Text & documents
Numbers & maths
Data & formats
Security
Development & DevOps
Artificial Intelligence
Finance
Health & Wellness
Productivity
Games & Entertainment
Multimedia & design
Business
How to use
The five tabs are ONE journey

They are not five separate calculators: it is the same cell seen five times. It holds a bit (noise margin), it has to be read without being destroyed, it has to be fed current, it has a shape on silicon, and all of that decides how fast it can run. Two buttons cross from one tab to another: «take this resistance to the grid» and «slow the logic down with the IR drop».

Everything here is a first-order model

And every screen says so in place. There is no design-rule checking here, no parasitic extraction, no place and route: that is professional software and it does not fit — nor should it — in a browser tab. Use it to see why a number moves when you move a control, not to sign anything off.

The link button shares the exact settings

It copies an address with the tab and every value exactly as you have them. Use it to paste a specific case into an article or send it to someone: they will open exactly what you were looking at, not the starting screen.

Why it is measured with a square

Each inverter has its curve; crossing them gives the butterfly. The side of the largest square that fits inside is how much noise the cell survives before flipping — past that, positive feedback does the rest. It is Seevinck's method (1987): everyone cites it and almost nobody lets you touch it.

The three modes, and the 0 that is good news

Hold is the cell on its own: the largest margin. Read opens the word line and the access transistors pull up the node holding a 0 — that is why the margin narrows, and it is the number-one confusion on this topic. Write gives margin 0, and that is fine: writing means flipping the cell on purpose.

From the transistor to the chipFrom a cell's noise margin to the clock: SRAM, read path, IR drop, standard cell and setup/hold
From the transistor to the chipFrom a memory cell to how the current reaches it and what shape it takes on silicon

An SRAM cell holds a bit with two inverters feeding each other. The harder they hold on to one another, the more noise they survive without flipping — and you measure that by plotting each inverter's curve against the other's: you get a butterfly, and the side of the largest square that fits inside is the static noise margin (SNM). Move the controls and watch it shrink.

Word line open with both bit lines precharged: the access transistors pull up the node that holds a 0. That is why reading narrows the margin.

The cell, transistor by transistor

VddGNDBLVddBLBVddWL = VddQQBM2M1M4M3M5M6

M1/M2 and M3/M4 are the two inverters; M5 and M6 are the access transistors, which only conduct while the word line is active. Hover over each one to see its role.

The butterfly and the square

0.000.000.250.250.500.500.750.751.001.00V(Q) — voltsV(QB) — volts
244mVnoise margin (SNM)

Opening the word line eats 39% of the margin: from 398 mV holding down to 244 mV reading. That is the read disturb.

Lobes: 244 and 244 mV — the smaller one wins, because the cell flips on its weak side.
2
Drives READ stability: a stronger pull-down better resists the tug of the bit line.
1.8
Drives WRITING: if the pull-up is too strong, the cell resists the 0 on the bit line and refuses to be written.
1
Lower it and watch the square shrink: this is why the static SNM «stops being enough» at modern nodes.
0.3
How hard you have to push the gate before the transistor conducts.

From a margin to a memory that works

A 200 mV margin does not say much on its own: what decides is how many standard deviations fit inside it. Every cell on the chip comes out slightly different, and it only takes ONE failing to spoil the word. That is why memory people talk about 6σ rather than «enough».

30
6
8.1σof margin per cell
2.1 · 10^-16one cell fails
2.1 · 10^-10the whole array fails

This is an illustration, not an analysis. It assumes the margin follows a bell curve and that cells fail independently. Real work uses Monte-Carlo with the foundry's variability models, and the real tail is not a bell curve. Use it to understand why array size matters; not to decide anything.

First-order model

All three panels use first-order models and say so in place. There is no design-rule checking here, no parasitic extraction and no automatic place and route: that is professional software and it does not fit — nor should it — in a browser tab. For real layout: Magic, OpenROAD or KLayout. And if what you want is to draw the silicon layer by layer and see it in cross-section, SiliWiz does exactly that, and does it well.

The whole thread:the atom and why there is a thresholdthe transistor and the CMOS inverterthe cell, its power and its shape (you are here)