The Cairn Research Optoscan Monochromator. Rack controlled Monochromator

Transcription

1 The Cairn Research Optoscan Monochromator Rack controlled Monochromator Technical Manual 2.0 Aug 2003

2 Important Information I. Important - Please read before installation As with all our equipment, we have tried to impose as few constraints as practicable on the flexibility and performance of the Optoscan Monochromator. For maximum reliability we would recommend using the equipment within certain guidelines, but with care the Optoscan can be driven somewhat harder than this. If in any doubt, then please feel free to contact our technical support department ( tech@cairnweb.com). The following points should be considered when using the Optoscan: 1. Xenon arc lamps are pressurised and a potential hazard. If using the Optoscan with our light source than please refer to the relevant section of the manual when changing lamps. If using a third party product then please refer to the manufacturer's documentation. 2. Use suitable protective eyewear when focussing the light. The short wavelengths generated by the monochromator are potentially hazardous, so care should be taken avoid direct exposure to the beam. 3. The diffraction grating is galvanometer driven and can change between wavelengths very rapidly (less than a millisecond for most practical purposes). This level of performance is attainable in prolonged repeat sequences provided that either the steps are very small, such as in scan mode, or if there are sufficient periods of inactivity within a cycle. If the grating is overdriven then both the galvanometer and the drive electronics will cut out transiently. These measures should not be regarded as a full protection against damage, and if the instrument does start to behave erratically it should be switched off and allowed to cool down before re-use. Please refer to the control interface section of this manual (section 3.1) for suggested protocols. 4. The slit changers are the most vulnerable part of the instrument to mechanical damage. When operating the galvanometer controlled slits, it is important to consider not only the points in note 3 above, but also the physical implications of rapidly opening and shutting a pair of sprung blades. If operating fast or for long periods it is advisable to keep the degree of movement relatively low. Particular care should be taken if driving the galvanometer slits from an external source because an inappropriate applied voltage could destroy the blades by over opening or closing the slits. We recommend a maximum opening of 3mm. 5. If either or both slits are micrometer controlled then care should be taken not to over open the blades. It is however acceptable to unwind the micrometer head to below zero in order to provide an overlap in the blades for shuttering. 6. The output light guide connector should be fitted very carefully so as not to push the guide into the exit slit mechanism. 7. The lamps we supply for use with our light source have a built in filter to cut off the short wavelengths which generate ozone, so there is little or no ozone emitted from the light source. Page i

3 Contents II. Table of Contents I. Important - Please read before installation... i II. Table of Contents... ii 1 Introduction The Optoscan system controller System overview General Notes Software Menu System Principles of monochromator operation Controlling the wavelength Controlling the bandwidth Speed, optical aberrations and optical throughput Choice of grating Use of signal control modules with the Optoscan General Notes PMT supply module Input amplifier and PMT amplifier Output module Gain / offset module Ratio module Metering module Rear panel connections on Optoscan control rack The Optoscan 37-way control interface The monochromator 25-way connector Technical Support...40 Page ii

4 Introduction 1 Introduction The Cairn monochromator system comprises of four main components, namely a light source, a power supply for the light source, the monochromator box itself, and a programmable system control box. If you have the rack mounting version of the control system it will also contain support for our photometry modules. In addition, components for connecting the monochromator output to a fluoresence microscope via an epifluoresence port will usually be supplied. A key design requirement for our monochromator was to ensure compatibility with other Cairn products. It has been possible to achieve this without any performance compromises as the design of our other equipment - in particular our fluorescence photometry modules - anticipated the introduction of such an instrument. As with all our equipment there is considerable versatility in its configuration and mode of use, allowing it to be optimised for a specific application. For initial setup of the Optoscan, and basic operation and control of the monochromator please refer to the Optoscan user guide. This manual contains detailed technical information for users who want to utilise advanced control features available within the Optoscan, or those of an inquisitive nature. If you are using third-party software to control the monochromator then please refer to their software manual for wavelength selection and control. Details on how to connect the Optoscan to your PC can be found in the installation section of the user guide. Page 1

5 Optoscan Control 2 The Optoscan system controller 2.1 System overview The microprocessor system is actually a small PC in its own right, and although the control unit is available in several different versions, the microprocessor system is the same in all cases. Most versions of the control unit include a keypad and a display that can operate in both text and graphic modes. However, the keypad and display functions can be provided by a PC emulator program in all versions, operating via a standard serial (RS232) link. To use this facility, the microprocessor unit should be connected to any available COM port on the PC via a standard serial (i.e. null-modem) cable and the optoscan.exe program should be installed and run. From this point onwards things should be self-explanatory, but please note that when the emulator program is running, the microprocessor keypad can still be used if preferred. On the PC, either the corresponding keyboard keys can be used, or the mouse point and click facilities (ugh!) can be used to operate the virtual keypad on the PC display. The serial link can allow certain program and control information to be transferred directly between the PC and the microprocessor, to assist in integration of the microprocessor system with third-party software. In addition to the 25-way D connector for the monochromator itself, the system also includes a 37-way D connector for interfacing with other equipment. This connector provides both input and output signals. All inputs are held low by internal 10Kohm resistors, and their polarities have been chosen so that a logic low level corresponds to the default state for each input. Therefore, if the associated facilities are not being used, there is no need to connect anything to these inputs Although the microprocessor system provides completely generalised facilities, it includes specific provision for the control of our fluorescence photometry modules. It is therefore also available in versions which can accommodate either a basic set or a complete set of these modules. They are designed for operation with wavelength-changing systems (including our rotor as well as the Optoscan) that can select between up to eight different wavelengths, and this is the reason why the basic Optoscan functions are also designed to support eight wavelengths rather than some other number (operation of the modules is described in a later section). However, up to four independent wavelength sets can be programmed at any one time, and up to four wavelength-scanning modes for the generation of continuous spectra are provided as well. These facilities will now be described in more detail. The Optoscan control system is actually a small PC in its own right, which is controlled via the keypad on the system box. It can also take its input from a standard PC using the Optoscan terminal emulator, which is provided with your system. To program the controller via this route, the serial port on the rear of the Optoscan control system must be connected to any serial port on your PC. The description here describes data input from the keypad, but the process is identical via the PC. 2.2 General Notes The keypad (or PC keyboard) is used both to select the numbered menu options and to enter data as requested. All entered values are range-checked, so it should not be possible to make inappropriate settings. Note that in some cases fractions are permitted. For example, wavelengths can be entered in increments of 0.5nm, optical slit widths in increments of 0.1nm, and times in increments of 0.1 msec. The keypad does not have a decimal point key, but fractions can be added or subtracted during numeric entry by using the up and down arrow keys respectively. On the PC keyboard, fractions can be entered either this way or by typing the number with a decimal point as usual, but in Page 2

6 Optoscan Control that case fractions will be rounded to the permitted increments. Pressing the clear key on the keypad or the backspace key on the keyboard erases the last digit entered, and subsequent presses erase previous digits, so a complete entry can be cleared if required. Users of an unusually curious disposition will discover that once numeric entry has started on either the keypad or the keyboard, it can continue only via the same device. The system can write both to its own display and to the PC via the serial link at the same time, but it does so only if appropriate. In systems lacking the built-in display, this situation is detected at switchon, and the associated display and keypad code is disabled. Systems that DO have a keypad and display can be forced into this state if required, by holding down any of the keypad keys for about 1 second at switchon (but in this case the keypad will also be disabled). Similarly, if the PC terminal program isn t running, the system will not send any information down the serial link, which allows the built-in display (if fitted) to be updated somewhat more quickly. However, data output via the serial link can be enabled or disabled by the user if required, by pressing the enter and clear keys respectively. 2.3 Software Menu System In describing the operation of the microprocessor control unit, it will be most convenient to follow the same sequence as used by the software main menu. The monochromator functions in two principal modes, namely step and scan, which are referred to as wavelength programs and scan programs respectively. On power on you are presented with the main system menu. The first four menu functions control the wavelength position programs, where the monochromator switches between discrete wavelengths of defined bandwidth. Function five configures how the Optoscan responds to external control signals, and the remaining functions control the scan programs, where the system operates as a spectrophotometer with user selected wavelength ranges. The use of each of these functions is detailed in this section. For an overview of the menu system refer to the system user guide. 1. Set wavelengths and times This sub-menu allows up to four independent sets of eight wavelength positions to be programmed. Each set of positions is referred to as a program, and the user can switch between these programs at any time. The programs can be thought of as an eight position filter wheel where you can define the centre wavelength and bandwidth of each filter Associated with each wavelength position is a variable time. When the Optoscan is running a wavelength program (which allows any combination of individual positions to be run as a repetitive Page 3

7 Optoscan Control sequence), this parameter sets the time spent at that position. Our photometry modules integrate the signal obtained while the monochromator is at each position, and most CCD cameras operate in this way as well, so in both cases a longer sampling interval give a greater output. Therefore, having an independently variable time for each position is a useful way of controlling the relative sensitivity (but see also the following explanation of slit width control). Programming any particular position with menu function 1 does NOT necessarily include it in the wavelength program, as any combination of positions can be excluded if preferred. However, even if (say) only two positions are included in a program, all eight positions in each of the four sets are accessible from the manual and external wavelength control functions (menu functions 4 and 5 respectively, as described below). Each position also has independently programmable input and exit slit widths, which together control the optical bandwidth as described previously. As also explained, doubling the widths of both slits doubles the optical bandwidth but gives four times the optical throughput, so this can be a particularly effective way of increasing the optical signals. Together with the centre wavelength and the time, there are thus four parameters per position altogether, plus the choice of whether or not to include it in the associated wavelength program. Each wavelength program can include only the (up to) eight positions in the associated wavelength set, but since it is possible to change between programs during runtime (see menu function 3), then in practice any combination of up to 32 wavelengths can be supplied during any given experiment, which we hope will be sufficient for most purposes. The programming of so many possible positions may seem a daunting task, but since the memory is preserved when the system is switched off, this should not need to be done very often. As a further precaution, the PC terminal software allows a copy of the microprocessor memory to be stored in a data file, allowing easy restoration if anything does go wrong. Alternatively, if a system is shared between several users, each user can use this facility to load the system with their own individual programs whenever they want. The final point to discuss for this menu option is the real time adjustment facility, which is intended primarily for our photometry modules, but may also be useful under other circumstances. If this facility is on, then the grating and slits are at the settings specified for the currently selected position, and internal timing pulses are generated to integrate the optical input signal for the specified period and to send it to the (analogue) BNC output for that position. Changes to any of these parameters are immediately implemented, allowing the user to experiment to find the most appropriate settings. If the real time option is NOT selected, both slits will be fully closed, and no timing signals will be generated; this may not sound so useful, but it is intended to avoid unnecessary illumination of the sample (with possible consequent photobleaching and other undesirable effects). 2. Show wavelengths and times Menu function 1 includes an option to show the programmed settings for the four sets of positions, but menu function 2, while also including this option, allows much more detailed information to be displayed. In any case, the advantage of using menu function 2 for showing program information is that it is for display purposes ONLY, so there is no risk of accidentally changing anything by pressing the wrong keys. This function provides two additional facilities. The first of these is to show the actual slit widths for each position. As explained in the theoretical section of this manual, the relationship between optical bandwidth and mechanical slit width is somewhat wavelength-dependent, and it is also different for the two slits. For users who do NOT have the galvo-driven slits (or for those who are just curious), information on the actual slits widths for a particular optical bandwidth allows the most appropriate mechanical slit widths to be set for a particular experiment. Of course a compromise is likely to be involved here, as the same mechanical slit widths are going to apply for ALL positions, but Page 4

