16
\$\begingroup\$

Suppose I have an idea for a product, where if it were successful, would be in production for 5-10 years. How can I make decisions on what parts to use now such that they will still be available to order in the future?

I'm not old enough to know, but I have heard from those that have been around long enough that some parts, such as some based on the Motorola 68HC11, have been able to stand the test of time and are still available today in pin-compatible and (roughly) code compatible packages and variations.

ARM has looked extremely attractive to me lately, a lot of Cortex-M's seem to fit perfectly for an idea that I have but what guarantees do I have that a pin-compatible variation of a particular micro-controller will still be around in 5 years? Or 10 years? How do I even begin to evaluate this? What are the key factors and does anyone keep data on a chip's production lifetime?

\$\endgroup\$
  • \$\begingroup\$ I'm going through the same process now, and would be interested in what you end up doing to solve this. \$\endgroup\$ – RCProgramming Oct 16 '11 at 20:05
  • \$\begingroup\$ @RCProgramming , I was just curious about what other experienced users here knew about this problem and how they solved their problem. I haven't had to deal with this quite yet. The answers here have been excellent and something that I will revisit when the time is right. \$\endgroup\$ – Jon L Oct 20 '11 at 19:53
13
\$\begingroup\$

There are no guarantees, not even for the 68HC11. Ask your supplier what the policy on obsoleting parts is. Usually they send a notification with a last buy date. You'll have to buy sufficient parts to cover the remaining production years, or at least until you have a redesign ready. Depending on your relationship with your manufacturer (read: how many parts you purchase per year) you may get an early warning.

On one occasion, for a custom IC the manufacturer didn't discontinue the part, but raised the price to such ridiculous levels that we decided ourselves to stop production. This was an IC produced with an older process which didn't have much production anymore.

Like Olin says, look out for second sources. If your supplier discontinues a part you may still find it at other manufacturers. But scrutinize datasheets. Sometimes second sources aren't exact copies, and the details may need a engineering changes in your design. If you're lucky this is just a resistor value, if you're out of luck this could be an extra resistor.
Also, changing manufacturers may also imply a different price, and a (much) higher price for the part may be a reason for a redesign, especially if you're running large production.

edit
Mike mentions the Flash memory market as notorious for its volatile availability. This is probably due to continuous advances in the field, especially in memory size.
Also expect short life times for emerging technologies, like OLED. I've had OLED modules becoming obsolete before we finished our design!


Further reading
NXP sample product discontinuation notice

\$\endgroup\$
  • \$\begingroup\$ Some years ago I had a visit from a technical sales representative of a major EPLD manufacturer who recommended a part for my application, assuring me that it would remain in production 'for the forseeable future'. He couldn't give me a price at the time and suggested I 'phone their sales office. After he left I did exactly that and was told 'Sorry, that part is obsolete' :o( \$\endgroup\$ – MikeJ-UK Sep 14 '11 at 11:09
  • \$\begingroup\$ @Mike - Ouch, that's pretty bad. I always ask for this kind of information in writing, so that they're forced to check their information. At least if they want to keep their job. That said, my experience with FAEs (avoid the sales guys!) from major distris (EBV, Avnet) is that they're well informed and more open than sales people. \$\endgroup\$ – stevenvh Sep 14 '11 at 11:18
  • \$\begingroup\$ The Flash memory market -- NAND, NOR, serial, whatever -- is notorious for volatile availability. You may need to do "lifetime buys" at the outset if you can't get parts that fit a standard that is likely to be followed, like ONFI or most serial flashes (i.e. non-Atmel-DataFlash). We've had a lot of trouble with this. \$\endgroup\$ – Mike DeSimone Sep 14 '11 at 13:03
  • 2
    \$\begingroup\$ Don't assume that they will send YOU the last time buy notification. While they might try, errors do happen and balls get dropped. It's up to your purchasing people to keep on top of things to avoid a crisis. \$\endgroup\$ – user3624 Sep 14 '11 at 14:32
  • 1
    \$\begingroup\$ @stevenvh All I'm saying is that despite everybodies best intentions, mistakes happen. I've had exactly that happen to me, several times. And while it would be nice to be able to rely on other mfgs/reps/distys/FAEs/Etc to give you timely information on your parts, if your company's health depends on availability of those parts then your company should be pro-active in monitoring that. This goes doubly-true if you are buying from a small company, or you are a small company (I.E., your volumes are low). \$\endgroup\$ – user3624 Sep 15 '11 at 13:10
5
\$\begingroup\$

As Steven said, there are no guarantees, but there are things you can look at to make some judgements.

First, how mainstream and multi-sourced is the part? Something like the 74xxx series logic chips have been around since silicon was discovered, are produced by various vendors, and will probably be around in some for at the end of time along with cockroaches and smelly socks.

