Editor's Note: Semicondutor Insights got a flood of responses from readers of this Special Report. One of the recurring questions was why Semiconductor Insights did not go through authorized distribution channels to ensure it didn't put itself at risk. That seems logical enough, but in a follow-up piece, SI explains how this can happen to any designer. See: Counterfeit parts Part 2 -- Baiting the trap.
Semiconductor Insights (SI) helps other companies protect their intellectual property (IP) and document-infringement rights. But we have foundas countless designers have discoveredthat counterfeit parts can really put a damper on our business. Indeed, we've been seeing more counterfeit devices coming in over the past few years, which could indicate an alarming trend.
In the business of analyzing devices, SI purchases thousands of components each year. Sometimes they are bought directly from the manufacturer, sometimes through well-established channels; often they are purchased from other countries, including China. In the past year, SI has seen an increase in the number of counterfeit parts from China.
Semiconductor sales are projected to reach $252 billion in 2007, according to the latest numbers from the Semiconductor Industry Association. As consumer interest grows in electronic products, so too does the volume of counterfeit devices.
In consumer electronics, cost is king; lowering the bill of materials even by a few cents can drastically affect margins and revenue. Suppliers dealing with counterfeit components are willing to offer the parts at prices that are lower than legitimate competitors can match. The low prices are very appealing to OEMs, particularly smaller companies that don't have an agreement with the manufacturer itself. The drawback is that the parts may not function as well as they should.
Unfortunately, selling counterfeit parts is relatively easy. Someone buys a bunch of parts. They find out what the customer needs, etch off the current markings and then restamp the parts with legitimate part numbers and logos. With today's technology, even putting the devices back into a reel to replicate official production is not too difficult.
Counterfeiting components has a significant negative impact on the semiconductor industry. First, it can hurt the reputation of a manufacturer. For example, a company buys a device and tries to design it into a system. Unbeknown to the company, the part may not be legitimate. When the part doesn't work correctly, the company will either blame its design or blame the component manufacturer. Revising the design can delay time-to-market and drain engineering resources. That can have a big impact, especially on smaller companies, which are typically the ones that would buy parts on the open market as opposed to directly from the vendor.
Counterfeiting can go way beyond components. In May 2006, for example, EE Times ran an article headlined, Fake NEC company found, says report. Instead of creating fake devices, someone counterfeited an entire company with about 50 products, including home entertainment systems, MP3 players, batteries, microphones and DVD players.
In addition to copying NEC products, the bogus company had developed products of its own that were not in the legitimate NEC lineup.
Counterfeit devices also pose a threat to consumer safety, since they are not likely to have been tested to the standards of the industry. If a semiconductor overheats, for instance, it can destroy the product. If a counterfeit DVD player overheats, it could start a fire and destroy a home. What if a counterfeit part was used in a car? The potential impact of a failing component here could be catastrophic.
SI is always scouring the world for the latest and greatest semiconductor components. In the NAND flash market, that would mean sub-60-nanometer technology. We have successfully sourced both Samsung 51-nm and Toshiba 56-nm devices, but it took a few tries before we were able to get legitimate components.
The first part we expected to receive was the Toshiba 56-nm 16-Gbit multilevel-cell NAND flash. We received a purchase order for the part with a part number that was identical to the one we had requested, which had Toshiba markings. The part arrived, but we noticed something fishy on the package marking. The part number was TC58NVG4D1DTG00. A quick Web search showed hits on the 16-Gbit 56-nm MLC NAND flash. The Toshiba title looked a bit odd, but not so unusual that it was flagged. A decap of the device was done and photos of the die and die marking images were taken, per standard SI protocol.
To our dismay, the die markings were not Toshiba's, but Samsung's. After comparing the die markings (K9G4G08U0A) in our database, we were able to match the markings to the Samsung 4-Gbit 65-nm MLC NAND flash.
If we had ordered the part in question for system design purposes, our system would have failed, since the device was only a 4-Gbit model and not the 16-Gbit version we had ordered and expected.
While the number of pins was the same on both the legitimate and the counterfeit parts, the pin configuration was different. The counterfeit parts, which use the Samsung die, have a 2-kbyte page size. In contrast, the Toshiba device has a 4-kbyte page size. Designing-in the counterfeit part would only result in partial functionalityif it would even work at all, given the pin configurationwhen trying to utilize the entire expected memory space.
Speaking from design experience, there is nothing more frustrating than knowing that your design should work, checking countless times, and still having things consistently fail.
The next part we were expecting was IM Flash Technologies' 50-nm NAND flash. We first analyzed this device back in July 2006 and wrote an article about it in September (50-nm device revs NAND race). The Micron marking on the package was definitely odd compared with the original samples we received, but the part number was accurate. We do not see pictures of the devices prior to their being sent in, so while the marking was odd, there was no way to determine that it was counterfeit when we placed the order.
Analyzing the die in this package revealed Samsung yet again. This time, it was the 8-Gbit version of the same device that was found in the counterfeit Toshiba part.
(Click on image to enlarge)
The remaining part we expected was Samsung's 51-nm 16-Gbit MLC NAND flash. It was received from a different supplier than the previously mentioned counterfeit parts. This time, it was a bit easier to flag the part as having an issue, but the problem could have easily been overlooked. There was a single character missing from the package marking. The counterfeit part bore the part number K9GA08U0M; the legitimate part is the K9GAG08U0M.
More interesting were the X-ray images that were taken. The nine parts that were received all had multiple dice inside. So not only did we get counterfeit parts, but it seems someone grabbed whatever was lying around, retagged all the pieces as Samsung devices and sealed them in a strip.
Organizations such as the National Electronic Distributors Association , or NEDA, have been formed to prevent and control counterfeit devices in the industry. They recognize that counterfeiting is an escalating threat to all supply chains, especially for electronic components.
A quote from a discussion with a Toshiba executive serves as an appropriate close. "We realize that there is an ongoing issue with counterfeit chips," said Scott Nelson, vice president of memory products for Toshiba America Electronic Components Inc. "It's a 'buyer beware' situation, in which we encourage customers to make sure that they are really receiving the product they think they're buying.
"One way buyers can help protect themselves from counterfeit parts is to purchase products only through manufacturers' authorized distribution channels."
Gregory Quirk (gregoryq@ semiconductor.com) is technology marketing manager at Semiconductor Insights (Kanata, Ontario). He has a bachelor's degree in systems engineering and is working on his MBA at the University of Ottawa.
Fake NEC company found, says report
50-nm device revs NAND race
Atmel memory embeds crypto for counterfeit busting
Consider OTP memory for IC antipiracy