The short answer
A single G figure is one number squeezed out of a whole curve. Two tests can report the same figure and shake a package in completely different places.
Ask for the profile, not the level. The curve, the frequency range, the time per axis and which way the box sat are what make a result something you can compare.
A test report says the package passed at 0.5 G. That reads like a specification, and it is closer to a summary.
Behind it sits a curve that says how much energy went in at every frequency. The single figure is what falls out when that curve collapses into one value, and the collapse throws away the part that decides what breaks.
Which matters when two quotes arrive with the same number on them. They can describe tests that load your product nowhere near the same way.
This reference covers where that figure comes from and what a full profile states. It also covers why the standards leave the severity to someone else, and what to ask for so you can compare one result with another.
What a Single G Figure Leaves Out
Start with where the number comes from. The root mean square acceleration is the square root of the area under the spectral density curve.
Area is the key word. Two curves can enclose the same area with completely different shapes, so they report the same figure.
That is not a technicality. One shape can pour its energy into a band where your product has a resonance, and the other can put the same total energy somewhere harmless.
A vibration testing paper puts the cost of ignoring this plainly. Comparisons made on that single figure ignore the spectra shapes, which can have a great effect on test results.
So the figure is useful shorthand between people who already share a profile. On its own, it does not let you compare two labs.
The level tells you how much. Only the curve tells you where.
What the Profile Itself Says
The curve has a name and a unit, and the unit is the thing to look for first. For an acceleration signal measured in G, the spectral density units are G squared per hertz.
The per-hertz part is doing real work. Normalizing to a unit bandwidth is what lets you compare two signals at all, instead of each one depending on how finely the measurement sliced the frequency axis.
A profile is then written as a short list of points on that curve. Each one pairs a frequency with a density, and straight lines between them give the shape the shaker follows.
So a complete specification is a small table, not a single value. If a report gives you one number, the table behind it is what you are missing.
| What to ask for | Why it changes the result |
|---|---|
| The breakpoints, in G²/Hz | Fixes where the energy goes, which decides what resonates |
| The frequency range | A band that stops short of your resonance never tests it |
| Minutes per axis | Severity is time as well as level |
| Which axes ran | Levels may legitimately differ between them |
| The named profile and its revision | Truck levels changed, so old and new figures are not comparable |
The Standards Name the Method, Not the Severity
This is where a lot of quotes quietly lose meaning. ASTM D999 covers vibration tests of filled shipping containers, and it carries four methods.
Two are repetitive shock, in vertical and rotary motion. The other two look for resonance, one on a single container and one on a palletized, unitized or vertical stack load.
Then comes the part worth reading twice. The standard says test frequency ranges and test durations are set by the performance specification rather than by the document itself.
Read that against a line like “tested to D999” and it names a method with no severity attached. The performance specification is the document that decides whether the test was hard or gentle.
Pre-set sequences work the other way around. ISTA 3A arrives with its levels already chosen, which is why a parcel test is quotable in a way a bare method reference is not.
The random method works the same way. ASTM D4728 covers random vibration testing of filled shipping units, and asks that the tests rest on representative field data.
Which means the profile should trace back to measurements of a real lane. Asking whose data it came from is a fair question, and the answer is sometimes nobody’s in particular.
What Sits on the Table
A level means little until you know what the shaker was carrying. D999 splits its methods along exactly that line.
Two of them work on a single container. The other two look for resonance, one in an individual container with its interior packaging, and one in a palletized, unitized or vertical stack load.
That last case is a different problem, not a bigger one. The bottom carton of a stack carries mass above it, so it can resonate in ways the same carton tested alone never will.
A box compression test measures what that static mass costs. The resonance methods ask what happens once it starts moving.
The random method follows the same logic, since its scope is filled shipping units rather than loose cartons. A unit load is the thing that ships, so a unit load is a fair thing to test.
Which gives you a plain question for any report. If it does not say whether a single box or a full pallet sat on the table, it has not told you which problem it tested.
There are two honest answers to where a profile came from. One is a published profile that someone else measured, and the other is measurements from your own lanes.
The second route is the stronger one. The recommendation is to measure a number of shipments that apply to your particular situation, then compile that data into customized profiles.
Published profiles get built the same way, from real measurements turned into an envelope that covers them. Knowing which lanes went into that envelope tells you whether it covers yours.
Random and Sine Do Not Convert
A tempting shortcut is to treat a sine result and a random result as two readings of the same thing. The random method rules that out directly: there is no direct equivalence between random vibration tests and sinusoidal vibration tests.
The testing paper goes further on what sine is for. Sine tests are not environmental simulations, so tying them to transport modes or distances misreads them.
That does not make them useless. A swept sine finds the frequencies at which a product amplifies input, which is exactly what the resonance methods in D999 exist to do.
The mechanics are easy enough to picture. Sine sweeps run across a range of frequencies to find where a product and its components resonate, and the test then dwells at those frequencies for extended periods.
A dwell is the harsh part. Holding a product at a frequency it amplifies is a fair way to find out what that amplification costs over time.
The split is clean once you see it. Sine locates the weak frequencies, and random asks whether the package survives a lane that contains them.
Duration, Axis and the Three Truck Levels
Time belongs in the specification, not in the assumptions. For random testing the paper treats the clock as real: one hour of test stands for one hour of transport motion.
Shortening a test means raising the level, and that trade has a stated ceiling. A time-compression ratio no greater than 5 to 1 is the recommendation for keeping the result valid.
Raising intensity also does something specific to the curve. The shape stays put and the whole profile translates up the plot, which is why a compressed test is not a different test.
Orientation is equally explicit. Random vibration may run in any axis, vertical or horizontal, and in any package orientation. Levels may differ by axis, depending on which environment the test stands in for.
One more thing dates a figure. ASTM has updated the truck profile used in ASTM D4169, and there are now three profiles with similar shapes but differing levels, run in sequence as low, medium and high.
So a number quoted against the older single profile will not line up with one quoted against these. Any comparison needs the profile name and its revision alongside the level.
Common Questions
Is a Grms figure enough to compare two vibration tests?
No. It is the square root of the area under the profile curve, so two differently shaped profiles can produce the same value while loading a product in different frequency bands. Comparisons made on it alone ignore the shape, which has a large effect on what fails.
What should a vibration test report contain?
The profile breakpoints in G²/Hz, the frequency range, the time run per axis, which axes ran, and the named profile with its revision. A level on its own gives you nothing to check against another lab’s result.
Can a sine test result be converted to a random one?
No. The random vibration method states there is no direct equivalence between the two, and sine tests are not environmental simulations. Sine finds the frequencies where a product amplifies input, and random asks whether the package survives a representative lane.
How short can a vibration test be made?
Shortening it means raising the level, and the published recommendation is a time-compression ratio no greater than 5 to 1. Beyond that the result stops standing in for the transport it is meant to represent.
How to weigh this page. It gathers what published test methods and technical papers state, so it can tell you which questions to put to a lab and what a complete answer looks like. It cannot tell you which level your product needs, because that comes from the lane, the product’s own fragility and the acceptance limits in a performance specification. Test standards are also revised periodically, so confirm the current profile and revision with the lab before comparing figures.
Key takeaways
- A single G figure is the square root of the area under the profile curve, so two different shapes can share it.
- The unit to look for is G²/Hz, and the profile is a short list of frequency and density breakpoints.
- ASTM D999 and D4728 set methods and leave frequency range, duration and level to the performance specification.
- There is no direct equivalence between random and sinusoidal results, so the two cannot be converted.
- Time counts as severity, with a recommended compression limit of 5 to 1, and truck levels now run as three sequential profiles.