9.5 A Sub-1V, 50mV Dropout LDO Using Pseudo-Impedance Buffer with Phase-Margin Improvement Design

As technology continues to scale down, the supply voltage <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(V_{\text{IN}}]$</tex> of systems has gradually decreased. At the same time, devices require a larger load current <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(I_{\mathrm{L}}]$</tex> due to a higher performance of the system. Accordingly, it is important that a low-dropout regulator (LOG) operates at <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{low}-V_{\text{IN}}$</tex> (specifically sub-1V) and drives heavy <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$l_{L}$</tex> at the same time. Digital LOGs have been used in sub-1V applications. However, they suffer from a large output voltage <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(V_{\text{OUT}}]$</tex> ripple [1] even with an off-chip output capacitor <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(C_{0})$</tex> because of a limit cycle oscillation. On the other hand, the analog LDOs are free from the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$V_{\text{OUT}}$</tex> ripple. When an analog LDO uses a <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$C_{0}$</tex>, a multi-stage structure usually has a narrow bandwidth (BW) and becomes unstable under light load conditions. This is because Miller compensation is not effective in this case. Therefore, to stably handle light load or even no-load conditions while maintaining sufficient BW at 10w <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$V_{1\mathrm{N}}$</tex>, the LDO needs to be designed as a low <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$-\vert /_{1\mathrm{N}}$</tex> structure, comprising a single-stage error amplifier (EA), a buffer, and a pass transistor <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(M_{\mathrm{P}}]$</tex>, avoiding the use of Miller compensation, as shown in Fig. 9.5.1 (top). In low <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$V_{1\mathrm{N}}$</tex> analog LDO s, a current mirror (CM) buffer that composes the CM with the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$M_{\mathrm{F}}$</tex> has been commonly used because this buffer properly operates with a small voltage headroom and has a relatively low resistance of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$1/\mathrm{g}_{\mathrm{m}}$</tex>, which widens BW. This CM buffer can be driven by a simple EA [2] or a transconductance <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(G_{\mathrm{m}})$</tex> boosting cell [3]–[4], as shown in Fig. 9.5.1 (top-left and -middle), respectively. Since the EA cannot adopt a cascode structure due to the voltage headroom, an LDO that uses the EA exhibits a low DC gain, particularly when supplying a large <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$l_{L}$</tex>. Moreover, the output pole of the EA <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\omega_{\text{EA}})$</tex> limits the BW under the heavy load condition. On the other hand, the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$G_{\mathrm{m}}$</tex> -boosting cell can have a higher gain than that of the simple EA while it does not have a low frequency <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\omega$</tex> EA because the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$G_{W}$</tex> cell commonly comprises several current mirrors of which ratio is larger than one. However, LDO s with not only the EA but also the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$G_{\mathrm{m}}$</tex> -boosting cell have limitations in supplying a wide range of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$l_{L}$</tex> because of the CM buffer. To supply a large <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$l_{L}$</tex>, the ratio (K) between a CM buffer transistor <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(M_{\mathrm{c}})$</tex> and the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$M_{\mathrm{P}}$</tex> must be high. However, with the high <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{K}(\mathrm{K}_{2})$</tex>, the LOG cannot properly supply a small <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$l_{L}$</tex>. This is because the bias current of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$M_{\mathrm{c}}(/_{\mathrm{c}})$</tex> becomes considerably small in this case, which increases the buffer resistance, and finally shifts the gate pole of the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$M_{\mathrm{F}} ((\lambda)_{\mathrm{G}})$</tex> to the low frequency. This phenomenon could be severe when the LDO uses the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$G_{\mathrm{m}}$</tex> -boosting cell because the conventional <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$G_{\mathrm{m}}$</tex> cell has a bias current related to the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$l_{L}$</tex>. On the other hand, when the ratio K is low <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{K}_{1})$</tex>, the LDO can supply a small <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$l_{L}$</tex>. However, this limits supplying a large <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$l_{L}$</tex> due to the limited source-gate voltage <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(V_{\text{SG}})$</tex> of the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$M_{\mathrm{P}}$</tex> by the limited voltage headroom, especially in sub-1 V <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$V_{1\Uparrow}$</tex> LDO s. Because of these reasons, previous LDO s using the CM buffer cannot cover a wide range of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$l_{L}$</tex>, as shown in Fig. 9.5.1 (top-right), and usually focus on the light <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$l_{L}$</tex> to cover the no-load condition [2]–[4].