Principles of Fluorescence Spectroscopy

PRINCIPLES OF FLUORESCENCE SPECTROSCOPY 115
Figure 4.22. Constant fraction discrimination in TCSPC. Top: Timing
error due to pulse height variations using leading edge discrimination.
Bottom: Operation of a constant fraction discriminator.
4.5.2. Amplifiers
Amplifiers can be used after the start and stop detectors in
order to obtain adequate signal levels for timing. The present
trend is to avoid such amplifiers, 65 which can result in
additional difficulties. In general, the most noise-free
amplification can be obtained within the detector (PMT or
photodiode). The cable connecting the detector and amplifier
can act as an antenna, resulting in amplification of the
RF noise. The use of amplifiers was necessary when MCP
PMTs first appeared because the pulses were too short for
the CFDs available at that time. Amplifiers were used to
broaden the pulses sent to the CFD. 40 This is no longer necessary
with newer CFDs. If amplifiers must be used, they
should be positioned as close as possible to the detector.
4.5.3. Time-to-Amplitude Converter (TAC) and
Analyte-to-Digital Converter (ADC)
The role of the TAC in Figure 4.8 is to generate a voltage
proportional to the time between the excitation pulse and
the first arriving emitted photon, which are the start and
stop pulses, respectively. This is accomplished by charging
a capacitor during the time interval between the pulses.
Typically the capacitor is charged from 0 to 10 volts over a
nanosecond to microsecond time range. For instance, if the
chosen range is 50 ns, the capacitor is fully charged at 50
ns. If a stop pulse is received at 25 ns, the charging is
stopped at 5 volts. The voltages are calibrated to time delays
using delay lines or optical pulses with known time separation.
If a stop pulse is not received, the TAC is reset to zero.
After the start and stop pulses are received the voltage is
connected to a digital value by the ADC. This method of
measuring time delays is indirect but provides higher time
resolution. At present, inert measurements of time delays
cannot be performed accurately enough for nanosecond
timescale delays.
In general, the TAC is a rate-limiting component in
TCSPC. A certain amount of time is needed to discharge the
capacitor and reset the TAC. Prior to about the year 2000,
the reset time for most TACs was about 2 ms. This was not
a problem with flashlamps where a 50-kHz rate results in
start pulses every 20 µs. However, with a high-repetitionrate
laser source at 1 MHz, the TAC will be overloaded due
to continuous start pulses. The TAC will be instructed to
reset before it has completed the previous reset.
The solution to this problem is relatively simple, which
is to operate the TAC in reverse mode. 68–70 In this mode of
operation the first photon detected from the sample serves
as the start pulse, and the signal from the excitation pulse is
the stop signal. In this way the TAC is only activated if the
emitted photon is detected. The decay curves can appear
reversed on the screen of the multichannel analyzer (MCA),
but this is corrected by software. The reverse mode of TAC
operation is not needed with flashlamps because of their
lower repetition rates.
The reset time of a TAC is also called the dead time. A
TAC with a 2-:s dead time has a saturated count rate of 0.5
MHz. Photons arriving within the dead time cannot be
counted, so the counting efficiency drops. The photon count
rate for 50% counting efficiency is sometimes called the
maximum useful count rate, which for a 2-:s dead time is
250 kHz. Modern TCSPC electronics have dead times near
125 ns and a saturated count rate of 8 MHz. These electronics
can efficiently process and count photons at MHz rates.
An important characteristic of a TAC is its linearity. If
the voltage is not linear with time, then the data will contain
systematic errors, resulting in difficulties with data analysis.
One way to test the linearity of a TAC is by exposure of
the detector to a low level of room light, and still use the
pulsed-light source to trigger the TAC start signal. Since the
photons from the room lights are not correlated with the
start pulses, the stop pulses should be randomly distributed
across the time range, which is a horizontal line in the mul-