Accuracy of datasheets?

[analogo]

Datasheet are not only technical literature but also a kind of marketing material, companies have an incentive to show better-than-reality results.

In your experience, how accurate are the graphs in datasheets?

Do people mistrust datasheets and re-characterize components to find their real power efficiency or response times, for example?

[Rerouter]

front page claims you should not trust at face value.

If something has an asterix *, read the sub point,

If it has a min/max tolerance you can be confident that under those exact stated conditions it will meet them

For the charts, read exactly under what conditions they are testing, in that exact use case you can generally trust it. outside of it your on your own short of contacting them.

Also sometimes they leave out gotcha's, like a random capacitor that appears on all there example circuits, but never mentioned in the rest of the datasheet, half the time its there for a good reason, the rest of the time someone was paranoid. IF you follow the example circuits under the same conditions you can expect the charts to be truthful.

Yes it is marketing, but If you bought a 2A 5V power supply brick and it didn't meet what the datasheet promised, would you personally ever use that company again?, same for if you got burnt by something they didnt document?, this drives them to be possibly misleading as all heck, but never an outright lie.

[coppice]

Most component datasheets try to be fairly realistic. Its not like selling one offs, where product information typically strains the truth, or outright lies. Selling components only makes a profit if you can sell a steady stream month by month to a customer. If your datasheet is not realistic about parameter spreads, it might get you more design ins, but the flow of parts to production will usually soon cease if they are not consistently within spec.. Then you make no money. When a datasheet doesn't represent the production components its usually that the vendor was overly optimistic, rather that dishonest. Most respectable vendors will revise their datasheet information, or provide errata information, when that happens.

Vendors don't always specify their parts in the same way. Beware of parameters which carry a typical figure, but no maximum or minimum. Some will only quote parameters at 25C, and be rather vague about what happens at the extremes of temperature. These are issues of the incompleteness of datasheets, rather than honesty. You need to be careful than you do not rely on a performance parameter which is not claimed in the datasheet.

Always check for errata documents before committing to a part.

[SiliconWizard]

Datasheets from reputable vendors usually don't lie. That wouldn't get them very far.

What you always have to be careful about though are the conditions under which the given figures have been characterized. The test setups are usually described (at least roughly), so it's up to you to figure out how the figures can apply to any specific use.

[pelule]

the graphs in datasheets

are not confirmed as liable data, just typical.
Supplier don't lie, if the promote the best features of their product, they just don't tell you everything they know (typ. human behaviour).
Specifications are (at given conditions including notes) are trustable, as the supplier is liable for.
BUT you have to use the newest published data sheet.
/PeLuLe

[David Hess]

If the datasheet does not include at least a simplified schematic with input and output details, then I often end up making my own measurements and I generally do not trust these manufacturers, like TI.

For instance on an operational amplifier without input bias current cancellation, which direction does the input bias current flow?  If there is no schematic, then there is no way to know without testing because there is no standard way to specify this.

Some companies like TI have a long history of lying on datasheets.