized in geometry. It was originally developed for room T operation, but it was able to work very well down to 60 K temperature.

IEEE TNS Conf Reco 2011 N40-5

T=300 K1 s shaping


High impedance Cryogenic Detectors (9)CUORE is an array of large mass TeO2 crystals held at about 10 mK. The thermal increase of T generated by the particle is measured by a NTD Ge thermistor. CUORE will study the 0

of Te.

Signals are very, very slow and here we have an example:

Signals evolve in about 0.5 s, with a rise time of close to 100 ms, the bandwidth is about 10 Hz. This is due to the large heat capacitance of the crystal that give rise to large thermal constant.


High impedance Cryogenic Detectors (10)NTD Ge thermistors of CUORE have impedance in the hundred of M

range.

Parasitic capacitance in the order of a few hundred pF does not degrade the slow signal although the large value of the thermistor impedance.

This permitted to locate the readout at room temperature and to optimize the connecting link to both radio-purity and thermal conductance.

2009 IEEE NSS Conf Rec, N13-042 JLT, V 151, p.964, 2008.

Pre

X 1000

http://pessina.mib.infn.it/Biblio/Biblio_Articoli/2009 IEEE NSS Conf Rec N13-042 pp 389 395 2009.pdf

http://pessina.mib.infn.it/Biblio/Biblio_Articoli/Journal of Low Tem Phy V 151 p 964 970 2008.pdf


High impedance Cryogenic Detectors (11)In CUORE parallel noise from both the preamplifier and load resistors operated at room temperature was taken into account. A pair of 30 G

load resistors (optimized

for small LF noise) is employed and the gate current of the input JFET is below 100 fA.

RL

ZDPRE

PB+

PB-RL

Refrigerator

TD =10 mK

OUT

IEEE TNS, v. 49, p. 1808, 2002


High impedance Cryogenic Detectors (12)A completely different situation is the case with small mass bolometric detectors for which the thermal relaxation time is much smaller. The signal is relatively fast, 10 to 100 of s and the parasitic capacitance that shunts the thermistor must be minimized.

Temperature of operation of the detectors is in the order of tens of mK , the load resistors are at the same temperature while the input stage of the front-end is at its optimum temperature of operation.

The electrical connecting link between the thermistor and the gate of the input transistor is particularly important:

IIS

VDRL

DET. BIASREFRIGERATOR

VOL. GAIN

1

LOW NOISE II STAGE

VE

TB =10 mK

TB 100 K



From one side it must be short to minimize stray, but it must show a small thermal conductance for not injecting heat to the detector.

This is of particular concern when the readout is with a Si JFET (almost always as I know) working at 130 K since in a few cm the heating power must be driven to the heat sink.

This application should get a big benefit from the possibility of using a HEMT or a device from III-V compounds.

Importance of the electrical link:

IIS

VDRL

DET. BIASREFRIGERATOR

VOL. GAIN

1

LOW NOISE II STAGE

VE

TB =10 mK

TB 100 K


High impedance Cryogenic Detectors (14)Here an example of front-end of an array of small mass bolometers operated in space, D.McCammon.

A Si-JFET having a small input capacitance (although this is not mandatory since the signal is not so fast) buffers the detector and drives the line to room temperature, where a second stage with very low noise amplifies the signal to the DAQ.

The Astroparticle Journal V 576 p 188 2002


High impedance Cryogenic Detectors (15)These array of bolometers showed less than 15 eV FWHM energy resolution on average. Limitation was essentially given by LF noise of the thermistors, which prevent to reach the intrinsic limit.

This means that another possible application for HEMT was available.

The Astroparticle Journal V 576 p 188 2002


High impedance Cryogenic Detectors (16)The possible use of HEMT in this readout could help in minimizing the heat injection. LF noise must be small, of course, but this is not the only requirement.

A good dynamic performance can be obtained if the drain to source impedance of the HEMT is adequate (although this can be obtained by circuit design).

This seems the case for HEMTs from CNRS, in the current range below 0.5 mA.


High impedance Cryogenic Detectors (17)So far we have considered the detector DC biased.

