HRV Calculator: RMSSD & RR Interval Analysis

Heart rate variability (HRV) measures how much the interval between one heartbeat and the next changes over time. It's not the average heart rate: two people can have the same average heart rate (e.g. 60 bpm) and very different HRV, because HRV describes the variability of those intervals, not their average.

The heartbeat isn't a metronome: the interval between one beat and the next constantly changes under the influence of the autonomic nervous system (its sympathetic and parasympathetic branches), breathing, posture and many other factors. HRV is the tool used to quantify that variability numerically.

This page describes exclusively the mathematical calculation of HRV from data you provide. It is not a diagnostic tool, it does not classify the result as "good" or "bad", and it does not estimate biological age, recovery state or cardiovascular risk.

For a broader guide on what HRV values mean and how to interpret them over time, read the dedicated article →

RR interval: the time between two consecutive R peaks on an ECG trace (the R peak is the sharpest deflection of the QRS complex, corresponding to ventricular contraction). It's the technically correct term when the data source is an electrocardiogram or a chest strap with ECG electrodes.

IBI (Interbeat Interval): a more generic term, often used by devices that detect the heartbeat with an optical PPG sensor (photoplethysmography, typical of wrist watches and rings) instead of ECG electrodes. A PPG sensor doesn't measure a true R peak; it estimates the timing of the beat from the change in blood volume in the tissue: the value it returns is conceptually an interval between beats, not strictly an RR interval, even though it's often called that in practice.

NN interval (normal-to-normal): an RR/IBI interval after the series has been cleaned of ectopic beats, motion artifacts, and missed or false peaks, i.e. after only intervals between "normal" sinus beats remain. Clinical standards (ESC/NASPE Task Force, 1996) define the classic HRV metrics, including SDNN, on NN intervals, not on raw RR intervals.

This tool works on the values you enter exactly as they are: it does not automatically distinguish RR from IBI, and above all it does not apply any cleaning to turn them into NN intervals. If your data contains ectopic beats or artifacts, they remain in the calculation: see the Limits section below.

RMSSD (Root Mean Square of Successive Differences) is the square root of the mean of the squared differences between consecutive RR intervals:

RMSSD = √[ Σ(RRᵢ₊₁ − RRᵢ)² / (n − 1) ]

where n is the number of RR intervals in the series, so there are exactly (n − 1) successive differences. Squaring each difference before summing it gives more weight to abrupt beat-to-beat changes than to a slow trend, which is why RMSSD is considered particularly sensitive to the high-frequency/parasympathetic component of variability, and is the metric recommended by the ESC/NASPE standards when working with short recordings.

This tool requires at least 2 valid RR intervals to calculate RMSSD (at least one successive difference is needed). With only 2 intervals, RMSSD equals the absolute value of the single available difference.

This tool also calculates the sample standard deviation of the intervals you entered, with denominator (n − 1): Bessel's correction, standard when the entered data is treated as a sample:

Standard deviation = √[ Σ(RRᵢ − RR̄)² / (n − 1) ]

where RR̄ is the arithmetic mean of the valid intervals. The (n − 1) denominator, not n, is stated explicitly because the choice changes the result (especially on small samples) and different sources/tools use different conventions without specifying it.

This value is labeled in the results as "standard deviation of the entered intervals", not "SDNN": see the next section for why.

SDNN is, by clinical definition, the standard deviation of NN intervals, i.e. of RR/IBI intervals AFTER removal of ectopic beats and artifacts (see above). SDNN reflects the overall variability of the recording, both the sympathetic and parasympathetic components, and is particularly sensitive to recording duration: an SDNN calculated over 24 hours is not comparable to one calculated over 5 minutes, per the explicit guidance of the ESC/NASPE standards.

This tool does not automatically remove any suspicious interval (see Data quality/Limits): it calculates the standard deviation on exactly the values you entered, artifacts included if present. Calling it "SDNN" without qualification would imply a data cleaning step that doesn't happen here.

The term SDNN is appropriate when: (1) the series has actually been cleaned of ectopic beats/artifacts using a stated method, and (2) the recording duration is known and consistent with the comparison you want to make (short-term with short-term, 24-hour with 24-hour). If both conditions hold for your data, you can legitimately call the value this tool labels "standard deviation of the entered intervals" SDNN.