8 Optoscan Control on the other hand the information provided by this facility should allow the best compromise to be selected. The second facility is potentially very useful under ALL circumstances, as it allows the actual bandpass characteristics for each position to be displayed. This is particularly valuable when experimenting with different input and exit slit widths. A very important guide is that the optical throughput at each position is proportional to the area enclosed by the bandpass characteristic. In general, for maximum throughput the input and exit slit widths (defined optically, as we do, rather than mechanically) should be the same. The bandpass characteristic is then triangular, with the bandwidth, as defined between the 50% points (full width half maximum, FWHM), being half the total width of the spectrum along its baseline. If one slit is made wider than the other, then a flat top appears on the bandpass characteristic, and the FWHM bandwidth more closely approaches the total baseline width of the spectrum. Such an effect may sound desirable, and under some circumstances it may well be so, but comparisons will readily show that it is achieved at the expense of reduced throughput compared with the equal-slits condition. This topic has already been well described in the theoretical part of the manual, so for more information, please (re)read that section. Here the emphasis is very much on the practical side to see the effect of a particular slit width combination, one just needs to program it in and look. 3. Run wavelength program This function allows any of the four wavelength programs to be run. The first option selects the wavelength program, or it allows the program number to be controlled externally using the two program control input pins on the 37-way D connector, thereby enabling the user to switch between programs at any time. A program consists of a series of up to eight positions, repeated either for a specified number of cycles or indefinitely. Each position (if included in the program) is visited in numerical order, but since each position can correspond to any wavelength, the actual wavelengths can be visited in any desired sequence. When a program is running, the run/stop output pin is high, and the two program output pins encode the currently running program number ( or 11 for programs 1-4 respectively). Similarly, the three position output pins encode the current position ( for positions 1-8 respectively). Since the slits are open when a program is running, the shutter output pin is also high (see next section for further explanation). However, the slits remain closed at other times, in order to avoid unnecessary illumination of the sample. In addition, an analogue output (the user DAC output) is available on the D connector. This has a range of 0-5V, and defaults to representing the wavelength for the current position, with a scale of 1V per 100nm, starting with zero at 300nm. However, it can be programmed to follow any other wavelength-dependent function, which may be useful for controlling other equipment. Setup and use of this facility is dealt with in the description of the wavelength scanning mode. A program can either be run immediately, or it can start on receipt of a trigger input pulse via the run/stop input pin. Operation in this mode is particularly flexible. The program will start as soon as the run/stop pin goes high, and it will continue either for the specified number of cycles, or until the run/stop pin goes low again, which will cause the program to stop at the end of the current cycle. If the run/stop pin is taken high again, then the program will be run again, starting from the first cycle, but if the program had stopped with the run/stop pin high (i.e. because the specified number of cycles had been executed) the run/stop pin must be taken low and then high again to restart. Note that the action of the run/stop pin in this mode is mirrored by the go input pin, as the two inputs are ORed together. This is to allow either of two independent devices to control program execution. However, if only one control device is used, we recommend that it is connected to the run/stop pin rather than to the go pin (which can be left disconnected as it is held at logic low level by an internal resistor). Page 5

9 Optoscan Control When a program is running, and the external program control option has been chosen, then the logic levels on the two program control input pins determine which of the four possible programs is being run at that time. Specifically, these pins are monitored at the transition to the last position in each cycle, and the corresponding program number will be run, starting from the first position of the next cycle. However, the cycle counter will not be reset by switching programs in this way. This facility can be used to interpose either drastic or quite subtle changes to a wavelength sequence, depending on exactly how different the individual programs are. This may sound like a solution in search of a problem, but we foresee a number of useful applications, and we expect that users will be able to devise many others. The other major feature here is the control and display of transition times between positions. The galvo drive electronics provides three logic-level signals to inform the microprocessor system when the grating and slits have reached their new positions. These signals are ANDed together by the microprocessor to produce an overall ready output. As well as being available on the 37-way D connector, it is also sent to the internal backplane (if fitted) for the system modules, where it inhibits input signal integration when in the low (not ready) state. It is important to switch off signal detection during wavelength transitions, since the grating will be scanning across intermediate wavelengths during this time. The effects of acquiring data while slit widths are changing are likely to be less serious, but in general the slits are likely to reach their new positions before the grating does, so there is usually no additional time penalty for waiting for all three ready signals. We realise that some applications may also require receipt of a ready signal from external equipment. We have therefore provided an external ready input on the 37-way D connector, which is combined with the three galvo inputs to derive the overall ready signal. This pin is an exception to the rule for the other inputs, in that it is held at a logic high level if not connected. However, as this corresponds to the ready state, the rule remains the same as for the other input pins, i.e. nothing needs to be connected here if this facility is not required. The transition time between positions clearly depends on how much each galvo needs to move, so it will be different for each transition in the program cycle. However, the time for any given transition will be essentially constant from one cycle to the next. A key question now arises, as to whether or not the transition time should count as part of the total time spent at each position. The decision is up to the user, as the software allows either mode of operation, although our preference is to count the transition time as part of the total time. This has the advantage that the overall cycle times are entirely regular and predictable, allowing the various outputs to be used as accurate timebases for controlling other equipment. On the other hand, there will clearly be problems if the transition time (however specified) exceeds the total time at any position, since it is impossible for the system to operate correctly under these conditions. It is the responsibility of the user to avoid such a situation! We have therefore also provided the possibility of keeping the transition times independent from the sampling times, so that the system will wait for the transition to be completed before acquiring photometric or other data for the ENTIRE specified sampling time at each successive position. What does this choice mean in practice? Imagine that the monochromator is programmed to spend 10.0msec at each of three positions, and that the transition times to reach each position are and 0.6msec respectively. If we include the transition times within the time at each position, then the actual sampling times are 9.7, 9.5 and 9.4msec respectively, and the total cycle time is 30.0msec. On the other hand, if the transition times are not included, then the sampling times at each wavelength are all exactly 10.0msec, but the total cycle time is extended by the sum of the transition times, to 31.4msec. As stated above, we prefer the first option, but purists may feel unhappy with either. We have therefore provided the additional facility of transition time extension. This allows a minimum transition time value to be specified, which is the same for all positions. In the above case, we could Page 6

10 Optoscan Control specify a minimum transition time of 1.0 msec. Depending on whether or not transition times are included within the times at each position, this would give either 9.0msec sampling time at each wavelength and a 30.0msec cycle time, or 10.0msec sampling time at each wavelength and a 33.0msec cycle time. However, in both cases both the sampling times and the cycle time are now explicitly specified. Although this would appear to give the best of both worlds, it does have the disadvantage that one has thrown away a total of 1.6msec of usable data during the cycle. No method is perfect! Minimum transition times of up to 50msec can be specified, to allow more slowly-responding external equipment, reporting via the external ready input, to be treated in the same way. The smallest value is 0.1msec, which is to allow time for our photometry modules to perform the necessary signal processing for switching between positions. The above discussion has presumed that there is some way of determining the actual transition times, so that an appropriate value for the minimum transition time can be specified, and indeed there is. After a wavelength program has run, there is a menu option to show the actual transition times for each position change, to microsecond precision. These times will of course include the specified transition time. Transition times of up to at least an hour can be displayed, and our polite recommendation is that if external equipment has not responded within this time, then in fact it could be time to replace it. However, if the transition time is not included in the sample times, the microprocessor will continue to wait patiently until this is done. The minimum transition time is also available as an output on the new position pin on the D connector. A pulse of 100µsec to 50msec duration will be appear here when the system changes positions. There will also be a high level signal on the new cycle output pin, throughout the period corresponding to the first position in each cycle. Since this signal is very useful for synchronisation purposes, it is also available on a BNC socket in those versions of the system that can accommodate our photometry modules. 4. Manual wavelength control Operation of this menu should be largely self-explanatory. The choice of positions is the same as that specified for menu function 1, i.e. four sets of eight. A position does not have to be included in a wavelength program in order to select it from this menu, so all eight positions in each set are always available. Please note, however, that when this function is selected, the slits will initially be closed. As already described in the theoretical section, the "slits closed" state actually corresponds to the two edges of each slit overlapping slightly so that they form a very effective optical shutter. They can be opened and reclosed from this menu, regardless of which wavelength is selected, by using the keypad (or PC keyboard) arrow keys. Their condition is indicated by the slits status output on the 37-way D connector, as well as on the display. Within a given program, positions 1-8 are selectable directly using keys 1-8 respectively, and the currently selected position is also indicated by a marker at the appropriate position on the display. During manual wavelength control, the microprocessor system generates the necessary switching waveforms for controlling the photometry modules and/or external equipment. The integration period is set by the time entered for that position, so the signal values will correspond with those that would be obtained while running a wavelength program. If one were to change positions during the integration period, then the last signal value would be low, which is clearly undesirable. The system therefore waits until the end of the current integration period before changing positions. Of course, a short time may be lost from the first integration period for the new position, corresponding to the transition time for changing positions. However, if that is a problem, the solution is to use the minimum transition time facility, as explained above for menu option 3, since that remains operational in this mode. The system also behaves in exactly the same way Page 7

11 Optoscan Control as specified in the wavelength program menu, in respect of whether or not the (minimum) transition time is included in the sample time. Since this is specifically a manual control option, none of the inputs on the 37-way D connector (apart from the external ready input) have any effect in this mode. However, the outputs representing position and program number are all active, and the user DAC output is also available. 5. External wavelength control In this mode, control of the system is intended to be by external equipment, rather than from the keypad or from the internally programmable facilities. However, the keypad (or PC keyboard) allows any of these controls to be overridden, which can be useful during setup and testing. The display also shows the current program number, position number and slit status. The first three menu options determine whether the program number, position number and slits respectively can be controlled externally, the default condition being yes for all three. In external control mode, the system can also be treated as having 32 directly accessible positions, rather than four sets of eight as for the previously described modes. This is because position number and program number can be changed at the same time. To go to a new position and/or new program, the appropriate signals are applied to the corresponding program and position input pins on the D connector, and the go input is then taken high. The system will immediately go to the new position and/or program number. Since the go input is active only on its rising edge, it can remain high indefinitely without any further effect. If preferred, the run/stop input can also be used, as these two signals are ORed together by the software. However, the go input will give a slightly faster response (albeit by only a few microseconds), so this is the recommended choice. Even if the program and position numbers are under external control, they can still be changed by the keypad, since the external inputs are read only on the rising edge of the go (or run/stop) input. Any changes made via the keypad will therefore remain in effect until the next go command is received. However, the slits control input is active all the time, so any attempt to open or close them from the keypad when they are under external control will immediately be overriden. Note that the slits input is high to close; this polarity has been chosen so that the slits default to open if nothing is connected to this pin. Receipt of a go command (or making a change via the keypad) will also generate a new position signal, and the system will behave in exactly the same way as for the other control modes. Since the go command is implemented by a software interrupt, the software execution time should be constant (and short in any case), so the overall timing accuracy will depend only on the accuracy of the signal used to drive it. (The run/stop input is polled, so the software execution time in this case will be somewhat longer and more variable, but the effect is negligible compared with the timing problems of the PC environment). 6. Set scanning parameters Menu functions 6 and 7 control the scanning mode of the monochromator, in the same way that functions 1 and 3 control the wavelength program mode. The scanning mode has been implemented in a simple but very powerful way, which allows it to serve a variety of functions. Although it can be made to approximate arbitrarily closely to a continuous scan, it actually divides the scan range into a linear array of individual wavelengths, which it steps through in sequence. This means that the operation in scan mode is effectively identical to running a wavelength program from menu function 3, except that there may be many more wavelengths. Page 8

12 Optoscan Control Scans can be programmed to run in either direction, anywhere within the wavelength range nm, with a wavelength interval as low as 1nm. Alternatively, much broader wavelength intervals can also be selected, as can a much narrower scan range, so a scan may contain anywhere between only a few and several hundred individual measurements. Scanning can also be triggered by an external logic-level signal in exactly the same way as for the wavelength program, and up to four different scan programs can be entered and run. In fact the main difference between the wavelength programs and the scan programs is that in the wavelength programs each wavelength has independently-specifiable slit widths and times, whereas in the basic scan programs the (optical) slit widths and times are the same for all wavelengths in the scan. However the mechanical slit widths will change during the scan to maintain the specified optical bandwidth. While this is a very desirable arrangement, it may not be adequate to deal with the conditions encountered in practice, so additional facilities have been provided to modify the scan parameters in a variety of different ways.. A full scan can cover the range nm, and although a xenon light source produces a usable output right across this range, wavelength-dependent losses in the rest of the system and the wavelength-dependence of the detector sensitivity can combine to give a response characteristic that varies substantially with wavelength. In many scan applications, however, one is looking for small CHANGES in the spectrum, and these can be detected and recorded much more accurately if one can arrange for the initial response to be relatively constant across the spectrum, so that the signal levels can be high at all wavelengths. This can be done by appropriately modifying the slit widths and/or times at each scan wavelength, as described below. Another requirement may be to adjust other equipment during a scan. The most important example is where microscope images are involved, as many microscope objectives particularly those designed to work at UV wavelengths show substantial focussing shifts across the spectrum. It can therefore be useful if not essential to provide some form of programmable control voltage to make the necessary compensatory adjustments via a motorised focussing unit. Since this is a very similar requirement to that of adjusting the detected signal levels across the spectrum, it is handled in exactly the same way, which is as follows. As explained above, a scan may be comprised of literally hundreds of individual wavelength positions. Although the parameters could in principle be entered individually for each one, a much simpler method is clearly required. The method we have chosen requires far fewer entries, and there is also no need to re-enter the values whenever the actual scanning parameters are changed. It will be described using programming of the user DAC output as the example, which is done via option 8 of the scan setup menu function. All these adjustments have the feature that the necessary compensations are likely to change relatively smoothly across the spectrum at least in the sense that there will be no sudden discontinuities. Under these circumstances it is sufficient to enter the adjustments at a series of spot wavelengths, and the system can then calculate the adjustments at any given wavelength by linear interpolation between the values for the adjacent pair of spot wavelengths. However, we have set the spot wavelengths relatively close together, at 25nm intervals, so that even relatively complicated characteristics can be followed with good accuracy. In the case of the user DAC output, the spot wavelength values are entered directly as voltages, in the range 0 to +5V. Voltages at each wavelength can be entered with a precision of 1mV, but since the DAC resolution (12 bits) is slightly lower than this, displayed values are rounded to the nearest even mv, although the equivalent DAC value closest to the voltage actually entered is always used. When used for applications such as focussing control, where the required characteristic is unlikely to be known in advance, the DAC values at each spot wavelength need to be determined by trial and error, for which the real time adjustment facility (as described for menu function 1 above) is extremely useful. Page 9

13 Optoscan Control Option 8 of the user DAC programming submenu allows real time adjustment to be selected for this purpose. Entered values can all be seen together by selecting the show DAC values option. They are stored in nonvolatile RAM, so they are preserved when the system is switched off. As an additional safeguard, the backup and restore functions in the PC terminal program allow copies of ALL programmed information to be kept in a PC file. The default case for the user DAC output is for it to give a voltage proportional to wavelength, as described previously. Adjustments are therefore made with the system initially in this state, and the restore default values option returns it to this state if required. The system also reverts to this state automatically if the modified values have been lost for some reason (such as the electromagnetic pulse from switching on another company s light source...). As already mentioned, there can be up to four scanning programs resident in the system. The main reason for this is to allow each scanning program to have independent control tables for the user DAC, slit width and scan step times. In the case of the user DAC focussing application described here, it means that an appropriately different set of values can be entered for each of several different microscope objectives. Although the features described here were developed specifically for scanning, the focussing application may also be useful for wavelength programs. There is therefore a use for wavelength program option in the user DAC programming submenu. If this is selected, then the user DAC values appropriate to the programmed wavelength for each position are generated while the wavelength program is running. Note that only ONE of the four possible sets of DAC values can be used at any one time, and all four wavelength programs will use that same set of values. The rationale behind this is that is one is unlikely to want to change objectives during a wavelength program, but one might possibly want to change the wavelength program number in order to introduce alternative wavelength sequences during the program. Similar tables for controlling slit widths and times during a scan are also provided, and they can be programmed via the spectral adjustments submenu. Either or both adjustments can be active during a scan. Varying the slit widths has the advantage that the time spent at each wavelength remains the same, so in a recorded file the time at which each sample is recorded is a direct linear representation of the wavelength. The obvious disadvantage of this method is that the optical bandwidth changes across the spectrum. However, in practice this may not be a problem, particularly if things are arranged so that the bandwidth is reduced (compared with what may be required) for brighter spectral regions, rather than increased for darker ones, and this is how our system is designed to operate. Another point to remember is that, as explained in the theoretical section, the optical throughput is proportional to the square of the slit widths (i.e. optical bandwidth), so the optical bandwidth actually changes only with the square root of the relative sensitivity adjustment in this mode. The slits adjustment table allows slit values to be entered at each spot wavelength, in the range of 10%-100% of the default values. The two slits are adjusted together (although their relative widths, if different, remain constant), so this therefore corresponds to a 100:1 adjustment range. The default state of this table, which can be restored at any time if required, is 100% for all wavelengths. There is a different table for each of the four scan programs, and the corresponding table is ALWAYS used for controlling each scan. Therefore, if no slit adjustments are required, this is achieved by using or restoring the default values. In contrast, changing the sensitivity by varying the time spent at each wavelength allows the optical bandwidth to remain constant across the spectrum, but in this case the wavelength and time axes will no longer correspond, so additional measures must be taken to keep track of the wavelength. One way is to record a single point (or image) each time the microprocessor instructs the Page 10

14 Optoscan Control monochromator to go to the next wavelength in the scan, but this may not be convenient - or even possible - if the recording system is acquiring other data at a constant rate. In practice it may therefore be easier to record the user DAC output signal on another channel (assuming that the programmed output characteristic is still wavelength-dependent!), record at a constant rate, and then sort everything out at the data-processing stage - or to use the variable slit width option instead. The time adjustment table works in the opposite way to the slits adjustment table, in that times (and hence optical signals) can only be increased rather than decreased from the default step time value. In this case the adjustment range is 1:50, and here of course the effect is a linear one. Just as for the other tables, there is a different one for each scan program, and if time adjustments are not required, the corresponding table should be left in or restored to its default state. 7. Run scanning program This menu option has exactly the same choices as for option 3, for running a wavelength program, but there are some differences. For example, there is no facility to change the program number while a scan program is running, since that is unlikely to be of so much practical use. As explained above, different scan programs are designed to be used in conjunction with, say, different microscope objectives, which are only going to be changed between rather than during data acquisition periods. However, the control of scanning programs by an external trigger pulse is exactly the same as for the wavelength programs. During a scan, the various internal and external output control signals correspond to the system being permanently in position 1. The other important point to mention about scanning is that within a scan, all the wavelengths are visited in a linear sequence, and in the case of multiple scans, the grating must then jump back all the way to the start wavelength again. Clearly the time for this transition is going to be significantly greater than that for stepping between the individual scan wavelengths. However, the minimum transition time facility can be used to iron out this difference if required. Alternatively the option of not including transition times in the sampling times is also available here. Actual transition times cannot be displayed for the scanning program, but the transition time for jumping back - or for any other wavelength change - can easily be found by running a wavelength program that includes a jump between the appropriate pair of wavelengths. To ensure that the behavior is the same both for single and multiple scans, all scan programs start with the grating at the end wavelength, so even for a single scan, or for the first of a series, the program must still begin by jumping from the end to the start wavelength. However, in applications where short and precise sampling times are of critical importance, the situation is most easily handled by programming one more wavelength than is actually required at the start of the scan, and then just ignoring it. 8. Restore default settings The parameters entered for menu functions 1-7 are stored in non-volatile memory for indefinite re-use, but menu function 8 allows these to be replaced by the default values with which the system was originally supplied. This function may not be of much practical use, but the main reason for providing it is so that it can be run automatically if system memory is lost for any reason. Memory is normally preserved by a backup battery, which is recharged whenever the microprocessor control unit is powered. A full charge should be sufficient to last for many months, but if the battery does become discharged, or if the memory is corrupted for some other reason (such as severe electrical interference), the microprocessor will detect this when the unit is next switched on and it will automatically load the default settings. However, there is a possibility that a partial memory corruption will not be detected automatically, so if such a condition is suspected, then menu function 8 can be run to correct the situation. Page 11

15 Optoscan Control 9. Diagnostic mode Finally we come to menu function 9. This allows third-party software developers to access some of the program facilities without having to go through the menu system. A list of functions provided so far is given in the appendix of the technical manual. The diagnostic mode also allows full access to the microprocessor program, although that is primarily of use only to ourselves. However, for diagnosis of possible faults, we may ask users to perform some specific tests while the system is in this mode. Page 12

