Monitoring obsolescence of electronic circuits

[gunter]

Today electronic circuits have a life-cycle of 36..72 months. But several electronic product segments like automotive or automation need availability of the circuits for 10 or even 20 years (for extensions and repair).

At my company we seek an early warning system to monitor our whole range of components to avoid bad surprises and to launch redesigns early. Information like component status and years left until end.of-life acc. EIA724 would fit our need, but most of our suppliers do not provide this info. Commercial obsolescence tools only cover a percentage of our active circuits, and passive components like capacitors, coils, and electromechanical parts like connectors are not covered by any tool we have seen.

I am looking for colleagues working in this topic to establish an exchange of info. If you know any solution to this problem please let me know.

 

[buzzp]

The only thing I have found to address this issue is to NOT sole source components. During the design, it is worth the time to test it with different components. The more, the better. This method may not work for some items such as unique connectors, etc. It would be best to standardize these with the industry standard. Thats all you can do besides having a good relationship with your distributor and keeping abreast of changes in technology that will affect the parts you are using.

 

[electricuwe]

Since I'm designing electronics for industrial control, I'm faced to the same problem. Some weeks ago I took the opportunity to test the obsolence manager from precience (

http://www.precience.com/ObsolescenceManager.asp),

which is a very expensive tool. But, unfortunately the database they use seems to be not very reliable. I got wrong information on quite a high percentage of componenents, e.g. stating a component as obsolete while the manufacturer informed me via email that the part is still in production and will be manufactuerd for the next years.

So we decided not to buy that service but to continue the approach we followed in the past. Listen careful to the news forwarded through the supply chain. Since we only manufacture in small volume we usually buy parts to stock them if a part is going to become obsolete.

 

[repoman]

If you are sole sourcing ANYTHING, then you are setting yourself up for disaster, ESPECIALLY IF YOU WORK IN CONTROLS!!! Use a different design approach. Component Obsolescence should NEVER BE A PROBLEM. It sounds like you do A LOT of analog and NOT A LOT of PLCs or PLD's or other firmware. Remember, the only thing that firmware costs is time to develop it. Pick a ubiquitous chip like a motorola 68000 or a z80 or 8051, these are very cheap, and made by a multitude of manufacturers, how about the 6502, or PIC?

 

[electricuwe]

repoman,

unfortunately in the real world of electronics a MCU alone is quite useless. Of course we use a lot of PLDs in our systems but we also need ADC, Dac, optos, OP-amps,etc.

 

[designchain]

With the rapid decrease of component lifecycles (we have seen the frequency of EOL notices from the component supply base increase linearly for many years, and accelerate dramatically over the last few years), many OEMs have been caught off-guard with this problem.

There are a couple services that provide obsolescence and product change notice information, if you’re unable to get them from your suppliers or contract manufacturers directly. The ones I’m aware of are

http://www.totalpartsplus.com,

which is the base for the Precience product mentioned above, and

http://www.pcnalert.com.

The former also claim to have an obsolescence prediction algorithm. I have no experience with it and can’t vouch for it’s accuracy but would recommend a healthy dose of skepticism for many component types. Neither provide the breadth needed for most systems, missing many passives, interconnect, and smaller semiconductor suppliers. But if they can get you 80% of the way there, you only have to deal directly with the remaining 20% of your supply base. While it is absolutely critical that you receive, analyze, and determine appropriate action on each and every one of them for every component you receive one on, depending on these alone is reactive and invariably results in fire-fights and interdepartmental resource hassles to resolve the problems caused by them in the astonishingly short 90 days most give you.

The basic problem is that in order to design a product that has a lifecycle greater than the lifecycle of the components used to design it, you need to understand both and plan accordingly prior to undertaking the design. Understanding component lifecycles takes time and effort. But understanding that some will go obsolete, and what’s most likely to and what they’ll be replaced by is important. So a key to implementing an “early warning system” is to understand the technology and product roadmaps of the parts you use and design products so replacing parts does not required a complete redesign and code rewrite. The recommendation above by repoman to use a common processor core is sensible but isn’t close to the whole answer, as electricuwe points out. Get periodic (quarterly/annual) roadmap presentations from your key suppliers; talk to their marketing and sales people about how they determine when and why to obsolete parts. Also talk to their competitors and understand the space. This is a key strategic activity for your component engineering group.

You need to plan for, and budget for, sustaining engineering to review market and supply base changes on a periodic basis and determine what, if any, incremental design and part changes are required. This also gives you the periodic opportunity to cut in cost reductions and bug fixes.

Yes, single and sole sourcing tends to entail greater risk than multiple sourcing but more often than not these days many important parts will be single or sole sourced. Having a good, strong risk assessment and mitigation program in place when high risk parts (and suppliers) are being identified and selected for a new product is important to plan potential courses of action to take should the part become unavailable (“made of unobtainium”, as a former manager of mine likes to say), or to reject the proposed part entirely. And that’s where a good supplier selection and qualification process comes in to play. You shouldn’t have to worry (too much, anyway) about getting in a bind if you select the right suppliers, are a good, strategic customer to them (few small companies realize that they can in fact be a strategic customer – it’s not just volume based), and negotiate a mutually acceptable contract. One think I keep seeing over and over are companies that view sourcing as a technical problem and try to resolve it only by a technical approach; also taking a business approach can minimize technical problems and help speed their resolution when they occur.

