Dumbest "equivalent schematic" from a datasheet - how many errors can you find?

[magic]

Datasheet schematics happen (albeit not always) and so do small errors or sometimes deliberate omissions.

But "small" apparently isn't enough for TI. Here's their (simplified) schematic of TS321, a single operational amplifier equivalent to half of LM358.
One might think it should have been a job easy enough...

[fourfathom]

(#1) PNP output transistor base-emitter short.
(#2 Trivial) Missing Vcc connection dot.

[SiliconWizard]

Ah well, shit happens.
The not-so-good part is that it never got fixed (just checked). It's still in the current datasheet.

[magic]

I did take it from the current datash*t
But you aren't wrong, it goes back all the way to 2005.

Still one blatant error to go. Hint:
- are polarities of the inputs even right?
- would it be stable without a whole bunch more compensation?

OTOH, I just noticed that the connection of the current sink to output PNP base is not actually technically wrong, because the base is shorted to the output anyway. So perhaps one less than I thought

[bsfeechannel]

LM358 equivalent circuit for comparison.

[jpanhalt]

The ST schematic seems to have it right: https://www.st.com/resource/en/datasheet/ts321.pdf

Shame on TI.

[srb1954]

I did take it from the current datash*t
But you aren't wrong, it goes back all the way to 2005.

Still one blatant error to go. Hint:
- are polarities of the inputs even right?
- would it be stable without a whole bunch more compensation?

OTOH, I just noticed that the connection of the current sink to output PNP base is not actually technically wrong, because the base is shorted to the output anyway. So perhaps one less than I thought

Even if the base emitter short was removed there is nothing actively driving the output PNP transistor. That transistor would just be permanently turned on by the current sink on its base and short the output to GND.

[fourfathom]

Still one blatant error to go. Hint:
- are polarities of the inputs even right?
- would it be stable without a whole bunch more compensation?

The + and - inputs are indeed swapped. (pretty blatant hint!) 

Compensation?  That's above my pay grade.

[tooki]

So report the error to TI and see what they do. When I reported an error in the data on the website, they fixed it within days and wrote back to let me know. It’s eminently possible that nobody has ever reported the error.

[T3sl4co1l]

Are you sure this isn't a comparator?

So report the error to TI and see what they do. When I reported an error in the data on the website, they fixed it within days and wrote back to let me know. It’s eminently possible that nobody has ever reported the error.

I've pressed the "report" button in datasheets many times and haven't seen anything change, that I know of... but there's always hope.

Tim

[mc172]

I've pressed the "report" button in datasheets many times and haven't seen anything change, that I know of... but there's always hope.

I thought I was the only one!

TI have got to be the worst culprit - I swear they use unpaid interns to write their datasheets. The last one I noticed had a suggested value of 2k in place of what was meant to be 2R in the application circuits section.

[schmitt trigger]

Doesn’t anyone senior proofreads a datasheet? 

[magic]

Are you sure this isn't a comparator?

Of course, high DC gain and positive feedback at AC, how could I miss it

Doesn’t anyone senior proofreads a datasheet? 

That's why it isn't right to report errors past a certain threshold of severity.
The datasheet says as much about the manufacturer as about the product itself.
Other customers deserve to know

BTW, check out the schematic in OP07 datasheet from the same guys.
Been that way since 2014 apparently and no one cares.

[dmills]

Given that TI appear to have forgotten how to make chips, there may well be nobody there who still understands how to read a schematic!

I have a TI regulator in a design that only started working after I added two mysterious extra resistors found on the app note circuit (but nowhere on the datasheet!), most unimpressed.

Where are the Pease/Williams/Widlar trained ones?

[MegaVolt]

The AD8367 inverts the signal. There is no mention of this in the datasheet

https://www.ti.com.cn/cn/lit/ug/tiduc87a/tiduc87a.pdf
Figure 3 not correct. Although the article below is correct. I reported the bug 3 years ago
https://e2e.ti.com/blogs_/b/industrial_strength/posts/a-modern-approach-to-solid-state-relay-design

[Zipdox]

So.... is anyone gonna email them to have it fixed?

