Becker W. Advanced Time-Correlated Single Photon Counting Techniques

Подождите немного. Документ загружается.

7.6 Safety Considerations 315

Signal Cable

Housing

Detector

Copper Foil

wrapped around the cable

and soldered to the housing

Fig. 7.55 „Emergency repair“ of a poorly designed RF shield. A copper foil wrapped

around the cable forms a capacitor with the cable shield and diverts a part of the noise

current

7.6 Safety Considerations

PMTs are operated at a voltage in the range of 800 to 3,500 V, and the operating

voltage of avalanche photodiodes can be as high as 300 V. Therefore it is neces-

sary to obey the usual safety rules for handling high voltage. An extremely impor-

tant but often neglected issue is the ground return path to the power supply. A

typical situation is illustrated in Fig. 7.56.

HV Connector

broken

PMT Housing

PMT

HV cable

Signal Cable

GND (ground conductor)

HV Power Supply

HV OUT

Danger!

TCSPC

Module

Fig. 7.56 Effect of a broken ground return path to the high voltage power supply

The cable that connects the voltage divider of the PMT to the high voltage

power supply has a broken outer conductor. The inner conductor still connects the

„hot“ side of the voltage divider to the high voltage. However, because the outer

connector is broken, the current through the voltage divider resistors cannot flow

back to the power supply. Therefore the current flows through the TCSPC module,

the computer, and the ground conductor of the power socket back to the power

supply. Nothing happens, but the setup is not safe: When the signal cable is dis-

connected, there is a high voltage between the signal cable and the TCSPC mod-

ule. Fortunately the resistors in the PMT voltage divider limit the current so that a

broken HV cable will probably not deliver a fatal shock; but nevertheless the ef-

fect can be surprising.

316 7 Practice of TCSPC Experiments

The safe way to use a HV power supply is to provide an auxiliary ground return

path via a separate conductor from the PMT ground to the power supply ground;

see Fig. 7.57. Nevertheless, broken cables

must be repaired or replaced.

HV Connector

PMT Housing

PMT

HV cable

Signal Cable

GND (ground conductor)

HV Power Supply

HV OUT

TCSPC

Module

Auxiliary Ground

Return Path

Fig. 7.57 An auxiliary ground return path makes the system safe even if the HV cable

should break

The high voltage used in the detectors may also cause possible damage to am-

plifiers or CFD inputs. Figure 7.58 shows what can happen if a signal cable from a

PMT has no reliable connection to the load.

100 V

PMT

-HV

Load

open

connected

-100 V

-50 V

Signal Cable

Fig. 7.58 Connecting an open PMT output cable to the load

If the load is unconnected, the output current of the PMT charges the signal ca-

ble until the voltage reaches approximately the voltage at the last dynode. This

voltage can be as high as several hundred Volts. When the load is connected, the

cable discharges into the load. For a load impedance equal to the characteristic

impedance of the cable, the amplitude of the resulting pulse is half the voltage to

which the cable was charged. The pulse duration is twice the cable transit time.

The pulse can be enough to destroy an amplifier or a CFD.

For this reason, do not connect or disconnect a photomultiplier to or from the

load when the high voltage is switched on. Do not use switchable attenuators be-

hind the PMT output. Do not use cables and connectors with bad contacts. The

same rules should also be followed for photodiodes that are operated at supply

voltages above 20 V.

The problem can easily be avoided by connecting a resistor of about 100 k

from the PMT anode to ground. However, in practice the PMT module often can-

not be opened. The only way to avoid damage is to be careful.

7.7 Setting the TCSPC System Parameters 317

7.7 Setting the TCSPC System Parameters

7.7.1 Optimisation of the CFD in the Detector Channel

Pulse Shaping

The general function of the constant fraction discriminators (CFDs) of a TCSPC

system is described in Chap. 4, page 47. The CFD contains a pulse shaping net-

work and two discriminators (Fig. 7.59). The shaping network changes the unipo-

lar input pulse shape into a bipolar one. One discriminator picks up the zero cross-

ing of the shaped pulse, and the other one enables the output circuitry of the CFD

when the input pulse amplitude exceeds a reference voltage. The reference volt-

ages of both discriminators are adjustable. Although the structure of the CFDs of

individual TCSPC modules may differ in detail, the general effect of the adjust-

ments is the same as shown in Fig. 7.59.

Zero cross

DEL1

Input difference

voltage of C1:

DEL2

Threshold

Input

Zero Cross

Enable

Input pulse

Difference

Threshold

Zero

cross level

Input pulse

Delayed

Difference

Threshold

Small

Large

pulse

DEL 2:

Zero

cross level

Delayed

pulse

Fig. 7.59 Effect of the CFD threshold, CFD zero cross level, and delay used in the pulse

shaping network

To optimise the CFD for different width of the detector pulses, the pulse-

shaping network can usually be changed. In general the transit time of the delay

line DEL 2 should be slightly longer than the leading edge of the input pulse.

