EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH

Bipolar versus Unipolar Shaping

of MDT Signals

Werner Riegler, Martin Aleksa

Harvard University, CERN

Abstract

The MDT frontend electronics scheme, as presented in the TDR, was optimized for fastlinear gases like Ar=N2=CH491=4=5 [1], but since these gases contain hydrocarbons which

seem to be responsible for aging problems, the current baseline gas is Ar=CO293=7 which

shows nice aging properties but is slower and signicantly nonlinear. The nonlinearity and

reduced speed have a signicant impact on the pulse shapes: the trailing edge resolution is

reduced and we get multiple threshold crossings per signal with the current shaping scheme.The dierent pulse shapes as well the fact that trailing edge information and double track

separation information were not found to be very useful made a complete review of the front

end electronics specication necessary.

1

2

1 Introduction

At the time of the TDR [2] the baseline gas for the MDTs was Ar=N2=CH491=4=5 (550ns

maximum drift time) and after detailed test beam studies of this mixture the following front

end electronics specication was worked out (g. 1):

Wire diameter: 50um

Tube diameter: 3cm

Gas gain: 2-5x10^4

Preamplifier peaking time: 10-15ns

Gas pressure: 3bars

Phase Calibration System

1

2

3

LVDSPreamp Diff Amp Shaping Amp

Discriminator 1

Discriminator 2

Leading Edge Trailing Edge

Threshold 1

Threshold 2

Charge

ADC

Logic

Control Logic (JTAG Interface)

MDT

1M380

HV

ΩΩ

Figure 1 : MDT frontend schematic and operating parameters as presented in the TDR.

Preamp peaking time of 10ns to 15ns. Although the resolution is better for short

peaking times a 'slower' preamp is preferred for reasons of stability and multiple hits.The dierence in resolution between a 5ns and 15ns peaking time preamp is about

10m and can even be reduced by performing a time slewing correction.

Short gate ADC encoding the charge information in a pulse width. Measuring

the signal charge with an ADC for 10-20ns after the threshold crossing time allows to

correct for time slewing. It also provides very useful for monitoring purposes.

Unipolar Pulse shaping. Although unipolar pulse shaping requires an active base-

line restoration circuit to avoid baseline uctuations due to high rates (up to 400kHz/wire)

it is preferred to bipolar shaping since

1. We get one threshold crossing per signal.

3

2. The trailing edge of the signal which is correlated to the bunch crossing can be

measured to t 25ns and can be used to eliminate out of time events to increase

pattern recognition eciency and reduce fake track rate.

3. The information of a second discriminator with high threshold can be encoded in

the same channel.

Programmable lter time constants. In order to adjust the tail cancellation circuit

for dierent gases the lter time constants have to be programmable.

Discriminator hysteresis is important to avoid multiple threshold crossings.

A second discriminator with high threshold allows to identify muon signals

that are piled up to background signals. The separation eciency was shown to beabout 60% for a leading edge separation of > 100ns. Encoding the high thresholddiscriminator information in one channel with all the other hits gives on average 1.7

hits/signal.

There are three modes of operation:

1. Time over threshold mode giving leading and trailing edge, one hit per signal.

2. ADC mode giving leading edge and charge, one hit per signal.

3. ADC+high threshold mode giving leading edge, charge and double track infor-

mation.

Four 24-channel TDCs are daisy chained into one front-end link.

In the meantime, many things were studied in more detail or have been changed:

Aging problems with gases containing hydrocarbons resulted in switching to the gas

Ar=CO293=7 which has a maximum drift time of 700ns and is very nonlinear. Adjusting

the tail cancellation time constants such that we arrive at one hit per signal, the trailing

edge resolution is strongly degraded (t 80ns) which makes this information not very

useful.

The double track separation information was shown not to be useful for pattern recog-nition which is mainly due to the fact that, using the high threshold discriminator, we

get multiple hits even for single tracks. Also the decreased resolution of the recuperated

hit ( 160m) is a major drawback.

The trailing edge information (even with a resolution of 25ns) was never used or found

useful in simulation [3].

