Notes on digi­tal filter

The cri­ti­cal­ly dam­ped fil­ter is pre­set at the fac­to­ry for all GEMAC sen­sors with an adjus­ta­ble digi­tal fil­ter (sen­sors of the Refe­rence-, Clas­sic- and BasicLI­NE). The reason for this is the uni­ver­sa­li­ty of this fil­ter. It has a 4 times shorter reac­tion time or delay than the But­ter­worth fil­ter (see Fig. left). The But­ter­worth fil­ter, in turn, has a hig­her atte­nua­ti­on, even at fre­quen­ci­es that are slight­ly abo­ve the set cut-off frequency.

So when is which fil­ter to use?

Basi­cal­ly, users should first test with the default fil­ter set­tings, sin­ce a shorter respon­se time is usual­ly a major requi­re­ment. If the signal respon­se, i. e. the tilt signal, is “too noi­sy”, then the cut-off fre­quen­cy should first be redu­ced until the limit of the reac­tion time has been rea­ched. The respon­se time of the fil­ter respon­se for the cri­ti­cal­ly atte­nu­a­ted fil­ter in[s] is 1/​adjusted cut-off frequency[Hz]. If the respon­se time is not rele­vant, a cut-off fre­quen­cy of 0.1 Hz should be tes­ted, which repres­ents the lowest adjus­ta­ble cut-off frequency.
The But­ter­worth fil­ter should only be tes­ted and used in appli­ca­ti­ons whe­re no or only slight or rare input shocks are expec­ted and whe­re the sen­sor does NOT con­trol a loop, as the fil­ter cha­rac­te­ristic reacts to input jumps with over­shoo­ting in the signal respon­se. The But­ter­worth fil­ter is espe­ci­al­ly useful when the­re are very low-fre­quen­cy, inter­fe­ring vibra­ti­ons in the appli­ca­ti­on. In this case, the much stron­ger sepa­ra­ti­on bet­ween pass­band and cutoff fre­quen­cy ran­ge is used in a tar­ge­ted manner.