The engineering spent 3 weeks debugging a failed circuit. It looked all good on paper – nice clean schematic and a perfect looking PCB layout. In the end the ICs were found to have been sanded down to bare die and then reprinted with fake IC markings. All this for the amount of $40 saving for the engineer. Three weeks of his life for the amount of money for purchasing 3 additional ICs of the correct specification.
Three weeks. For a purchasing decision that probably saved someone forty dollars.
So another crucial point is that the choice of a single part can rapidly outweigh the importance of other design parameters and so must be worked out very carefully. By definition the part has been specified by the engineering team; it is then the job of the purchasing team to go out and buy it. However they are generally working under extreme pressure to purchase on time, within budget etc. and the temptation to buy the cheapest possible part that just about meets the specified parameters is very hard to resist.
Specifications aren’t suggestions — but let’s be honest about how we treat them
Of course, even when you have locked down all the relevant parameters, it is possible that there are parts from different manufacturers that have the very same part number and different characteristics. For this reason, it is essential to read the datasheet very carefully before designing your product to determine the specific characteristics that your chosen part will have in your application. Once you have chosen a part and ordered it, the same part number no longer has the same rules apply to it. The 40 dollars that the purchasing department saved can potentially threaten the rest of the engineering team’s design.
LOCKPLACEHOLDERZZ0
- Supply voltage range and whether you need LDO tolerance or tight regulation
- Operating temperature range, especially for automotive, industrial, or outdoor deployments
- Package type and whether your PCB footprint actually matches (you’d be surprised how often this bites people)
- Revision or silicon version, because manufacturers sometimes change behavior between die revisions without a fanfare announcement
Just because you are substituting parts for a reason does not mean that the substitution will have no downstream effects. You need to document the reasons for your substitutions just as the designer of the part did.
The counterfeiting problem is worse than most people want to admit
Yes, it is a problem. But the problem of counterfeit ICs is a lot more widespread than just affecting the odd designer who’s been looking for a part in error. They flood the market through perfectly normal channels. Especially for parts that are in high demand or are obsolete (discontinued). The profit from fake parts can be too tempting to very undesirable individuals. A simple repainting of markings on a IC (sanding off the original markings and then reprinting new ones) is child’s play for these types of suppliers. More sophisticated would be remarked parts. Which are normal IC dies but of a lower grade then what the part number would indicate for a normally qualified part. And then there are non functioning dies that are packaged in very realistic looking IC part packages.
However, most ICs are not obvious counterfeits. As a rule of thumb, parts that are verified by a third party before they are used in critical systems are best, regardless of whether they were procured from a normal distributor.
Some red flags worth tattooing somewhere accessible:
- Pricing that’s dramatically below market, especially for parts that are currently hard to find
- Date codes that don’t match the claimed production run
- Inconsistent marking fonts or finish quality across a batch
- Sellers with no traceable distribution history
Availability and lead times will humble you
The semiconductor shortage of 2020–2022 was a very educational demonstration of the fragility of semiconductor supply-chains that most engineers have never even imagined. ICs would be listed as “in stock” one day and then 52 weeks later you could find out why your product launch was delayed: because a single part (a single microcontroller) that someone assumed would be there, had not materialized.
Engineering needs to manage the process of procurement of parts as much as possible, in order to avoid surprises. A good starting point for managing the purchase of parts for a project would be designing a circuit that can be built with alternative components that are usually in stock or are available shortly if the first choice part is not available. In order to avoid being caught off guard by a long queue to purchase a part, maintaining a good relationship with several distributors and not letting stock of a single part dwindle to zero before it is time to restock, could help avoid trouble. Once all stock of a part has been used up and a “safety stock” of zero has been in place for some time, it is likely to be discovered that the item is no longer available due to long queues of other customers.
A rough comparison of where you might actually be sourcing from:
| Source type | Typical reliability | Risk level | Best for |
| Authorized distributor | High | Low | Production runs, traceability |
| Manufacturer direct | Very high | Very low | High-volume, long-term supply |
| Specialty electronics supplier | Generally good | Low to medium | Prototyping, hard-to-find parts |
| Gray market broker | Variable | High | Last resort only, with verification |
Where you buy matters as much as what you buy
I have found that one of the areas that engineers are not informed of is the variations of parts that have identical specifications and are packaged identically. Many engineers assume that a part is a part, regardless of where it is purchased. However, in reality, there can be many variations in terms of physical and electrical characteristics. There are many cases where high quality parts have been used in design and the part functions perfectly on the engineers bench, but then fails in the field. The reason for this failure is often due to the engineers not being aware of the distribution of electrical characteristics of the ICs that they are using.
(Yes, this is similar to stating ‘buy from authorized distributors’. However, this is obvious advice and is therefore frequently ignored. This section is different to that).
The process of selecting ICs for your next project is similar to selecting components for your next meal. Look for suppliers of high quality parts. When purchasing parts look for suppliers who can provide documentation on the origin of the parts. This will allow you to confirm that the parts are what you asked for and were manufactured by a reputable company. Also look for suppliers who can provide traceability of the parts back to the manufacturer. This will allow you to confirm that the parts were manufactured correctly and within the manufacturers stated quality control process. In addition to the parts, look for suppliers with a good return policy. There are times when you receive parts that do not function as expected. In these cases it is imperative to be able to return the parts for a full refund or replacement. In addition to the return policy look for suppliers with a good dispute resolution process. This will allow you to resolve any issues in a fair and timely manner. The above features are not “frills” and should not be looked at as additional cost that can be cut to save money. These features are “load bearing” and must be included as part of your normal procurement process.
The boring paperwork is actually load-bearing
Retention of paperwork for buying integrated circuits (ICs) will save your butt when something fails in the field months later. All details should be retained for such parts. For example: Part numbers, lot numbers, supplier’s invoice, and date codes. Records for such purchases are tedious to retain; however, when a failure occurs in the field retaining such documentation will expose root cause of failure in shortest possible time.
The worst part of ICs failing is that they don’t fail dramatically. They don’t fall apart in your hands. They don’t leave a trail of smoke behind them. Instead, they slowly begin to fail over weeks. And, of course, the worst part is that they fail to perform in a way that is not immediately apparent. In other words, you can’t just look at an IC and tell if it’s good or not. You have to test it. And even then, the results may not be immediately obvious. That’s why good records are so important. Without them, you’re left staring at a chip with a loupe, trying to read off sanded down markings. It’s a slow and painful process. But with good records, the problem can be solved in an instant. It won’t make your job any more glamorous, but it will make it possible for you to work at all.