However, many parts, although popular today, are single sourced. Microcontrollers are like this since every vendor has their own sets of wrinkles to add. All you can do is look at the company's history. Individual microcontrollers due to the nature of their technology go obsolete quickly. See how the company has dealt with that over the years. Some are significantly better than others. You can still today buy the long obsolete Microchip PIC 16C54 or 16F84. There are newer parts that cost less, do more, and fit in the same footprint, but Microchip understands the nature of long term products and these are still available.

Also keep in mind that ARM is just a architecture, not a particular part. There will likely be ARM-based products for a while yet to come, but that is of little relevance to your design. You need the specific individual part you are using from that vendor to still be available in 10 years. That said, if the part does go obsolete in 5 years there will more likely be a reasonably compatible replacement if the original part was from a major family.

So the short answer is you have to do your homework and study up on how well the vendor has dealt with this issue over the last 20 years. If the part you are considering is from a new vendor that doesn't have at least a 15 year history, then pick one that does if you're worried about longevity. In the stock market they say past performance is not guarantee of the future success. That's true, but it means more in the semiconductor business where each company has evolved a sort of corporate philosophy.

\$\endgroup\$
3
\$\begingroup\$

(in some sort of order of effectiveness)

  • Don't design to the edges of the specs of your part. If you design for your NPN transistor to have a Beta of >100 and a Icmax of 100 mA you will be able to switch to a different part very easily. You might even explicitly design for more than 1 type of part.
  • Look for components that are manufacturered by different companies (second sourcing). I think I can make a safe bet that the 555, 7805, ULN2803 end even the 741 will still be available 10 years from now. But be aware of subtle differences of the 'same' part from different sources.
  • Look at the track record of a company. For instance, Microchip's PIC chips are not the nicest CPU designs, but the company has a great record for keeping old parts available. Motorola, Atmel and Dallas have a very different reputation.
  • Make your design modular, so when one component needs to be changed you don't have to re-design the total product. For software you can do a lot to make it more easily portable over let's say 32-bit microcontrollers (ARM, cortex, PIC32, AVR32, ...)
  • For a cheap component that is vulnerable to becoming unavailable you could consider stocking it yourself.

Note that you won't get something for nothing: these measures will cost you, so you might need to quantify your need for long term availability, and weight it against the cost of (partial) re-design. (IME re-design is not that costly, re-testing, re-rectification etc. are the real costs)

\$\endgroup\$
  • \$\begingroup\$ 741 will be there for long time, but it's a long time that it sucks :) \$\endgroup\$ – clabacchio Apr 25 '12 at 8:24
2
\$\begingroup\$

Talk to the sales reps of the different companies and ask about their track record for part lifecycles.

Some companies have a nasty history of discontinuing parts, or particular package styles. Others -- like TI and Microchip -- are pretty stellar in having long life cycles. You can still buy PIC16F84 microcontrollers -- and I can remember buying them in college in 1995 (16 years ago!) when they were fairly new.

\$\endgroup\$
  • 1
    \$\begingroup\$ No, don't ask the sales rep about their company's track record. Of course they're going to say everything is great and how wonderful the company is. In the off chance one tells you about real problems, there might be some little information in that although they'll also tell you how that's been fixed and how wonderful everything will be going forwards. Asking a sales rep for a honest answer about company problems is a really dumb idea. \$\endgroup\$ – Olin Lathrop Sep 14 '11 at 14:38
  • 1
    \$\begingroup\$ I don't ask sales reps about problems, I ask them about part lifecycles and obsolescence. From my experience, what they say jives with reality. The companies that are good at keeping parts around can back up what they say with data. With the other ones, the reps tend to sidestep the question. Of course, you have to read between the lines a little bit, and if you're new to the game this is harder to do. \$\endgroup\$ – Jason S Sep 14 '11 at 16:35
  • 2
    \$\begingroup\$ Don't talk, write. If you talk it on the phone, there's no record of the conversation (unless you record it, of course). That way, they have time and incentive to make sure that they're correct, instead of making educated guesses. \$\endgroup\$ – Kevin Vermeer Sep 14 '11 at 20:13
1
\$\begingroup\$

Also consider how popular a part is, and what markets it is used in. Manufacturers respond to demand, so if demand dries up, so will supply. For example, a processor used in a smartphone could be obsolete next year, as sales of that smartphone end. However a processor used in a popular range of industrial controllers could be around for decades, as that market is slow to change

\$\endgroup\$
0
\$\begingroup\$

Take a look at industrial parts ranges.

In the industrial parts range, the parts tend to hang around for a relatively long time. They cost more, and are generally a couple of generation behind on technology, but they are going to be available for a while.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.