Analog/Linear EOL wave incoming?

[floobydust]

I asked Maxim about an IC's life-cycle, for design-in. It was a CMOS MUX, datasheet for consumer headphone switching, but I needed the low ON resistance for an industrial application. Would it be going EOL? Be around a few years? They wouldn't give me an answer. I just went with another make of IC that was newer, to end that worry.

I think semiconductor acquisition mania has made, as always, a huge mess. Redundant portfolios means they will EOL parts, usually the wrong ones. Jack up the prices to appease Wall Street. Bungle the fab forecast and run out of stock, long lead times. Massacre datasheets and app notes but get that new logo there. Fire all the "redundant" people, the ones who innovate and actually know what they are doing. Existing Mr. Big Co. staff gets stressed out and overloaded by the acquisition's extra workload. Bring in cheap, inexperienced staff. Contact ADI on a tech issue and find out for yourself, many layers of noobs there, I was shocked and in the end their IC failed at low temps.

I could go on, I've been employed through these acquisitions and watched the train wreck that they are.
It always ends up a fiasco. Integrating the new company, its corporate culture, quality and CADD systems, etc. is very difficult and usually impossible, although the big CEO and Wall Street could care less. Maximum profit is paramount. This is modern business.

[julian1]

Problem is all the "success" is only fueled by price increases. LT1028 (SOIC-8) is CAD $25, LT1115 (SOIC-16) is CAD $21, LT317 $13 FFS.
ADI I would say has out-priced themselves, literally 1.75x what they used to be. Very difficult to make a cost competitive product using their parts.

If ADI sold off LT's fab assets - then can these (process node/fab tied) products even be manufactured again in future?
Or are they selling out of an inherited and fixed inventory that will eventually be depleted to zero?

Perhaps ADI intends to keep increasing price as inventory goes to zero, with the expectation designers will switch over/substitute ADI's part catalogue.
 

[twospoons]

Having been hit by the EOL hammer a few times this year, we've now started running our BOMs through Silicon Expert* to get a heads-up on parts that are, or may be near EOL.  Its flagged a few things for us so far, so we can design them out before mass production.

*not affiliated with this company in any way apart from being a customer.

[rsjsouza]

If ADI sold off LT's fab assets - then can these (process node/fab tied) products even be manufactured again in future?
Or are they selling out of an inherited and fixed inventory that will eventually be depleted to zero?

Perhaps ADI intends to keep increasing price as inventory goes to zero, with the expectation designers will switch over/substitute ADI's part catalogue.

The ability to manufacture a part to exact specifications is quite dependent on the fab itself and can cost quite a lot of money in tweaks, runs and prototypes until it gets fully qualified in another fab. However, what some companies are realizing is that a specialized manufacturer such as TSMC is quite capable in doing this job with certain advantages (larger wafers and higher precision and yield).

Such decisions come with careful considerations regarding the existing fab's ability to keep churning enough different part numbers to justify its cost. For the most profitable parts, this qualification process most probably already started in another fab and, for the parts that do not sell well, indeed they will keep selling whatever wafers they have in stock until depletion.

I can only wonder if the profit of specific parts such as the LTZ1000 will justify its qualification efforts on the new fab... Or they will be extinct.

[julian1]

I can see that digital/logic/HDL can be moved/mapped between PDKs/ processes relatively more easily, because the design mostly works at an abstraction above the physical process.

But do fabs like TSMC offer customer tuned high-voltage bipolar processes?
LTs reputation and interesting parts are mostly analog.
Analog is such a niche field, I wonder if it wouldn't be regarded as a distraction from a business point-of-view.

Eg. For a specific op-amp, the datasheet pre-specifies the parameters for input bias current, capacitance, input noise etc - and these all relate back to the input super-beta bjts/jfets and the bipolar process used.
It's hard to see how these products could be duplicated well-enough to match the datasheet spec, without exactly duplicating the underlying physics of the manufacturing lab.

It is true that some parts have second/multiple sourcing options available.   
For example - lt1013 is supposedly made by both TI and LT/ADI.
But I (no industry knowledge whatsoever) would still suspect it is more likely an aggreement/arrangement was in place between the two companies - to share the fab/process run - 
but with enough legal separation - to allow each company to represent themselves as an independent manufacturer.
At least enough to win contracts where alternative sourcing was deemed a requirement.

[coppercone2]

they licensed production. This is normal in America. You can also license sales.

