β: This English translation is in beta — the Traditional-Chinese original is the authoritative version.

Design Issues in Cross-Coupled Inverter Sense Amplifier

Prerequisites (recommended reading): this page is not a prerequisite for the ISF. Its only connection to this course is the regeneration / positive feedback of the cross-coupled pair; to see how that mechanism becomes "oscillator start-up," first read oscillator_phase (limit-cycle start-up) and tank_Q_and_energy_restoration (negative resistance compensating losses). To learn the ISF itself, go straight to paper_001.

Let's be clear up front: this paper has nothing to do with ISF / phase noise / jitter. It is an ISCAS 1998 short paper (4 pages) on the design of a cross-coupled inverter sense amplifier, covering regeneration speed, offset voltage caused by device mismatch, and a figure of merit for offset. It appears in this list purely because it is in the source folder and shares the author Hajimiri. This page honestly flags that mismatch (claim C12) and offers only one conceptual bridge.

Why write a page anyway: Section 9 of the authoring conventions requires "[P5] must always be honestly described as a sense-amplifier paper unrelated to ISF." We do not pretend it relates to the ISF, nor force equations onto it; we simply state its actual content and point out its only legitimate connection to this course.

Citation

[P5] A. Hajimiri and R. Heald, "Design Issues in Cross-Coupled Inverter Sense Amplifier," Proc. IEEE International Symposium on Circuits and Systems (ISCAS), 1998. (file Hajimiri_ISCS_98.pdf, paper_005, 4 pages)

One-sentence contribution

Analytic design of a CMOS cross-coupled-inverter sense amplifier: analyzes the effect of the equilibrating transistors and the tail current source on sensing speed, the offset caused by mismatch, and proposes a figure of merit for offset — unrelated to oscillator phase noise / ISF (claim C12).

Why this paper matters (for this course: essentially not at all)

For the ISF course, this paper's importance is ≈ zero. It solves problems in memory and datapath circuits: a sense amplifier must quickly and reliably amplify the tiny voltage difference appearing on the bitline into a full-swing digital 0/1. Its concerns are:

• Regeneration speed: the cross-coupled pair uses positive feedback to exponentially separate the two node voltages from the metastable point — the faster the separation, the faster the sensing.
• Offset: the two sides' transistors can never be identical (mismatch); this asymmetry makes the sense amp favor one side even at zero input difference, causing read errors. The paper analyzes the mismatch sources and gives a figure of merit.
• How the gradual switching of the equilibrating device and the tail current source degrades the above performance.

These are digital/memory circuit topics — no limit cycle, no excess phase, no ISF, no phase noise spectrum.

Main assumptions

Per paper_metadata (paper_005.assumptions):

• Small-signal regeneration analysis around the metastable point; mismatch-based offset model.
• (As stated in the paper) current has flowed through the transistors long enough, and the equilibrating device can be treated as an ideal switch — two simplifying assumptions that are challenged one by one in the later sections.

Key equations (not transcribed; outside the scope of this course)

Per paper_metadata (paper_005.important_equations): the equations of this paper (regeneration time constant, offset voltage expressions) are outside the scope of the ISF course and are therefore not transcribed verbatim.

⚠️ TODO: equations not transcribed because this PDF is unrelated to ISF/phase noise.

However, to make the single conceptual bridge clear, we point out only its core mechanism (regeneration), with a maximally simplified small-signal model showing "how positive feedback amplifies exponentially" — the same mechanism is also what allows an oscillator to start up:

Regeneration of the cross-coupled pair (simplified small-signal): for two mutually fed-back inverters, the differential voltage $v_d=v_1-v_2$ near the metastable point satisfies

$$C\frac{dv_d}{dt}=(G_m-G_0)\,v_d\quad\Rightarrow\quad v_d(t)=v_d(0)\,e^{\,t/\tau_{regen}},\quad \tau_{regen}=\frac{C}{G_m-G_0}.$$

