Manual Layout Study

The automatic layout work was preceded by a manual study of selected analog layout techniques using a two-stage operational amplifier.

The purpose of the study was not to identify one universally optimal layout style. Instead, it was used to determine which physical effects different layout techniques target, how those effects should be evaluated, and which information should later be preserved by an automatic flow.

Reference Layout Situations

The reference amplifier contains several recurring analog layout situations:

a matching-critical differential pair;
current-mirror structures;
output-connected devices;
a Miller compensation capacitor;
source and bulk connection issues;
substrate/well isolation structures.

These situations later become either generated module families or explicit physical-design metadata.

Controlled Layout Variants

Four manual variants were used to separate the main layout decisions.

Variant A. Clustered baseline without guard rings. Layout boundary: 36.7 µm × 29.6 µm.

Variant B. Baseline with module-level guard-ring structures. Layout boundary: 37.9 µm × 29.6 µm.

Variant C. Matching-oriented variant using interdigitation. Layout boundary: 42.9 µm × 29.6 µm.

Variant D. Matching-oriented variant with a common-centroid differential pair. Layout boundary: 42.9 µm × 33.2 µm.

The comparisons are organized as controlled changes:

A → B: guard-ring insertion;
B → C / D: matching-oriented placement.

The schematic topology remains unchanged while the selected physical layout decision is varied.

Spacing Sensitivity

A spacing sweep was used to test whether increasing the physical separation between selected layout groups strongly changes the amplifier response in the evaluated setup.

Metric	Pre-layout	1.0 s₀	3.0 s₀	Observation
GBW	9.751 MHz	9.574 MHz	9.574 MHz	Post-layout shift, then essentially flat
Phase margin	70.609°	70.092°	70.095°	Post-layout shift, then negligible sweep change
Positive slew rate	27.042 V/µs	26.133 V/µs	26.120 V/µs	Small monotonic sweep change
Selected critical parasitics	—	baseline	< 0.5% change	Spacing is not dominant here

Interpretation. For this circuit and extraction setup, the evaluated spacing range has only a limited influence on the selected performance metrics. This is a useful negative result: spacing should not automatically be treated as the dominant layout variable when stronger physical effects are present.

Guard-Ring Isolation

Guard-ring insertion targets substrate and well-domain coupling. The two manual layouts are therefore compared using a controlled body-domain disturbance model.

AC disturbance-transfer comparison with and without guard rings — AC disturbance-transfer comparison. The guarded variant exhibits a smaller output response under the adopted body-domain sensitivity model.

Transient disturbance comparison with and without guard rings — Transient comparison under a pulsed aggressor. The main disturbance appears around pulse transitions, while the guarded case shows a smaller response.

The model uses normalized disturbance-transfer factors to distinguish the guarded and unguarded body domains.

Model boundary. This experiment is a comparative body-domain sensitivity study. It is not a calibrated substrate-extraction model and the numerical attenuation should not be interpreted as an silicon-accurate prediction of guard-ring isolation.

Matching-Oriented Evaluation

Matching-oriented layout techniques are evaluated statistically because their intended benefit appears primarily in mismatch sensitivity, distribution spread, and tail behaviour rather than as a large shift of the nominal operating point.

Conceptual clustered, interdigitated, and common-centroid matched-device organizations — Conceptual matched-device organizations used in the manual study: clustered placement, interdigitation, and common-centroid placement. Dummy devices are included at the array boundaries where applicable. The patterns are schematic and are not drawn to scale.

Empirical cumulative distribution of absolute input offset for clustered, interdigitated, and common-centroid layouts — Empirical cumulative distribution of absolute input offset for the three matching-layout variants.

Metric	Clustered B	Interdigitated C	Common centroid D
std(\|Voffset\|)	2.67 mV	2.45 mV	2.44 mV
p90(\|Voffset\|)	7.64 mV	7.23 mV	7.27 mV
p10(CMRR)	62.57 dB	63.02 dB	63.03 dB
p10(PSRR)	61.77 dB	62.16 dB	62.17 dB

The mean values remain close. The clearer difference appears in spread and distribution tails. Variants C and D also remain close under the adopted simplified model.

The Monte Carlo model is comparative rather than process-calibrated. Behavioural threshold-voltage perturbations represent residual random, gradient, and edge-environment related mismatch terms.

From Manual Study to Automation

The manual study provides the basis for later automation decisions.

Manual observation	Automation implication
Current mirrors require compact ratio matching	Shared-diffusion interdigitation
Differential pairs benefit from two-dimensional symmetry	Common-centroid array generation
Dummy devices define the local edge environment	Generator-level dummy policy
Guard rings are extended physical domains	Explicit guard-ring geometry and separate routing
PMOS well relations constrain legal placement	Well-domain metadata and post-placement repair
Buses and capacitor plates provide extended routing access	Explicit access geometry and access-aware routing

The important transition is therefore not from one manual layout directly to one generated layout. The manual study first identifies the physical purpose of each technique; the resulting knowledge is then abstracted into reusable policies for generation, placement, and routing.

Next: the conceptual framework .