Typical pulse rise times are given in the table below.

R3809U MCPs 100 ps

R5600, R7400 miniature PMTs 700 ps

H5783, H5773 photosensor modules 700 ps

H7442 modules 700 ps

R928, R931 side-window PMTs 1.5 ns

R5900 multianode PMTs 1 ns

XP2020 linear-focused PMTs 2.5 ns

318 7 Practice of TCSPC Experiments

In practice the effect of changes in the pulse shaping network is often hard to

predict. A longer DEL 2 may result in an unfavourable zero transition shape but a

higher bipolar pulse amplitude (Fig. 7.59, right). However, the bandwidth of the

discriminator input stage is usually unknown, and how the discriminator actually

„sees“ the pulse shape is not predictable. The only way to find it out is to try

different delays.

Detector modules with internal discriminators, such as the Hamamatsu H7421

PMT modules or the Perkin Elmer SPCM-AQR single photon APD modules,

deliver stable output pulses without amplitude jitter. The timing performance is

defined by the internal discriminator of the detector module, not by the CFD of the

TCSPC device. Thus changing the CFD configuration does not improve the time

resolution of these detectors.

Discriminator Threshold

The CFD threshold determines the minimum amplitude of the input pulses that

trigger the CFD. The threshold of the CFD in the detector channel has a consider-

able influence on the efficiency of a TCSPC system. As described under Sect. 6.2,

page 222, the single-photon pulses of a PMT have a strong amplitude jitter. The

general shape of the amplitude distribution is shown in Fig. 7.60, left.

Pulse amplitude CFD Threshold

Count

Rate

Pulses

per

second

Amplitude distribution

regular photon

electronic noise

pulses from dynodes

electronic noise

pulses from dynodes

regular photon pulses

Optimum CFD

threshold

Optimum CFD

threshold

Rate vs. threshold

pulses

Fig. 7.60 Amplitude distribution of the PMT pulses (left) and count rate vs. CFD threshold

The pulse amplitude distribution consists of three major components. There are

the regular photon pulses, i.e. the pulses originating from electron emission at the

photocathode, which form a wide peak at relatively high amplitudes. Thermal

emission, photoelectron emission, and reflection of primary electrons at the dyn-

odes forms a secondary peak at lower amplitudes. At very low amplitudes elec-

tronic noise, either from the preamplifier or from the environment, causes a third

peak of extremely high count rate.

The TCSPC system should record the regular photon pulses, but not the dynode

pulses and the electronic noise background. Therefore, the optimum CFD thresh-

old is in the valley between the regular photon pulse distribution and the peak

caused by dynode emission.

7.7 Setting the TCSPC System Parameters 319

Unfortunately, in a normal TCSPC system the photon distribution can only be

measured indirectly by observing the count rate for different CFD thresholds.

Some TCSPC devices have an option to run a CFD threshold scan; in other de-

vices the threshold must be changed manually. The function of the count rate

versus the threshold is the integral of the pulse amplitude distribution, see

Fig. 7.60, right. The correct CFD threshold is in the plateau formed by the regular

photon pulses. Of course, the whole pulse amplitude distribution is stretched hori-

zontally as a function of increasing PMT gain or preamplifier gain. Therefore the

optimum CFD threshold changes depending on the operating voltage of the PMT

and the preamplifier gain.

In practice there may be a considerable overlap of the regular and the dynode

pulse spectrum. The dynode spectrum can be less pronounced or not visible at all

in a CFD threshold scan. Insufficient gain or poor detector shielding may even

lead to an electronic noise background extending far into the regular pulse spec-

trum. Both the dynode pulses and the electronic noise impair the timing accu-

racy. Dynode pulses can cause a prepeak in the IRF; the electronic noise broad-

ens the IRF. In these cases, which are not uncommon in practice, the „plateau“

of the counting characteristic may be poorly defined, and a compromise between

counting efficiency and IRF shape must be found. Therefore, the IRF should be

checked while the threshold is being changed. If prepulses appear at low thresh-

olds or if the IRF broadens substantially, it may be impractical to reduce the

threshold further.

As a practical example, Fig. 7.61 shows a CFD threshold scan for an

XP2020RU PMT. The count rate versus discriminator threshold is shown left, the

IRF right. Both figures are in linear scale.