• Meaning: as long as the effective transconductance $G_m$ exceeds the node leakage conductance $G_0$, the differential voltage grows exponentially (positive feedback), rapidly amplifying a tiny input difference to full swing — this is regeneration. The smaller $\tau_{regen}$, the faster the sensing.
• Dimension check: $[\text{F}]/[\text{S}]=[\text{C/V}]/[\text{A/V}]=[\text{C/A}]=[\text{s}]$ ✓.
• This comes from simplifying the paper's Sec. 2 pair of cross-coupled differential equations ($dv_1/dt$, $dv_2/dt$); this is the minimal form we wrote to explain the bridge — the paper's full expressions (including equilibrating / tail effects) are more complex and outside the scope of this course.

Key figures

The paper has sense-amp schematics and small-signal equivalent diagrams (Fig. 1 etc.), but they are irrelevant to the ISF course; this site neither cites nor redraws them (paper_metadata: important_figures is empty).

Design insights (for the sense amp, not for the ISF)

• At the regenerative node, the design goal is to minimize the time constant $\tau_{regen}$, not to blindly enlarge the initial voltage difference (the paper explicitly points out this trade-off).
• A complete offset analysis must consider the cell and bitline structure together, not just the cross-coupled pair itself.
• The gradual switching of the equilibrating device and the tail current source significantly degrades speed and offset and must be included in the analysis.

These are useful to SRAM / sense-amp designers, but carry no transferable design rules for ISF / phase noise.

Limitations (for this course)

Per paper_metadata (paper_005.limitations):

• Entirely outside the scope of ISF / phase noise / jitter. This site treats it as a corner-case deep-dive, honestly flags it as mislabeled, and offers only the "regeneration → oscillation" conceptual bridge.

Relationship to other papers (the only bridge)

[P5] has no theoretical continuity with [P1]–[P4]. The only legitimate connection is one mechanism:

The regeneration (positive feedback) of the cross-coupled pair is also the foundation on which latch-based and LC oscillators "start up on their own and sustain a limit cycle."

• In a sense amp: positive feedback amplifies a tiny input difference once to full swing, then settles into a steady state (the latch locks).
• In an oscillator: the same negative resistance / positive feedback supplies energy to compensate the tank's losses, letting the oscillation persist without decaying — this is exactly the physical origin of a stable limit cycle ([P1] assumption 2). The $-G_m$ pair of a differential LC-VCO and the cross-coupled stage of a latch-based ring share the same start-up mechanism as this paper's cross-coupled pair.

So remember it this way: the same cross-coupled positive feedback, when it stops, is a latch/sense amp; when it cannot stop (keeps oscillating), it is an oscillator. But the moment we get to "the oscillator's phase response to noise," that is ISF territory — unrelated to this paper.

For the details of oscillator start-up and limit-cycle geometry, see oscillator_phase; for the ISF itself, see paper_001.

Further reading / corresponding teaching pages

[P5] is unrelated to the ISF, so there is only one bridge here — we do not pretend there are more:

Which part of this page	Corresponding teaching page	Why this link
Regeneration / positive feedback of the cross-coupled pair (the only bridge)	paper_001 and the paper-by-paper deep-dive guide	The same positive feedback: "stops = latch/sense amp, cannot stop = oscillator"; the oscillator's limit-cycle start-up comes from it, and only then does the ISF enter the stage (claim C12)

Honesty note: this page does not link to any core ISF theory page or design page, because [P5] is outside the scope of ISF / phase noise (see the mismatch statement above). To learn the ISF, return to paper_001; for the full map of the five papers' roles, see the paper-by-paper deep-dive guide.

What to remember

• [P5] is a sense-amplifier paper, unrelated to ISF / phase noise / jitter (claim C12) — do not cite it as ISF literature.
• It is in the list only because it is in the source folder and shares an author; this site honestly flags it as mislabeled.
• The only conceptual bridge: the regeneration / positive feedback of the cross-coupled pair, which is also the foundation of latch and LC oscillator start-up (stops = latch, cannot stop = oscillator).
• To learn the ISF, return to [P1] (paper_001).