This allows for someones own marketing (i.e. field specific, hard to get through to certain people). Its kinda shady but there are companies so dependent on other companies they will only buy somethign with that companies name plate on it because their locked into their ISO stuff and support. So if you sell the design to someone, they can sell it to them, no one else can.

it allows for your own R&D into manufacturing (making it cheaper). And you have a fallback ,that is, you can ask the original makers if its good (a semi independent review process.. so not ideal, but you get the 'experts' in their own domain) to assure customers

it redirects tech support to the company that made it (they might be in a position to offer better tech support, or have a market segment with demanding customers that you don't wanna deal with because their like that). You can also use this to isolate a specific market segment from your general 'reputation' to some extent.


it allows for modified versions to be made that are not really convenient for the original maker (i.e. say someone wants a 150C version or whatever, special package, etc)


While it seems kinda stupid, with how RIDICULOUS some company internal procedures and requirements can be, its just another way to sell a product... and it almost makes sense.

Getting accreditation for something or another can just be very difficult, unpleasant for your company, etc. It can change the entire work environment, culture,  etc. At some level the certifications are basically like... life philosophies... not all are compatible.... when it has to do with perception of risk (which is actually dependent on situation). For instance, certain qualifications.. if you have it setup for certain styles... won't let you have any machine around, even in R&D, that does not have say a training, use, quality document... it CAN be useful in some chaotic crappy company, but with responsible people... its enough to say "don't touch that". It starts to feel like someone has EXTREME ocd when it comes to some 'directives'. 

So imagine you have a happy work force and a well functioning company with a good history but someone won't buy from you unless you get some kinda management-related cert that seems generally preposterous. If your smart you can license the base design to someone, and have them figure it out. They might even turn a profit at the end of it. You make money doing nothing. If your psycho you can just try to implement it yourself and sink the company to get a small market segment that might decay into nothing within 5 years. (i.e. in the case where it goes from "strict enough and well proven" to "paranoid kafka-esque nonsense".

Or something with the military.. like.. they want everyone to have clearance, including the mail room, because their scared of something. Its so ridiculous to make that step that it just makes sense for someone else that can do it to do it. win win. You get the biggest candidate pool and highest efficiency by hiring the most qualified. Someone else can figure out how to do it with a reduced labor pool that has some kind of special requirement. 

or legal, if they want everyone like your techs to have a PhD. In that case, license it to some whackos and forget about it. 

You need to watch sales people because its a "hot shot" move to promise some kind of huge profitable simple looking deal, until you read the fine print, that it requires nude virgins in a clean room....


And I think making crazy sales requirement is a buyout scare tactic by people looking to aquire, they make you feel like your never gonna be "at their level". Thats when you need to have some guts.. thats why I sometimes start feeling like somethings up when people start talking about ESD requirements. you don't need to walk around with a lighting rod up your ass.

[magic]

It is true that some parts have second/multiple sourcing options available.   
For example - lt1013 is supposedly made by both TI and LT/ADI.
But I (no industry knowledge whatsoever) would still suspect it is more likely an aggreement/arrangement was in place between the two companies - to share the fab/process run - 
but with enough legal separation - to allow each company to represent themselves as an independent manufacturer.

I doubt it. TI also makes OP07 and theirs is definitely a different chip than the original from PMI/AD. They have their own masks and it looks like they process them in their own fabs.

[julian1]

That's interesting, the successor to OP07, the OP27/37 are both ADI/TI too.

There are quite a few interviews and historical retrospectives with the founders of Linear-Tech on youtube.
Early on, much emphasis was placed on having one's own fab/processes, in order to achieve specialization, and to help carve out a protective moat, against competitor copying.
But perhaps some of that was over-stated, and intended more as a nice story for an investor audience.
And the reality behind-the-scenes was fierce reverse engineering/ of competitor processes, not to lose an edge, and even to enable duplicating other company's products if needed.

It's also possible that the dual-source mandate for govt/military contracts, could have led to quiet co-operation/ cross-licensing/ sharing of fabrication details.

LT1013/LT1006/LT1413 still seems pretty boutique/unique - bjt single-supply, with large input pnp long-tail-pair.
I searched but couldn't find if Noppy took die pictures of lt1013 variants.

[magic]

Everything's here, there is only LT1013 from Linear.

[David Hess]

I doubt it. TI also makes OP07 and theirs is definitely a different chip than the original from PMI/AD. They have their own masks and it looks like they process them in their own fabs.