4

In some high rate regions in the muon system it is not possible to daisy chain four

TDCs into one front end link, but in order to keep the number of cables to the MDTs

low it was important to daisy chain as many TDCs as possible into one link, which

resulted in a non uniform scheme. Since the ReadOutDrivers (NIMRODs) have now

been moved from the cavern to the USA15 and the front end links can not be 100m

long, a dierent scheme was adopted.

A data concentrator card will gather the TDC signals from one chamber and transmit

them to the NIMRODs through an optical ber, so there is no need for daisy chaining

TDCs. Each TDC is connected directly to the concentrator card.

In the previous scheme, the maximum allowed data rate was 400kHz per tube with an

average of 1.7 hits per signal. Using bipolar shaping for the new baseline gas results inan average of 3 hits per signal, so this solution would not be acceptable with the oldreadout scheme. However, the new scheme might impose dierent requirements on the

maximum number of hits per signal, so for each front end scheme we have to study theimplications on the readout.

All that made a complete review of the frontend electronics scheme necessary. Especially

the question of unipolar versus bipolar signal shaping has to be reconsidered since bipolarshaping would not require an active baseline restorer as well as programmable time constants,

which would simplify the whole frontend scheme drastically.

5

2 MDT performance for dierent shaping schemes

Most of the signals in the muon system result from background photons. These photons

create high energy scattering electrons, some of them able to penetrate several tubes. The

photon spectrum and other background characteristics are given in [4].

The average charge deposit per tube from a photon interaction is about 32keV which is two

times more than the charge that muons leave in an MDT. This is mainly due to the fact

that muon tracks are perpendicular to the wire while electrons created by the background

photons are isotropic.

All the simulations were done using GARFIELD to calculate the induced current signal and

a stand alone program for electronics simulation.

2.1 Frontend electronics parameters

The transfer function for preamp+lter for bipolar shaping was assumed to be

g(s) =n!s

(1 + s)n+2! f(t) = (1 t=

n+ 1)(t=)net= tp = (n+1pn + 1) (1)

Since the delta response integrates to zero

Z1

0

f(x) =exxn+1

n+ 1j10= 0 (2)

every signal convoluted with this transfer function integrates to zero.

For unipolar signal shaping a preamp delta response

g(s) =n!

(1 + s)n+1! f(t) = (t=)net= tp = n (3)

together with two pole zero lters

g(s) =s+ 1=1

s+ 1=2! f(t) = (

1

1 1

2)e

t

2 + (t) (4)

was assumed. Sending a signal s(t) = exp(t=1) through a pole/zero lter with time

constants 1 and 2 results in a signal s(t) = exp(t=2). Since wire chamber signals don'thave an exponential form but are / (t+t0)

1 one usually ts a sum of several exponentials tothe signal and uses several pole/zero lters to cancel each of these exponentials [5]. Usually

two lters are sucient.

The transfer functions for both shaping schemes together with the response to a single

ionization electron can be seen in g. 2.

A photon signal together with the output of a time over threshold discriminator for both

shaping schemes is shown in g. 3. From these gures we can infer that the standard

6

0 20 40 60 80 100 120 140 160 180 200-0.6

-0.4

-0.2

-0

0.2

0.4

0.6

0.8

1

1.2

Frontend Deltaresponse

Bipolar Shaping

Unipolar Shaping

Time [ns]0 50 100 150 200 250 300 350 400 450 500

-0.6

-0.4

-0.2

-0

0.2

0.4

0.6

0.8

1

1.2

Time [ns]

Response to a singleIonization Electron

Unipolar Shaping

Bipolar Shaping

Figure 2 : Frontend delta response for unipolar and bipolar shaping and response to a singleionization electron. The dotted lines show the signal at the dierent lter stages of the unipolarscheme. The 'high' dotted line shows the preamp output, the 'lower' dotted line shows the signalafter the rst pole/zero lter, the 'upper' solid line shows the output of the whole chain. Theunipolar signal is slightly bipolar since the time constants are adjusted assuming only the ion-tailof the signal. The slight undershoot is caused by the electron part of the signal.

unipolar shaping scheme neither gives appropriate trailing edge resolution nor creates asingle threshold crossing per signal for this specic gas.

Introducing a large discriminator hysteresis helps to reduce the number of hits for the unipo-lar scheme (g. 4). We can also cancel the signal tail 'less strongly' in order to reduce the

hit number, but as shown in g. 4, the dead time increases. However, since for bipolar

shaping it takes additional time for the signal to return to the baseline from the undershoot,the dead time for both schemes turns out to be the same as shown later (g. 8).

