Thought Leader Article, Dave Huntley, PDF Solutions
In an ideal world where ideal products are developed, we would never have any issues resulting from any step in the design, fabrication, assembly, test and field operation processes. But in the real product development world that we inhabit, it is a fact that some level of undesired quality and reliability issues will occur. We continually look for novel ways to address these manufacturing issues especially as the components and the PCBs where they are populated continue to increase in terms of density, complexity and cost.
In the semiconductor world, the technology to inspect wafers and capture outliers has been around for more than 30 years. Given the volume of manufacturing associated with semiconductor development, the ability to reduce errors and identify excursions (aka bad parts) and improve overall manufacturing efficiency by just a few percentage points can result in savings that run into the billions of dollars.
But similar technology solutions have not been available for PCBs. Now, through standards being developed in the industry, the reach of the semiconductor technology is now encompassing PCB manufacturing, test and traceability. Essentially, the ability to determine end-to-end or “cradle-to-grave” quality for both semiconductors and the PCBs on which they are populated is a significant breakthrough not previously available to the industry. A leading provider that can provide “Single Device Traceability” (SDT) technology is PDF Solutions, and the visionary within the company for this effort is Dave Huntley who is responsible for business development of Exensio® Assembly Operations and has been at the forefront of the standards development efforts associated with this technology.
This article describes the technology developed by PDF Solutions; how that technology is being expanded from traditional semiconductor manufacturing and test to include PCBs and the future of the technology going forward. For this article, in addition to Dave Huntley, I received the help of David Park, who is Vice President of Marketing. David Park is a former client and long-term friend of mine and the contribution of his efforts along with those of Dave Huntley were pivotal in the creation of the following article.
About PDF Solutions
PDF Solutions, a publicly-traded company, was founded in 1991. The two co-founders, Drs. John Kibarian and Kimon Michaels remain with the company today (this long-term affiliation in itself is an anomaly in high tech). Initially, PDF Solutions was a specialized consulting company helping semiconductor companies improve their yield ramp in manufacturing. PDF Solutions devised new ways to generate and collect manufacturing data that enabled its customers to significantly accelerate yield learning, which in turn accelerated time to market and time to volume production.
The foregoing was how the company operated for the first 20 years but then the company repositioned itself for the “more than Moore era”, to identify ways beyond transistor scaling to increase performance and reduce power consumption. To this end, PDF acquired a number of different companies in the test operation, assembly, and packaging space.
Today, PDF Solutions is effectively a big data analytics company supporting end-to-end data from semiconductor manufacturing to test operations, assembly and packaging and field performance. Prior to PDF’s approach, the data generated from semiconductor manufacturing, test and assembly have essentially been ‘data silos.’ You put your data in your silo and you analyze it in your silo but you don’t necessarily share it outside of your silo. One of the things PDF Solutions has done is to build an end-to-end solution that encompasses 50 different semiconductor data types from the fab to assembly, packaging and end use. All of this information is collected in a semantic data model that allows the user, once he or she has access to the data, to do the kinds of anything-to-anything correlations that can identify the opportunities to make improvements in manufacturing test and root cause analysis across the entire semiconductor supply chain.
Today’s PCBs have more than just a chunk of memory, a processor and a graphics processor, they also have multichip modules (MCMs) or Systems in Packages (SIPs) that have numerous devices inside of them. What you don’t want is to have a situation where a very expensive MCM fails and you don’t have a way to find out why except at the monolithic level. The desire to have traceability for every component in your MCM or SIP is becoming more important. This has been important for regulated industries such as automotive, medical and aerospace but even there the traceability is not 100%. Now with our mobile phones we are holding portable supercomputers, which are ubiquitous and also demand more precise traceability to accelerate yield ramps and prevent supply chain glitches. That’s why it’s critically important for product development companies not just from the quality perspective but from a returned material authorization (RMA) management perspective to have traceability for every component in an electronic device
About Dave Huntley
Dave Huntley who is responsible for business development for Exensio Assembly Operations at PDF Solutions, came to the company as the result of his company KINESYS’s flagship product, the Assembly Line Production Supervisor (ALPS), being acquired by PDF in 2017. ALPS is now the core of the assembly Operations module of the Exensio Analytics Platform, and enables complete traceability, including individual devices and substrates, through the entire assembly and packaging process. The product maintains an accurate link between a final device and the precise, original location of the source die from the original silicon wafer as well as any materials, such as gold wire and epoxy used in the die attach and wire bonding process.
