TI Stellaris Cortex-M3s obsoleted

[andersm]

I was surveying possible micros for a project when I noticed that TI have removed all Stellaris Cortex-M3 products from their MCU pages, and the few product pages I looked up all had the status changed to NRND. Looking around on their forum I found this thread which confirmed that this was not just some bug on their website.

This seems like a bizarre move on their part. The LM4F series is just getting into production, and is not even close to replacing the LM3S lineup, and the MSP430 and C2000 aren't very good substitutes either. This is also the second mass cull of the Stellaris series. They might just as well include links to their competition in their part selection tables.

[McMonster]

It doesn't surprise me with the size of erratas of some of their M3s. Maybe they decided it's not worth making any more chips if noone wants to buy them and started working on a new M3 series?

[TheDirty]

I think they inherited a lot of crap from Luminary Micro, which was the first to the CM3 game.  Now they've decided to cut their losses in the CM3 arena and bet everything on CM4F.  I think the CM3 market is saturated and there's too much competition for them to make any money there.

I just noticed this as well when I got my Stellaris Launchpad.  As nice as the LMF120 CM4F chip is, these chips are just too big for most of the things I do.

[PuterGeek]

Regardless of whether they inherited 'crap' from Luminary or not, I'm glad we didn't pick a Stellaris part for our last design. NRND parts aren't in parametric search and only five (5!) of the M4 parts are available. None of the active or preview parts include Ethernet, none will be physically smaller than LQFP-64 and they won't be pin compatible with M3 parts (per their 'official' statement on the forum).

NXP was able to make their LPC17xx parts backward compatible with the older LPC24xx but TI can't with Stellaris?

I would hate to have production using a Stellaris part right now. No official word on availability, long lead times for available parts, no pin compatible replacements. And how would TI like me to explain this to my boss? Do they really expect me to choose their parts going forward?

I've been considered a rebel without a clue by employers but never a fool! What would posses me to even consider TI given this fiasco?

TI was historically considered highly reliable if a somewhat stodgy supplier. Jack Kilby, a TI researcher, was one of the inventors of practical integrated circuits. TI parts were used in military and aerospace projects for decades and at one time they were the largest IC manufacturer in the world (lost the title to Motorola and then Intel).

Now they are a sad shadow of their legacy and falling further. It seemed they had found their niche in semiconductors but I guess that was just good marketing.

[nctnico]

NXP was able to make their LPC17xx parts backward compatible with the older LPC24xx but TI can't with Stellaris?

A quick side note: NXP reps like to say the LPC1000 series is backwards compatible with the LPC2000 series but they are not. Most of the peripherals are similar but there are differences like pull-up being enabled by default, peripherals shut down by default and some peripherals are changed radically. Overall the peripherals improved but you'll need to check your software for compatibility issues and you can't simple replace an LPC2000 with a LPC1000 even though it has the same package.

[andersm]

It doesn't surprise me with the size of erratas of some of their M3s.

If you've got an issue with erratas it doesn't look like you'll like their M4s much either.

[PuterGeek]

I was just using NXP as example than pin compatibility is possible, TI just decided not to do it.

The NXP parts are LPC17xx pin compatible with the 24xx parts but do require software changes. The LPC17xx parts are Cortex M3 (Thumb instructions only) and like you say the peripherals are different. The LPC24xx parts are ARM7 (mixed Thumb and ARM instructions).

I'm not sure if TI's reasons were

• an attempt to reduce the design cost,
• a marketing decision and/or
• nobody thought about it.

I could understand number 1 and I would be ticked about number 2. Number 3 would be evidence they fired the people with experience and ignore their customers.

Unfortunately, I'm guessing all three were at work but in the reverse order.

[nctnico]

Looking at TI's website it seems they have renamed some parts to 'Hercules'. TI likes to rename parts...

[andersm]

The Hercules parts are a separate line of MCUs for safety-critical applications. They've been around for several years.

[westfw]

What would posses me to even consider TI given this fiasco?

Well, you might be paying attention to the financial state of NXP and STM as companies...

[Bored@Work]

Well, you might be paying attention to the financial state of NXP and STM as companies...

But what does this help when it is TI behaving like dickheads^(TM), pulling a Maxim, and thereby creating similar uncertainty?

I'm not sure if TI's reasons were

• an attempt to reduce the design cost,
• a marketing decision and/or
• nobody thought about it.