--
Mike Kirschner
Design Chain Associates, LLC

http://www.designchainassociates.com

 

[Chuckles159]

Good advice. Stick with standard parts as much as you can. 74XX and 40XX logic. For op amps, use parts with industry standard pinout equivalent to the LM324 and LM358. These parts will be in production forever. Buy CPUs from established families. Keep expertise available to port your firmware to a newer chip on a month's notice. Same for FPGAs. Passive components are not as big a problem as these only change about every 25 years or so.

 

[designchain]

Not sure that's the best approach, Chuckles159. More 74xx and 40xx series parts have gone EOL than have remained in production and the majority of top tier suppliers have bailed out on those technologies all together. The only ones that have an even remote chance of staying for the long haul are the standard gates and functions: 00, 04, 08, 32, 74, 86, and so on. Look for ones that keep being designed in new processes and new technologies like CMOS and 3.3V (and lower). Bus drivers/xcvrs/latches/registers like 240, 244, 245, 373, 374 and so on are also a good bet. Another problem with 40xx series (and the bipolar 74xx series) is that we're seeing more and more being outsourced to third tier foundries running on old, fully depreciated equipment. So reliability is in jeopardy as is maintainability of said fab equipment.

Passives no longer can be considered "stable" either. The 1206 ceramic cap hasn't been the "sweet spot" for a decade; 0805s are being displaced by 0603s so there's plenty of movement in the passive space (especially capacitors). There's lots of new passive technology and form factors available now - they have to make a profit now too and can't be depended on to keep the old stuff around forever.

I agree completely with your assertion about CPUs; this is one of the most dangerous areas. Coding in a high level language helps with porting, but it's not always efficient.

--
Mike Kirschner
Design Chain Associates, LLC

http://www.designchainassociates.com

 

[Warpspeed]

Thank goodness I am now retired as a hardware design engineer. This obsolescence business is becoming truly frightening these days.

My advice as Chuckles says is to stay with simple devices with industry standard pin-outs wherever possible. At least that way you have a chance of engineering in something else.

The parts suppliers simply look at from the supply and demand aspect and profit. They will suddenly stop supplying components as soon as the demand falls below the profitability threshold.

So the tip is to use whatever parts are high volume parts, and add to that volume. Things like 555 timers in 8 pin dip packages, are really obsolete dinosaur parts, but the demand is still there, so they are still readily available.

 

[Chuckles159]

As I stated previously, I find no problem with 4000 and 74 series parts. Even though many devices have been obsoleted, they are usually replaced with an equivalent higher performance device with the same pin out and a variation of the old part number. However, you cannot be religious about the exact part forever. For example, many 74xx parts are not available any more but you can replace them with a 74ACxx part that will generally work just as well. However, I have not run across a 40xx series part I used in 1978 that I cannot still buy now in one form or another. You just have to stay agile and stay away from sole source parts as much as possible.

 

[designchain]

Then is it a good idea to design a complex (or even simple) system with 7400 and 4000 series gates from the ground up? Is management going to allow that? Is it even possible? Just how is your product going to compete in the market when your competitor is using a single chip to achieve functionality that requires 50 or 100 (or more) discrete logic ICs (and the attendant added layout work and more expensive PCB) and has not only a smaller form factor but higher field reliability (the discrete design has more solder joints plus more silicon plus lower-rel silicon, as mentioned above), lower power requirements, better manufacturing yield, lower assembly cost, possibly lower overall product cost, and lower total cost of ownership? Availability is just one piece of the puzzle...an important piece to be sure but be careful about basing your design on it alone.

By the way, 74 series TTL and 74AC series CMOS have a few different characteristics. Not only is the VIH/VIL different (try 74ACT for better compatibility there) but in higher speed boards you might find signal integrity - related problems due to edge-rate differences and decoupling problems related to dynamic power consumption differences. Propogation delays are also different; I've seen plenty of poor async designs that depend on prop delays alone to achieve a certain timing. On the other hand the swap out is relatively simple if the design is low speed or static. There are also plenty of other, more recent, logic families available besides AC/ACT and HC/HCT that might have better properties for replacement of existing parts in current and future designs.

Staying "agile" has lots of different meanings and implications. Could you explain a bit more about your view of an "agile" design and how it's achieved by using 35 year old technology?

Staying "away from sole source parts" is a noble goal but near impossible these days if you have to build a competitive and differentiated product. Understanding and evaluating your risks during product design and component selection, working closely with the supplier, and planning mitigation strategies is part of how your competition is dealing with the problem. My dissertation above is really proposing that OEMs take a business approach that encompasses and comprehends technical issues instead of just treating it as a technical issue and address it by trying to avoid it.

--
Mike Kirschner
Design Chain Associates, LLC

http://www.designchainassociates.com

 

[Chuckles159]

Good points but for simpler designs, I still maintain that 4000 series logic is a good choice. I have had no problems getting all the parts I used in the 70s now and from major suppliers like TI. For more complex designs here "discrete" logic is not practical, you will have to assume you will have to design in a newer part every 5 years or so. Plan on it, document all the code so someone else can step in and port it to a newer device and don't be too offput by the idea that you have to re-engineer the product every few years to use currently available devices. After all, can you predict the feature set you will need for sure 3 years from now let alone 20 years. After all, I'm designing products now with DSPs that weren't even science fiction 10 years ago. However, I look forward to even more feature rich DSPs that will be available 5 years from now. If nothing else, look on the need to update a design every few years as job security.