Huntley explains, “With KINESYS, we had the complete genealogy from the raw die on the wafers and the raw materials into the final package. PDF took that traceability and connected it into their Exensio® analytics platform. This enabled our customers to have the ability to correlate wafer test data to final test data on a die-by-die basis where previously it was only available on a lot basis.” This can be done without depending on an ECID programmed into the die and has the further advantage that it provides visibility into the increasingly important assembly processes.
“Now we can correlate what happened at final test vs. wafer test which is obviously a benefit,” he continues. “In addition, when there is a final test excursion or some sort of reliability event, you can drill down on that and bring that back through equipment history as well as FDC, parametric and metrology data to help you find root cause.” (More about this below).
At PDF Solutions, Huntley continues to be involved with strategic marketing and business development for Single Die Traceability within the semiconductor and electronics ecosystems, and the continuation of the ALPS vision as the Assembly Operations module of the Exensio® Analytics Platform. In addition, Huntley has been involved with SEMI standards development since 1989 starting with the cluster tool standards that led to the GEM 300 standards suite. He was co-chairman of the Sort Map task force responsible for the SEMI E142 Specification for Substrate Mapping Standard, approved in 2006, that enables single device tracking through the assembly process steps. In 2018, he was appointed co-chair for the Single Device Traceability task force working on traceability for the supply chain and counterfeit prevention standards. In 2019, Dave was appointed co-chair for the Advanced Backend Factory Integration task force.
Huntley received a first class honours degree in Electrical and Electronic Engineering from Bristol University, UK.
How PCBs Factor into The Equation
As Huntley explains, “50% of the failures that are being seen in the industry are due to packaging. This is because packaging has become an increasingly functional and important final part of the product. Think about how thin the printed circuit board is in your smartwatch for example. There is a lot of supply chain pressure being put on assembly technology. Product developers tend to design in such a way that makes it difficult to manufacture with high reliability. You can get failures from the epoxy that you put down when you tap the die down into a package. You can have problems caused by just one tube of epoxy on one machine that will create a long-term field failure. It will successfully pass through final test and everything. But then it will break apart before the end of warranty period and if it is a high volume, or expensive device you can find yourself in real trouble with RMAs and angry customers.”
“With these aforementioned failures, when you get down to the root cause, it’s getting less and less often that it is coming in from the wafer. The problems are happening after wafer fab.”
He adds, “The advanced packaging that is taking place at Offshore Semiconductor Assembly and Test (OSAT) facilities can be a source of problems. But, the PCB can equally have just as many problems with their process as well. That’s why we are going end-to-end. Traditionally, we have been very strong in wafer fab but we are increasingly in OSAT and packaging assembly. We’re now penetrating into the PCB world and capturing their product and assembly data and doing the same thing that we have done with semiconductors. The methodology is slightly different but it’s the same basic concepts. “
In tracing failures, a large part of the motivation is to be able to isolate the failures to as small of a population as possible. Huntley states, “There was a situation involving a medical application where a company was off-shoring their assembly work to get economies of scale, and their lot size went up from a couple hundred devices to 5,000 devices. There were some critical device failures and the company had no choice but to recall an entire lot, or 5,000 units, because there was no way to isolate the root cause within a lot. Because this was a medically implanted device, some patients had to go back into surgery to replace the device. Broad recalls due to electronics components in medical devices is never a good thing.”
