EEVblog 1752 - WTF Texas Instruments BIG NE5532 CHANGES

[EEVblog]

Use this thread to discuss the actual NE5532 changes or general conversation:
Forum thread: https://www.eevblog.com/forum/chat/ti-ne5532-audio-opamp-changes/

Use this thread for comment related to the video.

WTF were Texas Instruments thinking changing the design and major specs of the 45+ year old jellybean NE5532 audio opamp WITHOUT changing the part number?!

New datasheet: https://www.ti.com/lit/ds/symlink/ne5532.pdf
Old datasheet: https://www.mouser.com/datasheet/2/405/ne5532a-556776.pdf
Change Notification: https://mm.digikey.com/Volume0/opasdata/d220001/medias/docus/5726/PCN20231114002.1.pdf
Onsemi datasheet: https://www.onsemi.com/pdf/datasheet/ne5534-d.pdf

[Unixon]

This is totally unacceptable. Such changes require changing the part number. Doing it the way they did is short of a fraud.

[Unixon]

I guess old parts are now a gold equivalent of sort 

[ITDH]

I've got caught out last year when TI removed the Offset adjustment pins on the TL071!!!!! LIKE you don't get more "Jelly Bean" then the TL071 !!!

You can still purchase the "Original" TL071 version from TI but you really have to take care with the last letters of the part number!!!

TI don't (or didn't at the time) make it clear on there TL071 Datasheet, so take care...

[ArdWar]

This being TI I fully expected them to change the process into CMOS or something.

Silver lining it's still BJT.

[Psi]

Speculating a bit.

It's probably being driven by an extremely fast push to bring newer process nodes online to take advantage of AI demand.
To do that you normally close your oldest nodes to make room, or at least the node making the least amount of money now and going forward.

So they have to transition existing parts to new nodes quickly so they can close the node for upgrades and they're probably doing that way too fast and making mistakes, or just not taking enough care with it, or pushing staff to get it done to unreasonable deadlines.

[The Chump]

RC4580 inside, from the horse's mouth... in a PM of course, not on the actual public TI forums

[Agent24]

Well that takes TI instantly from a trusted brand to no better than whatever grab-bag you get from AliExpress, if they're going to change part designs like that and claim it has no functional difference, WTF indeed!

It's probably being driven by an extremely fast push to bring newer process nodes online to take advantage of AI demand.

Wouldn't surprise me, everyone seems to be falling over themselves to jump on the AI idiot-wagon

[floobydust]

Dave I appreciate you bringing this corporate malfeasance to light. It really is a mess.
TI needs to go outside of their corporate bubble and realize they are unquestionably on Bullshit Road.

[Kleinstein]

RC4580 inside, from the horse's mouth... in a PM of course, not on the actual public TI forums

It would need to be a slightly improved RC4580, to get the same noise specs, but otherwise most specs fit.

I guess old parts are now a gold equivalent of sort 

There are plenty of alternative manufacturers for the NE5532 that still produce the old design. Ti was anyway only a 2nd source - the original was Phillips/Signetics that is now part of ONSemi.
Just avoid the Ti part, as you don't know what you get. A new part may be bad, but it is worse when you don't know what you get. There can be excactly the same label on quite different chips. It needs the box label to tell the difference.

Ideally the distributors would just drop the Ti Ne5532 - with the surprise package they are pretty much junk and there are easy alternatives (other manufacturers).
Maybe this way Ti would get the message from the mess they make.

[golden_labels]

We know the what. But what is the why? Some ideas coming to my mind.

Fabs offering older processes are closing and TI is forced to move to a newer process? This would be the worst option. In this scenario Texas Instruments just happened to be the first and with time other manufacturers will follow.

TI is under a major financial stress and forced to reduce production costs? As long as the electronics community recognizes in time something is off, not tragic. Sad, but there comes a time for everybody. We may promise setting 555-based candles for them.

TI joins the investors-are-the-customer crowd? If that’s true, we are left with little more than a hope for tides to change. Switching to another manufacturer or part number isn’t going to work, as the next company may do the same.