It is always the same story. Some MBA punched some probably fake numbers in an Excel sheet^*, and convinced his boss that "something must happen". And they did this "change in promotional status only", believing their F4 vapourware will now sell like hot cake.


^*80% of Excel sheets contain serious errors

[poorchava]

long lead times for available parts

ever purchased (erm... tried to) anything in high volume from Maxim? TI is not that bad

[PuterGeek]

Well, you might be paying attention to the financial state of NXP and STM as companies...

Freescale is on the ropes too.

Maybe TI is having trouble digesting their acquisitions also. They lowered their expected EPS for the fourth quarter last week.

http://investor.ti.com/releasedetail.cfm?ReleaseID=725954

None of the majors will survive trying to consolidate and reduce expenses. I'm can't recall one of the large deals working in the long run.

[free_electron]

TI is pulling out of a number of markets ...

There is a big reshuffle happening right now in semiconductorland... It's a reshuffle that was long overdue. too many were gunning for the same pot. Now everyone has found the pot he wants to pursue so they are dropping the rest... Markets have changed dramatically the last 2 . 3 years.

[jerry507]

Like many have said, their offerings were pretty bad. I did do a design recently using a Stellaris part and it was pretty damn hard to find parts. It seems like most of that entire line hasn't been in production for a long while and only a few are available.

[nctnico]

After the collapse in late 2008 TI quickly closed many manufacturing plants because they expected the semiconductor industry to take a big hit. They couldn't be more wrong and they lost a lot of customers because of that. For a few years it was impossible to use TI parts in a design because they simply where not available. And maybe Stellaris wasn't very attractive. Over the years I've looked at the Stellaris parts for several projects (even when they where still sold by Luminary) but none of the parts was a good match for any project.

[jerry507]

I spoke with our sales reps at Arrow today and the story I was given is that they're all still in production as long as people are buying them. They won't be putting any more money into fixing any of the errata for the vast majority of the M3 products though. There are apparently a few new M3's coming out in some specialized areas though, but I guess they were more or less ground up redesigns.

This is good because I did a design with a TI Stellaris M3 within the last year and they'd like to run it for at least 4-5 years of production.

[PuterGeek]

I spoke with our sales reps at Arrow today and the story I was given is that they're all still in production as long as people are buying them.

I would take that with a grain of salt unless your volumes are tens of thousands annually.

TI [i]may[/i] continue to make them if there is high volume demand.

Since the M4 parts are made with smaller geometry than the M3s. I'm sure TI would prefer to close or upgrade the larger geometry production.

And I would expect Arrow's contact to assure them the parts will remain in production. They have no reason to say otherwise until obsolescence is announced. They could miss a high volume project or they could get stuck with inventory if they scare their customers.

The more I think about the way this was done, it seems clear TI wants the M3 parts to die. They reduce demand by 'hiding' the chips which avoids running off existing customers. And when the demand drops they can say 'we would have kept making them if there was any demand'.

I'm hating the business weasels more everyday!

[jerry507]

The intent is clear: new customers use M4 products. They're just not going to continue to piss money down the drain on poorly designed chips with a team that can't make them work right. Errata after errata was getting pretty old. Trying to get started with one of the M3's was a history lesson where you first learned what all was broken, hopefully before you did your design.

[nctnico]

Come to think of it... Its not the first time TI bought a bunch of 'broken' products. The formerly Chipcon RF chips have a long errata sheet as well.

[westfw]

TI may continue to make them if there is high volume demand.

Of course, the reality is that this is essentially true for all parts provided by all vendors, regardless of whether there is a "NRND" tag attached or not.  Some vendors have better reputations than others.  (Try sourcing, say, an AT90S2313 and a PIC16C84 (which were approximately concurrent when "new."))

[rsjsouza]

TI may continue to make them if there is high volume demand.

You are right; the NRND status is usually very far from stopping the production of the device... And sometimes you can either have only certain variants being obsoleted, while others are still in production - as an example check this 1990s DSP that has been in NRND since 2000/2001 and some of its variants are still in production...

They have no reason to say otherwise until obsolescence is announced. They could (...) get stuck with inventory if they scare their customers.

In all fairness, that is all anybody can say at this time... Also, I don't think their main concern is about stock, since they usually only have the most popular parts anyways...

The intent is clear: new customers use M4 products.