“From a cost point of view, each one of these incidents can cost a company tens of millions of dollars at a minimum. And that doesn’t even begin to take into account the PR costs involved to do damage control with public perception.” (Think Exxon Valdez). “And depending on the end market and how the device is used in the final system, the costs can become stratospheric. There was a production problem at a manufacturing plant in Thailand that was part of a global automotive supply chain. The recall this supplier had to do, cost them billions of dollars.”
When it comes to determining when and how things fail in the field, Huntley notes, “It’s very challenging to get industry statistics on failures and recalls because companies are very protective about this information and for very good reasons. Any public recalls can result in lost customers and design wins that can have ramifications for years. As a software vendor supplying analytics and traceability solutions to many companies in the semiconductor and electronics supply chain, we have a pretty good idea of what typical RMA rates are for companies, and everyone is going to have them. No one is immune to them. But the challenge is, how quickly can you identify the root cause of the RMA and find all the other potential future RMAs that have the same signature so you can proactively address them? That is the real value of Single Die Traceability.”
The Role of Standards in Failure Tracking and Tracing
As noted above, Huntley has been very active in the standards that are being developed to support the entire industry for failure tracking and tracing. His first effort was with the SEMI E142 standard. At the overview level this standard applies to the substrate types, wafers, frames, strips and trays associated with semiconductor manufacturing. The purpose of the standard is to define the data items that are required to report, store and transmit map data for the foregoing items.
Huntley explains, “The E142 standard allows you to track when you pick a die off a wafer and place it into a package on a strip. (Note: here the word “strip” refers to the lead frames on which the packages are assembled). A die from a wafer is picked up and placed in a particular location in a particular package on a strip. It has an X-Y location which identifies which package it is being placed in and a layout path that identifies where in the package it is being placed. This wafer to strip transfer information is recorded by the equipment and uploaded to the server in the factory. Traceability is now in place. As long as you know the strip ID + X-Y location + layout path, you can find every die that went into a package, which wafer it came from and its original X-Y location.”
“There are all sorts of device IDs that are inside the silicon. Those IDs are also marked outside the package. All of this information being captured is in a format to describe electronic assemblies. It will give you a framework to say this is the data associated with this particular part in this particular package on this place on a strip of devices.”
Associating the Information with PCB Operations
The tracking and tracing activities noted above are now being expanded into the PCB world. Huntley notes, “The data is being widely used in offshore test and assembly houses to assemble many different types of packages. The packages go through a distributor and then off to the PCB house where they are picked from the tape and reel and placed on the PCB. Picking up a package from tape and reel and placing it on a PCB is analogous to picking it from a wafer and placing them into a package. It’s the same basic story.”
Similar to component devices, PCBs are identified with a 2-D code. “Many companies use some form of surface recognition technology which is a unique, immutable, unclonable identifier for a specific PCB,” Huntley adds. “E142 is capable of recording all this information. The role of the standard is to create the data framework that allows you to capture this traceability information in a standard way.”
Obviously, having easy, but secure access to all of this tracking and tracing information is a critical part of the equation, as is the ability to combine the data that comes in from the various players. Huntley states, “Exensio Assembly Operations is able to take any packaging and assembly data that comes in.
We also clean the data and align it so that it is ready for analysis. The Exensio data management capabilities alone can save customers up to 80% of the effort in finding the root cause of a problem through traceability. This is why PDF Solutions supports the E142 standard. The standard provides a structured uniform way for companies to record their data which facilitates traceability when they need it.”
The Cost of Pushing the Envelope
As noted in many of my previous articles as well as just about any Internet research you might do, one of the challenges facing the industry is that we are continually pushing the envelope in terms of what technology can do. An offshoot of this is that we don’t have much “wiggle room” when it comes to developing new electronic products.
Huntley cites, “The semiconductor and electronics industries have a much higher requirement for reliability these days. The stuff we are creating is not going into a something trivial like a toy or music player. These devices are going into mission-critical applications like medical, automotive, aerospace and smart factory applications where reliability is paramount.”