Figure 5 shows that the signal shape also strongly depends on the position of the muon track

along the tube.Now that we have a feeling for the problem we can start to look into the statistics of the

problem i.e. we want to nd the number of hits per signal for dierent electronics schemesand operating conditions. The hit multiplicity for both schemes was calculated by creating

1000 signals at random distances for both gases Ar=N2=CH491=4=5 and Ar=CO293=7 with

GARFIELD for photons and muons.

2.2 Bipolar Shaping

The hit multiplicity versus threshold for dierent particles, gases, tube lengths and hysteresis

settings is shown in gure 6. The Ar=CO293=7 gas shows more hits due to the longerdrift time and nonlinear space-drift time relation. Photons show a higher hit multiplicity

compared to muons which is due to the increased charge deposit. Looking at the bipolar

7

0 200 400 600 800 1000 1200

Pu

lse

he

igh

t (e

lect

ron

s)

-100

-50

0

50

100

150

200

250

300

Photon Signal (Garfield)

Unipolar Shaping

Time [ns]0 200 400 600 800 1000 1200

Pu

lse

he

igh

t (e

lect

ron

s)

-100

-50

0

50

100

150

200

250

300

Photon Signal (Garfield)

Bipolar Shaping

Time [ns]

Figure 3 : Photon signal for Ar=CO293=7. The standard unipolar shaping scheme shows a hitmultiplicity very similar the bipolar scheme.

pulse in g. 3 one can infer that for a larger charge deposit more signal spikes 'will reach upthe threshold'. The fact that every spike of the signal is bipolar also explains why dierent

hysteresis settings and tube lengths have almost no eect on the hit multiplicity.To conclude we can say that for a threshold of 20e- we nd (on average) 2.6 hits for photonsfor Ar=CO293=7. Since the signal rate is dominated by photons we have to adjust the

readout for these numbers. However the hit multiplicity increases strongly when we increasethe gas gain (and keep the threshold xed), so if we want our system to work also for a

higher gas gain we even have to consider 3-4 hits on average.

2.3 Unipolar Shaping

Although it is obvious from g. 3 that for the standard unipolar shaping scheme, i.e. using

two pole/zero lters that cancel two exponentials of the signal tail according to [5], will givemore than one hit per signal, it will always be possible to arrive at one hit per signal if

we cancel the signal tail less strongly. Fig. 7 shows that introducing large discriminator

hysteresis also helps to reduce the hit number which also can be seen in g. 4.

It is therefore possible to have one hit per signal even for a unipolar shaping scheme, however

this is only possible if all the lter constants and the hysteresis are very carefully adjusted.In a standard tail cancellation scheme with two pole/zero lters [5] one adjusts the time

constants for dierent gases by changing the capacitors and therefore leaving the pole/zero

ratios constant (this is the baseline scheme as presented in the TDR). Adjusting the ltersin order to minimize the number of hits per signal requires programmable pole/zero ratios

(ratio = 1 to 2.5).

8

0 200 400 600 800 1000 1200

Pu

lse

he

igh

t (e

lect

ron

s)

-100

-50

0

50

100

150

200

250

300

Photon Signal (Garfield)

Unipolar Shaping

Hysteresis = 15e-, Time over Threshold

Time [ns]0 200 400 600 800 1000 1200

Pu

lse

he

igh

t (e

lect

ron

s)

-100

-50

0

50

100

150

200

250

300

Photon Signal (Garfield)

Different Filter Time constants

Figure 4 : Introducing Discriminator Hysteresis and using dierent tail cancellation time constantsone can reduce the hit number but one introduces dead time (the increase is however small).

0 200 400 600 800 1000 1200

Pu

lse

he

igh

t (e

lect

ron

s)

-100

-50

0

50

100

150

200

250

300

Photon Signal (Garfield)

Unipolar Shaping

Different Tubes and Positions

Time [ns]

Figure 5 : The same muon signal for dierent tubes and tube lengths. The MDT transfer functionaects the signal shape signicantly.

