General > General Technical Chat

Do semiconductor datasheets suck?

<< < (10/12) > >>

shapirus:
Here's a fresh example, the one I'm reading right now.

TI's CD74HCT132: https://www.ti.com/lit/ds/symlink/cd74hct132.pdf

It's either that I don't know how to read and understand datasheets, or it has not been QA'ed and is a result of some quick copy-pasting.


section 9.2.1.2 Input Considerations


--- Quote ---The CD74HCT132 has CMOS inputs and thus requires fast input transitions to operate correctly, as defined in
the Recommended Operating Conditions table. Slow input transitions can cause oscillations, additional power
consumption, and reduction in device reliability.

--- End quote ---

Fair enough. Now, how fast should "fast" be? Let's see the mentioned table... only to find that it does not contain any specifications on the required input transition timings.

And the very next paragraph says:


--- Quote ---The CD74HCT132 has no input signal transition rate requirements because it has Schmitt-trigger inputs.

--- End quote ---

Of course, the Schmitt-trigger inputs are in the product title. But... why does the first paragraph that I quoted above state that the device requires fast input transitions? Looks very much like an overlooked copy-paste from another datasheet to me.


Next,


--- Quote ---Unlike what happens with standard CMOS inputs, Schmitt-trigger inputs can be held at any valid value without
causing huge increases in power consumption. The typical additional current caused by holding an input at a
value other than VCC or ground is plotted in the Typical Characteristics.

--- End quote ---

...unless I fail to see it there, the mentioned plots are not present in the Typical Characteristics section.


Finally,


--- Quote ---Refer to the Feature Description section for additional information regarding the inputs for this device.

--- End quote ---

But there is no section named Feature Description (do the datasheet writers not use tools such as LaTeX that can substitute the section names, links, and page numbers dynamically, referenced by internal id?).

Additionally, there is no specification for the minimum time during which an input must be held high or low to be registered as such, whereas it is one of important characteristics of logic ICs.

This one is definitely not among the better part of the datasheets that I've ever read.

Curiously, it does contain a "submit document feedback" link. I'll try it!

SiliconWizard:
Yes, it definitely looks like some half-baked copy-pasting from other datasheets. (And to be fair, for people who have written many datasheets, who has never done that, and forgot to adapt/remove some sentences that would not apply?)
Reviews should catch that though, but given the budgets allocated to technical documentation these days, not too surprising they haven't.
What sucks is when there is no real feedback channel from customers that would allow fixing documentation when it's incorrect. Often there is very little of that, and still for the same reason, budget.

But yeah, it's hard to write 100% correct technical documentation without any strong process to check for correctness. At least when it comes to "bugs" that lead to a tangible defect (software or hardware bug), observing the issue is not hard (fixing it may), but when it comes to strictly abstract design (which a doc is), it's very hard to catch unless you have a strict process for that. Imagine writing software code that never gets compiled or tested, shiping it to customers and aiming for 100% correctness. Nobody would expect that. Just to give some perspective.

So let's just give design teams some budget for proper docs.

T3sl4co1l:
What's funny is, the ancient datasheet (databook scans) that they replaced, probably contained all that, and more if you include the logic family introduction (which is basically what the application section is doing now).  Actually, they may well be rewriting (OCR, copy-paste, transcribe, whatever) that content, into the new format, and just not parsing it to find discrepancies like this.  Which, yeah, all comes down to department budget (and responsibility of management given that budget).

On the upside, NXP is a prominent competitor to TI in the logic space, and I haven't noticed gross defects in their documentation.  Sounds worthy of awarding some design wins eh? :)

Tim

SiliconWizard:
Which is Nexperia actually, isn't it? Yeah, same boat. And yes, their datasheets are usually clean. And most of their parts have Schmitt trigger inputs, by the way. ;D

T3sl4co1l:
Ah yeah, it is Nexperia with logic. Think NXP datasheets still turn up quite often...

Tim

Navigation

[0] Message Index

[#] Next page

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod