Back to projects
Power / RegulatorFeaturedProject completed

Capacitorless Linear Regulator with Gain-Peaking Suppression and Transient-Response Enhancement

Studied and implemented a capacitorless linear regulator using damping-factor control to suppress gain peaking, then added a transient-accelerator path to improve load-transient behavior. The project included small-signal analysis, bias and reference design, PVT-condition sweeps, and worst-condition Monte Carlo simulations.

Status
Project completed
Project domain
Power / Regulator
  • Analog LDO
  • Capacitorless Regulator
  • Damping Factor Control
  • Gain Peaking
  • Transient Accelerator
  • PVT
  • Monte Carlo
  • HSPICE
  • Schematic-level design and simulation study
  • No completed full-chip layout
  • No PEX or silicon validation
  • Completed undergraduate project with documented limitations

Quick Summary

30-second project summary

Project objective
This project was based on a literature-proposed DFC architecture. I implemented and evaluated the circuit, performed small-signal and PVT/Monte Carlo verification, and added and iterated the transient-accelerator path.
My contribution
Completed the literature review, circuit implementation, small-signal and PVT/Monte Carlo verification, and transient-accelerator design iterations.
Strongest verified result
Load-transient overshoot · 189 mV → 126 mV
Main limitation / redesign focus
Worst heavy-load phase margin was approximately 55°, below the original greater-than-60° target.
Verification level
Schematic-level design and simulation study

Target Specifications and Final Results

Pass or fail is assigned only when the original report defined an acceptance threshold. Metrics without an original threshold are presented as reported results without post-hoc specifications.

Except for the phase-margin target of greater than 60°, the original report did not define numerical acceptance thresholds for most performance metrics. Only phase margin is evaluated against an explicit acceptance threshold; the remaining metrics are reported without post-hoc limits.

Design Baseline

Process
U18 0.18 µm CMOS
Regulated output
1.0 V nominal; tolerance not originally defined
Input-voltage range
1.2 to 1.8 V
Temperature
-40, 50, and 125 °C
Load range
100 µA to 100 mA
PVT sweep
810 conditions
Monte Carlo
1000 runs with process and mismatch enabled at selected worst-case conditions

Target Specifications and Final Results

MetricTargetResultScopeStatusNote
Phase margin> 60°No-MC worst = 66.157° heavy / 66.305° light; MC worst ≈ 55° heavy / ≈ 64° light810-condition PVT + 1000-run MCPartialThe non-MC and light-load MC results meet the target, but the worst heavy-load MC result falls below 60°.
Output-voltage range1.0 V nominal; tolerance not originally definedApproximately 0.95 to 1.07 V under MC1000-run MCReported only
Loop gainNot originally definedMC worst ≈ 64 dB heavy / ≈ 60 dB light1000-run MCReported only
Gain-bandwidth productNot originally definedMC worst ≈ 0.8 MHz heavy / ≈ 0.9 MHz light1000-run MCReported only
Line regulationNot originally definedMC worst ≈ 0.4 mV/V1000-run MCReported only
Load regulationNot originally definedMC worst ≈ 8 mV/A1000-run MCReported only
Load-transient overshoot / undershootNot originally definedMC worst ≈ 185 mV overshoot / ≈ 660 mV undershoot1000-run MCReported only
PSR @ 1 kHzNot originally definedMC worst ≈ -36.2 dB heavy / ≈ -35.8 dB light1000-run MCReported only
PSR @ 10 kHzNot originally definedMC worst ≈ -31.3 dB under both loads1000-run MCReported only
Reported total powerNot originally definedMC worst ≈ 200 mW heavy / ≈ 1.7 mW light1000-run MCReported onlyPresented as total power according to the original report and not reinterpreted as quiescent power.

Visual overview

Key metrics

Operating conditions
810

non-Monte-Carlo

Monte Carlo
1000

process-and-mismatch samples

Overshoot
189 → 126 mV

without / with TA

Worst heavy-load PM
≈ 55°

Visual overview

Verification coverage

  • ExecutedDFC gain-peaking suppression
  • ExecutedTA overshoot improvement
  • Executed810 operating conditions
  • ExecutedSelected worst-condition Monte Carlo
  • Future verificationFull layout and PEX
  • Out of scope for this phaseSilicon measurement

Transient-accelerator comparison

Load-transient overshoot

Without TA189 mV
With TA126 mV

Problem & Design Goal

This project was based on a literature-proposed DFC architecture. I implemented and evaluated the circuit, performed small-signal and PVT/Monte Carlo verification, and added and iterated the transient-accelerator path. The work focused on trade-offs among stability, transient response, power, and circuit complexity.

My Contribution

  • Reviewed LDO compensation and gain-peaking literature
  • Compared a conventional LDO with a DFC-based architecture
  • Analyzed the error-amplifier output pole, output pole, complex poles, and damping factor
  • Implemented and simulated DFC and gm-boosting paths
  • Designed and evaluated constant-gm bias and bandgap-reference blocks
  • Added and iterated a transient-accelerator path
  • Compared circuits with and without DFC and with and without the transient accelerator
  • Evaluated 810 non-Monte-Carlo operating conditions
  • Ran 1000-sample process-and-mismatch Monte Carlo simulations at selected worst conditions
  • Analyzed stability, loop gain, GBW, power, regulation, transient response, and PSR
  • Studied layout matching, common-centroid placement, and physical-design principles

Architecture & Method

Simplified architecture: this diagram shows only the approved high-level signal path and is not a complete schematic.
  1. Error Amplifier / Gain Stages
  2. Pass Device
  3. VOUT
  4. Feedback
  • DFC path for damping control
  • TA path for transient acceleration
  • Bias / reference blocks

Design and verification method

The study first compared a conventional LDO with the literature-based DFC architecture and analyzed the error-amplifier output pole, output pole, complex poles, and damping factor. It then implemented the DFC, gm-boosting, constant-gm bias, and bandgap-reference paths. After the TA was added, AC and transient behavior were compared with and without DFC and with and without TA. The final study covered 810 non-Monte-Carlo operating conditions and 1000-sample process-and-mismatch Monte Carlo simulations at selected worst conditions.