9

Threshold (e-)0 10 20 30 40 50 60

Ave

rag

e N

um

be

r o

f H

its

0

1

2

3

4

5

6

Shaping: Bipolar

Mode: Time Over Threshold

Hysteresis: 0

Tube Length: 2m

Photons Ar/CO2 93/7

Muons Ar/CO2 93/7

Photons Ar/N2/CH4 91/4/5

Muons Ar/N2/CH4 91/4/5

Threshold (e-)0 10 20 30 40 50 60

Ave

rag

e N

um

be

r o

f H

its

0

1

2

3

4

5

6

Shaping: Bipolar

Mode: Time Over Threshold

Hysteresis: 0

Tube Length: 5m

Photons Ar/CO2 93/7

Muons Ar/CO2 93/7

Photons Ar/N2/CH4 91/4/5

Muons Ar/N2/CH4 91/4/5

Threshold (e-)0 10 20 30 40 50 60

Ave

rag

e N

um

be

r o

f H

its

0

1

2

3

4

5

6

Shaping: Bipolar

Mode: Time Over Threshold

Hysteresis: 0.5 x Threshold

Tube Length: 2m

Photons Ar/CO2 93/7

Muons Ar/CO2 93/7

Photons Ar/N2/CH4 91/4/5

Muons Ar/N2/CH4 91/4/5

Threshold (e-)0 10 20 30 40 50 60

Ave

rag

e N

um

be

r o

f H

its

0

1

2

3

4

5

6

Shaping: Bipolar

Mode: Time Over Threshold

Hysteresis = Threshold

Tube Length: 2m

Photons Ar/CO2 93/7

Muons Ar/CO2 93/7

Photons Ar/N2/CH4 91/4/5

Muons Ar/N2/CH4 91/4/5

Figure 6 : Hit multiplicity for dierent gases, particles, tube lengths,thresholds and hysteresissettings for bipolar shaping. One can see that the hit multiplicity is not aected by hysteresissettings or tube lengths. The number of hits is higher for the Ar=CO293=7 gas due to the longerdrift time and nonlinear rt-relation. Photons cause more hits due to the increased charge deposit.

10

Threshold (e-)0 10 20 30 40 50 60

Ave

rag

e N

um

be

r o

f H

its

0

1

2

3

4

5

6

Shaping: Unipolar standard

Mode: Time Over Threshold

Hysteresis: 0

Tube Length: 5m

Photons Ar/CO2 93/7

Muons Ar/CO2 93/7

Photons Ar/N2/CH4 91/4/5

Muons Ar/N2/CH4 91/4/5

Threshold (e-)0 10 20 30 40 50 60

Ave

rag

e N

um

be

r o

f H

its

0

1

2

3

4

5

6

Shaping: Unipolar standard

Mode: Time Over Threshold

Hysteresis: 0.7 x Threshold

Tube Length: 5m

Photons Ar/CO2 93/7

Muons Ar/CO2 93/7

Photons Ar/N2/CH4 91/4/5

Muons Ar/N2/CH4 91/4/5

Threshold (e-)0 10 20 30 40 50 60

Ave

rag

e N

um

be

r o

f H

its

0

1

2

3

4

5

6

Shaping: Unipolar

Mode: Time Over Threshold

Hysteresis: 0

Tube Length: 5m

Photons Ar/CO2 93/7

Muons Ar/CO2 93/7

Photons Ar/N2/CH4 91/4/5

Muons Ar/N2/CH4 91/4/5

Threshold (e-)0 10 20 30 40 50 60

Ave

rag

e N

um

be

r o

f H

its

0

1

2

3

4

5

6

Shaping: Unipolar

Mode: Time Over Threshold

Hysteresis: 0.7 x Threshold

Tube Length: 5m

Photons Ar/CO2 93/7

Muons Ar/CO2 93/7

Photons Ar/N2/CH4 91/4/5

Muons Ar/N2/CH4 91/4/5

Figure 7 : Hit multiplicity for unipolar shaping. The rst gure shows the standard unipolarshaping scheme with a time over threshold discriminator and no hysteresis. The second gureshows that we can reduce the hit number by introducing hysteresis but we still get more that twohits for the Ar=CO293=7 gas. The third and fourth gure show numbers for 'soft' tail cancellation.We can see that soft tail cancellation and large hysteresis helps to arrive at a single hit per signal.

