A ultra low power CMOS pA/V transconductor and its application to wavelet filters

Abstract

Two compact ultra low-power CMOS triode transconductor topologies denoted VLPT-g m and Delta-g m are proposed. In both circuits, input transistors are kept in the triode region to benefit from the lowest g m /I D ratio. This allows achieving a small-signal transconductance g m down to hundreds of pA/V, making such transconductors attractive for the synthesis of g m -C filters with cut-off frequencies in the range of Hz and sub-Hz. The g m value is adjusted by a well defined aspect-ratio (W/L) and drainsource voltage V DS , the latter a replica of the tuning voltage V TUNE imposed as drain-source voltage of input devices. VLPT-g m reaches a minimum g m of 1 nA/V, whereas Delta-g m exhibits a g m as low as 400 pA/V. Inputreferred noise spectral density is typically 12.33 lV/Hz 1/2 @ 1 Hz and 93.75 lV/Hz 1/2 @ 1 Hz for VLPT-g m and Delta-g m , respectively. In addition, setting their g m equal to 1 nA/V and arranging them as first-order lossy integrators, Delta-g m presents higher bandwidth with respect to VLPTg m . Cut-off frequencies are 1.33 kHz and 24 kHz for VLPT-g m and Delta-g m integrators, respectively. Finally, as an application example, both transconductors were used as building blocks to realize a 6th-order wavelet g m -C filter. For both approaches, THD was kept below 1% for signal swings up to 200 mV pp .The design complies with a 1.5 V supply and a 0.35 lm CMOS process and features an overall power consumption of 51 and 114 nW, respectively for VLPT-g m and Delta-g m filters.

Design of Gm-C wavelet filter for on-line epileptic EEG detection

“…Recently, WT has been applied to on-line spike detection in wearable ambulatory EEG (WAEEG) system for epilepsy diagnosis [7][8][9]. To realize on-line operation, analog implementation of WT has been investigated owing to the better performance in low power design required by battery powered device such as WAEEG [10][11][12][13][14][15][16][17][18][19][20][21][22]. So far, the popular method for implementing WT in analog domain mainly involves the rational approximation of wavelet base (i.e.…”