“MEMS is in need of standards.”

I’ve heard that declaration many times over the last few months. But just try to search the Internet for evidence of standards or even standardization efforts, and you’ll find… stuff, but you have to look hard and follow many fruitless threads to get there, and it’s anything but conclusive. Discussions with the actual people involved paint a very different picture from that which emerges from the Internet.

This article is an attempt to capture the major activities underway, at least to the extent that I’ve become aware of them. Obviously this whole field is moving even as I type, so specifics will become dated. In addition, many standards have their origins in hallway hellos and cocktail conversations; it’s impossible to say which such musings might at this very moment be fermenting into formal activities. What follows is, then, clearly a snapshot.

Standards are a touchy topic for industries experiencing rapid growth. There’s a sense that manufacturers need to be free to try any and all new ideas to harness their potential, that freezing elements of technology at some arbitrary time, before there is proof that further change is unlikely, will stifle innovation (a word I used guardedly since it tends to be the cop-out get-out-of-jail-free card for “I want to do what I want to do”) and will ultimately hurt the industry.

Conversely, there comes a time in the growth of an industry when the chaos stops acting as quantum turbulence spawning wild new ideas and starts to act more like thermal vibrations that get in the way of progress. That’s when purveyors and purchasers alike begin to clamor for some order in the form of standards.

The trick is in figuring out when the timing is right. Done too early, the standards will simply be ignored as running counter to the interests of important players. Done too late, not only will the industry undergo suffering, but entrenched ad hoc positions will increasingly harden, driving up the disruption that must occur when the time comes to harmonize – or creating the risk that ignoring the standard will prove less painful than changing to meet it.

Ultimately, there has to be a long-term economic benefit to the work involved in creating and adopting standards. Customers can often rationalize benefits from standards – especially if they don’t have to spend any effort in helping set the standards. Vendors, on the other hand, aren’t so sure, and different vendors will judge the right time for standards differently.

All of that said, there appears to be relatively broad consensus that the time is right for judicious standard setting. In fact, there are already some standards in place – it’s just not clear who – if anyone – is using them. Of course, many generic standards put out by organizations like JEDEC and IEEE will affect MEMS devices for things like package dimensions and communication protocols. The discussion that follows focuses only on standards established and intended specifically for MEMS and sensors.

Starting at the start

Believe it or not, there are ten existing standards set by SEMI, an organization that focuses on the processing of semiconductors and other related technologies. There are four active areas of effort: microfluidics, materials characterization, wafer bonding, and terminology. They are managed by a dedicated MEMS/NEMS committee co-chaired by Mark Crockett, Janet Cassard, and Win Baylies.

Metrology-related standards have originated to a large extent by disagreements between equipment manufacturers and buyers about the performance of various processes. For example, wafer bonds must achieve acceptable strength, but equipment makers were measuring one way and their customers were measuring another; eventually they realized they had to agree on a methodology – hence the need for a standard.

The following are the metrology-related SEMI standards:

• MS1: Guide to Specifying Wafer-Wafer Bonding Alignment Targets
• MS2: Test Method for Step Height Measurements of Thin Films
• MS3: Terminology for MEMS Technology
• MS4: Standard Test Method for Young’s Modulus Measurements of Thin, Reflecting Films Based on the Frequency of Beams in Resonance
• MS5: Test Method for Wafer Bond Strength Measurements Using Micro-Chevron Test Structures
• MS8: Guide to Evaluating Hermeticity of MEMS Packages
• MS10: Test Method to Measure Fluid Permeation Through MEMS Packaging Materials

The general approach here is one of starting with the most basic of elements, like a thin-film measurement. Get that working, and, as success is achieved, move further up the manufacturing chain, identifying further opportunities.

Meanwhile, the microfluidics industry is, at present, a fragmented medical razor blade business. As described by Henne van Heeren, a consultant in this arena, a given system will consist of a piece of equipment, related support equipment, and then a range of consumables (also called “disposables”). And the consumables are where the money is. This suggests an ecosystem, and, right now, each company tries to build and maintain its own closed ecosystem so that it can own and control all aspects of the business – and capture all of the revenues and profits.