11

3 Bipolar versus Unipolar Shaping

It is instructive to compare our situation to the ATLAS Transition Radiation Tracker (TRT)

which foresees a unipolar shaping scheme with active baseline restoration. Since the drift

distances in the TRT are very short and the drift velocity is high (pressure 1bar) the signals

are very short. Using a bipolar scheme would double the dead time since the signal under-

shoot has the same duration as the signal itself. Since in the MDT case the duration of the

undershoot is short compared to the whole signal length and since we have to use 'soft' tail

cancellation in a unipolar scheme to avoid multiple hits the dead time is essentially equal

(g. 8).

ns0 100 200 300 400 500 600 700 800 900 1000

0

5

10

15

20

25

30

35

40

Figure 8 : Deadtime spectrum for Ar=CO293=7 for unipolar and bipolar shaping. The dierenceis insignicant.

At this point it would seem natural that unipolar shaping would be the preferred solution,

however we have to keep in mind that this requires active baseline restoration, large discrim-inator hysteresis and programmable time constants for dierent gases. Even if we have all

these options they must be very carefully adjusted and we must work exactly at the dened

operating point.This is a signicant complication compared to a bipolar scheme where we use only one lter

with a xed time constant, no baseline restorer and no large hysteresis and where we don'thave to adjust anything.

The biggest concern for the bipolar scheme are multiple hits. An average of up to 3 hits persignal increases the data volume signicantly. Simulations of the TDC behavior showed that

for a rate of 400kHz with 1.7 hits per signal on average there is no loss of leading edge hits.If we have 3 hits on average, the Level1 Buer on the TDC has to be extended. Whether

the front end link can handle this rate has to be studied.

12

However even if we are able to read out all the hits there is a dierence between the unipolar

and bipolar scheme:

In case of a level one trigger all hits within the maximum drift time window (plus some

propagation time) are read out. In case of unipolar shaping this window would contain

background hits and muon hits, in case of bipolar shaping we would have after pulses in

addition. Since we don't know whether a hit corresponds to a real leading edge or just an

after pulse, the pattern recognition program has to treat all the hits with equal priority.

However it was shown that this can even reduce the eciency, so we have to nd some other

algorithm.

At a rate of 300kHz, the average time between two hits is 3s, the average time betweentwo after pulses is much smaller (100 200ns). From this we can conclude that the rst hit

in the drift time window is most probably a real leading edge. So the pattern recognitionprogram would use only the rst hit in the time window. However, this way we eliminate

also a small number of real hits and we reduce the single tube eciency. The impact of thisis shown in the next chapter.

1us

PhotonMuon

Figure 9 : Using only the rst hit in the drift time window introduces articial deadtime. Howeverusing all the hits in the time window can create problems in case of bipolar shaping due to multiplehits.

In case of unipolar shaping we know that every hit corresponds to one particle, but in thebipolar scheme we don't know any more whether the hits correspond to the leading edge ofa particle or whether it is just an after pulse. The point is that we lose information with the

bipolar scheme. However, as shown in the next section, this eect is small.

13

4 How to decide on the shaping scheme

It is of course desired to have a scheme that guarantees the same pattern recognition per-

formance that is advertised in the TDR and also doesn't introduce major changes to the

TDC or the readout scheme. We will estimate the impact of the dierent options on the

pattern recognition eciency by looking at the single tube ineciency for several schemes.

The results are shown in Fig. 12 and discussed below. It is however important to perform

dedicated pattern recognition studies for the dierent schemes.

The single tube ineciencies were calculated by randomly overlaying photon signals to the

muon signal (according the given rate) and looking if the muon leading edge was lost (single

events are shown in g. 11). To get a feeling for the dierence between Ar=CO293=7 and

Ar=N2=CH491=4=5 g. 10 shows the rt-relations and the deadtime spectra for both gases.

Although the average drift time is almost equal for both gases (248ns for Ar=N2=CH491=4=5and 255ns for Ar=CO293=7), the 'photon induced deadtime spectrum' diers by almost 200ns(330ns for Ar=N2=CH491=4=5 and 520ns for Ar=CO293=7). The reason for that is the fact

that photon interactions create high energy scatter electrons that either originate from thetube walls or end in the tube walls. Therefore the deadtime is dominated by the drift velocity

close to the wall which is very low for the Ar=CO2 mixture.

