The timing accuracy and stability of the system are of fundamental importance when attempting to obtain high-quality time-resolved data. This is because the initial temporal calibration of the system must be assumed to remain valid throughout the imaging experi ment, which in the case of continuous clinical monitoring may be up to several hours. There
are numerous causes of temporal drift and jitter, and various measures have been invoked to eliminate or reduce these to acceptable levels.
30
T im e [m ln ]
Figure 7-6 Graph illustrating small temporal drift and Jitter in 32 channels over a 60 min period after the instrument had been allowed to stabilise.
The method employed here to evaluate the temporal stability of the system is to repeatedly perform a detector calibration measurement as outlined in section 8.1.1 below. The pulse arrival time (defined by the peak position, which is derived from a polynomial fit to the recorded time curve) is repeatedly measured for all channels, and plotted as a function of time. Figure 7-6 shows such a plot, recorded over a 1 h period after allowing the instrument to stabilise. The delay has been arbitrarily set to 0 ps at time = 0 min, and is recorded relative to this reference for each channel. Data points were recorded every 90 sec, with each acquisition lasting 10 sec. The system can be seen to be highly temporally stable in this experiment. When it is in optimal working condition, the drift (slow variation in delay) is typically around 5 ps/h, and the jitter (erratic point to point variation in delay) is about the same magnitude.
I 1 B -50 : O -100 : -150 30 25 15 20 10 5 0 T im e [m in ]
Figure 7-7 Graph showing strong drift and jitter in 32 channels over a 30 min pe riod. Note the different x- and y-axis scales as compared to Figure 7-6.
Before the system had been optimised, we observed drift and jitter over comparable periods of time of the order of tens to hundreds of picoseconds. Eliminating this problem was a major task and has been the focus of a considerable effort for nearly a year.
A drift plot showing typical (but not worst case) behaviour when the system was in an unstable state is shown in Figure 7-7. In this experiment the system had been switched on for only 1.5 h, which partly explains the relatively strong drift in all chaimels of -25 ps in the 30 min period. While some channels are relatively stable, others exhibit strong erratic jitter. The measures taken to minimise these problems and achieve the current level of
stability are outlined below.
7.3.1 System warm-up
Figure 7-8 shows the air temperature in MONSTIR and room temperature recorded over a 20 h period immediately following start-up of the system. While the graph suggests that MONSTIR requires approximately two hours to reach a steady operating temperature, drift tests, and advice from EG&G engineers, indicate that a minimum warm-up time of approximately 10 h is required to fully stabilise the system. Ideally it should be left switched on permanently. 28.0 26.0 24.0 22.0 20.0 50.0 45.0 40.0 35.0 30.0 25.0 20.0 240 0 120 360 480 600 720 840 960 1,080 1,200 O) a> -a T im e [min]
F igure 7-8 Ambient (room) and MONSTIR temperature (measured above the CFD rack) after start-up at Time = 0. The vertical line on the right-hand side indicates the switching on of the air-conditioning system. If MONSTIR, the MCP-PMT chiller units and the laser system are all switched on, the room temperature can rise up to about 31°C - despite air-conditioning.
7.3.2 Temperature fluctuations
Fluctuations of a few degrees in the ambient temperature can cause measurable temporal drift or jitter that is believed to be primarily due to phase delays in the coaxial signal cables and delay unit (which itself contains long cables). For instance, the RG-58A/U coaxial cable used in MONSTIR has a phase delay specification of -150 ppm/°C, corresponding to -1 ps/°C per meter of cable. Keeping temperature fluctuations during the experiment to a minimum by ensuring a relatively stable system temperature, and shielding or enclosing the cables and electronics in a stable unit, helped reduce drift and jitter. Furthermore, minimis ing cable lengths reduces these effects. In a future design it may also be advisable to choose cable types that have been designed to exhibit small phase delay variations with tempera ture.
7.3.3 PSU stability and noise
Power supply stability is important since supply voltage variations have been experimen tally shown to directly affect the timing properties of the pulse processing electronics^'*. The system incorporates a combination of linear and switch-mode Power Supply Units (PSUs). Switch mode PSUs are inherently very noisy, and high frequency bursts at a repetition rate of -100 kHz, with tens of mV amplitude, were measured at the pre-amplifier output (= CFD input). Though it is possible to filter the noise by raising the CFD threshold, this may compromise the optimum operation of the discriminator in terms of the IRF shape/width and count rate. It was found that this noise signal, which exhibits very large variations in time and at different cable positions, was primarily picked up by the power supply and signal cables that run between the MCP-PMT cooling housing Peltier units and the corresponding linear PSU/Controller unit. In order to reduce its effect, the switch-mode PSU output cable lengths have been minimised. Furthermore, screening is employed and ferrite filters are fitted throughout the system to reduce the radiation and pick-up of high- frequency noise that may interfere with the pulse processing equipment.
As far as the MCP-PMT HV supply is concerned, Hamamatsu quote a voltage dependence on the transit time of their detectors of only -0.1 p,s/V. This is believed to be negligible given the stability of the Canberra HV PSUs described in section 6.2.2.7.
7.3.4 Wiring and grounding
Tlie system has been rewired so that the mains/power cables are located on one side, and the signal lines are located on the other side of the rack where possible. A rigorous grounding scheme for all components has been installed.
7.3.5 Electronic connectors
Significant jitter of up to several hundred ps can be caused by poor electronic connections between components resulting, for example, from loose fitting and dry Joints. The system employs a combination of LEMO, BNC and SMA connectors. Cable flexure, especially when the cable is fitted with a poor coimector, can also have a measurable effect.
7.3.6 Laser pulsing stability
Laser pulsing instability (random variations of the laser pulse width from the normal ~1 ps up to several hundred ps) causes significant jitter, as well as a broadening of the IRF. Moreover, laser instability usually demands re-alignment of the laser optics, which in turn may necessitate re-calibration of the imaging system.
7.3.7 Reference photodiode
Optimum reference photodiode operating conditions (supply voltage and illumination level; see also section 6.2.1.10) are important. It was found that a small drift in the incident laser power (a few %/h) can translate into a measurable temporal drift of the reference signal.