Fig. 2: Evaluation of system sensitivity.
a Boxplot of gm for nine randomly selected probes from three different wafers produced in independent batches. The yield in terms of gm above 70% of the median is indicated. b Boxplot for Vgs-rms measured in the 1–10 Hz frequency range, plotted for the same neural probes evaluated in part a. All probes consisting of 64 g-SGFETs. The boxes extend from the lower to the upper quartiles, with a line at the median. The whiskers extend 1.5 times the inter-quartile range and the data points beyond the whiskers are indicated by a dot. c Histogram of gm for the 64 g-SGFETs of probe #3 (labeled in panel a) and Gaussian fit of the histogram excluding the outliers shown in panel a. d Histogram of Vgs-rms for the 64 transistors in probe #3 (see panel b) and log-normal fit of the histogram excluding the outliers shown in panel b. e Equivalent circuit of the wireless headstage. f Power spectral density (PSD) of the noise from DC channels (black) and AC channels (orange) in probe #3. The 1/f dependence is represented by the solid red line. The vertical orange line indicates the hardware high-pass filter applied to AC channels at 0.15 Hz. The quantization noise of the DC and AC channels is indicated by the horizontal dashed red lines. g Representation of the Vgs-rms for all g-SGFETs in probe #3 shown for different bandwidths; 0.05-0.5 Hz band for the DC-channels (left), 1–10 Hz band (middle), and 20–200 Hz (right). The position of the g-SGFETs on the array connected to DC-channels of the headstage is indicated by the red squares. h Time domain representation of the noise spectra shown in part f and g (DC-channels filtered in the 0.05–0.5 Hz band and AC channels in the 20–200 Hz band). Signal from eight channels are overlapped.