Maxim also offered their own versions of the OP-07 and OP-27 series of operational amplifiers, which I assume they made themselves.

These parts are all based on a common 44 volt NPN only process, which allows low performance lateral PNPs, and is not too different from the process for the 741 or 324.  The secret of the OP-07 and other precision parts is more about design and layout and using more area, although I bet the alternates are built using finer geometry and take less space allowing smaller packages and duals and quads.

I suspect the alternative OP-07s actually perform much better than the original from PMI, which suffered from low open loop gain for a precision operational amplifier leading to poor linearity, but this would still allow the later improved parts to technically meet the original specifications.

[floobydust]

I've seen (e.g. TI) will acquire a company and take those IC designs and do a die shrink plus conversion to their own in-house standard fab processes. So of course the op-amp is changed. Their response is "it meets the datasheet specs" so anything off the datasheet or ill-defined can get corrupted, worsened and they feel it's acceptable.

The trend towards fabless semiconductor companies, I see it is still MBA popular. It's too hard to manage them, requires effort and nobody wants that whaaa need to get my game of golf in.
I still laugh the US government wants to bring back semiconductors to the USA, and here we have ADI close that LT fab in California, sell off the land, layoff the staff etc. Just like a can of soup, outsource it! "We aren't in the soup making business anyway".

I don't think semi's are the commodities that the bean counters and Wall Street view them as.

[rsjsouza]

I can see that digital/logic/HDL can be moved/mapped between PDKs/ processes relatively more easily, because the design mostly works at an abstraction above the physical process.

But do fabs like TSMC offer customer tuned high-voltage bipolar processes?

I don't know about TSMC capabilities for higher voltage, but Taiwan certainly has a significant presence in the higher µm nodes as shown in the article below - typically the nodes used by bipolar and analog ICs.

https://www.mckinsey.com/industries/industrials-and-electronics/our-insights/semiconductor-fabs-construction-challenges-in-the-united-states

I've seen (e.g. TI) will acquire a company and take those IC designs and do a die shrink plus conversion to their own in-house standard fab processes. So of course the op-amp is changed. Their response is "it meets the datasheet specs" so anything off the datasheet or ill-defined can get corrupted, worsened and they feel it's acceptable.

If it is not well defined or not present on a datasheet, then it is a non-spec anyways. Sure, your design might rely on this to meet a specific parameter, but this can be severely affected across wafer lots anyways - depending on the popularity of the part, they can be manufactured years or even a decade later than the original. I have seen this happen in the same part with severely different datecodes.

[coppercone2]

I've seen (e.g. TI) will acquire a company and take those IC designs and do a die shrink plus conversion to their own in-house standard fab processes. So of course the op-amp is changed. Their response is "it meets the datasheet specs" so anything off the datasheet or ill-defined can get corrupted, worsened and they feel it's acceptable.

The trend towards fabless semiconductor companies, I see it is still MBA popular. It's too hard to manage them, requires effort and nobody wants that whaaa need to get my game of golf in.
I still laugh the US government wants to bring back semiconductors to the USA, and here we have ADI close that LT fab in California, sell off the land, layoff the staff etc. Just like a can of soup, outsource it! "We aren't in the soup making business anyway".

I don't think semi's are the commodities that the bean counters and Wall Street view them as.

Have you ever seen what kind of bedlam occurs when you out source simple containers?   

That is another delusion I see caused by too many layers of abstraction and assumptions, that containers are easy. Outsourced might fool the people in some 30 minute board room meeting, that's it. There is a value to every correct bend

The other thing is that they are too stupid to realize how the other country is going to scam them on simple things, that is, the machinery will deteriorate, it will not be repaired, spec will be loosened and that it will all be justified because someone else is going to try to get rich off it like you did. Then you core business is basically at the mercy of some guy skimming small amounts of profit from a company with the usual tricks.

If your product is simple enough that they can get the heavy machinery fixed in Pakistan, it might maintain quality,. As soon as some poor place has to buy spare parts from Switzerland your going to have a problem. I just imagined what would happen if you told some cheap place to "keep using Motorex", when the initial quantity runs out they would switch to used motor oil, cleverly sold as a 'motorex alternative'.


