1 Introduction

Recently, the SAR ADC has been frequently used for biomedical applications due to its digital feature. The capacitor digital-to-analog converter (DAC) dissipates a large portion of the total power. Thus, many capacitor switching schemes [17] have been demonstrated for better power efficiency. Compared to the conventional switching scheme, the monotonic switching technique [1] achieves 81.26% less switching energy. The charge-average switching scheme [2] obtains 93.5% saving. The Sanyal and Sun [3], the Tong [4] and the Xie [5] switching schemes achieve low power by introducing the minus energy from the forth bit cycle, but they still consume much power during the third bit cycle. The bidirectional single-side (BBS) switching technique [6] generates the minus energy during the LSB cycle, but it is not energy efficient. The hybrid switching technique [7] dissipates no switching power during the first three bit cycles, and the Zhang switching scheme [8] introduces the minus energy during the second bit cycle. However, both are less energy efficient from the forth bit cycle because the monotonic switching method is used and double-side capacitors are switched. In this letter, an ultra-low-power capacitor-splitting switching algorithm with minus energy is proposed which consumes zero switching energy from the reference during the first three bit cycles and only single-side capacitors are switched from the forth bit cycle. The dummy capacitor is also reused to add one more bit, which further reduces switching energy. Therefore, compared to the Sanyal and Sun switching scheme, the proposed switching scheme reduces the switching energy by 94.19%.

2 Proposed capacitor-splitting switching scheme with minus energy

To reduce switching energy, several methods are often used, such as top-plating sampling, three voltage references, and making full use of the dummy capacitor [9]. But these techniques still consume much power from the voltage reference. The Sanyal and Sun scheme introduces the minus energy, which means that the DAC gives energy back to the reference voltage sources [3]. Thus, an ultra-low-power capacitor-splitting switching algorithm with minus energy was proposed. The capacitor-splitting method is to reduce the switching energy, because only single-side capacitors are switched from the third bit cycle.

To illustrate the proposed switching method, a 4-bit DAC is shown in Fig. 1. The C/2 is two Cs in series. First, top-plating sampling is conducted with bottom plates of the most significant bit (MSB) capacitors and the LSB capacitors reset to ground and V REF, respectively. Then, the sampling switches are off and the first comparison is performed. The MSB is achieved, but no power is dissipated. If MSB = 1, the LSB capacitors on the V DACP side are switched from V REF to ground with other capacitors unchanged. Otherwise, the LSB capacitors on the V DACN side are switched from V REF to ground. Another comparison is performed, and the MSB-1 is obtained. It does not consume switching energy, too. From the third bit cycle, only single-side capacitors are switched, which is based on the MSB. If MSB = 1, only capacitors on the V DACN side are changed. Otherwise, the same case occurs on the V DACP side.

Fig. 1
figure 1

Proposed switching scheme of 4-bit DAC

The switching detail of the third bit cycle is shown in Table 1. When the MSB and MSB-1 are the same, the MSB capacitors are switched from the ground to V CM; when the MSB and MSB-1 are different, the LSB capacitors are switched from V REF to V CM. In this bit cycle, the minus energy is introduced, which is not non-physical. It means the capacitors discharge and give energy back to the reference voltage sources [3]. For an N-bit ADC, the switching energy during the third bit cycle is

$$\begin{aligned} E_{3} & { = } - V_{\text{CM}} \cdot \left[ {\left( {V_{\text{DACP(N)}} \left( 2 \right) + \frac{{V_{\text{REF}} }}{4} - V_{\text{CM}} } \right) - \left( {V_{\text{DACP(N)}} \left( 2 \right) - 0} \right)} \right] \cdot 2^{N - 4} C \\ & \quad - \,V_{\text{REF}} \cdot \left[ {\left( {V_{\text{DACP(N)}} \left( 2 \right) + \frac{{V_{\text{REF}} }}{4} - V_{\text{REF}} } \right) - \left( {V_{\text{DACP(N)}} \left( 2 \right) - V_{\text{REF}} } \right)} \right] \cdot 2^{N - 4} C \\ & = - 2^{N - 7} CV_{\text{REF}}^{ 2} \\ \end{aligned}$$
(1)

