General > General Technical Chat

Do semiconductor datasheets suck?

<< < (7/12) > >>

T3sl4co1l:

--- Quote from: Kleinstein on November 29, 2023, 12:03:41 pm ---
--- Quote from: magic on November 28, 2023, 11:36:37 am --- There is the TL072H shoved into TL072 datasheet even though all its specs are separate from the latter.

--- End quote ---

The mess with TL072H is not just a data sheet problem. It starts with naming that is causing more confusion than good.  It is still another bad step to have a combined datasheet, as the old TL072 and new CMOS TL072H have not much in common. Not hat the TL072H is a bad part, it just got a bad name.
Combined datasheets that handle multiple parts are often a good thing, if the parts are really the same core  (e.g. 1/2/4 fold OP-amps or different voltage regulators).

For the OP-amps it would be good if they directly write if the use bias compensation or not. At least the newer DS give the signs and one finds out.

The main 2 points where I have seen repeatedly wrong values is with the current noise of AZ OP-amps and the DC SOA of MOSFETs. The wrong SOA curves are especially nasty, as this can (but one may be "lucky" with a prototype) cause failures that cause additional damage.

--- End quote ---

This is probably the most egregious example, being same-source.

Similar hazards can be found with new substitutes from other companies.  A recent one that comes to mind:
http://file.3peakic.com.cn:8080/product/Datasheet_LM2903A-LM2901A.pdf
The things that most jump out to me are:
- Input bias current is greatly improved
- Output current and saturation voltage are improved
- Response time is a little slower
- Waveforms are far too symmetrical: response time looks like a double-pole CMOS amp (decomp'd, of course), output is clearly CMOS type
- Input ESD diodes include supply (original allowed "over the top" inputs!)

It's a lie; it's definitely a completely different type, design, architecture.  It does look pretty good, in and of itself -- aside from the somewhat slower response, I don't have any particular problems with its specs, though it would be nice to have more performance graphs.  They also specify input reversal conditions (one input beyond Vicm still gives correct result; both outside, indeterminate).  It's literally just one thing: an LM2901/3A, it is not!

I had looked at a few other products of theirs, I think which seem alright mostly, but catches like this can be as simple as tripping up the less-careful employees in purchasing.

Tim

magic:
There are similar LM358 coming from China too.
It will be particularly "fun" when it breaks and somebody replaces it with the original part, but the design relied on lower bias or something.

Thankfully, nobody repairs anything anymore :phew:

T3sl4co1l:

--- Quote from: magic on November 29, 2023, 12:45:07 pm ---There are similar LM358 coming from China too.
It will be particularly "fun" when it breaks and somebody replaces it with the original part, but the design relied on lower bias or something.

Thankfully, nobody repairs anything anymore :phew:

--- End quote ---

Or relies on the class C output stage to...

wait. :-DD

([further?] implying that random CMOS versions are likely to be, if not universally, then in most respects, a strict improvement upon the original)

Tim

coppice:

--- Quote from: magic on November 29, 2023, 12:45:07 pm ---There are similar LM358 coming from China too.
It will be particularly "fun" when it breaks and somebody replaces it with the original part, but the design relied on lower bias or something.

Thankfully, nobody repairs anything anymore :phew:

--- End quote ---
This happens all the time. None of the look alike parts work exactly like the originals. Usually they work a little better than the original, as they were designed not to cause trouble as a substitute. If you design and test around the second source part, the original may cause some nasty surprises. The snag is often trying to find which of the alternatives has the worst spec in its data sheet, so you can design around that.

VK3DRB:
Most Chinese datasheets suck. Missing key information in almost every datasheet, but providing useless manufacturing inspection pass/fail criteria. You have to contact the Chinese manufacturers to get the info you need. Recently, after asking for 3D step files and pricing for a connector, the female sales person hunted my phone number down on WhatsApp and seem to want to set up a relationship as friends. Not bloody interested, I don't care how pretty she looked. I just want data, not a date.

American datasheets are pretty good generally. But not long ago I found a subtle bug in a Texas Instruments datasheet where they had cut and pasted info from another datasheet and forgot to change the data. Just this week, I found TI has been shipping a certain IC with the WRONG part number on them. One of the recipients has been Mouser. The chips don't work because they are in fact a different part! TI has quality assurance problems which seem to have been getting worse in recent times. Lets' hope things improve within TI.

Worst American datasheets by far are Honeywell sensor datasheets. Wrong data causing fatal damage to the device when used as instructed. A company that has no competent documentation QA. Sensirion is a far safer company to use and they have excellent datasheets and sensors.

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