In order to achieve acceptable speed performance from their confocal system the Glasgow group had developed a prototype preamplifer, which they connected to the photomultiplier in their microscope, to form the basis of an independent data acquisition system. It integrated the photomultiplier signal for a series of brief periods, each of which corresponded to a single pixel in the line image, and the signals were recorded and analysed uisng a software package developed for this purpose by Dr. John Dempster of Stratchclyde University. This worked well, but there was some loss of data between pixels, during the periods when the integrator was being reset, and our assistance was requested in order to circumvent this.
We came up with a three-phase integration system, which has proved to be a very effective solution to the problem. At any one time, one of three integrators is acquiring the current pixel. At the start of the next pixel, a second integrator is selected to accumulate the signal for that one, while the first integrator remains at its final value and its output is sent to the data acquisition system. For the following pixel a third integrator is selected, the output of the second is now sent to the data acquisition system, and the first integrator is reset in preparation for the pixel after that. This configuration therefore gives a continuously available data output with no data loss, and it can operate at pixel clock speeds of up to at least 1MHz. Of course, the characteristics of the three integrators have to be closely matched, but this turned out not to be a problem.
Why use integration rather than just digitising the photomultiplier signal at fixed intervals? The problem here is to choose an appropriate bandwidth prior to data acquisition. If the bandwidth is too high, the acquisition system will be unduly affected by the noise on the optical signal, so the sampled values are no longer the best estimates of the average signal level for a given pixel. On the other hand, if the bandwidth is too low, the recorded value for a given pixel will be more greatly affected by those for the previous one(s), giving a smearing effect to the acquired images. The noise arises from statistical fluctuations in the arrival rates of the individual photons that constitute the overall optical signal, so it would appear that there is no ideal solution to this problem. However, an integrator gives a well-filtered signal for data acquisition, and the resetting between pixels prevents any smearing effects. In fact, to all intents and purposes (bar one small consideration which we leave to our more expert readers to point out) an integrator provides the analogue equivalent of photon counting, which some may argue is what everyone should be doing anyway. However, in practice the maximum count rate of the system has to be at least ten times higher than the average photon arrival rate in order not to miss a significant number (say 10%) of counts, which is likely to require compromises in other important areas. In our opinion (and experience), an integrating system using a high quantum efficiency photomultiplier is hard to beat.
As well as being suitable for use with current confocal systems, it can also give a new lease of life to older designs, especially where their own data acquisition is limited to 8-bit precision. We recommend a 12-bit system for our preamplifier, and John Dempster’s software (available free of charge from either him or us) supports this.