16 Principles of Operation 3 Principles of monochromator operation 3.1 Controlling the wavelength Although many detailed variations are possible, most monochromators are of the same basic design, which is generally known as the Czerny-Turner configuration, and ours is no exception. Its operation is quite easy to understand, but this topic is not well covered by standard optical texts, and we therefore give a basic description here. This description will also explain the particular features of our own design, and will also explain the very important interaction between bandwidth and optical throughput, which our real-time slit width control system can fully exploit. The standard Czerny-Turner configuration is shown in Fig. 1. Light from an appropriate source (a xenon arc in our case) is focussed onto an input slit, and light passing through this slit is collimated by a concave mirror, which also reflects it onto a diffraction grating. The grating in turn directs the light onto a second concave mirror, which reflects and focusses it onto an exit slit before it leaves the instrument. Mirrors are used rather than lenses because they do not introduce any chromatic aberration, but they do introduce other aberrations which limit the resolution of the instrument, as will be discussed shortly. Since the redirection of the light beam by the grating is actually a diffraction rather than an ordinary reflection, the grating disperses the beam, i.e. different wavelengths leave the grating at different angles. By rotating the grating about its central axis, we can vary the range of wavelengths which can be reflected and focussed by the second mirror onto the exit slit. Figure 3.1 : Standard Czerny-Turner configuration Page 13

17 Principles of Operation Fig. 2 shows how a reflective grating works. This is a horizontal section through the grating, which resembles a saw blade. Viewed face on, the teeth are actually grooves, and the two slits in the instrument are parallel with them. (Slits can be used rather than circular apertures because the grating disperses light only in the horizontal direction, so by lengthening the apertures in the vertical direction we can get more light through the instrument without losing any resolution). Imagine a light beam arriving at a direction perpendicular to the grating. If the grating just acted as a mirror, the beam would be reflected straight back again, and indeed, some proportion of the light is reflected in this way, but in this case it is referred to as the zero-order diffracted beam. However, the grooves are shaped so that most of the light leaves the grating at a different angle, corresponding to the first-order diffracted beam. This angle corresponds to the one where reflections from adjacent teeth give optical patch length differences of exactly one wavelength. Since the angle is therefore wavelength-dependent, the diffracted beam actually forms a spectrum, from which we can select the required wavelengths by focussing them onto the exit slit. Figure 3.2 : Principle of operation of reflection grating In the Czerny-Turner configuration, this is done by rotating the grating so that the required wavelength is always reflected at the same angle. This means that the angle of the incoming light also changes, but that doesn t have any major impact apart from making the calculations somewhat more interesting. The angles of the incoming and outgoing beams are conventionally referred to as α and β respectively, and although they both change when the grating is rotated, the difference between them, conventionally referred to as D, remains constant. Note that α and β are both measured with respect to the grating normal, and they will be of opposite signs if they are on opposite sides of the grating normal. The basic grating equation is given by sinα + sinβ = 10-6 knλ (1) where λ is the wavelength in nm, k is the diffraction order and n is the number of lines per mm for the grating. Page 14

18 Principles of Operation This can be rearranged in terms of the wavelength to give 10-6 knλ = 2sin[(β+α)/2]cos[(β-α)/2] (2) For a given instrument the deviation D is fixed and given by: D = β-α (3) so we can express β in terms of D and α and then solve for α to obtain: α = sin -1 [10-6 knλ/2cos(d/2)] - D/2 (4) For reference, n for our standard grating is 1200 lines/mm, k is always 1 (i.e. first-order diffraction) and D in our instrument is 20 degrees. The sensitivity of the electrical input which controls the grating angle is 0.25 volts per degree. Users who wish to make their own arrangements for driving the monochromator can therefore substitute these values into equation (4) in order to calculate the appropriate drive voltage for a given wavelength. However, users who also have our microprocessor control box do not need to become involved with any of this, as the microprocessor performs these calculations itself, so the required wavelengths can be specified directly. There are also maxima corresponding to path length differences of two or more wavelengths, giving second- and other higher-order diffracted beams (this series continues up to the maximum possible diffracted angle of 90 degrees), and there is another complete set of diffracted beams at the same angles on the other side of the incoming beam. However, by appropriately shaping the grooves in the grating, which is termed blazing, it is possible to direct most of the diffracted light into a particular ONE of these several destinations, over a reasonably wide range of wavelengths. The wavelength at which this occurs most efficiently is known as the blaze wavelength, and we use a grating with a blaze wavelength of 400nm in order to obtain highest grating efficiency (of around 70% into the required first order diffracted beam) in the near UV, for optimum results with indicators such as fura2 and indo1. However, the efficiency is still above 50% over most of the visible spectrum too. Although a number of other configurations have been described, nothing else seems able to beat the performance of the Czerny-Turner, particularly in respect of the (for biological fluorescence applications) requirement for high optical throughput, which we discuss below. In our opinion the only viable alternative is one in which the grating itself is concave, so that the two concave mirrors are no longer required. However, the substantially higher cost of such gratings (which must be custom designed) makes the instrument more expensive overall. Conventional plane gratings are much cheaper, and our suppliers can provide gratings with almost any required characteristics at the same relatively low cost. The standard grating we supply has 1200 line lines/mm, blazed at 400nm, since we believe it represents the best overall compromise for this type of instrument. However, just about any other required characteristic can be supplied to special order. 3.2 Controlling the bandwidth There is more to a monochromator than just selecting a particular wavelength. If the input and exit slits were arbitrarily narrow, and the internal optics were arbitrarily good, then one could indeed approach the ideal of selecting just one particular wavelength. Unfortunately, this goal can only be approached at the expense of lamentably low optical efficiency, for two reasons. The first and obvious one is that the narrower the range of wavelengths required, the smaller the proportion of the light entering the instrument that will be within that range. The second one is not so obvious but in practice is equally important. High spectral resolution requires BOTH slits to be arbitrarily narrow, and if the input slit is very narrow, then in practice only a small proportion of the light from the light source will be able to ENTER the instrument. To anticipate the following more detailed analysis, this means that the total light output tends to increase with the SQUARE of the slit widths, whereas as we shall see, the optical bandwidth increases LINEARLY with the slit widths. Page 15

19 Principles of Operation In practice, therefore, there is a very important trade-off to be made between bandwidth and light intensity. This is particularly relevant for biological fluorescence applications, where for fluorescence excitation, high light intensity is generally more useful than narrow bandwidth, but nevertheless some compromise must ultimately be made. Most monochromators make some sort of provision for varying the slit widths, although this may just be as crude as providing (or offering to provide ) a set of interchangeable slits of different widths. This is hardly user-friendly, and since the slits must be both manufactured and located with some degree of precision, it is clearly undesirable for the user to become involved with such things in any case. Those who are somewhat more fortunate with their choice of supplier may get some kind of calibrated mechanism that allows the slit widths to be varied, as provided on the lowest-cost version of our instrument. However, it is not generally appreciated that the relationship between mechanical slit width and optical bandwidth is wavelengthdependent, and is different for the two slits, so setting the correct slit widths for a given optical bandwidth is not as trivial a matter as might have been thought! Furthermore, the whole POINT of using a monochromator is because one wants to switch wavelengths, so even to maintain the SAME bandpass characteristic it would be necessary to readjust both slits. On the other hand, the best compromise between bandwidth and light intensity for a given application may well DIFFER according to the currently-selected wavelength. Consider the ph indicator BCECF, for example. This is a dual-excitation ratiometric indicator, which is normally excited at 440 and 490nm. The excitation spectrum for this indicator is shown in Fig. 3. The excitation bandwidth at 490nm must be fairly narrow, to avoid overspill into the emission region for this indicator, which starts at around 510nm, but on the other hand the fluorescence excitation at 490nm is relatively efficient. Figure 3.3 : Excitation and emission spectra for ph indicator BCECF Excitation at 440nm is much less efficient, but we can improve the situation considerably by using a greater bandwidth here (remembering that the improvement is with the square of the bandwidth as explained above). The ability to change bandwidth as well as wavelength is thus of great practical value, so if one is interested in changing the wavelength rapidly (i.e. on a millisecond timescale), then one should really be able to change the slit widths with similar rapidity. We believe the Cairn Optoscan to be the first monochromator which allows you to do exactly that. Furthermore, the microprocessor option makes this facility extremely easy to use, since for each wavelength the optical bandwidth can be specified directly, and the microprocessor looks after the chore of calculating and setting the required mechanical slit widths. However, as for setting the wavelength, users who wish to provide their own electrical signals to set the slit widths will need to perform equivalent calculations, and this information is provided here. Page 16

20 Principles of Operation Referring back to Fig. 1, imagine for a moment that the input slit is arbitrarily narrow, so that the light incident on the grating is completely parallel. The angular dispersion of the grating in radians, dβ/dλ, is given by: dβ/dλ = kn10-6 /cosβ (5) We need to convert this to a linear dispersion at the position of the output slit, taking into account the effect of the intervening concave mirror. This mirror focusses an image of the input slit onto the output slit, although of course one will only observe a clear (i.e. nondispersed) image if the grating is illuminated by a monochromatic light source. The output slit must therefore be located at the principal focus of the mirror, and the linear dispersion dλ/dx, in nm per mm at the output slit position is: dλ/dx = 10 6 cosβ/knl o (6) where L o is the focal length of the mirror in mm. In the Optoscan this mirror is actually an offaxis parabolic mirror, for reasons that will be discussed later, but the effective value of L o in our configuration (and also of that of the input mirror, L i ) is 54.4mm. Since cosβ is wavelength-dependent, this gives rise to the need to vary the output slit width in order to maintain a constant optical bandwidth across the spectrum. However, it turns out that we need to vary the input slit width as well. So far we have only considered the case where the input slit is arbitrarily narrow, but as pointed out already, this is in practice a useless condition because then arbitrarily little light can enter the instrument. We therefore now have to consider the practical case of a significant input slit width. As we shall see, the best compromise is generally to arrange that the image formed by the input slit on the output slit (for monochromatic light of the appropriate wavelength!) is the same width as the output slit itself. This gives a spectral response that declines linearly (to a reasonable approximation, which ignores the wavelength-dependence of cosβ over this range) on either side of the centre wavelength, which of course corresponds to the wavelength at which the image of the input slit exactly aligns with the output slit. What may not be immediately obvious is that the grating has an anamorphic effect on the input slit image. Unlike a simple mirror, the incident and (in this case first-order) diffracted beams are at different angles to the grating normal, so they are of different widths (but still the same height), and this affects the width of the input slit image. Specifically, the magnification factor M is given by: M = (L o cosα)/(l i cosβ) (7) This just reduces to (cosα)/(cosβ) in our instrument since L o = L i. The only other necessary parameter is the relation between the input control voltage and the mechanical width of the two slits, and for both slits it is 1.592V per mm. (This somewhat strange number arises because the operating mechanism actually works by an angular rotation, and for our calibration purposes it is more convenient to have a simple relationship between the input voltage and the rotation angle, which is 0.5V per degree). The setup procedure is thus first to calculate the grating angle α (and hence also β) for the required centre wavelength using equation (4), then to calculate the widths of the two slits using equations (6) and (7). Our microprocessor control unit uses a 12-bit DAC with a unipolar output between 0 and +5V for driving each of the three inputs, and there is no advantage to be gained from using either a wider range or a higher resolution. In practice, there is no need to perform all these calculations every time. Our microprocessor software uses lookup tables for greater simplicity and speed, and copies of these are given in the appendix. Sufficient accuracy is obtained by having table entries at 20nm intervals, with intervening values calculated by linear interpolation. For users who are merely curious as to what effect the wavelength has on the slit widths in practice, there is no need to work through these equations. The Optoscan microprocessor software Page 17