Higher pulse amplitudes in general give lower timing jitter because the influ-

ence of the background noise is smaller and the influence of the amplitude jitter on

the timing is reduced. Therefore TCSPC users often increase the CFD threshold

Fig. 7.61 CFD threshold scans for an XP2020RU PMT at –2.5 kV. Left: Count rate versus

threshold from –400 mV to –12 mV. Right: IRF versus threshold from –270 mV to –23 mV

320 7 Practice of TCSPC Experiments

substantially above the level for optimum efficiency. The method may be accept-

able if one does not reject more than 80% of the pulses and takes into account the

corresponding loss in efficiency. However, it should be noted that the efficiency,

and to a smaller degree also the IRF, will be less stable than within the plateau of

the counting characteristics. Moreover, since a large part of the detector current is

caused by rejected photons, the maximum safe count rate will be considerably

reduced, in particular for MCPs, which have a very limited output current.

Excessively high CFD thresholds in combination with low detector gain can

almost entirely suppress the detection of single-photon pulses. If the light intensity

is increased, eventually multiphoton events are detected. Particularly at low pulse

repetition rates, this may remain unnoticed because high peak intensities can be

applied without getting an exceedingly high average detector current. A typical

example of multiphoton detection is shown in Fig. 7.62.

Fig. 7.62 Detection of multiphoton events at low detector gain. Left: Correctly recorded

signal, recorded at a detector gain of 10

and a count rate of 37

-1

. Centre and right:

The same signal recorded at a detector gain of 10

and 10

. The light intensity was in-

creased until the PMT signal triggered the CFD.

The detector was an H577320, connected to the CFD input via a 20 dB pre-

amplifier. The light signal was a pulse from an LED, with a repetition rate of

1 MHz. The left curve was recorded at a detector gain of 10

and shows the true

shape of the light pulse. The curves in the middle and right were recorded at a

detector gain of 10

and 10

, respectively. At a gain this low a single photon does

not trigger the CFD. However, if the light intensity is increased, the PMT signal

eventually turns into a continuous signal representing the shape of the light pulse.

If the amplitude of the signal reaches the CFD threshold, the CFD triggers. The

count rate steeply increases from zero to the signal repetition rate, and extremely

narrow pulse shapes are recorded. Operating the PMT in this multiphoton mode

results in an extreme distortion of the recorded signal waveform. An extremely

short IRF of less than 20 ps duration can be obtained. This IRF is, of course, en-

tirely useless for any waveform recording. It can, however, be exploited if TCSPC

is used for distance measurement.

7.7 Setting the TCSPC System Parameters 321

Multiphoton detection is a frequent source of errors in attempts to use standard

avalanche photodiodes as single photon detectors. If the diode is not really oper-

ated in the breakdown region, a detectable output pulse is obtained only if several

photons are detected within the impulse response time of the diode, with a similar

results as shown in Fig. 7.62.

Zero Cross Level

The zero cross level adjustment minimises the timing jitter induced by amplitude

jitter of the detector pulses. The zero cross level is therefore often called „walk

adjust“. In early TCSPC systems the walk adjust had an enormous influence on

the shape of the instrument response function (IRF). In newer, more advanced

systems the influence is smaller. The reason is probably that detectors with shorter

single electron response are used and the discriminators in the newer CFDs are

faster. Therefore, the effective slope of the zero cross transition is steeper, with a

correspondingly smaller influence of the zero cross level. Figure 7.63 shows the

IRF for an XP2020UR linear-focused PMT and an H5773

20 photosensor module

for different zero cross levels.

Fig. 7.63 IRF of an XP2020 (left) and an H577320 (right, with 20 dB preamplifier) for

different zero cross settings. Supply voltage of XP2020 –2.5 kV, gain control voltage of

H577320 0.9 V

Theoretically the best zero-cross level should be expected exactly at zero. How-

ever, in practice the zero-cross discriminator has an offset voltage of a few mV.

Moreover, the intrinsic delay of the discriminator depends on the amplitude and the

slope of the input signal. The corresponding amplitude-induced timing jitter of the

discriminator can be compensated for by slightly offsetting the zero-cross level

from the signal baseline. Therefore the best zero cross value can be some tens of

mV above or below zero. A zero-cross level extremely close to the signal baseline

can cause problems. In that case, the zero-cross discriminator triggers due to spuri-

ous signals from the synchronisation channel or to noise from the environment; it

may even oscillate. Of course, spurious triggering and oscillation stop when an

input pulse arrives, but some after-ringing may still be present and modulate the

trigger delay.

322 7 Practice of TCSPC Experiments

Extremely large zero cross settings in combination with small threshold settings

should be avoided. In those cases it may happen that the threshold discriminator

triggers, but the zero cross discriminator does not. Consequently, there are no

output pulses from the CFD, although the threshold discriminator indicates a trig-

ger rate.