[Yansi]

It is obvious, that the newer the product, the worse the spec. We have already likely been at the peak of engineering one or two decades ago. Products are no longer improving, they are getting back shittier, just for a short term profit.

And I mean, not only individual components, but whole products and their accompanying services and software/firmware.

We have already lost numerous irreplaceable components. From small signal JFETs, linear high voltage transistor, through robust power transistors, up to more complex ICs with amazing specs, that seemingly can't be recreated with modern fab technology.

[nfmax]

We have already lost numerous irreplaceable components. From small signal JFETs, linear high voltage transistor, through robust power transistors, up to more complex ICs with amazing specs, that seemingly can't be recreated with modern fab technology.

We have been here before. Companies like Germanium Power Devices in the USA and Semelab in the UK made a good living buying up obsolete production lines and keeping them ticking over to meet continuing demand. Someone needs to start doing the same now!

[Rafiki]

As it is not mentioned here - TI also destroyed their (actually Nationals) LM317. Be careful if you're using it and think about changing to OnSemi. For me National (TI) was second source for decades in different designs. Now deleted from all my BOMs.

To bad. National and TI both were quite good (exceptional good in case of National) suppliers over decades. But now it seems they wanna kill them self.

[GandalfDerGraue]

Isn't TI an american company? Fits exactly into my playbook for the USA these days.
Seems, the new way of thinking is "Who cares about contracts, agreements, or specs? Let's just do it totally different and look if we can get away with it."
This whole story proves my feeling right, that USA is no longer reliable in any way.

Maybe we should find a label, warning about reduced quality and reliability? Let's say  "Made in the USA"? 
Oh, this label already exists, you say? Well, fine - let's keep it that way!

[hans]

The new datasheet also presents conflicting information.

6.3.1 Unity-Gain Bandwidth
The unity-gain bandwidth is the frequency up to which an amplifier with a unity gain can be operated without greatly distorting the signal. The NE5532, NE5532A, SA5532, and SA5532A devices have a 10MHz unity-gain bandwidth.

[...]
6.3.3 Slew Rate
The slew rate is the rate at which an operational amplifier can change output when there is a change on the
input. The NE5532, NE5532A, SA5532, and SA5532A devices have a 9V/ms slew rate.

Meanwhile the spec table says 12MHz and 9V/μs. The older datasheet has the exact same chapters 6.3.1 and 6.3.3.

Also it mentions +/- 22V as potentially damaging in 7.2:

Supply voltages outside of the ±22 V range are able to permanently damage the device (see Section 5.1)

Clicking through to Section 5.1 then mentions +/-18V

This may be very nitpicking from me, but some copy-cats can get this kind of detail right:
https://www.lcsc.com/datasheet/C700139.pdf

Although most datasheets I saw don't mention this stuff in text at all. IMHO it doesn't really present significant new information/insight. To me it feels like filler to make the datasheet longer.. but then listing specs in 2 different places is a perfect source for ambiguity..

As a designer, I'd love to see what kind of test circuit these specs were measured. Meanwhile test conditions and max specs are also removed from the datasheet. If you're building 100k+ gadgets per year, you may want to know at what kind of sigma those max specs are presented as, and how that translates into your design. Plus of course, how they are tested.
Just having a "typ" spec is kinda pointless. A lot like the uCurrent offset voltage problems I remember seeing a few years ago..

Maybe they should just started listing their dozens pages on packaging after the typical performance graphs. /S

End of rant.

As for the other parts and spec changes.. maybe an alternative explanation is because some stuff gets over-specified? Like was the original part ever that good in the first place? I don't mean this to seed conspiracy.. but there are lies, damned lies and statistics. If you measured an anomaly from spec, have you measured 10+, or 100+, or more to proof it? Some specs aren't very easy to measure, like homeopathic levels of THD or -130dB crosstalk. And then the design also needs to be correct for that to work out.
In other words: plenty of room for debate, especially if the conditions are implicit (which is why I like to see them listed in the spec tables).