Signal modulation obtained by AC bias could help in suppressing LF series noise since the signal is shifted at high frequencies.

RF

RBO

Vb sin(t)

SIGNAL MODULATION

VO

VO =RL

RL + RBORBO sin( )t

Series pream.noise

Signal spectrum MODULATION

Signal spectrum

Series pream.noise

AFTER DEMODULATION THE SERIES 1/f NOISE OF THE PREAMPLIFIER IS ATTENUATED.


High impedance Cryogenic Detectors (18)AC biased is applied by the EDELWEISS collaboration for the readout of both thermal and ionization channel.

JLT (2012) 167:645–651

AC bias vs DC bias what is the best? Difficult to say. For sure there is a limit in AC bias that is reached when the thermistor has a large impedance.

Other parameters to take into account are stability, biasing amplitude accuracy, frequency stability, …


High impedance Cryogenic Detectors (19)There are also some suggestion and solution to the room T stage. For instance the implementation of a pseudo differential stage for the suppression of the ground disturbances.

Pseudo differential is meant here on the optimization of the number of JFET at the amplifier inputs with respect to the impedance of the node to be read.

IEEE TNS, V. 53, pp. 2861, 2006.IEEE TNS, V 58, p 3204, 2011.


http://pessina.mib.infn.it/Biblio/Biblio_Articoli/IEEE NSS V 58 pp 3204 3212 2011.PDF


High impedance Cryogenic Detectors (20)Background reduction can be achieved also adding some signals which help in discriminating the events.

In this respect light produced in a crystal by an

particle (one of the main sources of background) is easily distinguished from , electrons and the like.

The new generation of experiments, or upgraded experiments, on

decay are all trying to exploit this channel. Examples are LUCIFER, LUMINEU,...

The crystal is coated with a reflecting foil which drive photons towards a thin sheet of Ge or Si which absorb them increasing its T. A thermistor or a whatever other sensor converts the increase of T in an electrical signal.

Depending on the properties of the crystal the light can be Cherenkov or scintillation.


High impedance Cryogenic Detectors (21)Results obtained so far are quite promising.

The light signal can be fast and a cold stage is welcome that can be either Si- JFET or, better, HEMT MOS,…


• Preamble;

• Technology;

•





Low impedance Cryogenic Detectors (1)

This class of detectors include TES, MMC, MKID, …

All this sensors are characterized by being metal and/or superconducting: very, very small impedance sensors.

For most of the only way to readout is by interfacing them with a SQUID, a sort of quantum transformer that allow matching with the preamplifier input.

The SQUID has a low impedance output and can be easily readout with a room T operated amplifier. Here a beautiful example:



Squid readout provides also for biasing the SQUID.

Stabilization can be DC or AC.



Circuits with SQUID are easily multiplexed.



A readout exception to the use of SQUID is with MKID, KID.

MKIDs are in essence superconducting stubs bias with a wave signal whose wave length is multiple of the stub in order to have resonance.

As soon as the MKID is subjected to a T increase its impedance change, inducing a change in both the amplitude and phase of the applied signal.



The readout for a KID consists in a RF signal input and the resulted output.

A cryogenic LNA (Low Noise Amplifier) works at such large frequency (order of GHz) to readout the MKID.

This is the domain of III-V compounds (SiGe is also used).



Changing the length of the stub allows to tune the resonance at a different frequency: MKID with different lengths can be connected to a single line excited with a superimposition of waves.

It is therefore possible to develop array of sensors multiplexed on a single line.



Here a very beautiful scheme from CNRS-IN2P3-LPSC Grenoble:



Here a more elaborated scheme from Open Source Readout for Kinetic Inductance Detectors, CALTECH:



LNA for MKID and also antenna sensors are a cascade of common source stages AC coupled at high frequencies and DC coupled at DC. The circuit works open loop for the AC signals.



Here an example of a SiGe biased with DC mirrors



Here a cascode solution


Conclusions

Thanks…

gianlugi pessina infn milano bicocca università di milano ...pessina.mib.infn.it/wolte2014/tutorial...

Documents