7.7.2 Reference CFD

In most TCSPC applications the pulses in the reference channel have less ampli-

tude jitter than the PMT pulses. Therefore the delay lines and the zero cross and

threshold settings of the reference CFD have little influence on the time resolu-

tion.

Problems can arise from reflections within the reference signal cable. The input

impedance of a CFD is usually far from being an ideal termination of the signal

cable. Therefore some reflection of the input pulses must be expected. The detec-

tor that drives the cable is normally a current source. The reflected pulse runs

backward and forward along the cable and can trigger the CFD again after twice

the cable transit time. For the PMT channel this happens within the dead time of

the TCSPC module and does not cause any problems. However, false triggering in

the reference channel causes ambiguous time measurement in the TAC. Typical

effects of false triggering are shown in Fig. 7.64. Similar effects can arise from

afterpulses in lasers with pulse pickers. Usually the problem can be solved by

selecting an appropriate CFD threshold in the reference channel.

Correct Signal

Double Signal caused

by false SYNC triggering

Correct Signal

Signal shape

false triggering

caused by

Fig. 7.64 Effect of false reference triggering on the recorded signal shape. Pulsed signal

(left) and continuous signal (right)

The CFD in the reference channel often has a selectable frequency divider (see

section Sect. 4.1, page 47). The frequency divider ratio determines the number of

signal periods recorded; see Fig. 7.65.

7.7 Setting the TCSPC System Parameters 323

Fig. 7.65 High-repetition-rate signal recorded with different reference frequency-divider

settings

Values greater than one are convenient for finding a short signal in a longer

signal period. Furthermore, the frequency divider can be used to check or calibrate

the time scale by comparing the displayed pulse distance with the known pulse

repetition rate (see Sect. 7.10, page 345).

For recording final results the reference-frequency divider should not be used.

Fluorescence data analysis programs expect data for only one signal period. Re-

cording several periods and analysing only one means wasting at least 50% of the

recorded photons. Moreover, unnecessary recording of several signal periods

wastes memory space. This can be an issue for TCSPC imaging or sequential

recording. It should also be noted that reference-frequency division reduces the

effective stop rate and thus increases the pile-up effect. Moreover, the time from

the photon detection to the next stop pulse adds to the effective dead time (see

Sect. 7.9.2, page 338).

7.7.3 Adjusting the Delay in the Detector and Reference

Channel

The transit times in the optical path, the detector and the signal lines of a TCSPC

system are often not exactly known. Therefore the photon pulses are detected with

an unknown temporal shift from the reference pulses. If the signal is recorded

without a reference frequency divider, it usually happens that the wrong part of the

signal period is recorded. For a signal period of the order of the recorded time

interval, the result may appear shifted in phase (Fig. 7.66, top). The phase of the

signal can be corrected by changing the optical or electrical path length in the

reference path (Fig. 7.66, bottom left) or in the detection path (Fig. 7.66, bottom

right). One meter of 50-

: cable adds approximately 5 ns delay. Please note that

the signal may wrap around the signal period, as shown in Fig. 7.66, bottom right.

324 7 Practice of TCSPC Experiments

Fig. 7.66 Recordings of a high-repetition-rate signal with different transit times in the

signal lines. Top: Original recording. Bottom left: Correction by adding delay in the refer-

ence path. Bottom right: Correction by adding delay in the detection path

Reversed start-stop operation of TCSPC requires a reference pulse at the end of

the signal period or at the end of the recorded time interval. For high-repetition-

rate lasers it is not always clear which laser pulse actually stops the time meas-

urement. It can happen that the stop pulse is not the same laser pulse that excited

the detected photon but a pulse from a period before or after. Stopping with a

pulse from a different period is no problem if the laser pulses have a constant

period and no pulse-to-pulse jitter. This is certainly the case for a Ti:Sapphire

laser. Diode lasers, however, may have selectable pulse periods. If the reference

pulses come from the wrong signal period the position of the recorded signal in

the TAC range changes when the laser period is changed. Moreover, the clock

oscillator of a diode laser may have a pulse-to-pulse jitter of some 10 ps. This

jitter adds to the transit time spread of the TCSPC system if the TAC is not

stopped with the correct pulse.

To stop the TAC with the correct laser pulse, the reference signal must be de-

layed so that the reference pulse arrives

after a photon pulse from the same period.

The correct delay in the reference channel is the detector transit time, plus the

width of the recorded time interval, plus a few ns for the TAC start delay. The

relation of the detector and reference delay is shown in Fig. 7.67.