[chilternview]

It's probably being driven by an extremely fast push to bring newer process nodes online to take advantage of AI demand.

I don't think so. AI etc is on digital processes (finfet), which are probably on completely different fab lines.

I would say it's all about cost reduction under pressure from the bean counters, with die shrinks to get there, but with die shrinks the breakdown voltages get reduced.

[floobydust]

Why would TI make low margin jellybean parts at the most expensive place, the Texas fabs? https://www.ti.com/lit/ml/szzo116b/szzo116b.pdf
"Direct Funding $1.7B", "100s of millions of analog and embedded processing chips manufactured daily"
Have to look at the CHIPS and Science Act and I'd bet the number of chips produced is the metric, never mind their complexity.

[Xenon]

There are more silently changed TI parts.
I'm working at a medical facility as a service engineer for medical accellerators. About two years ago a driverboard controlling motors to position shielding blocks in the radiationhead started failing by oscillating or completely dropping out. These pcb's normally die from radiation damage after about 5 years. Now they failed after a few months months or sometimes even minutes, depending on where in the machine this pcb was used. It took the manufacturer a long time to find out a TI part (don't know which part, that information is not shared) had changed spec and caused these problems. They had to redesign the board, put it through expensive testing, and supply new boards to all affected customers.
They did mention the TI logo looked slightly different on the failing part, because it was now manufactured in another country.

Long story short: if you have a known reliable product  that's suddenly weirdly failing, and there is a TI part on it, suspect that part.

[Alex Nikitin]

As I’ve already suggested in the other thread, it might be good to create a pinned thread to keep a list of counterfeit TI parts with information on dates, ways of testing and possible replacements. NE5532 is a very visible example, however other not so glamorous parts could be considerably more dangerous to fake, such as INA133, used extensively in industrial electronics.

Cheers

Alex

[Psi]

It's probably being driven by an extremely fast push to bring newer process nodes online to take advantage of AI demand.

I don't think so. AI etc is on digital processes (finfet), which are probably on completely different fab lines.

I would say it's all about cost reduction under pressure from the bean counters, with die shrinks to get there, but with die shrinks the breakdown voltages get reduced.

I'm not necessarily meaning GPUs/RAM but all the other ICs that go into AI superclusters.
There will be a shit-ton of mosfets and dcdc regulators needed in multiphase configurations to pump all that current into the GPUs.
Probably other stuff as well for driving all the high speed interconnects

[santiall]

I've got caught out last year when TI removed the Offset adjustment pins on the TL071!!!!! LIKE you don't get more "Jelly Bean" then the TL071 !!!

You can still purchase the "Original" TL071 version from TI but you really have to take care with the last letters of the part number!!!

TI don't (or didn't at the time) make it clear on there TL071 Datasheet, so take care...

be very careful with the TL07x as they changed them from JFET to CMOS technology and the noise spec is much worse now, 37nV/sqrtHz instead of the usual 15nV
At lest they changed the suffix to H but still very tricky.

[Kleinstein]

be very careful with the TL07x as they changed them from JFET to CMOS technology and the noise spec is much worse now, 37nV/sqrtHz instead of the usual 15nV
At lest they changed the suffix to H but still very tricky.

They initially had the seprate making with the H for the CMOS version. However they now make it hard to tell if you get the 40 V max CMOS or 36 V max JFET version. With the destinction in the maximum specs the new parts are now all CMOS except for SO wide cases and the military version.
There was also a confusing PCN on the markings and versions. With the old fab changed they should only have a limited supply of the old JFET version.

They really need to add a chapter on how to tell the different chip versions apart.

[Jeroen3]

I showed this video to a quality manager of a competing chip vendor and his reaction was "yeah.. TI does that. I don't allow that". 

[pope]

As I’ve already suggested in the other thread, it might be good to create a pinned thread to keep a list of counterfeit TI parts with information on dates, ways of testing and possible replacements. NE5532 is a very visible example, however other not so glamorous parts could be considerably more dangerous to fake, such as INA133, used extensively in industrial electronics.

Cheers

Alex

+1