Rt-Relations

0

100

200

300

400

500

600

700

800

0 0.2 0.4 0.6 0.8 1 1.2 1.4

Ar/N 2/CH4 91/4/5

Ar/CO 2 93/7

ns0 100 200 300 400 500 600 700 800 900 1000

0

100

200

300

400

500

600

Figure 10 : Rt-relation and dead time spectrum for both gases. The average deadtime is 330nsfor Ar=N2=CH491=4=5 and 520ns for Ar=CO293=7.

In the following we will present three dierent options.

4.1 Unipolar shaping

We use a unipolar shaping scheme, widely programmable lter time constants, active base-

line restoration, large discriminator hysteresis and carefully adjust all these parameters in

14

order to get on hit per signal.

The single tube ineciency for a rate of 300kHz (Safety Factor 5) is 21% for the Ar=CO2

gas and 16% for the Ar=N2=CH491=4=5 gas (g. 12).

4.2 Bipolar shaping, hit reduction in the TDC

We use a bipolar shaping scheme which makes the front end electronics very simple. There

are no parameters that have to be adjusted even if we change the gas. We have up to 3 hits

(on average) per single signal. The TDC LVL1 buer must be extended. The capacity of

the readout scheme must be increased by a factor 3.

Assuming that the pattern recognition program can only use the rst hit in the time window

the single tube ineciency for a rate of 300kHz (Safety Factor 5) is 23% for the Ar=CO2 gasand 20% for the Ar=N2=CH491=4=5 gas (g. 12).

If the TDC would have the option to only give the rst hit in the time window the readoutcapacity would not have to be increased.

4.3 Bipolar shaping, articial deadtime

We use a bipolar scheme and introduce articial deadtime in the discriminator to reducethe number of hits. The ADC information will be appended to the deadtime pulse. It isimportant to notice that the deadtime has to be equal or very close to the maximum drift

time. If e.g. for Ar=CO293=7 (700ns maximum drift time) we use a xed deadtime of 500nsit can happen that a late cluster, arriving at 600ns, triggers another 500ns pulse which would

in that case increase the deadtime to 1:1s.

The single tube ineciency for a rate of 300kHz (Safety Factor 5) is 25% for the Ar=CO2

gas and 20% for the Ar=N2=CH491=4=5 gas (g. 12).

15

Time (ns)-1500 -1000 -500 0 500 1000 1500

Sig

na

l (e

-)

-100

-50

0

50

100

150

200

250

MuonMuon+Background

Time (ns)-1500 -1000 -500 0 500 1000 1500

Sig

na

l (e

-)

-100

-50

0

50

100

150

200

250

MuonMuon+Background

Time (ns)-1500 -1000 -500 0 500 1000 1500

Sig

na

l (e

-)

-100

-50

0

50

100

150

200

250

MuonMuon+Background

Time (ns)-1500 -1000 -500 0 500 1000 1500

Sig

na

l (e

-)

-100

-50

0

50

100

150

200

250

MuonMuon+Background

Time (ns)-1500 -1000 -500 0 500 1000 1500

Sig

na

l (e

-)

-100

-50

0

50

100

150

200

250

MuonMuon+Background

Time (ns)-1500 -1000 -500 0 500 1000 1500

Sig

na

l (e

-)

-100

-50

0

50

100

150

200

250

MuonMuon+Background

Figure 11 : Single events for a background rate of 500kHz. The solid line indicates the muonsignal, the dashed lines shows the photon background. The left gures show the unipolar scheme,the right gures the same event for the bipolar scheme. In case of a trigger, the TDC would sendout all the hits in the time window from -100ns to 800ns.

