The TCSPC system response for a Hamamatsu R3809U-50 MCP [27] is shown in fig. 20. The MCP was illuminated with a femtosecond Ti:Sa laser, the response was measured with an SPC-630 TCSPC module. A HFAC-26-01 preamplifier was used in front of the SPC-630 CFD input. At an operating voltage of -3 kV the FWHM (full width at half maximum) of the response is 28 ps.

The response has a shoulder of some 400 ps duration and about 1% of the peak amplitude. This shoulder seems to be a general property of all MCPs and appears in all of these devices. The width of the response can be reduced to 25 ps by increasing the operating voltage to the maximum permitted value of -3.4 kV. However, for most applications this is not recommended for the following reason:

As all MCP-PMTs, the R3809U allows only a very small maximum output current. This sets a limit to the maximum count rate that can be obtained from the device. The maximum count rate depends on the MCP gain, i.e. of the supply voltage. The count rate for the maximum output current of 100 nA as a function of the supply voltage is shown in fig. 21. To keep the counting efficiency constant the CFD threshold was adjusted to get a constant count rate at a reference intensity that gave 20,000 counts per second. The R3809U tubes have a relatively good SER pulse height distribution which seems to be independent of the cathode type - possibly a result of the independent manufacturing of the channel plate and the cathode. Therefore a good counting efficiency can be achieved.

Due to the short TCSPC response and the absence of afterpulses the R3809U is an ideal detector for TCSPC fluorescence lifetime measurements, for TCSPC lifetime imaging, and for combined lifetime / FCS or other correlation experiments. Recently Hamamatsu announced the R3809U MCP with GaAs, GaAsP, and infrared cathodes for up to 1700 nm. Although these MCPs are not as fast as the versions with conventional cathodes they might be the ultimate detectors for combined FCS / lifetime experiments.

The flipside is that MCPs are expensive and can easily be damaged by overload. Therefore the R3809U should be operated with a preamplifier that monitors the output current. If overload conditions are to be expected, i.e. by the halogen or mercury lamp of a scanning microscope, electronically driven shutters should be used and high voltage shutdown should be accomplished to protect the detector.

The H7422 incorporates a GaAs or GaAsP cathode PMT, a thermoelectric cooler, and the high voltage power supply [28]. Hamamatsu delivers a small OEM power supply to drive the cooler. However, we could not use this power supply because it generated so much noise that photon counting with the H7422 was not possible. Furthermore, we found that the H7422 shuts down if the gain control voltage is changed faster that about 0.1V / s. Apparently fast changes activate an internal overload shutdown. Unfortunately the device can only be re-animated by cycling the +12 V power supply.

Therefore we use the Becker & Hickl DCC-100 detector controller. It drives the cooler and supplies the +12 V and a software-controlled gain control voltage to the H7422. Furthermore, the DCC in conjunction with a HFAC-26-1 preamplifier can be connected to shut down the gain of the H7422 on overload. If the H7422 shuts down internally for any reason, cycling the +12 V is only a mouse click into the DCC-100 operating panel.

The FWHM of the system response is about 300 ps. There is a weak secondary peak about 2.5 ns after the main peak, and a peak prior to the main peak can appear at low discriminator thresholds. The width of the response does not depend appreciably of the discriminator threshold. This is an indication that the response is limited by the intrinsic speed of the semiconductor photocathode.

The afterpulsing probability of the H7422-40 can be seen from the histogram of the time intervals of the photon (fig. 24). For maximum gain the afterpulse probability in the first 1.5 µs is very high (fig. 24, red curve, control voltage 0.9V). If the gain is reduced the afterpulse probability decreases considerably (fig. 24, blue curve, 0.63V). The timing resolution does not decrease appreciably at the reduced gain, fig. 25.

The H7422 is a good detector for TCSPC applications when sensitivity has a higher priority than time resolution. A typical application is TCSPC imaging with laser scanning microscopes [18,29]. The high quantum efficiency helps to reduce photobleaching which is the biggest enemy of lifetime imaging in scanning microscopes.

The H7422 can also be used to investigate diffusion processes in cells or conformational changes of dye / protein complexes by combined FCS / lifetime spectroscopy. Although the accuracy in the time range below 1.5 µs is impaired by afterpulsing, processes at longer time scales can be efficiently recorded.

Another application of the H7422 is optical tomography with pulses NIR lasers. Because the measurements are run in-vivo it is essential to acquire a large number of photons in a short measurement time. Particularly in the wavelength range above 800 nm the efficiency of H7422-50 and -60 yields a considerable improvement compared to PMTs with conventional cathodes.