“One manufacturing scenario that is becoming more high-profile is in wire bonding,” he continues. “It’s not unusual with the super fine pitch wire bonds used in advanced packages that you can get a connection that is good enough to pass final test but might fail out in the field. At PDF Solutions, we are looking into applying advanced FDC (failure detection and classification) to analyze waveforms for every single wire bond. There are hundreds of wires on a single package. Each wire has two bonds—one on each end. By capturing the waveform data of the process on both ends of the wire, we can analyze the waveforms to look for anomalies. We are applying big data analytics and machine-learning algorithms that have the ability to handle very large data rates and can analyze the waveforms and recognize anomaly patterns that affect quality and reliability.”
The whole cost/account problem is another major factor that comes into the tracking and tracing process. Huntley notes, “There are very high profile failures that happen every year that result in a huge financial cost for companies. In addition to E142 we are working on a standard that includes not only traceability but chain of custody. This involves inserting a ‘fingerprint’ or other unique identifier into a chip. This would give you an immutable ID for the chip which gets recorded in a blockchain. Then this chip is assembled into a more complex package that has its own unique identifier that also gets put into the blockchain. Finally, all of these packages end up as components on a PCB, but now you have a chain of custody for every component on the PCB.”
All of the foregoing information gets recorded on a blockchain. “This gives you a genealogy between members of the supply chain,” Huntley states. “In my work at SEMI, we are trying to get the industry to agree on how to cooperate on all that. Several companies have built things like this for their own supply chains but there’s a mish mash of them and they don’t seem to cooperate. SEMI is collaborating with other standards bodies such as ISO and IPC, and the Global Semiconductor Alliance (GSA) Trusted IoT Ecosystem Security (TIES) to drive industry consensus. The goal is to define a standard like E142 but for the chain of custody. We aren’t there yet, but that is the goal.”
“Once you pinpoint the problem and contain it, you get to the next evolution of the industry from the data analytics point of view,” Huntley adds. “The first approach is to react to the problem. ‘Let’s fix it.’
The next phase we are in now is predicting the problem. We need to come up with a signature analysis that looks at all the data available and can predict a recurrence of the problem. If that is the case, you can pull it out of the line and put it on hold. That’s predictive.”
“The final phase for the industry is prescriptive. This is where we say, ‘OK so we are getting these failures, let’s go back and figure out what we’re doing wrong and change the design because we are making an unreliable product. Can we identify what aspect of the design needs to be changed or fine tuned to eliminate or reduce the current set of failures we are seeing in test operations or in-field use?”
Obviously, there are some real challenges involved in moving to both the predictive stage and the prescriptive stage. Huntley says, “The primary challenges is how do you get all this data into the system to determine where failure originates? You have various suppliers and third-party companies involved. How do you get everyone to cooperate and give you the data you need to solve the problem?”
“This is the big industry challenge right now,” he continues. “There needs to be some sort of contract between the end-product owner, the semiconductor company, and the other members of the supply chain that gives eco-system members the authorization to access this data. Assuming that they have authorization and we have a secure network in place to move the data around, then it comes down to the amount of data you get and the quality of that data. If we can get to this point, the value of a semantic data model and an end-to-end analytics platform becomes self-evident. Instead of optimizing ‘local minima’, the semiconductor and electronics supply chain can optimize the entire product lifecycle for cost, quality, and reliability.”
The traceability we have today in the semiconductor and electronics industries are not up to the task at hand. We need a comprehensive Single Device Tracking (SDT) in a common semantic data model based on standards all the way from wafer fab, package assembly and downstream PCB processes of manufacturing, assembly, test and repair. Given the complexity of today’s packaging and assembly, the ability to have an end-to-end genealogy of all the elements contained on the final PCB can be a huge benefit for locating and resolving design issues that reduce cost, improve reliability and ensure the integrity of crucial time-to-market windows.
The design tools in Altium Designer® contain everything you need to keep up with new technology. Talk to us today and find out how we can enhance your next PCB Design.