21 Principles of Operation can display the actual slit widths that it has set to obtain the requested optical bandwidth for each of the programmed wavelengths. This option also allows the user to set the most appropriate slit widths for the lower-cost version of the Optoscan, where the slit widths are set manually by calibrated micrometer heads. Before leaving this section, it may also be useful to consider the effect of making the slits of UNequal widths. Since the input and output slits are necessarily controlled independently, there is no reason why they cannot deliberately be made different. However, the question is whether this is ever likely to be of any advantage. Although for most applications the most appropriate condition will be to make them equal, there are nevertheless some circumstances in which it may be useful to have unequal slit widths. Fig. 4 shows what happens in this situation. First, consider the simple case in Fig. 4a, where the two slits are equal. As already explained, the bandpass response is triangular under these conditions. For light at the centre wavelength, the image of the input slit is exactly superimposed on the output slit, giving a maximum intensity at this wavelength. For light either side of the centre wavelength, the image of the input slit moves across the output slit, giving a (to a good approximation) linear decline in intensity with wavelength. The bandwidth of an optical system is conventionally measured between the half-maximum points, and it should be readily apparent that the bandwidth defined in this way corresponds to the values obtained using the foregoing equations, whereas the TOTAL bandwidth, i.e. the length of the base of the triangle, is twice as great. Figure 3.4 : Effect of varying slit widths on bandwidth Page 18

22 Principles of Operation As the width of both slits increases, the triangle becomes both wider and higher, as shown in Fig. 4b. Since the area of the triangle represents the total light output, this graphically illustrates the squarelaw relation between output intensity and bandwidth, which holds as long as the input slit is fully illuminated by the light source. (For reference, this relation is followed exactly for slit widths up to about 20nm in the Optoscan, and the departure from square-law only begins to become significant for slit widths above about 30nm.) But now imagine that we begin to reduce the width of one or other of the slits. For the sake of argument we shall assume that this is the input slit, but in fact the effect is the same for either slit, as a consideration of the other case would show. The image of the input slit can now move some way either side of the centre of the output slit while still falling entirely within it, so instead of a single peak there is now a plateau to the bandpass characteristic, converting it from a triangular to a trapezoidal shape as shown in Fig 4c. As the input slit width tends towards zero, the length of the plateau region approaches the entire width of the output slit. For intermediate conditions, it corresponds to the DIFFERENCE between the two slit widths. Similarly, the total bandwidth, i.e. the length of the base of the trapezoid, is reduced as the input slit width is reduced, being given by the SUM of the two slit widths. Thus we have w peak = abs(w o w i ) (i.e. always positive!) (8) w tot = w o + w i (9) w 1/2max = [wo + w i abs(w o w i )]/2 (10) Effectively this means that as the width of one or other slit is reduced, the bandpass response becomes more nearly square, which may well be desirable, but this is achieved at the expense of a loss in light output, which may well not be. The bandwidth between half-maximum points is now a more cumbersome expression, which as equation (10) shows, is now equal to half the sum of the slit widths minus half the difference. For fluorescence excitation, where one generally needs as much light as possible, our recommendation is always to use equal slit widths, so there is actually no need to bother with any of these complications. On the other hand, this is a totally general-purpose instrument, and in other applications the ability to control the shape of the bandpass response in this way could be quite useful. Although the above relationships are relatively simple, it is clearly useful to have some way of displaying the actual bandpass characteristic for a given slit width combination, and our microprocessor system can also do that. This is valuable even in the simple case of equal slit widths, since the graphical display clearly shows the important distinction between the total bandwidth and the bandwidth measured between the half-maximum wavelengths. 3.3 Speed, optical aberrations and optical throughput This section is provided primarily for reference, since apart from correctly setting up the light source, there are no user adjustments to be made. It is intended to provide general background information about monochromator design, and, for users who are sufficiently interested to open the box and see what is inside, it will explain why the design is the way it is. The traditional design goal for a monochromator is, for obvious reasons, to provide high spectral resolution. Specialist instruments are capable of providing very narrowband outputs, capable of giving spectral resolutions of 0.1nm or better, but for general laboratory use a resolution of around 1nm is usually more than adequate. For fluorescence excitation the demands are even less great in this respect, as the main requirement is to provide excitation light within a bandwidth of 10nm or possibly even more. However, here the key requirements are for high optical throughput and rapid wavelength- Page 19

23 Principles of Operation changing. Optimising the design in these respects can have adverse effects on the achievable spectral resolution, as explained below. This may not matter too much for fluorescence applications, but on the other hand our variable slit system also permits the easy selection of much narrower bandwidths, where such shortcomings would be much more obvious. We have therefore taken additional precautions with the design of the Optoscan, to ensure that it can provide the necessary brightness and speed for fluorescence applications, while maintaining 1nm resolution. We ll first consider the topic of high optical throughput. As already discussed, we need to get as much light through the input slit as possible, and whatever else we may do, it clearly helps to begin with a bright light source. We have for a long time appreciated the superior light-collecting abilities of ellipsoidal mirror light sources over condenser light sources, and we have made our own for some years now. Our original design, which we now call the "classic" light source, gives good results with the monochromator under most operating conditions, although it was not specifically designed for this purpose. For applications such as fast imaging, where the very highest achievable light intensity is required, we subsequently introduced the Cairn Optosource, the design of which has been optimised for the monochromator, although it is just as suitable as the classic source for other applications. The classic source takes 75W lamps, whereas the Optosource can be supplied in either 75W or 150W versions. For the reasons described below, the 75W version tends to give somewhat brighter illumination under most conditions, and is therefore our generally recommended choice, although the difference is not very great. Although both light sources use ellipsoidal mirrors, the configuration is somewhat different. In the classic light source, the lamp is aligned along the optical axis of the mirror, so that the mirror collects light radially in all directions around the arc, converging it towards the second focus of the mirror as shown in Fig. 5a. Compared with the commonly used alternative of a condenser lens system, this arrangement collects several times more light, even if the condenser system is augmented by a spherical mirror, as shown in Fig. 5b. Placing such a mirror behind the lamp, to produce a focussed image of the arc next to the arc itself, theoretically doubles the amount of light collected, although in practice the improvement is rarely more than 50% because of the less favourable optical pathway (for example, the reflected component has to pass through the lamp envelope three times rather than just once), so this may not always be worthwhile. When connecting light sources to other equipment, it is important to match their f numbers, as we explain further below. The f number of a converging beam is given by the ratio of its diameter at any point to its distance from the focal point (of course, many optical systems are designed to operate with collimated light instead, but in this case the beam diameters have to be matched). The comparisons made here are therefore all with reference to the same f number for the output beam, which is f/2 for best matching to our monochromator. Within this constraint, we need to collect as much light as possible. For a given output f number, the more light we can collect from the lamp, then the larger will be the size of the focussed spot. This might seem to favour the Fig. 5a arrangement, but in practice things aren't quite that simple. The reason is that the output beam in the ellipsoidal mirror configuration is deficient in light at angles close to the optical axis, as more careful examination of Fig. 5a should make clear. The lamp envelope and electrodes provide a significant obstruction to paraxial rays, giving the unfocussed beam a characteristic doughnut shape. The greater total light collection of the ellipsoidal configuration does indeed give a larger focussed spot size, but the point intensity WITHIN the spot is less on account of the obstruction. Therefore, for passing as much light as possible through a narrow aperture - such as the input slit of a monochromator - a condenser system will perform better if the spot size is still large enough to fill the aperture. However, to collect sufficient light it is usually necessary to use high-aperture optics, which in turn requires aspheric lenses, otherwise the focussed spot size will be substantially increased by spherical aberration. Unfortunately, we also need to collect UV light, which requires silica lenses, and aspheric silica optics are rare and expensive. Page 20

24 Principles of Operation For the Optosource we therefore took the somewhat different approach shown in Fig. 5c. It also uses an ellipsoidal mirror, but in this case the lamp is perpendicular to the optical axis, as in condenser systems. The only theoretical drawback of this configuration is that the lamp does still obstruct the output beam to some extent, but the effect is much less and can be further reduced by using a relatively larger mirror, located relatively further from the lamp, and in the Optosource the resulting light loss is only a few percent. And since there is only a single optical surface, compared with the multiple optical surfaces in a typical condenser system, the losses elsewhere are likely to be reduced, so in practice the Optosource may even perform a little better overall. This configuration can also have a very high optical aperture. The mirror in the Optosource collects light over an angle of +/-70 degrees, which is a numerical aperture (NA), given by the sine of this angle, of This is a very high figure by condenser standards, and gives the largest feasible size for the focussed output spot (trying to collect light over still larger angles would begin to run into problems caused by the lamp electrodes and envelope obstructing the light path once again). An alternative approach to achieving a larger focussed spot size for a given output f number is to use a more powerful arc lamp, for which the brightest region of the arc is larger, but it turns out that the point intensity of the arc is somewhat LESS in a more powerful lamp. In principle it is therefore better to use a lower-power lamp at a higher NA, and in practice we find that the 75W version of the Optosource does indeed perform somewhat better than the 150W one, although the difference is not great (quantification is actually quite difficult because of individual variation between lamps, but it seems to be 25-30% on average). On the other hand, the larger focussed spot size of 150W lamps does make the system somewhat more tolerant of focussing and alignment errors, and the lamps last longer too, so we decided that it was well worth offering a 150W version as an alternative. Fig. 5d shows the basic design of the monochromator in an extended form for clarity, i.e the two focussing mirrors have been replaced by lenses of equivalent f number. It should immediately be clear why the f number of the light source should be matched to that of the monochromator. If, for example, the monochromator had a smaller aperture than the light source, e.g. f/4, then light entering the input slit at the more extreme angles would be unable to pass through the subsequent monochromator optics. We could prevent this from occurring by converting the light source to operate at f/4 as well, which could be done either by using a different ellipsoidal mirror or by using appropriate intermediate optics. Now all the light that enters the input slit can pass through the subsequent monochromator optics. However, things are not necessarily that simple, since reducing the output aperture of the light source, i.e. increasing the distance to the point of focus, also has the unavoidable effect of magnifying the image of the arc, i.e. its intensity at all points will be less. If the arc image was already overfilling the input slit, then less light is now entering the instrument, so unfortunately we are back where we started. However, this may still be a preferable condition, since the square-law relation between slit width and light output (when BOTH slits are varied) now holds up to greater slit widths than before. There is clearly quite a delicate balancing act here, since under normal operating conditions say 10-20nm bandwidth for fluorescence excitation we want to overfill the input slit by a reasonable amount (so that it is relatively evenly illuminated) but not by so much that we are losing light unnecessarily. Page 21