+1. I guess the intent is to let people know this as early as possible. Also, their obsolescence page mentions the time span before stopping production of a device, which helps with existing designs.

[poorchava]

It also depends on your contract with particular manufacturer. A large company or a one that relies alot on long term availability of components (cough... automotive..cough...) can get a supply assurance for more than 25yrs. Of course if it happens that you are the last customer buying particular part, and you are doing so in say 1k/yr quantity you can expect MASSIVE price increase (up to the point when you invest your big bucks for design validation and redesign the product).

how this works (theory):
-manufacturer announces 'NRND' status
-customers do not use the chip for new designs
-demand for the chip decreases as more and more old designs are withdrawn from customers' production
-manufacturer ramps down production to satisfy needs of customers still buying the part and increases the price
-some of the customers see the price increase and trigger product redesigns, others do not
-manufacturer issues a withdrawal notice with time gap as agreed with customer
-customer has time to calculate demand for the chip until they can accomplish the redesign and buy that number of chips
-chip is withdrawn

how this works (practice):
-manufacturer announces 'NRND' status
-customers still use the chip for new designs because that's what their customer requires or because designers/programmers are familiar with it
-demand decreases only a little, because people still use the chip for new stuff
-manufacturer ramps down the production which causes price increase, but most companies will happily pay slightly bigger price since they usually make large profit on electronic product anyway
-price increases until it hits the point where it actually makes financial sense to do a product redesign and validation (validation can costs hundreds of thounsands of $$)
-manufacturer further ramps down the production and issues withdrawal notice
-customers' management has their ass on fire (they have agreement for continuous supply of product to their customer for like 20 more years, which is not unusual for some industries) and they want to secure a stock.
-this makes higher management at customers even more unhappy, because company has to suddenly buy MANY already expensive chips.
-this happens to all companies which causes terrible lead times for the chip (90 weeks, anyone?), and of course manufacturer does not ramp the production up
-customers' customers see that product may go unavailable so they want to make some safety stock which results in increased demand on product
-chip buyer has suddenly increased demand for the chip and they run out earlier than they have planned. There is a backlog with end product
-finally someone from management has their ass on fire hard enough that they buy the chip from obscure source (chinese brokers, old-new stock, etc)
-quality of end product hits the floor
-chip buyer gets mollested/sued by end customer for bad quality and they finally redesign the product (incorrectly, because there is a manager with a whip standing next to each designer)
-[20 years later]
end-customer: our agreement is expired
chip-buyer: ok, how many pcs have you bought in last 10 years?
e-c: ummmm....500?
c-b: we have put $5M into redesigns and were selling products to you for $30 a piece as agreed....
e-c: ok, let's talk about the product XXX which had similar availability problems 5 years after the previous one....


You get the idea

[westfw]

how this works (practice)

You left out the (important) event where the manufacturer, or perhaps one of their competitors, introduces a new chip that is almost completely compatible, technically superior in almost every way (faster, more memory, fancier peripherals, lower power consumption, enhanced instruction set...) and cheaper as well.  So the the "product redesign" step gets (theoretically (1)) easier and more attractive.
For example, in hobbyist circles, it is pretty common to complain that beginners are still using PIC16F84 microcontrollers (because it was one of the first easily available flash microcontrollers, and it got a lot of press in hobbyist magazines and web pages), when there are MANY newer PIC controllers with the same footprint, enhanced capabilities, and lower price.  (Microchip, and vendors, still sell the F84, though.  You can compare vs the Atmel AT90S2313 vs ATtiny2313 situation for ... insight(?))

(1)  "Theoretically..."  In reality, there are designs that are very difficult to redesign for reasons other than the purely technical.  The weight of the certification documentation exceeds the current company workforce.  The source code has been lost.  No one who understands how it works is still at the company.  Re-engineering of any kind is contractually disallowed by the client of the producing company.  (and this is why certain fields reliant on "high technology" don't follow the same lovely price reduction curve as "high technology" in general.  Sigh.)

[EEVblog]

-price increases until it hits the point where it actually makes financial sense to do a product redesign and validation (validation can costs hundreds of thounsands of $$)

That's nothing.
$100K is only 1000 man hours at any sort of minimum sensible internal engineer cost rate.
So if one lousy engineer spends 6 months on something full time, there goes your 100K.
In a big company with big overheads, say halve that time.
That often why manufacturers can increase their old chip price even ten fold and still have customers buy it.

Dave.