4. If your device performance starts degrading seriously beyond a point, just cut off the X-axis early in your graphs before that happens. For example, if the frequency response is really bad beyond 500 MHz, just stop at 450 MHz, so people will never notice this deficiency in the datasheet!
If it's sold as a 450MHz part, what's the problem? Those who rely on how a part behaves "overclocked" should not be surprised if a different batch doesn't work as well as they hoped.
Those repairing old equipment (especially audio) sometimes experience the opposite problem: the modern parts perform way better at high frequencies than the original causing the circuit to oscillate.
If the part starts behaving in a rather unexpected manner just beyond a point, the datasheet better to get that documented as a warning. For a hypothetical example, an amplifier can either gradually roll off, or start peaking wildly, which can be a problem even if you're not trying to drive it off the limit deliberately. Some vendors clearly show that in the datasheet for the benefits of their users, others do not...
What I wanted to say is that, as users, don't always assume vendors are all that friendly.
Well, that does not make much sense to me: where would one stop pushing parameters outside of the designed range to try to get a more complete picture of the behaviour in these conditions? Imagine if, in your example, the part was pushed up to 10% above its designed spec (495MHz)? Out of spec parts are anyone's game. A 50A transistor driving 55A blows up? Should this be published?
I only said, from the perspective of a customer, it would be "better", or to be precise, "nice to have", I did not say a vendor is "required" to do that.
Of course it's all subjective, and there's no technical obligation for a vendor to publish typical out-of-spec curves (even when it is published, it's usually "typical" and not parts of the guaranteed specs). But then, there's also no obligation for a vendor to write any application note, design checklist, FAQ, product selection guides, etc., either. But still, in practice, if a vendor is selling some parts, all of these (including putting a typical out-of-spec curves in datasheets) are often done (though not always) in goodwill to make their components easier to use by their customers, reduce the numbers of oversights and unexpected "gotchas", and in turn, ultimately helping the vendor to sell more parts in the future.
Back to the hypothetical amplifier example. An amplifier can either gradually roll off, or start peaking wildly outside their rated bandwidth. As a vendor there's absolutely no obligation to put any additional information in the datasheet in a peaking amplifier. If unsuspecting users selected that amplifier - thinking that the bandwidth is sufficient with no intention to exceed its spec - in a time-domain application, but later gets horrible overshoots and ringings in the step response, of course it's all their faults for this oversight. On the other hand, in this situation, the vendor may as well include the peaking part in the "typical" frequency response curve (or do something similar, such as putting a warning in an application note) as both a friendly gesture to their customer and a way to reduce the unnecessary workloads of their field application engineers, so it saves everyone's time - which is a pretty common thing do do.
What I wanted to say is just that - don't get too accustomed with it, don't always assume the documentation provided by the vendors is all written like world-class manufacturers like Linear or TI, who are often friendly enough to do the extra works outside their obligations - especially when the part is made by a smaller vendor.
I don't think there's anything controversial with that statement.
A 50A transistor driving 55A blows up? Should this be published?
If this is a common oversight often encountered in real-world applications (e.g. from common but unexpected transients in a particular topology in an application), and if the vendor happens to be known in the industry for providing good customer support and documentation, and if the vendor is still motivated to maintain this reputation, then yes, it should be published. Or rather, it would be nice to get that published.
I remember seeing a Linear application note, talking about how switching transients by slow diode turn-on time can exceed the absolute maximum of the DC-DC chip's rating in a rather surprising manner in some situations, and that can be detrimental to the lifetime of the chip. The note was completed with test circuits, scope traces, and references to primary sources. Did Linear have any obligation to write and publish this App Note. Of course not. But it was also what differentiated Linear as one of the most successful analog vendors from the rest of the industry.
Home
Help
Search
Login
Register