25 Principles of Operation Figure 3.5 : Basic monochromator configuration The aperture of a monochromator is determined by its optical design, and is usually within the range f/4 to f/10. A smaller aperture (i.e. approaching f/10) has the advantage for precision applications that optical aberrations of the monochromator are smaller, so its resolution is greater. However, this requires the concave mirrors to be of greater focal length, so the instrument must be bigger, which is not so convenient for a practical laboratory instrument. From equation (6), increasing the focal length of the output mirror increases the linear dispersion, so the slits must be wider for a given optical bandwidth. This is as it should be, because when we reduce the aperture of the light source to match the monochromator, the size of the arc image at the input slit is also correspondingly greater, so we can achieve the same optical throughput by making the slits correspondingly longer as well. However, if the slits are longer, then the dimensions of the mirrors and grating must also be increased in order to capture light from the top and bottom of the input slit as well as from the centre. Therefore the aperture of the mirrors must also be increased to some extent, until one approaches the limiting (and useless) situation where one is just scaling up the physical size of the instrument instead of reducing the aperture. The only way out of this loop is to accept a reduced throughput as the cost of reducing the aberrations, by limiting the slit height. A larger aperture is desirable for the Optoscan, since as the above discussion shows, this allows the slits and grating to be smaller in size. The advantages of that in terms of operating speed are enormous. Both the slits and the grating operate via a rotary motion, and the energy required for rotary acceleration is determined by the moment of inertia, which for a mass m at a distance r from the pivot is given by mr 2. If we increase the physical dimensions, then not only will r increase with the square but m will increase with the cube, so the overall relation is a fifth-power one. For a constant accelerating force, the time required to travel a given distance varies only with the square root of the inertia, but this still leaves us with a 2.5 th power dependence of time with size. In other words, the faster the optics, the faster the mechanics can be too. However, there is a countervailing problem on the optical side, because optical aberrations increase significantly as the aperture of the system is increased. Needless to say, this is an enormously Page 22

26 Principles of Operation complicated topic, but it is nevertheless possible to make some useful generalisations. For ease of manufacture, most optical components (be they lenses or mirrors) have spherical surfaces, but this is not an ideal situation, and the departure from the ideal introduces aberrations. The presence of aberrations means that light rays from a single point on the object are not all brought to a focus at the same point on the image, i.e. the image is blurred. In other words, the optics are operating in a somewhat nonlinear way, and although an exact expression of the nonlinearity takes the form of a power series, one can often make a reasonable approximation by considering only the first nonlinear term, which is a third-order one. Using this approximation, it is possible to analyse the aberrations algebraically in terms of spherical aberration, astigmatism and coma. (Higher-order terms cannot be analysed in this way, and one has to resort to numerical techniques, i.e. computerised ray-tracing programs, in order to take them into account). For our application, spherical aberration is the biggest problem. This is proportional to the cube of the aperture, and has the effect that a parallel on-axis beam is not focussed to a sharp point. Astigmatism occurs when the incoming beam is at an angle to the optical axis, and has the effect that the vertical and horizontal features of an image are brought to a focus at somewhat different distances, i.e. they cannot both be in sharp focus at the same time. The effect of coma is that point images formed away from the optical axis tend to have a somewhat elongated teardrop shape. Both coma and astigmatism increase only with the square of the aperture, and in a symmetrical optical system - such as the Czerny-Turner configuration - the coma tends to cancel anyway, although to be pedantic about it the anamorphic effect of the grating means that the cancellation can only be made complete at one wavelength. Astigmatism is more of a problem, because it increases significantly with the off-axis angle, and practical considerations usually favour a relatively large angle, as Fig. 1 should make clear. However, since we are dealing with slit images, the practical effects of astigmatism are greatly reduced, because it makes little difference if the input slit image is out of focus in the vertical direction. The spherical aberration problem tends to limit the maximum aperture of a spherical-mirror system to about f/4, where it limits the resolution to around 1nm or so, which is still OK. However at f/2 the aberration would be eight time worse, which is hopeless. Fortunately there is a solution to this, which is to use parabolic mirrors instead. This removes spherical aberrations completely, and the other aberrations caused by using the mirrors off-axis can be removed by using an appropriately off-axis SECTION of a parabolic surface instead, i.e. the image is brought to a focus at a point corresponding to the principal focus of the parent parabolic surface, of which the mirror forms a part. Using these components we obtain images of a sharpness that appears magical in comparison with those obtained using spherical mirrors. The actual aperture, which is to some extent determined by the availability of appropriate mirrors, works out to be f/2.2, which is a near perfect match to our f/2.5 light source, and in any case we are starting to hit other limits that preclude increasing the aperture much further. The problem here is that as well as aberrations, optical systems can have other shortcomings, i.e. various image distortions, and this one is no exception. Although our configuration gives an extremely sharp image of the input slit at the output slit, this image is somewhat curved. We estimate that this reduces the resolution approximately by the square of the slit height, but it still allows us to reach our 1nm resolution target. Of course, the problem could be removed by curving the output slit to match, but in practice that would not be easy, so it is far better to minimise it by other means. This involves using parabolic mirrors designed to work at a relatively small off-axis angle, and in practice it is also necessary for them to be oriented in the same sense. In the basic Czerny-Turner configuration, the input and output slits are on opposite sides of the light beams going to and from the grating as shown in Fig. 1, but to comply with the above requirement we need to move the input slit to the other side of the parallel input beam to the grating. In such a position it would obstruct the diffracted beam, but we prevent that by placing a plane mirror in the optical pathway between the input slit and the first parabolic mirror, which allows us to fold the input light pathway out of the way of the output one. There s no point explaining this any further, since interested users (and competitors too, no doubt ) can remove the lid and see the arrangement for themselves. Page 23

27 Principles of Operation 3.4 Choice of grating This is one area where a fair amount of independent choice can be made. Here we briefly describe the influence that the choice of grating has on the performance of any monochromator, and we explain the reason for our choice of the standard grating we supply. However, we must again emphasise that we can supply a WIDE variety of alternative gratings to special order. We also warn about possible shortcomings that are inherent in diffraction grating technology, and advise how to deal with them if they cause problems in practice (although they may well not). We have already discussed the importance of optimising the grating for highest efficiency at a particular wavelength, so that should be self-explanatory. Less clear is the choice of n, the number of lines per mm on the grating. Increasing n increases the dispersion, as shown in equations (1) to (4), but it also increases the angle by which the grating must be rotated in order to achieve a given wavelength change. Ideally one is likely to favour a high dispersion, because this allows the slits to be wider for a given optical bandwidth. Under conditions in which the light source is overfilling the input slit, this will allow proportionately more light into the instrument, and hence proportionately more throughput for the same bandwidth. As described previously, with our standard 1200 line grating this condition is met very closely for bandwidths of up to 20nm. Going to an 1800 line grating would reduce that bandwidth to 13nm, so in practice there would be an improvement in throughput for bandwidths in this range. However, we would not anticipate any useful improvement when the slits are opened widely say to 25nm or so and therefore where this is possible an 1800 line grating would be of no real advantage. Another factor to consider is that the Optoscan delivers huge quantities of light anyway, so the possibility of higher throughput may not be so useful in practice. The 1200 line grating has two other advantages. First, since it needs to rotate by a smaller angle for a given wavelength change, it will reach a new wavelength more quickly. Second, it can give wavelengths over the entire optical spectrum and beyond ( nm in our case) without difficulty. With an 1800 line grating, the angles α and β both become uncomfortably large at far-red wavelengths, and in practice we would not recommend using such a grating for wavelengths longer than about 600nm. Nevertheless, an 1800 line grating remains a very viable alternative for many applications, and in anticipation of some users preferring this choice, the software algorithms in the monochromator control box are already present in both 1200 line and 1800 line versions. On the other hand, we see little benefit in using gratings with fewer than 1200 lines. The next standard grating specification is 600 lines per mm, for which the dispersion is really becoming rather too small. However, such gratings could also be supplied if they do indeed turn out to be an appropriate choice for some particular application. Unfortunately, diffraction gratings are not perfect components. Compared with optical filters, their blocking of out-of-band light tends to be rather less, and there are also problems of breakthrough from higher-order spectra. In practice these problems can be controlled if necessary by use of appropriate additional filtering, but it is nevertheless important to be aware of them. One source of unwanted light is scattered light from the grating surface, and some (albeit small) proportion of this will inevitably reach the output slit. With reference to Fig. 2, one can appreciate that in practice the sawteeth are not going to be perfectly sharp, so light incident on the very tips of them may be reflected at an effectively random angle. This problem can be reduced by ensuring that the teeth are as sharp as possible, of course. However, an alternative solution is to ensure that the tooth profile, instead of being perfectly sharp, is instead some form of curved profile which is very constant for all the teeth, and such gratings are termed holographic on account of the procedures that are used to make them. However, holographic gratings are not necessarily any better in practice than good-quality ruled gratings, and in any case the presence of other scattering elements on the grating surface (of which dust and fingerprints are particularly good examples) can make this point academic anyway. Page 24

28 Principles of Operation The other source is inherent in how gratings actually work. As explained previously, light of a given wavelength is diffracted at any angle where the path difference between adjacent grating lines is equal to an exact number of wavelengths. As also explained previously, the grating profile can be optimised so that most diffracted light is first order (i.e. the path difference is one wavelength), but higher-order diffractions will nevertheless still occur. Unfortunately, a one-wavelength path difference at any given wavelength will correspond to a two-wavelength difference for light of half that wavelength, and so on. In practice though, this only tends to complicate matters at the red end of the spectrum, since there is unlikely to be a significant amount of light available at wavelengths shorter than about 300nm. A two-wavelength difference at 300nm corresponds to a one-wavelength difference at 600nm, so wavelengths longer than 600nm will be contaminated by second-order spectra. In practice though, the contamination may not be too serious, and in any case it can easily be dealt with by fitting a long-pass filter in the light path when making measurements at those wavelengths. Of the two problems, the light-scattering one is likely to have a far more significant impact on fluorescence measurements. The reason is that the fluorescence emission light intensity will always be very much less than the excitation light intensity, so the effect of the excitation light containing scattered components within the emission wavelength range will be correspondingly more serious. However, other components in the optical system particularly the dichroic mirror in the microscope can and do reduce this problem by rejecting much of the scattered light in the emission wavelength range, BEFORE it reaches the sample (afterwards is too late of course!). Our own experience with the Optoscan in a laboratory environment failed to detect any such problems in practice, but if there ever are any, then they should easily be dealt with by including appropriate additional filtering in the excitation pathway. The important general point to be aware of is that monochromators can never be as good in this respect as optical filters, and where better performance is required, it is actually common laboratory practice to use TWO monochromators in series, which reduces the scattered light problem to negligible proportions. Such drastic measures are not necessary here, however. Instead, one merely needs to bear in mind that under SOME circumstances, some additional excitation filtering may be useful. It is easy enough to check this, since if there is a problem, the emission photodetector will give a signal that remains present in the absence of any fluorescent sample. One final point to bear in mind is the lifetime of the grating itself. Optical filters tend to degrade with time when exposed to high levels of UV, as typically occurs in fluorescence applications. In our experience this is sufficient to warrant their replacement after maybe a couple of years of reasonably frequent use. However, from our experience with the monochromator over a similar period, there does not seem to be a comparable problem, and no such cases have been reported to us so far. In any case, it would not be a matter of any particular concern, as the cost of our gratings is similar to that of a couple of filters. What may well help here is that the thermal design of the Optoscan is such that the temperature of the grating itself never rises more than a few degrees above ambient. This is in contrast to some other monochromator designs, where our spies tell us that the grating may be heated substantially by its proximity to other components, so the risk of its deterioration may be greater. Page 25

