Working out the true cost of managing obsolescence

Paul Boughton

What do you do when a part you have designed into your product is no longer made by the vendor? For fast moving consumer designs, this is rarely a problem, but as digital electronics makes its way deep into industrial, transportation and military designs, it is a key issue that designers around Europe are having to face up to.

The Component Obsolescence Group (COG), which brings together component makers, specialist distributors and user, is expanding into France and Germany, with 23 companies joining the German chapter in just five months. This year (2005) will see a pan-European management body set up to oversee the different groups.
While there are many situations where multi-sourcing of components is common, and these are less likely to suffer, products from a single vendor can become less common as technology moves on and markets and companies change. Even with multiple sources, technology can move on to drive basic components such as particular discrete TTL logic parts of RF transistors out of the market.
There may be alternatives available, but there is a cost to making the switch, which, up until now, has been largely anecdotal. That has been a problem for engineers trying to persuade management that thinking about the types of components now, and even spending a little more now, can save thousands of pounds further down the line.
The true cost of managing obsolescence has been evaluated by a project survey by ARINC and the National Obsolescence Centre for the UK Ministry of Defence.
There have always been anecdotal reports of the costs of obsolescence, but the NOC report is the first time that real numbers have been determined. This will help in making the case to management to support an obsolescence strategy and avoid having to make cost estimates, says Gary Austin, business development manager at ARINC.
The survey found that the additional cost of managing obsolescence ranged from the most basic £100 for using existing stock that had been stockpiled in a last time buy, through £2400 for designing in a substitute part or £1300 for using a hard-to-find part reclaimed from another board. But the larger costs came from having to do a minor redesign to emulate the function of the obsolete part, at £75000, through to a major redesign of the systems, which cost on average £300000.
At that level, a major redesign is not done often. The majority of problems were solved by finding alternative parts, although a common alternative is to emulate the function of the obsolete part. This can be done with other discrete parts, or by implementing the function in afield programmable gate array, although both of these approaches need a re-design of the original sub-system.
FPGAs from Actel and QuickLogic with metal to metal links are popular for this, as they can only be programmed once, and there is no penalty for using small volumes. But even these sometimes have to be bought up in alast time buy' programmes, where the manufacturer offers customers the opportunity to order enough parts for the lifetime of a product, the alifetime buy'. These can be stored as packaged parts, which is expensive and needs a tightly controlled storage facility.
Another option is for the manufacturer to sell wafers, which are easier to store, often to specialist distributors. The wafers are then diced up, tested and packaged when required. The specialist distributors can then hold adie banks' of obsolete parts. This also gives full traceability, a vital element for military and aerospace designs, and something that can be difficult with a reclaimed or alternative part.
An even more extreme approach, pioneered by companies such as Richardson Electronics, is to buy up the mask sets of the devices as part of that last time buy so that it can even make new parts to the original design. However, this accounts for just 6 to 7percent of the business.
The problem is most acute with long range, long lead time projects in defence. Rolls Royce Submarines provides the control and instrumentation for the nuclear reactors in the UK's submarines. The designs for the second generation Vanguard class submarines started in 1974, and the submarines came into service in 1986 and the 1300 separate designs will have to be maintained until 2010. The design for the next generation started in 2000 and won't come into service until 2010 or 2015. Making sure that the components designed in today will be available in 30 or 40 years time is a huge challenge.
Modern technology offers a higher obsolescence risk than older technology said Brian Porter from the design team at Rolls Royce Submarines. For instancea Quicklogic FPGA was designed into a unit seven years ago and recently went to alast time buy'. We think there's a case for a life time buybut the unit hasn't gone into service yet so predicting the service requirement is difficult."
And it is not just components coming to the end of their product life that can be a problem. He points to the resistance thermometer in the primary reactor. This was made by an external supplierwhichafter a buyout of the companymoved production. The new site was in another country that was not acceptable to the UK MoDso an alternative had to be found. In the endthis specialist unit was re-designed for another manufacturer in Francebut with a design that was deliberately portable.
Another timeto avoid a potential supply problemRolls Royce bought up the tooling and intellectual from a supplier that was going bust and took that to another supplier. "Don't automatically accept a component is obsolete said Porter. By pushingpeople will make it in the end."
The problem is not just in military and aerospace systems. Semelab ships 90percent of its RF MOSFETs to industrial customers.
A radio handset was using a particular RF transistor. After two years in designwhen the first part was made obsolete after 18 monthsthe second alternative was made obsolete just nine months after the handset went into production with a planned six year product life. This is when Semelab stepped in with an alternative part. "It is more expensivebut in 14 years we have never made an RF transistor obsolete said Cliff Robins, sales and marketing director at Semelab.
Obsolescence is becoming much more of an issue as chip making technology speeds up. Devices made on 0.5µm and 0.35µm are reaching the end of their product lifetimes and going aend of life', and that raises questions for industrial designers on how to manage that. Having a strategy to tackle the obsolescence of parts in three to five years is vital, and the latest survey from the National Obsolescence Centre gives engineers, for the first time, the metrics they need to argue the cost savings to put such a strategy in place.