[tooki]

The AD8367 inverts the signal. There is no mention of this in the datasheet

https://www.ti.com.cn/cn/lit/ug/tiduc87a/tiduc87a.pdf
Figure 3 not correct. Although the article below is correct. I reported the bug 3 years ago
https://e2e.ti.com/blogs_/b/industrial_strength/posts/a-modern-approach-to-solid-state-relay-design

”Reported” by leaving the comment on the Web version, or reported by using the Submit Documention Feedback link inside the PDF?

[MegaVolt]

”Reported” by leaving the comment on the Web version, or reported by using the Submit Documention Feedback link inside the PDF?

Leaving the comment on the Web version.

[tooki]

”Reported” by leaving the comment on the Web version, or reported by using the Submit Documention Feedback link inside the PDF?

Leaving the comment on the Web version.

I wouldn’t expect that to be monitored, especially not for an error in the PDF. Use the Submit Documention Feedback link inside the PDF.

[MegaVolt]

I wouldn’t expect that to be monitored, especially not for an error in the PDF. Use the Submit Documention Feedback link inside the PDF.

I took advantage of this option. Thanks, I didn't notice that link in the pdf.

I don't know what to do with Analog Devices. Access to their site is closed for my country. Perhaps someone else will contact AD.

[EPAIII]

RE: "#2 Trivial"

IMHO, there is nothing trivial about this.

I have never liked the convention of drawing schematics without using the overpass style wire crossing where there is no connection. This NON-trivial error is the best argument for that dislike. And the fact that it has apparently survived over a long period speaks against any arguments that will be made against my dislike.

Not only was it an easy error to make, not only does that error render the circuit non-functional if followed, but apparently no one sees any real need to correct it even when told about it.

This style of schematic should not exist. It should have never existed. All existing such schematics should be properly updated with the overpasses ASAP by the original creators. And the original creators should not be paid a single penny for do this. And copies should be distributed by those creators IN PERSON, taking the time to insert them in place of the original, defective schematics in whatever form those originals may have taken.

Rid the world of electronics of this scourge! NOW!

(#1) PNP output transistor base-emitter short.
(#2 Trivial) Missing Vcc connection dot.

[T3sl4co1l]

Meh, overpasses are old fashioned (and still subject to transcription errors all the same).  Current best practice is to never use 4-way junctions, except for crossings.

Which, the above circuit seems to take advantage of otherwise: the output is dog-legged from the current sense B-E shorting wire, for instance.  And then the capacitor and IN+ cross in an orderly fashion as we expect.  But then Vcc goes into a damn current source, and all bets are off.

(Another reason against overpasses is the implication of orientation, ordering, or z-height, an irrelevant detail -- a minor one to be sure, but it does technically add to the visual noise.  Conversely, one might argue the 3-way junction rule, harnesses visual noise instead: such connections tend to be less symmetrical, arguably uglier -- which highlights their importance, in a way.)

Tim

[magic]

Which is funny because TI is normally very religious about avoiding four-way junctions.
I gather that they really meant to connect VCC to that damn current source rather than everything else


They are similarly obsessed with never drawing current mirrors with a single line crossing through all bases (like everyone else) but snaking around those transistors and branching off to each base

[Zero999]

- are polarities of the inputs even right?

I admit, the only error I saw was the inputs reversed. I can't believed I missed the missed the base-emitter short on the output transistor.

LM358 equivalent circuit for comparison.

The drafting errors are more obvious and indeed understandable, after seeing the original, nut why didn't they simply copy and paste it? It's probably even the same die as the LM358, just with only one half connected.

[Ed.Kloonk]

I did take it from the current datash*t
But you aren't wrong, it goes back all the way to 2005.

Still one blatant error to go. Hint:
- are polarities of the inputs even right?
- would it be stable without a whole bunch more compensation?

OTOH, I just noticed that the connection of the current sink to output PNP base is not actually technically wrong, because the base is shorted to the output anyway. So perhaps one less than I thought

Fascinating thread.

My spidey sence makes me wonder if they might of done that to try and catch counterfeiters.