5.5 Performance

The amplitude responses and group delay are shown in Figures 5.20 and 5.21 and indicate that they are very close to ideal. Figure 5.22 shows the passband of the filter with no feedthrough equalisation and this indicates the severity of high-frequency peaking due to non-reciprocal feedthrough capacitances and the effectiveness of the feedthrough equalisation technique.

Figure 5.20: Simulated amplitude response

Figure 5.21: Simulated passband response and group delay

Figure 5.22: Simulated passband response without feedthrough equalisation

The 1 dB compression point occurred with a differential output swing of 1.3 V peak to peak (at 1 MHz), but the maximum differential voltage swing was restricted to 1 V peak to peak, corresponding to the maximum wanted Bluetooth signal of − 20 dBm and required an input current swing of 40 μ A peak to peak. Figure 5.23 shows an output NPSD in the passband of aproximately 90 nV/ √ . The signal-to-noise ratio, found by comparing the power of the maximum signal with that of the noise integrated over 100 MHz bandwidth, was 68.2 dB.

Figure 5.23: Simulated output noise spectral density

For the IM3 test, sinusoids at 4 MHz and 7 MHz with amplitudes 2 dB below the maximum signal were applied and the resulting simulated output spectrum is shown in Figure 5.24. The third-order product at 1 MHz was at a level of − 86 dBV and this gives an IIP3 of 34.2 dBV.

Figure 5.24: Simulated third-order intermodulation

For the supply noise intermodulation test, first V _dd and then V _ss were modulated by a 1.5 MHz sinusoid while the input was driven with a 0.5 MHz sinusoid, each with amplitudes of 100 mV peak. These signals were both chosen to be in the filter's passband because this gave greatest intermodulation. The simulated output (nearly identical for either V _dd or V _ss excitation ) is shown in Figure 5.25 and indicates intermodulation products at 1 MHz and 2 MHz at 47.7 dB and 48.7 dB below the direct signal at 0.5 MHz. This attenuation of about 48 dB is maintained at lower input signal levels and is even greater for out-of- band supply noise.

Figure 5.25: Simulated power supply-signal intermodulation

The total current drain for the whole Gm-C filter was 512 μ A giving a power consumption of approximately 1 mW. The estimated chip area (including automatic tuning) is approximately 0.18 mm ² . The results with nominal processing, at temperatures of − 20 °C, 27 °C and 80 °C, are summarised in Table 5.2 and indicate good stability of performance.

With the transconductors deliberately skewed to emulate a 20 per cent mismatch between the transconductor's pMOS and nMOS transistors (the pMOS transistors were made 10 per cent wider and the nMOS transistors 10 per cent narrower), there were only minor changes to the typical performance as shown in Table 5.3. This justifies our claim that matching between the pMOS and nMOS transistor parameters is not critical.