hardware//electronics//analog filter

An analog filter lets some frequencies of a signal through and weakens others before the signal is digitised, and every measurement chain uses one to keep the band it cares about, to shed noise outside it and to stop fast components from corrupting the samples. Frequency counts cycles per second, in hertz (1 Hz is one cycle a second).

An analog filter lets some frequencies of a signal through and weakens others before the signal is digitised, and every measurement chain uses one to keep the band it cares about, to shed noise outside it and to stop fast components from corrupting the samples. Frequency counts cycles per second, in hertz (1 Hz is one cycle a second).

Any signal can be seen as a sum of sine waves of different frequencies, which is Fourier's result, and a filter acts on that sum: a low-pass filter keeps what is below a cut-off, a high-pass filter what is above, a band-pass filter what lies between two. Built from resistors and capacitors (and amplifiers), it works because a capacitor's impedance changes with frequency, so a divider made with one passes some frequencies and shunts others.

The sharper an event, the higher the frequencies it is made of.

A spike lasting a millisecond rises in a fraction of that time, so its sum of sines reaches several kilohertz, and a filter that cut at 1 kHz would round it into something else.

One signal can be split into bands that mean different things. A neural recording keeps a slow band (below a few hundred hertz, the summed field of many cells) and a fast band (a few hundred hertz to about 10 kHz, the spikes), each filtered and sampled at its own rate.

The anti-aliasing filter comes before the converter. It removes everything above half the sampling rate, which the analog-to-digital converter could not represent and would fold down into false slow components.

Filtering is part of the signal-to-noise ratio. Noise spread over all frequencies is cut by every band the filter removes, so the narrowest band that still keeps the signal's shape is the cleanest.