or

$$\begin{aligned} E_{3} { = } & - V_{\text{CM}} \cdot \left[ {\left( {V_{\text{DACP(N)}} \left( 2 \right) - \frac{{V_{\text{REF}} }}{4} - V_{\text{CM}} } \right) - \left( {V_{\text{DACP(N)}} \left( 2 \right) - V_{\text{REF}} } \right)} \right] \cdot 2^{N - 4} C \\ & = - 2^{N - 7} CV_{\text{REF}}^{ 2} \\ \end{aligned}$$
(2)
Table 1 Switching during the third bit cycle

From the forth bit cycle, related capacitors are changed according to the first two bits, too. When the MSB and MSB-1 are the same, the related capacitor in the MSB part is switched from V CM to ground or V REF based on the third bit; when the MSB and MSB-1 are different, the related capacitor in the MSB part is switched from ground to V CM or the related capacitor in the LSB part is switched from V CM to ground, which is also based on the third bit. The former switching procedure is repeated until the LSB is achieved.

The output waveform of the proposed switching scheme is shown in Fig. 2(a). The related common-mode voltage is shown in Fig. 2(b) which affects the ADC linearity [10].

Fig. 2
figure 2

Output waveform (a) and the related common-mode voltage (b) of the proposed switching scheme

2.1 Switching energy

The behavioral simulations of different capacitor switching schemes and proposed capacitor-splitting switching scheme for 10-bit SAR ADC were performed in MATLAB. Figure 3 shows the average switching energy for these switching schemes against the output code. The proposed switching technique consumes only 9.24 \(CV_{\text{REF}}^{2}\) average switching energy from the reference and achieves 99.32% energy saving compared to the conventional switching. In Table 2, the comparisons for these switching methods are made and the Sanyal and Sun method [3] is chosen as a reference. The Tong [4], the hybrid [7], and the Zhang [8] switching techniques consume 31.4, 15.88, and 11.22 \(CV_{\text{REF}}^{2}\) average switching energy, respectively. If the minus energy is considered, the switching energy is 1.24 \(CV_{\text{REF}}^{2}\), 94.19% less. Besides, 268 unit capacitors are required, 47.66% less compared to the Sanyal and Sun technique. Therefore, the proposed approach is more area efficient and energy efficient.

Fig. 3
figure 3

Switching energy against output code

Table 2 Comparison of different switching schemes for 10-bit SAR ADC

2.2 Linearity

The switching energy is directly proportional to the value of the unit capacitor C. Thus, C should be as small as possible. However, the value of C is usually determined by the capacitor matching. Assume that the unit capacitor is Gaussian-distributed, modeled with a nominal value of C u and a standard deviation of σ u . For a binary-weighted capacitor array, each capacitor is unit capacitors in parallel. The parasitic effect of top-plate parasitic capacitance is ignored because it just leads to a gain error with no effect on the linearity performance. The bottom-plate parasitic capacitance of C in C/2 is also ignored for simplicity because it is little. Figure 4 shows behavioral simulation results of 512 Monte Carlo runs of 10-bit DAC with proposed switching scheme. The DNL and INL (integral nonlinearity) curves are the root-mean-square (rms) values, and C is Gaussian random variable with standard deviation of 1% (σ u /C u  = 0.01). σ DNL,MAX and σ INL,MAX are 0.225 LSB and 0.228 LSB, respectively.

Fig. 4
figure 4

Standard deviation of DNL and INL at each output code

3 Conclusion

An ultra-low-power capacitor-splitting switching algorithm with minus energy and single-side capacitor switching for SAR ADCs is proposed. Compared to Sanyal and Sun switching scheme, the proposed switching scheme achieves 94.19% energy saving and 47.66% capacitor area reduction. Thus, the proposed capacitor-splitting switching algorithm is often used for low-power and small-area SAR ADCs.