Signal filtering
Noise comes from a variety of sources. The simplest source to consider is one that generates noise equally at all frequencies; for example, thermal noise generated by a resistor has this characteristic. Noise sources, being random, add in an rms rather than a linear manner. This has many implications, amongst which is that for a source that generates the same amount of noise at all frequencies (i.e. white noise in the usual jargon), the total noise increases with the square root of the bandwidth rather than linearly. However, for reasons which are well described by Sigworth (1995) in the single channels recording book, the internal noise generated by a patch clamp is NOT independent of frequency. Instead there is a noise component which increases linearly with frequency, and which tends to be dominant at frequencies above a few KHz, so the dependence of the rms noise on the signal bandwidth is much greater. This tends to favour use of multipole filtering. Four-pole filtering is often chosen as the best compromise between performance and complexity, whereas purists tend to favour eight-pole filtering, but there is general agreement that there are no practical advantages in having more than eight poles. We have compromised by providing four-pole filtering in our main filter, which is optionally extendable to eight poles by an internal accessory board.
Multi-pole filters, however they are implemented, tend to be quite complicated. Compared with a simple amplifier stage, they tend to be more susceptible to various forms of interference pickup, and they also tend to generate more noise of their own. Therefore, in a general-purpose gain and filtering circuit, it is preferable to put the gain stage first, so that the filter acts on the amplified signal. However, if there is a lot of high-frequency noise on the signal, this configuration introduces another problem. The gain stage also amplifies the noise, causing the amplifier to saturate at a possibly much lower gain than if the noise were not present. Siting the filter first would have avoided this particular problem. We therefore provide an additional filter prior to the gain stage. This filter can be much simpler than the output filter, so there are accordingly fewer performance compromises, and accordingly we provide a two-pole Bessel filter with corner frequencies of 3KHz, 10KHz and 30KHz, plus an "out" position. It is referred to as the prefilter in view of its location in the signal pathway, and we recommend that it is set to a frequency that includes the range of interest, which can then be defined more precisely by the main filter that follows the gain stage.
The base frequencies of the multi-pole output filter are 1Hz, 10Hz, 100Hz, 1KHz and 10KHz, set by the freq range selector, and they can be varied over the range x1 to x10 by the freq value selector. (x1, x1.5, x2, x3, x5 x7 and x10). The lowest frequency ranges are not appropriate for patch clamping, but they may be appropriate for other applications. In particular, they may be useful for filtering the signals generated by the capacitance measurement facility, and a switch is provided to connect the capacitance signals directly to the filter.