29 Control Modules 4 Use of signal control modules with the Optoscan 4.1 General Notes Yet another of the many features of the microprocessor control unit is that if it is supplied in a suitable electronics rack (either single-height or double-height), it can also control our established range of photometry modules. These were originally developed for fluorescence photometry with our spinning rotor system, where they can extract the signals obtained from each of the individual optical filters in the wheel, but they also included a much simpler form of operation that could be used when only a single excitation wavelength was required. However, modules supplied from 1992 onwards support an additional mode of operation which is particularly appropriate for use with asynchronous wavelength-changing systems such as the Optoscan. Since the subject of optical signal detection is a particularly important one, it is worth discussing in some detail in order to explain why our modules work in the particular ways that we have chosen. The simplest form of photometric detection is for the detector (usually a photomultiplier, but possibly a photodiode in some applications), which generates a signal-dependent current, to drive a current-to-voltage converter. This provides an amplified and low-impedance output signal for connection to subsequent signal-processing circuitry. The converter circuit is normally combined with a low-pass filter - which may be variable as on our PMT amplifier - in order to reduce the highfrequency noise on the signal. It is important to match the filter frequency to the subsequent data acquisition frequency, for the following reason. If (as is usually the case) the acquisition system makes each measurement at some effectively instantaneous point in time, as opposed to averaging it over a finite period, then successive samples of a noisy signal may differ substantially from each other, and in extreme cases one can just be measuring the noise spikes. Low-pass filtering reduces the instantaneous fluctuations, so that the samples increasingly represent signal rather than noise. On the other hand, excessive filtering has two disadvantages. First, the signal can now change only slowly compared with the sampling frequency, so this means that there is a lot of redundancy in the recorded file, as successive samples will always be so similar. However, this is not necessarily a problem, whereas sampling at too low a rate certainly is. The second disadvantage is more serious, since the filtering may be removing useful information as well as noise from the data. In a system with a single excitation wavelength, the effect is clear to see, and as long as the bandwidth and sample rate are both set to "reasonable" values, we see little advantage in doing things any other way. As a broad recommendation, single pole filtering, with a sample frequency at least 10 times the filter frequency, gives perfectly satisfactory results. We call this the "averaging" mode of the input amplifier. When this is likely to be the preferred operating mode, we may instead supply the input amplifier with a wider front panel, to make room for two additional controls (although the circuit board is identical). In this form we call it the PMT amplifier, the additional controls being the filter frequency (this is set by jumper links on the pcb when no front panel control is provided), and a variable DC offset control. For multiple-excitation applications, an offset control in this location is not appropriate, since different offsets may be required at different wavelengths, for which our gain/offset module should be used. However, averaging mode is NOT suitable for multiple-excitation applications, since the filtering would tend to smear out the signals for the different excitation wavelengths, causing them to "bleed" into each other. Under these circumstances, signal detection by integration is greatly preferable. Page 26

30 Control Modules Integration is effectively the analogue equivalent of photon counting - and in our opinion is a better choice than photon counting for this application, although we can also supply photon counting modules to users who prefer that method. In the integration method, signal currents charge a capacitor, so the change in capacitor voltage during a given sample period represents the integral of the current flow during that period. Some provision must be made to discharge the capacitor from time to time, and the most straightforward method is to do this at the start of each integration period, so the voltage at the end directly represents the integral for that period. An integrator is also a low-pass filter, albeit a rather special one as its corner frequency is zero, but it has the important advantage that successive integrals are completely independent. Therefore if integrator resets exactly coincide with excitation wavelength changes - which is very easy to arrange - then there can be no "bleed" between the individual wavelength signals, as there would be in averaging mode. Of course, one could argue that it is difficult to arrange for such exact coincidence, but in practice there is no need to do so. Instead one can disconnect the signal input to the integrator for a short while, to remove any timing uncertainty. This is particularly important when using a monochromator such as the Optoscan, since the grating must slew between successive excitation wavelengths, which even with our machine is not an instantaneous process. We have therefore provided a control signal from the Optoscan, which interrupts the integration while the grating is slewing. Only when the grating is at rest at its new position can acquisition of the next integral begin. Another possible question is how often the data acquisition system must sample the integrator output during the sampling interval. Fortunately the answer is only once! At the end of the integration period - which occurs automatically as soon as the grating is instructed to move to a new position - the capacitor voltage is copied onto a sample-and-hold amplifier just before the capacitor is discharged. The amplifier voltage can then be recorded by the data acquisition system at any convenient time before the next integral is acquired. Furthermore, a separate sample-and-hold amplifier can be used for each of the different excitation wavelengths, even though all the wavelengths are being acquired (in sequence) by the same integrator. This is what our output module does, so all the outputs can be acquired as a single operation, only once per excitation wavelength cycle. In wavelength scanning mode the situation is slightly different, since there may be many wavelength samples during the scan, and they all need to be acquired on the same data acquisition channel in sequence. In this case the output amplifier is no longer relevant, and the data can be acquired directly from the output of the input amplifier itself. The input amplifier also has its own sample-and-hold amplifier, so again one needs to sample only once for each scan wavelength, although in practice it may be easier to make multiple samples, especially if other data are being acquired at the same time at a higher rate. To activate the sample-and-hold amplifier, jumper J6 should be in the HOLD position, and monochromator systems will normally be supplied with the input amplifier(s) configured in this way. However, if preferred, J6 can be moved to the CONT (continuous) position, so that the integrator output can be observed directly (it will of course take the form of a series of ramps). In this case the sampled - i.e. final - integrator values for each scan wavelength can be read of from output 1 of the output amplifier. The mode of input amplifier operation that we have been describing here is called, not surprisingly, integration mode. One could also reasonably ask whether it offers any advantages over averaging mode when only a single excitation wavelength is used, and the answer is that it basically does, since there is no need to "oversample" the data, as there is in averaging mode. Instead, each integral needs to be acquired only once. The downside is that one must provide an appropriate control frequency to drive the integrator charge-sample-discharge cycle. In a simple fluorescence measurement system it may not be worth the trouble to provide such a signal, but the microprocessor control unit can easily do so. Software versions from 1.3 support this possibility, and also allow the control Page 27

31 Control Modules frequency to be varied. There is also the practical advantage that there is no need to change the operating mode of the input amplifier (set by J7) when changing from multiple-excitation to singleexcitation operation. The question of what control frequency to use is an interesting and important one. In principle, the control frequency for single-excitation conditions could be quite high, as there is no need to allow any tie for wavelength changing. When used with our spinning rotor system, the input amplifier works in just such a way, although that is not immediately obvious to the user, since it is implemented in a rather subtle way. The preferred method for the monochromator is also by integration, but in this case it IS obvious, since here the implementation is more conventional. What is the reason for this difference? Well, the rotor will normally be spinning continuously, so each of the (usually six) filters will be in the light path for the same time. However, the chosen rotor speed may vary substantially according to the required time resolution. 4.2 PMT supply module The output voltage can be varied continuously from 0 to approximately 1500V by the multiturn control on the front. The output voltage is displayed on the built-in meter. The unit is single-polarity, and the negative output version is standard (A positive output version could be supplied if required, but nobody has ever asked for one). Use of a negative supply is greatly to be preferred, since in this configuration the photomultiplier cathode is at the negative supply potential, and the anode is at or near ground potential. The anode is best connected to a virtual ground, and this method of operation can be selected for both our input amplifier and PMT amplifier modules. The output of the supply appears on a high-voltage MHV (out) socket, which is superficially similar to a BNC connector, but these two connector types are NOT compatible. A matching cable for connection to the photomultiplier is included with the unit, and further plugs, sockets and cables are available from us if required. We would recommend leaving the power supply in the off position when not in use to avoid accidental damage to the equipment through over exposure to light. When setting up, or performing a new series of experiments, we suggest turning the voltage potentiometer to a low level before switching on. For biological fluorescence measurements a value of between 700 to 1000 volts is usually best. When viewing paper during set-up procedures a much lower value (<600V) is often sufficient Input amplifier and PMT amplifier The input amplifier can be configured to measure positive- or negativegoing signals according to the setting of the jumper J1. This jumper sets the polarity of the first stage so that the output is always positive-going. The input impedance in either of these configurations is 1 megohm. J1 can also be used to select a virtual ground input stage configuration, in which the input stage acts as a current-to-voltage converter. This mode is intended for direct interfacing with a photomultiplier tube. The photomultiplier power supply should be negative with respect to ground, and the tube cathode connected to the negative supply. The tube anode is connected directly to the amplifier input on the front panel. In this mode the anode current is converted into a positive-going voltage in the input stage. The gain of the input stage is set by J2. In either the positive or negative voltage mode configurations, the gain is either x1 or x10, and in virtual ground mode the gain is either 0.1V or Page 28

