I have faced the problems of component obsolescence from my very early experiences of electronics design. In some cases it can be trivial — a transistor being discontinued and a suitable alternative easy to find. However, that is not always the case. Even a simple transistor replacement could result in PCB changes being necessary if the package/pinout is not the same as the old part. In the case of a low power RF transistor or other specialized device, a suitable alternative may not exist so a redesign will be required.
With some of the more complex designs, I have designed a custom IC to replace the obsolete one. I have even had components go obsolete before the product they are in has entered full production. With long approval times for aerospace and medical products and a short notification of discontinuation, it is not uncommon. Also, aerospace and medical products often have very long production times. The Boeing 737 has been in production since 1968 and is still being made (but not without several system redesigns). I know of at least a couple of products I have been involved in which are still in production after more than 20 years — in one case with the same custom device at its core and the other has been the subject of a re-target to a new semiconductor process as the old process was discontinued. I also know of companies holding stock of processed silicon wafers to cover future needs. It is cheaper than holding stock of packaged, tested devices but still has a significant cost and carries the risk of not having enough good devices.
Can the problems of obsolescence be prevented? In the past, there used to be second sources for many devices but with modern, complex devices (which can make designs simpler and more cost-effective) there often isn’t a second source. If you restricted yourself to designing only with devices with a second source you would make life very hard for yourself and even so, there is no guarantee of supply. I once created a design using a digital to analog converter (DAC) which had three independent sources. It was on an approved military parts list. All three sources discontinued it simultaneously.
So, I believe obsolescence is going happen so you need to decide how to react. You could buy yourself some time by stocking up when you are notified of the discontinuation. You could even buy enough devices for the lifetime of the product but you would then have to pay for them in advance and store them for possibly many years and you would have to guess at how many you would need.
I personally don’t see the option of searching the world for the last remaining stock or reclaiming parts from old systems as viable options. Maybe it could make sense for low volume military applications. I think you need to “bite the bullet” and get on with the necessary redesign if you cannot find a suitable alternative. Alternative parts may be available which are better than the original although you may have problems with higher voltage systems which use +/-15V supplies. The trend is for lower voltage devices so the number of devices that will work from 36V combined supplies or higher is reducing, not increasing.
Emulating old parts
One option may be to use multiple devices to replace a single device. One example I worked on was the obsolescence of the ICL8048 log amplifier.
In that case, the function was duplicated with a matched transistor pair and dual opamp so wasn’t too problematical. With many newer parts being in smaller packages it is often possible to add more circuitry and not take up extra board area. If space becomes a problem, a small daughter board may solve the problem.
One of the biggest problems I have found with replacing obsolescent parts is having a specification for the system. While you will presumably know the specification for the device that is obsolete, that isn’t the same as the “requirement” for the part. A simple example would be supply voltage. If the obsolete device can tolerate 36V total supply but the application only uses 10V, then the specification you need to use is 10V not 36V, and so you have a lot wider choice of replacement devices. That is a trivial example which is easily answered, but a more complex one might be noise. An opamp with low noise might be obsolete but how low does the device noise need to be in your system? The low noise opamp might have been over specified or selected for a different attribute and just happens to have low noise. Low current noise might be more important than voltage noise. Similarly, the bandwidth of an amplifier is likely to be higher than needed but how high is needed? Often a fair amount of time is spent reverse engineering the existing circuit to generate a requirement specification where none exists.
In some cases, the problem is made worse by incomplete, missing or inaccurate schematics and PCB layouts for old designs. I have even found that IC GDS II layouts don’t always agree with the schematics supplied making a process re-target even more tricky. It is always worth checking the validity of any information you have before you start your redesign to minimize any surprises later.
With component obsolescence also comes the higher probability that, in desperation, you might see a component from a non-authorized source. Counterfeit electronics is everyone’s problem but there are resources to turn to for guidance. Here are just a few:
2014 IEEE paper, Counterfeit Integrated Circuits: A Rising Thread in the Global Semiconductor Supply Chain, the authors concede that there are few test solutions and effective avoidance mechanisms to date to address the major challenge of counterfeit ICs. And, while they maintain that there are some counterfeiting and detection technologies that show promise, the implementation of those technologies is not without its own challenges.
“The Escape That Never Happens (A Leak in Authorized Distribution)” Henry Livingston is a Technical Director and BAE Systems Engineering Fellow at BAE Systems.
“Winning the Battle Against Counterfeit Semiconductor Products”, the Semiconductor Industry Association (SIA)