And honestly I don't expect the big companies to react favorably to any government regulation, they will keep doing what they are doing until you prevent them from doing it, its like trying to get a cat to stop eating lol. That plan for shutting down has been in the works for 5 years+, and the guy running the operation is thinking "maybe I can finish this before someone gets directly in my way". Unless the government made a specific deal with a company (i.e. it was formally addressed and negotiated), its just something they see with no bearing on their activities, unless they get scared for some reason. I think its like doing bad parking where you see the meter maid across the street. YOu know how long it will take for them to get to your car. Maybe the motormaid will get into a dispute with someone and get further delayed. Do you still go into the deli to buy a sandwich? Your hit the ground running buisness is war etc boss is gonna get the sandwich.


What is the immediate consequence for the company from the government actions? A few companies have long term goals and they might see a consequence strategically. Anyone else thinks "just 1 more quarter please". I think the only thing that could make them change their ways is if they think they can't make it through the quarter their in right now if they don't comply LOL ... something about stopping freight trains comes to mind too. I never get excited about government initiatives because usually they have a time line of several years (5+) to take into effect... chances are I will be dead or doing something else by then! They can grow new things pretty quick, with money, but usually their not stopping anything fast.. there is usually a huge consequence for making things illegal and immediate cease and desist.

[Oldtestgear]

Some second source agreements are a complete sham as all the parts are made on a single wafer line & shared between companies. This is usually because a key customer is demanding a second source for the part in question.  Look, company B also offers it. There is the second source. Utter rubbish of course but I know of major defence projects that swallowed the lie, right until the part went obsolete from both sources at once.

The deal between LTC & TI was done in about 1987 for a limited range of parts for tactical reasons. I thought that the IP was shared & not parts from the same fab but I might be wrong.

Company acquisitions rarely work for the end user, or many of the staff employed, but can look great for the next quarters earnings report.

Happily I have been out of the semiconductor industry for nearly 20 years & do not miss it.

Phil

[magic]

Noopy found an obscure NE1037 which was made with Linear's masks and possibly at Linear's fab, but the DIP package looks like typical Signetics from outside.

[David Hess]

I still laugh the US government wants to bring back semiconductors to the USA, and here we have ADI close that LT fab in California, sell off the land, layoff the staff etc. Just like a can of soup, outsource it! "We aren't in the soup making business anyway".

Trailing edge fabrication plants are often closed not because of anything to do with the fab itself, but because there is not enough volume to sustain the consumables so the supply chain feeding the fab dries up.

[floobydust]

LT and their fab was quite successful, lead to the $14.8B acquisition... volumes were fine enough. Then we have the ADI business model of outsourcing step in and shut it down.
"trailing edge fab" is perfectly acceptable in linear and PMIC etc. - this is not CPU's.

The true technology is in the ideas and creative world. When the fab is a short walk away from your office, you've got an idea you want to try and discuss, maybe to build it when they get a chance. It's like this when you are close to manufacturing, not an ocean apart.
Compare that to the mega-corp world of getting a project approval, budget approval, accounting cost-center codes, it's going overseas etc. takes many months - for a simple idea that might not pan out. They want a known sure-hit before even trying anything. So we have zero innovation possible because the value of trying and failing is intangible. The solution needs to exist beforehand- there can be no risk, no discomfort not knowing the idea is going to work. This is Wall Street.

When I worked with LT FAE's, they queried our endeavors. They heard the issues and ended up building IC's within about 1 year, and then expanding their portfolio into a whole lineup. I was quite impressed - that they were close to their customers, listened and thought if the application in an IC would benefit them, and they rolled out product. Customer's needs were a huge source for their product ideas.

Today, I look at the new (analog) parts coming out and they seem to be born in the boardroom. So a little off in the requirements but then the cost punch in the gut happens - really, you think the part is worth that much!? Cute part but not practical unless you are in a high margin industry. I would foresee low volumes then.

[Oldtestgear]

Noopy found an obscure NE1037 which was made with Linear's masks and possibly at Linear's fab, but the DIP package looks like typical Signetics from outside.

LTC sold many of their parts in wafer form, supposedly for hybrid manufacturers but a number of companies bought the parts for their own production. I worked for their European die & wafer distributor from 1988 - 2005.  We used LTC die to manufacture custom parts for end users that required either a special package or processing that LTC did not wish to offer.  These parts were NEVER marked as LTC because they did not control the full manufacturing process. Yes, LTC were aware that we were doing this & did not object as long as we did not infringe any trademarks or misrepresent the parts. It is not too surprising that some die found their way to a.n.other semiconductor manufacturer for either evaluation or some marketing activity.