32 Control Modules 1.0V per microamp. This signal is coupled to the next stage via the front panel potentiometer, allowing the effective gain to be varied continuously down to zero from these limits. The amplifier stage following the potentiometer has a gain of x1, x10 or x100, according to the setting of J3. This gives a maximum overall voltage gain of x1000 or, 100v per microamp in virtual ground mode, which should be more than sufficient for all applications. The module is normally supplied with J1 set to virtual ground mode, and J2 and J3 both set to x10. A further feature is that the input signal is briefly sampled by an additional amplifier during the dark period between each filter. This signal is filtered to average it over a number of rotor revolutions, to provide a reference against which the integration is performed. The input module thereby compensates for photomultiplier dark current or for any steady background illumination component. In practice this is very useful, as it greatly reduces the interference from room lighting. However, the room lighting will also contain components at the mains frequency and multiples thereof, which can still cause some interference (as explained previously). During system testing, it may be useful to disable the background subtraction, by moving jumper J5 to the off position. This allows the response of the system to a steady light signal to be evaluated (normally, a steady light signal would give a zero output because of the background subtraction). The integrated output voltage is automatically reset by a control pulse on the system backplane. Other control lines on the backplane ensure that the integrated output from each filter is stored in the appropriate sample-and-hold amplifier in the output module, just before the integrator is reset. The (out)put is also available on a BNC socket on the front panel, and when the system is functioning correctly, it will consist of a series of sigmoid curves corresponding to the integral of the signal from each optical filter in turn, with an abrupt reset to zero between each integrated signal. For testing, and/or as an alternative in normal operation, the BNC socket can instead be connected to the output of the preamplifier section, so that the (unintegrated) amplified input signal can be observed. This is done by plugging the lead from the BNC socket onto the SK2 pins instead of SK3. When the rotor is stopped, there are three possible operating modes, which are set by jumper J7. The module switches to the chosen mode automatically when the rotor is stopped, on receipt of a control signal on one of the lines on the system backplane. For use in simplified configurations (e.g. dual-emission only systems) in which this control signal may not be present, jumper J8 can be switched to simulate the permanent receipt of the control signal, thereby enforcing the selection of the chosen stopped mode. The stopped modes are intended to cover the situations where the system repetitively samples the same optical wavelength, or switches discontinuously between different optical wavelengths (in which case the different wavelengths may be sampled for different periods), or some combination of these alternatives. The first of the three stopped modes is the same pulsed integration system as described above, and is intended for sampling periods similar to those used for the rotor, i.e. about 1-100msec. The second mode is a conventional integration system, which is intended for signal measurement over longer periods, between about 100msec-10sec, as may be appropriate for a discontinuously-operating filter changer instead of a continuously-spinning rotor. This mode may be particularly useful for compatibility with camera imaging experiments, where measurements are often made on a similar timescale in order to achieve acceptable signal-to-noise ratios. The gain of the continuous integrator has deliberately been made much less than that of the pulsed integrator, on account of the expected longer measurement periods, and if required it can be decreased by a further factor of ten by changing the position of jumper J9. In both the integrating stopped modes, the output normally has a sawtooth appearance, and continuous outputs can be obtained from the output module, just as when the rotor is spinning. However, if the amplifier is sequentially measuring the same optical wavelength (i.e. no filter changing is taking place), then it is possible to obtain a steady output directly from the input amplifier, by Page 29

33 Control Modules activating its optional sample-and-hold stage. In this condition, the output is held at the final value of each integration until the next integration has been completed. This mode of operation is selected by moving jumper J6 into the hold position. The third stopped mode gives the average value of the input signal, low-pass filtered with a corner frequency given by J4. The corner frequency range is 1, 10, 100 or 1000Hz, according to the setting of the first jumper link, and the actual frequency can be reduced to 0.3 of the selected value (giving 0.3, 3, 30 or 300Hz) by removing the second jumper link. This mode is intended for repeated measurements at a single filter wavelength, and it has the advantage that (unlike the other two stopped modes) no external control pulses of any kind are required, making it particularly suitable for dualemission systems. The remaining jumpers, J9 and J10, affect the output signal from the module. The spectrophotometer can support up to four input modules, and there are therefore four alternative output lines available on the system backplane. These lines are labelled A, B, C and D, and the signal connection is set by J9 (which connects the output signal to pin A19, C19, A18 or C18 of the module DIN connector). The output signal is also available on the output BNC socket on the module front panel. When only one input module is used, it is recommended that J9 should be set to send the output to the A line, and this is the normal factory configuration. Similarly, a second input amplifier would normally have its output connected to the B line. The other jumper, J10, allows subtraction of a signal present on one of the other output lines. This facility is intended for making certain differential measurements between two photodetectors operating at the same wavelength, but under normal conditions this jumper should be left in the off position to disable the subtraction. When a computer or tape recorder interface module is installed in the system, it can place its own data on any of the four output signal lines. At the same time, a control signal disconnects all the input amplifiers from the backplane. This allows either the computer or the tape recorder interface to replay data back into the system, which can then be processed exactly as if it were coming from the input amplifiers. One final point to note is that the output from this module forms the input to other modules. Therefore, the A B C and D signal lines referred to above are known elsewhere in the system as the A B C and D input channels. 4.4 Output module This module consists of eight independent sample-and-hold amplifiers. These amplifiers retain the maximum signal (final integrated value) produced by the input amplifier for each filter position. The system can support up to four independent input modules (A B C and D), although no more than two (usually A and B) would normally be required. The output module is normally configured to operate with either the A or the B input module (or some combination of the two), and two output modules can be used at the same time. There is also a provision to accept signals from the C and D input channels if required. The outputs from the sample-and-hold amplifiers are available on the front panel, and they can also be sent to other modules via signal lines on the system backplane. For each of the eight amplifiers, there are two such signal lines available, each designated A and B to match their intended source. There are thus sixteen output data lines altogether, labelled A1-A8 and B1-B8 respectively. However, some additional configurations are possible as described below. Associated with each sample-and-hold amplifier is a signal input, a control input and a signal output. The amplifier samples the input signal when the control input is high, and retains the latest Page 30

34 Control Modules input signal when the control signal goes low. The control signals are generated by circuits in the main system enclosure. The digital control board within the enclosure has a jumper link that can be set to produce control pulses appropriate for rotors with two, three, four, six or eight filters. The normal configuration is six filters, which leaves two spare amplifiers in the output module; however, these can still be used under some circumstances. Selection of the input signals for the output amplifier is normally made between the A and B input amplifiers; however, the signals from the C input amplifier (if fitted) can be selected in place of the A input amplifier by jumper bank J1, and a similar substitution of D for B can also be made. These two substitutions can be made independently, but they affect all the sample-and-hold amplifiers together. Each amplifier has a jumper link to select whether its input is connected to the A or B input signal (which may have been diverted to C or D as described above), and these are grouped together as jumper bank J2. For normal operation these jumpers should be connected to the A signal line, and this is the normal factory configuration. Jumper bank J3 independently connects the outputs of amplifiers 1-8 either to the A or B output signal line for that filter position (note that there are no corresponding C and D output lines on the system backplane). The operation of this jumper bank is exactly as for J2, and the normal setting is to the A lines. Jumper banks J4 and J5 provide additional output configurations which may be of use under some conditions. Normally the pairs of pins comprising J4 are not connected, but if the shorting links for positions 5-8 on J3 are removed and replaced on the corresponding links of J4, then the effect is to connect sample-and-hold amplifiers 5-8 to output lines B1-B4 respectively. Of course, this would normally place the outputs for filter positions 5-8 on these lines, but jumper bank J5 (sampling diversion) can be reset to make these sample-and-hold amplifiers accept the signals for filter positions 1-4 instead. This is intended to allow one output module to provide all the outputs for two photodetectors, when the rotor has four or fewer filter positions. In this case, the 1A-4A signals are available on output sockets 1-4 and are sent to the 1A-4A lines on the system backplane. The 1B-4B signals are available on output sockets 5-8 and are sent to the 1B-4B lines on the system backplane. Jumper bank J5 also allows decoding of four filter positions from four input amplifiers when two output modules are used. In this case, J5 on both output amplifiers would be set to divert the sampling of amplifiers 5-8 to filter positions 1-4. In a typical configuration, J1 of the first output module would be set in the normal position, whereas J1 of the second output module would be set to select the C and D inputs instead of A and B. Output sockets 1-4 of the first output module would then carry the four decoded A signals, and sockets 5-8 would then carry the four decoded B signals. These eight signals could also be made available on backplane lines A1-A8. Similarly, output sockets 1-4 of the second output module would carry the four decoded C signals, and sockets 5-8 would carry the four decoded D signals, with all these signals also available on backplane lines B1-B8. Many other configurations are also possible, so please contact us for advice if your requirements are not met by any of the examples discussed here. 4.5 Gain / offset module This module has four identical stages, each of which is associated with its own front panel potentiometer. Jumper links on the board allow each stage to be configured so that the potentiometer provides either a variable gain or a variable DC offset adjustment. Each stage can receive its input from any of the 16 output signal lines on the system backplane (A1-A8 and B1-B8) and it can place its output Page 31

35 Control Modules on any other of these lines. This arrangement takes advantage of the fact that in any one system it is most unlikely that all 16 signal lines would already be in use. The input and output signal selection is done by two jumpers fitted to the long rows of pins immediately below each amplifier stage on the board. One jumper is placed to link a pin in the central column to the adjacent left-hand pin to select the input line, and the other jumper is place to link another pin in the central column to its adjacent right-hand pin to select the output line. From the bottom upwards, the pin order is for data lines 1B, 1A, 2B, 2A, and so on; this arrangement is labelled on the board. There is also a facility to link the input of the second stage to the output of the first, by connecting a jumper between the pins labelled LKl-2, and a similar facility is provided for the third and fourth stages by the LK3-4 pins. In this mode, the jumper selecting the input signal line for stage 2 (or 4) should be removed, and the jumper selecting the output signal line for stage 1 (or 3) can optionally be removed. The purpose of this facility is to allow one stage to provide variable gain and another stage to provide variable offset on the same channel if required, without using backplane signal lines to interconnect the two stages. Although the variable gain option clearly has its uses, it is not essential for most applications. In contrast, the variable offset facility is an essential requirement for ratiometric fluorescence measurements, to allow subtraction of a signal equivalent to the autofluorescence background at that wavelength. The module is therefore normally supplied with all four stages configured for variable offset, using signal lines A1-A4 as inputs and B1-B4 as the corresponding outputs. If required, a second module can be configured to perform the same functions for filter positions 5-8. The choice of gain or offset configuration is made by the pair of jumpers at the top left of each amplifier stage. As labelled on the board, connecting these jumpers between each central pin and the pin just above it provides the variable gain mode, whereas connecting each central pin to the one just below it provides the variable offset mode (in which the module is normally supplied). Immediately below this pair of jumpers on each amplifier stage is another jumper that sets either the maximum gain or the maximum value of the DC reference voltage in the offset mode. The maximum gain, as labelled on the board, is either x1 with the jumper connected between the central pin and the upper pin, or x10 with the jumper connected between the central pin and the lower pin; the corresponding DC reference voltages in the offset mode are 1V and 10V. To the right of this jumper is another double jumper link, with alternative jumper positions analogous to those that set the gain or the offset mode of operation. This is labelled "normal" and "invert" for the upper and lower positions respectively, and it should usually be left in the normal position as supplied. If this pair of jumpers is moved to the invert position, then in offset mode the DC offset voltage will be added to rather than subtracted from the signal voltage, whereas in gain mode the signal voltage itself will be inverted. This facility is included to make the module a fully general-purpose one. It should also be noted that other configurations of each stage, in which pairs of filter signals can be added together or subtracted from each other, are possible. Further information on these possibilities is available on request, but please note that we would normally recommend use of our combiner module for such applications. 4.6 Ratio module This module provides either a linear or a logarithmic ratio of any two filter signals. The two input signals are selected by the edge switches at the top right of the module, labelled N (for numerator) and D (for denominator) respectively. Each switch has ten positions, numbered 0-9. As explained elsewhere, there are two alternative output signal lines on the system backplane, labelled A and B, for each of the eight possible filter positions. A jumper link on the board determines for each position of each switch whether the A or B signal line is selected when that switch position is selected. Switch positions 1-8 correspond to filter outputs Page 32