Volts to Millivolts

Turning a Rail Spec into a Millivolt Budget

Pick a rail and the design work immediately drops a thousandfold. The schematic says 3.3 V, but the number you actually engineer against — the tolerance window, the ripple limit, the dropout headroom, the noise margin — lands in millivolts. One percent of 3.3 V is 33 mV, and 33 mV is what the oscilloscope cursor shows.

This converter turns those volt-level figures into the millivolt values you write into a power budget, a test limit or a review comment.

Conversion factor: 1 V = 1,000 mV, so multiply volts by 1,000. Worked example: a 3.3 V rail at ±5 % allows 3.3 × 0.05 = 0.165 V of drift → ±165 mV, a window of 3.135 V to 3.465 V.

Where the Millivolts Come From

Tolerance windows

A percentage on a datasheet is meaningless until you multiply it by the rail. The same ±5 % is 165 mV on 3.3 V but only 90 mV on 1.8 V.

Ripple and transients

Switching ripple, load steps and droop all eat the same window. They are measured peak-to-peak in millivolts, so the budget has to be stated that way too.

Headroom and thresholds

Regulator dropout and logic switching points are volt-level specs whose margins — a few hundred millivolts — decide whether the design survives a discharging cell.

From Datasheet Line to Scope Cursor

The workflow is the one you already use when reviewing a power tree: take the volt figure off the page, express it in millivolts, then compare it with what the bench measures.

Type the volt-level figure

Put the rail or the slice of it into the V field — 3.3 for the rail itself, or 0.165 for the ±5 % allowance you just calculated. The field starts at 1 and both boxes are live, so the millivolt side updates as you type.

Read the budget in millivolts

Results carry up to 8 decimals with thousands grouped by a space, and switch to scientific notation past 1e10 or below 1e-6 — useful when a leakage or offset term drags the figure into microvolt territory. A comma or a dot both work as the decimal separator.

Go the other way after measuring

The scope reports 42 mV of ripple and the spec sheet wants volts. Type into the mV box directly, or press the swap button (↔) to flip the direction — 42 mV becomes 0.042 V.

Drop the number into your budget sheet

Each field has its own copy button that puts the bare number on the clipboard — no unit, no spaces — so it pastes straight into a tolerance stack-up. Ctrl + C inside a field does the same.

Budget once, not twice: DC tolerance and AC ripple share the same window. If a rail is specified at ±3 % and the regulator alone is ±1.5 %, only the remaining 1.5 % is left for ripple.

Volt-Level Specs as Millivolt Budgets

These are the figures that turn up in most low-voltage designs, each shown as the millivolt number you would actually design and measure against.

Spec	Volt-level figure	Millivolt budget	On the bench
3.3 V rail, ±5 %	3.135 – 3.465 V	±165 mV	Total window
5 V rail, ±10 %	4.5 – 5.5 V	±500 mV	Generous — legacy USB/TTL style
1.8 V core rail, ±3 %	1.746 – 1.854 V	±54 mV	Tight
Switching ripple, 1 % of 3.3 V	0.033 V	33 mV p-p	Common self-imposed limit
ATX 3.3 V rail ripple limit	0.05 V	50 mV p-p	Published PC supply ceiling
FPGA core AC ripple allowance	0.017 V	17 mV p-p	Plus ~1 % DC regulator tolerance
LDO dropout, 3.3 V out	0.2 – 0.3 V	200 – 300 mV	Input must stay above 3.5 – 3.6 V
3.3 V logic V_IH / V_IL	2.0 V / 0.8 V	400 mV each side	Against driver V_OH 2.4 V / V_OL 0.4 V
Li-ion cut-off vs 3.3 V rail	3.0 V cell vs 3.3 V rail	−300 mV	Needs boost

Read the table as a single ledger. A 3.3 V rail with a 165 mV window that already spends 50 mV on ripple and 30 mV on load-step droop has under 90 mV left for regulator set-point error, feedback-divider tolerance and IR drop along the copper.

What the Converter Gives You

Budget in, measurement out

Both fields accept typing, so the same page serves the design direction (V → mV) and the verification direction (mV → V) without reloading.

Clipboard-clean figures

The copy button hands over the number alone, so a 165 mV budget pastes into a spreadsheet cell as a value rather than as text needing a cleanup pass.

Every voltage unit, both sides

Searchable dropdowns on each field cover all 12 units — kV and MV for the supply side, µV, nV and pV when an offset or noise term is the thing being budgeted.

Nothing leaves the browser

The maths runs locally in the page. Rail values from an unreleased design are never sent anywhere, which matters when the numbers are under NDA.

Millivolt Budget Questions

How many millivolts is 5% of a 3.3 V rail?

3.3 × 0.05 = 0.165 V, which is 165 mV in each direction — a legal band of 3.135 V to 3.465 V and 330 mV wide in total. That total is what everything shares: regulator set-point error, feedback-resistor tolerance, ripple, load-step droop and the IR drop between the regulator and the load. Tighter parts exist for a reason; the same ±5 % on a 1.8 V core rail is only 90 mV.

How much ripple in mV is acceptable on a 3.3 V logic rail?

A widely used working rule is 1 % peak-to-peak, so about 33 mV on 3.3 V. Published limits sit in the same neighbourhood: the ATX specification caps the 3.3 V rail at 50 mV p-p, while FPGA vendors budget far less for core and transceiver rails — AMD's Versal PCB guidance assumes a fixed 17 mV of AC ripple on top of a 1 % DC regulator tolerance, and transceiver supplies are often held to around 10 mV p-p. Measure with a bandwidth-limited probe and a short ground spring — a long ground lead can invent tens of millivolts that are not really there.

What does a 200 mV LDO dropout mean for battery life?

Dropout is the minimum input-to-output difference the regulator needs to keep regulating. At 200 mV, a 3.3 V output needs at least 3.5 V at the input; at 300 mV it needs 3.6 V. Since a single-cell Li-ion is usually treated as discharged at 3.0 V, the last 500 to 600 mV of that discharge curve becomes unusable — the rail sags out of regulation before the cell is empty. Dropout also grows with load current, so design with the figure at your worst-case current, not the headline typical — and that is why very-low-dropout parts (tens of millivolts) or a buck-boost stage earn their cost in battery designs.

How much noise margin in millivolts does a 3.3 V CMOS input have?

With standard 3.3 V thresholds — V_IH = 2.0 V, V_IL = 0.8 V — the margin depends entirely on how hard the driver swings. Against LVTTL-grade output levels (V_OH 2.4 V, V_OL 0.4 V) you get 400 mV on the high side and 400 mV on the low side. A rail-to-rail CMOS driver that reaches about 3.1 V and 0.2 V under light load leaves roughly 1,100 mV and 600 mV instead. Ground bounce, crosstalk and rail droop all subtract from those numbers, which is why a 50 mV ripple figure is not automatically harmless.

Why is op-amp offset specified in mV or µV when the rails are in volts?

Rails are set points and errors are absolute. An amplifier's input offset does not scale with the supply, so quoting it as a percentage of 3.3 V would be meaningless — it is a fixed few millivolts (general-purpose parts) or a few microvolts (precision and chopper-stabilised parts) referred to the input. It matters because gain multiplies it: 2 mV of offset behind a gain of 100 becomes 200 mV at the output, which is bigger than an entire ±5 % window on a 3.3 V rail. The same logic explains why drift is given in µV/°C — the number has to stay comparable no matter which rail the part runs from.