If you measure HRV with different devices on the same night, expect different numbers even under identical physiological conditions. The main causes, documented in the consumer wearable literature (see the JACC review cited in the sources):

Sensor technology: an electrocardiogram or ECG chest strap directly measures the heart's electrical activity; an optical PPG sensor on a wrist or ring estimates the beat from blood flow changes in the tissue, with a variable delay (pulse transit time) relative to the actual electrical beat, a source of noise that ECG doesn't have.

Proprietary cleaning/interpolation algorithms: each manufacturer applies different filters and corrections (not publicly disclosed in most cases) before calculating the HRV it shows: how many intervals it discards, how it corrects them, which time window it uses.

Reported metric: some devices show RMSSD, others ln(RMSSD), others SDNN, others still a proprietary composite score (e.g. a "readiness score" combining HRV with other signals), values with different scales and meanings that aren't directly comparable to each other.

Measurement window and context: an overnight measurement, a 1-minute spot-check while awake, and a recording during physical activity produce structurally different numbers, even on the same person on the same day.

Recording duration: very short recordings (a few seconds/a few beats) give statistically unstable estimates: this tool shows a warning when the sample is short, without blocking the calculation.

Posture: vagal tone changes between lying down and standing; measurements taken in different postures aren't directly comparable.

Time of day: HRV follows a circadian rhythm; the most reliable comparison is measurement against measurement taken in the same time window (typically overnight, the most standardized in consumer wearable practice).

Sensor: see previous section: ECG vs PPG aren't interchangeable.

Artifacts: movement, ectopic beats, and missed or false peaks strongly alter RMSSD, because the formula squares successive differences: a single large-amplitude artifact can dominate the sum and distort the result more than it would on a metric less sensitive to instantaneous changes.

Four RR intervals in milliseconds: 800, 810, 790, 805.

Successive differences: 810 − 800 = 10; 790 − 810 = −20; 805 − 790 = 15.

Squared differences: 100; 400; 225. Sum = 725. Number of differences = n − 1 = 3.

RMSSD = √(725 / 3) = √241.67 ≈ 15.55 ms.

Mean RR = (800 + 810 + 790 + 805) / 4 = 801.25 ms. Derived heart rate = 60000 / 801.25 ≈ 74.88 bpm.

Standard deviation (n − 1 = 3): deviations from the mean −1.25 / 8.75 / −11.25 / 3.75; squares 1.5625 / 76.5625 / 126.5625 / 14.0625; sum = 218.75; /3 = 72.92; square root ≈ 8.54 ms.

You can verify this example by pasting "800, 810, 790, 805" into the field above (unit: milliseconds).

It does not automatically remove or correct ectopic beats, motion artifacts, or missed/false peaks: the values you enter are used exactly as they are.

It does not calculate frequency-domain metrics (LF, HF, LF/HF ratio): only time-domain metrics (RMSSD, standard deviation, mean, derived heart rate).

It does not know your age, sex, health status, posture, time of day, or the source device of your data: it therefore cannot contextualize the result against any of these factors.

A single result says nothing about recovery, stress, fitness, or cardiovascular health: the literature this tool is based on (see Sources) describes HRV as most useful as a personal trend over time, measured under comparable conditions, not as an isolated value.

There is no table of "normal" HRV values by age or sex on this page, deliberately: the available evidence does not support absolute values that are universally applicable across different people, recording durations, and devices.

All calculations happen in your browser, in this page's JavaScript code. No RR interval, no result, and no input is ever sent to a server, including a FitMesh server: typing or calculating never generates a network request.

No value ever enters URL parameters, cookies, localStorage, or analytics events. "Copy results" copies text to the clipboard only after you click the button, an explicit action, never automatic. "Download CSV" generates a file locally in your browser: there is no upload.

The "Share" button opens your browser/OS's native share sheet with a text summary of the calculated results and this page's address - never the raw RR intervals you entered. No data ever enters the URL: sharing happens entirely locally, through the operating system, not through a FitMesh server.