The Hamamatsu H7421 is similar to the H7422 in that it contains a GaAs or GaAsP cathode PMT, a thermoelectric cooler, and the high voltage power supply. However, the output of the PMT is connected to a discriminator that delivers TTL pulses. The output of the PMT is not directly available, and the PMT gain and the discriminator threshold cannot be changed. The module is therefore easy to use. However, because the discriminator is not of the constant fraction type, the TCSPC timing performance is by far not as good as for the H7422, see figure 26.

The FWHM is only 600 ps. Moreover, it increases for count rates above some 100 kHz. Interestingly no such count rate dependence was found for the H7422. Obviously the H7422 is a better solution if high time resolution and high peak count rate is an issue.

The H5783 and H5773 photosensor modules contain a small (TO9 size) PMT and the high voltage power supply [30]. They come in different cathode and window versions. A ‘P’ version selected for good pulse height distribution is available for the bialkali and multialkali tubes. The typical TCSPC response of a H5773P-0 is shown in fig. 27. The device was tested with a 650 nm diode laser of 80 ps pulse width. A HFAC-26-10 preamplifier was used, and the response was recorded with an SPC-730 TCSPC module.

The response function has a pre-peak about 1 ns before the main peak and an secondary peak 2 ns after. The pre-peak is caused by low amplitude pulses, probably from photoemission at the first dynode. It can be suppressed by properly adjusting the discriminator threshold. The secondary peak is independent of the discriminator threshold.

The Becker & Hickl PMH-100 module contains a H5773P module, a 20 dB preamplifier, and an overload indicator. The response is the same as for the H5773P and a HFAC-26 amplifier. However, because the PMT and the preamplifier are in the same housing, the PMH-100 has a superior noise immunity. This results in an exceptionally low differential nonlinearity in TCSPC measurements.

A histogram of the times between the photon pulses for the H5773 is shown in fig. 28. The devices show relatively strong afterpulsing, particularly the multialkali (-1) tubes. Taking into regards the small cathode area of the devices the dark count rates are relatively high. Selected devices with lower dark count rate are available.

The H5783, the H5773 and particularly the PMH-100 are easy to use, rugged and fast detectors that can be used for TCSPC, multiscalers and gated photon counting as well. In multiscaler applications the detectors reach peak count rates of more that 150 MHz for a few 100 ns. The detectors are not suitable for FCS or similar correlation experiments on the time scale below 1 us.

The R7400 and the older R5600 are bare tubes similar to that used in the H5783 and H5773. There is actually no reason to use the bare tubes instead of the complete photosensor module. However, for the bare tube the voltage divider can be optimised for smaller TTS or improved linearity at high count rate. The TTS width decreases with the square root of the voltage between the cathode and the first dynode. It is unknown how far the voltage can be increased without damage. A test tube worked stable at 1 kV overall voltage with a three-fold increase of the cathode-dynode voltage. The decrease of the response width is shown in fig. 30.

The Hamamatsu R5900-L16 is a multi-anode PMT with 16 channels in a linear arrangement. In conjunction with a polychromator the detector can be used for multi-wavelength detection. If the R5900-L16 is used with steady-state and gated photon counters or with multiscalers 16 parallel recording channels, e.g. two parallel Becker & Hickl PMM-328 modules are required. For TCSPC application the multi-detector technique described in [9] and [12-15] can be used. TCSPC multi-detector operation is achieved by combining the photon pulses of all detector channels into one common timing pulse line and generating a ‘channel’ signal which indicates in which of the PMT channels a photon was detected. The Becker & Hickl PML-16 detector head [13] contains the R5900-L16 tube and all the required electronics.

The R5900-L16 has also been used with a separate routing device [12,31]. However, in a setup like this noise pick-up from the environment and noise from matching resistors and preamplifiers adds up so that the timing performance is sub-optimal.

The TCSPC response of two selected channels of the PML-16 detector head is shown in fig. 32. The response of a single channel of different R5900-L16 is between 150 ps and 220 ps FWHM.

The response is slightly different for the individual channels. Fig. 33 shows the response for the 16 channels as sequence of curves and as a colour-intensity plot. There is a systematic wobble in the delay of response with the channel number. That means, for the analysis of fluorescence lifetime measurements the instrument response function (IRF) must be measured for all channels, and each channel must be de-convoluted with its individual IRF.

The data sheet of the R5900-L16 gives a channel crosstalk of only 3%. There is certainly no reason to doubt about this value. However, in real setup it is almost impossible to reach such a small crosstalk. If crosstalk is an issue the solution is to use only each second channel of the R5900-L16 [31]. If the PML-16 is used with only 8 channels, the data of the unused channels should simply remain unused. If the R5900-L16 is used outside the PML-16 the unused anodes should be terminated into ground with 50 W .