The design phase-margin target was greater than 60°. That target is kept separate from the Monte Carlo observations below; neither one comparison case nor an approximate observation is treated as an all-condition signoff result.

This project re-implemented a published circuit architecture and completed device sizing, integration, and simulation-based verification. The figures below are author-generated simulation results; they do not imply that the published architecture was original to this project.

DFC stability and transient evidence

The AC comparison connects damping-factor control to the observed loop response, while the load step shows the practical difference between the compensated and uncompensated cases.

Light-load loop response

Light-load Bode comparison showing loop-gain magnitude and phase with and without DFC.

Load-transient comparison

Load-transient comparison showing reduced oscillation with DFC in the reported condition.

Design Challenges & Engineering Decisions

Case 01

Light-Load Gain Peaking and Stability Degradation

Improved
Challenge

Without DFC, pronounced gain peaking appeared under light load, with oscillation and poor phase margin in some load-transient conditions.

Diagnosis

At light load, lower pass-transistor gm moves the associated complex-conjugate poles toward lower frequencies. The resulting low-damping gain peaking forms near crossover and compresses phase margin.

Design Response

A damping-factor-control path was added. Its additional negative-gain path and compensation capacitor increase the damping of the complex-conjugate poles, with related device sizes adjusted from the small-signal analysis.

Outcome

DFC substantially suppressed gain peaking and removed the oscillation previously observed in the selected light-load transition condition.

Trade-off or Next Step

The observed worst heavy-load Monte Carlo phase margin remains approximately 55°, below the original greater-than-60° target. The result is therefore an improvement, not an all-condition pass or signoff.

Case 02

Transient Overshoot after Stability Improvement

Improved
Challenge

After DFC improved stability, load-transient overshoot and undershoot still required further improvement.

Diagnosis

The error amplifier has limited bandwidth and cannot charge or discharge the pass-MOS gate quickly enough during abrupt load changes, delaying the pass device conduction adjustment.

Design Response

A Transient Accelerator was added. Dynamic biasing detects changes in an error-amplifier branch and provides an additional momentary gate charging or discharging path.

Outcome

Load-transient overshoot decreased from approximately 189 mV to approximately 126 mV, while the undershoot improvement was more limited.

Trade-off or Next Step

Part of the Transient Accelerator discharge path does not fully turn off in steady state, leaving non-ideal leakage current. The case remains improved rather than fully resolved.

Verification Results

DFC comparison

  • Without DFC, the compared light-load case showed a phase margin of −36.52°
  • The DFC architecture suppressed the gain-peaking behavior observed in the comparison circuit
  • This was one comparison condition, not a result for every operating condition

Transient Accelerator

  • Load-transient overshoot improved from 189 mV to 126 mV
  • The improvement in undershoot was limited
  • Test condition: All Typical Case, Vin = 1.2 V, Temp = 50 °C

Transient-accelerator trade-off

Paired output traces show a clear overshoot reduction, while the undershoot responses remain close under the reported condition.

Overshoot

Output overshoot with and without the transient accelerator, showing reduced excursion with TA enabled.

Undershoot

Output undershoot with and without the transient accelerator; the responses remain close in this condition.

Condition sweep and Monte Carlo

  • 810 reported non-Monte-Carlo operating conditions
  • 1000 Monte Carlo samples with process and mismatch enabled
  • Worst non-Monte-Carlo conditions were used for capacitor, temperature, and supply selections
  • Heavy-load phase-margin worst value: approximately 55°; light-load: approximately 64°
  • Heavy-load loop-gain worst value: approximately 64 dB; light-load: approximately 60 dB
  • Heavy-load GBW worst value: approximately 800 kHz; light-load: approximately 900 kHz

All Monte Carlo values above are approximate observations, not precise signoff numbers.

Selected robustness evidence

This selected heavy-load distribution documents the simulated phase-margin spread; it is evidence for the reported study, not an all-condition signoff claim.

Heavy-load phase margin

Heavy-load phase-margin Monte Carlo distribution for the final compensated design.

Bandgap-reference robustness

This result supplements the bias/reference implementation by showing the simulated reference spread under one selected high-supply worst-case condition; it is not an all-condition signoff result.

High-supply worst-case condition

Bandgap-reference Monte Carlo distributions at the report's high-supply worst-case condition.

Engineering Insights

  • The original TA discharge path responded too slowly because of its multi-stage path
  • Simplifying the discharge path reduced control gain and made full turn-off more difficult
  • The transient accelerator improved overshoot substantially but had limited effect on undershoot
  • DFC, gm boosting, and transient acceleration had to be considered together with stability, power, and area

Validation Boundary & Limitations

  • The original phase-margin target was greater than 60°, but the reported worst heavy-load Monte Carlo result was approximately 55°
  • The transient accelerator retained non-ideal steady-state current or leakage
  • DFC, gm boosting, and transient acceleration increase circuit complexity, area, and power
  • Full layout, PEX, and silicon measurement were not completed
  • The results do not establish that every PVT or Monte Carlo sample met every specification

Next Focus

  • Redesign the TA discharge path
  • Reduce steady-state leakage
  • Recover heavy-load phase-margin margin
  • Co-optimize stability, transient response, power, and area
  • Complete layout, extraction, and post-layout verification

Tools or workflow

  • Synopsys HSPICE

Expanded engineering result