16

Ar/CO 2 93/7

0

5

10

15

20

25

30

35

40

0 100 200 300 400 500

Unipolar shaping

Bipolar shaping, all Hits

Bipolar shaping, first Hit

Bipolar shaping, 700ns fixed deadtime

Ar/N 2/CH4 91/4/5

0

5

10

15

20

25

30

35

40

0 100 200 300 400 500

Unipolar shaping

Bipolar shaping, all Hits

Bipolar shaping, first Hit

Bipolar shaping, 500ns fixed deadtime

Figure 12 : Single tube ineciencies forAr=CO293=7 andAr=N2=CH491=4=5. The 5% ineciencyat 0Hz is the ineciency due to delta electrons created by the muon in the tube wall [4]. The 'BipolarShaping, all Hits' curve has to be interpreted carefully: the ineciency is lower since some of thepiled up hits that are lost in the unipolar scheme are recuperated (see 'double track separationwith strong tail cancellation' [4]). However, the multiple hits and the fact that the resolution ofthe recuperated hits is lower leaves the pattern recognition eciency almost unchanged.

17

5 Summary

Because of high background rates we have to use either unipolar shaping with active

baseline restoration or bipolar shaping in order to avoid baseline uctuations.

The usual argument in favor of unipolar shaping - deadtime - does not apply in our

situation.

Because of our operating parameters - 3 bars and low gas gain - we have the peculiar

problem of multiple hits.

At the time of the TDR we decided on unipolar shaping because we get

one hit per signal

double track separation (eciency of 60% -80% for t > 100ns) [4]

trailing edge resolution of 20ns.

Double track separation was never found to be useful [3], the trailing edge was neverused and in addition the trailing edge resolution for Ar=CO2 is only 80 ns comparedto 25 ns for Ar=N2=CH491=4=5. In addition the drift properties of Ar=CO2 are such

that it is very hard to achieve one hit per signal.

The background photons create high energy scattering electrons that 'look like tracks'.

They deposit on average twice the energy of muons (muon tracks are perpendicular tothe wire).

The 'average' deadtime is 330ns for Ar=N2=CH491=4=5 and 520 ns for Ar=CO293=7.

Bipolar shaping is very 'simple' (no programmable time constants, no baseline restora-tion circuit, no ne tuning). We get up to 3 hits on average per signal.

The current TDC and readout can handle only two hits on average at a rate of 400kHz.

Using bipolar shaping we lose information compared to unipolar shaping since wedon't know whether a hit corresponds to a real leading edge or an after pulse (tube

ineciency goes from 21% to 23% for Ar=CO293=7). Essentially we can only use the

rst hit in the time window.

If we chose unipolar shaping we have to use 'non standard' tail cancellation with

programmable pole/zero ratios (not the baseline).

18

There are 4 possible frontend schemes:

Unipolar shaping ('complex'):

We need large discriminator hysteresis, programmable pole/zero ratios and an

active baseline restoration circuit. We get one hit per signal and a single tube

ineciency of 21% at 300kHz.

Bipolar Shaping ('simple'):

We get up to 3 hits per signal (on average) and a single tube ineciency of 23% at

300kHz. The TDC Level 1 buer and the readout capacity have to be extended.

Bipolar Shaping, rst hit TDC:

The TDC buers have to be extended, the additional function of a rst hit readout

has to be implemented. The single tube ineciency is 23% at 300kHz.

Bipolar Shaping, xed deadtime ('simple'):

We have to introduce programmable deadtime (0-800 ns). We get one hit persignal, the single tube ineciency is 25% at 300kHz.

In order to decide on the shaping scheme we still have to study pattern recognitioneciency, TDC occupancy and readout occupancy for the dierent schemes.

19

References

[1] M. Deile, J. Dubbert, N.P. Hessey et al., Testbeam Studies of the gas mixtures Ar :

N2 : CH4=91:4:5 ... ATLAS internal note MUON-NO-122 (1996), CERN.

[2] ATLAS Muon Spectrometer Technical Design Report, CERN/LHCC/97-22, ATLAS

TDR 10, (1997)

[3] Marc Virchaux, private communication

[4] W. Riegler: MDT Eciency, Double Track Separation, ATLAS Muon Note 173, CERN

1997

[5] R.A. Boie, A.T Hrisoho and P. Rehak, signal shaping and tail cancellation for gasproportional detectors at high counting rates, NIM 192 (1982).

[6] On the Number of Layers per Multilayer in MDT chambers Part I. Laporte J.F, Chava-

lier L, Guyot C , Virchaux M ATLAS Muon Note 126, CERN 1996

[7] Pattern Recognition with MDT Chambers, Shank J, Sliwa K, Taylor F, Zhou B, ATLAS

Muon Note 150, CERN 1997

20