Side window PMTs are rugged, inexpensive, and often have somewhat higher cathode efficiency than front window PMTs. The broad TTS and the long SER pulses make them less useful for TCSPC application or for multiscaling or gated photon counting with high peak count rates. However, side-window PMTs are used in many fluorescence spectrometers, in femtosecond correlators and in laser scanning microscopes. If an instrument like these has to be upgraded with a photon counting device it can be difficult to replace the detector. Therefore, some typical results for side window PMTs are given below.

The width and the shape of the TCSPC system response depend on the size and the location of the illuminated spot on the photocathode. The response for the R931 - a traditional 28 mm diameter PMT - for a spot diameter of 3 mm is shown in Fig. 35.

By carefully selecting the spot on the photocathode an acceptable response can be achieved [31,32]. A TCSPC response width down to 112 ps FWHM has been reported [32]. This short value was obtained by using single electron pulses in an extremely narrow amplitude interval and illuminating a small spot near the edge of the cathode.

The afterpulse probability for an R931 is shown in Fig. 36. The afterpulse probability depends on the operating voltage, and the afterpulses occur within a time interval of about 3 µs. The high afterpulse probability does not only exclude correlation measurements on the time scale below 3 µs, it can also result in a considerable signal-dependent background in high repetition rate TCSPC applications

Surprisingly, modern 13 mm diameter side window tubes are not faster than the traditional 28 mm tubes. The TCSPC response for a Hamamatsu R6350 is shown in fig. 37.

13 mm tubes are often used in the scanning heads of laser scanning microscopes. It is difficult, if not impossible to replace the side-window PMTs with faster detectors in these instruments. Therefore it is often unavoidable to use the 13 mm side-on tube for TCSPC lifetime imaging. Depending on the size and the location of the illuminated spot an FWHM of 300 to 600 ps can be expected. Although this is sufficient to determine the lifetimes of typical high quantum yield chromophores, accurate FRET and fluorescence quenching experiments require a higher time resolution.

CP 944 Channel Photomultiplier

The channel photomultipliers of Perkin Elmer offer high gain and low dark count rates at a reasonable cost. Unfortunately the devices have an extremely broad TTS. The TCSPC system response to a 650nm diode laser is shown in fig. 38. The FWHM of the response is of the order of 1.4 to 1.9 ns which is insufficient for typical TCSPC applications.

However, the Perkin Elmer channel PMTs have high gain, a low dark count rate and a surprisingly narrow pulse height distribution. This makes them exceptionally useful for low intensity steady state photon counting or multichannel scaling.

The Perkin Elmer SPCM-AQR single photon avalanche photodiode modules are well-known for their high quantum efficiency in the near-infrared. Unfortunately the modules have a very poor timing performance. The TCSPC response for a SPCM-AQR-12 (dark count class <250 cps) is shown in fig. 39.

The response was measured with a 405 nm BDL-405 and a 650 nm ps diode laser of Becker & Hickl. The pulse width of the lasers was 70 to 80 ps, i.e. much shorter that the detector response. The measurements show that the TTS is not only much wider than specified, there is also a considerable change with the wavelength, and, still worse, with the count rate. Therefore the SPCM-AQR cannot be used for fluorescence lifetime measurements.

Interestingly, an older SPCM-AQR had a smaller count-rate dependence. Fig. 40 shows the TCSPC response of an SPCM-AQR-14 (dark count class < 40 cps) manufactured in 1999. Although the shift with the count rate is still too large for fluorescence lifetime experiments, it is much smaller than for the new device.

The afterpulse probability of the SPCM-AQR is low enough for correlation experiments down to a few 100 ns, fig. 41. An inconvenience of the non-fibre version of the SPCM-AQR is that it is almost impossible to attach it to an optical system without getting daylight into the optical path. A standard optical adapter, e.g. a C-mount thread around the photodiode, would simplify the optical setup considerably.

The conclusion is that the SPCM-AQR is an excellent detector for fluorescence correlation spectroscopy and high efficiency steady state photon counting but not applicable to fluorescence lifetime measurements. This is disappointing, particularly because state-of-the-art TCSPC techniques allow for simultaneous FCS / lifetime measurements which are exceptionally useful to investigate conformational changes in protein-dye complexes, single-molecule FRET and diffusion processes in living cells. Currently the only solution for these applications is to use PMT detectors, i.e. the R3809U MCP, the H7422 or the R5900 which, of course, means to sacrifice some efficiency.

Full Text