But proliferation means driving these systems out of expensive labs with well-trained specialists and into local clinics, where they must share space with other equipment and files and doctors and patients, to be run by less-specialized clinicians. So having a completely different machine – and its ecosystem – for each different test just won’t work. A given machine will have more value if it can do more than one thing.

There’s also a move to separate out the sensing of a signal or result from the introduction of reagents. A clinic is a less well-regulated environment, and attempts to have the testing machine introduce reagents risk contamination and mistakes. Instead, the disposables should have all necessary materials (aside from the test sample) pre-loaded; the equipment then simply reads and delivers the result.

Put all of this together, and you need to start standardizing the way the equipment interfaces with the disposable. Hence a small, growing body of microfluidics standards – again, starting with basics. SEMI has so far contributed three standards:

• MS6: Guide for Design and Materials for Interfacing Microfluidic Systems
• MS7: Specification for Microfluidic Interfaces to Electronic Device Packages
• MS9: Specification for High Density Permanent Connections Between Microfluidic Devices

According to Mr. Crockett, MS1 and MS4 have achieved good uptake; the others aren’t completely inactive, but they lag behind.

Packaging and reliability are additional areas that have been identified as needing standards work within SEMI, but, at present, they don’t have permanent leaders in place (there’s an acting leader for the packaging area), and so they lie fallow for the moment. Mark Tondra, of Diagnostic Biosensors, noted that this industry isn’t really big enough to have captured the attention of NIST yet.

iNEMI, on the other hand, has a reliability project underway that is in the process of soliciting “founding members” (by February 1, 2013 – time’s almost up!*) Note that, as described by iNEMI’s Jim Arnold and Cynthia Williams, iNEMI doesn’t set standards; they take on research projects that may end up as inputs to standards efforts. The transfer of any such iNEMI results into the standards process happens through individuals that are members both of iNEMI and of whichever standards body has been anointed with the topic at hand.

I did ask a few MEMS device vendors whether there were any MEMS-related standards that they were following. Of course, none of the companies were microfluidics companies, so I wouldn’t have expected them even to be aware of those standards. But none were aware of the metrology-related ones – but, to be fair, it may be that those standards are low-level enough that they wouldn’t come to the attention of a marketing guy.

Helping out the customer

At the complete opposite end of the flow lies the customer. And they’ve noted that it’s really hard to compare different MEMS products because (a) the datasheets spec different parameters, (b) the parameters that are shared may use different test methodologies, or at least different test conditions, and (c) there may not even be agreement about basic terminology. What’s a mother to do?

It’s easy to imagine that, in a high-stakes, high-ante game like MEMS, the first guy out does things however he wants, and whether or not that’s sufficient becomes a matter of negotiation with the customer. When the second guy comes to market, the first guy tries to ignore him for as long as possible.

But we’re now to the point where there are numerous manufacturers of, for example, MEMS accelerometers, addressing a variety of different applications. At this stage of the market, customers are comparing datasheets, with little success.

As a result, efforts are now underway to standardize critical portions of the datasheets. The seed proposal was fashioned by a team consisting of Intel, Qualcomm, InvenSense, STMicroelectronics, Freescale, and Bosch, led by Intel’s Steven Whalley, and in collaboration with the MEMS Industry Group (MIG) and NIST.

Part of the discussion deals with achieving the right balance between not enough and too much. For example, Kionix’s Ron Chong noted that the proposal included a requirement that the datasheet include a table of power consumption for different operating modes. But, he notes, these devices are becoming so complex that such a table could end up with thousands of lines – benefitting no one.

A 0.9 revision is due for circulation any day now. An official rollout is expected before the end of Q1, 2013, according to Mr. Whalley and Analog Devices’ Rob O’Reilly.

What’s interesting is that this effort is more than just a cry for help from customers. It has come together from a number of efforts, not the least of which originated in work done by NIST to create some test structures for evaluating critical materials properties like Young’s modulus. This was a very specific project that might have been envisioned as enabling a standard testing methodology – both structure and measurement protocol. But it didn’t go anywhere.

According to Michael Gaitan, NIST’s Man About MEMS (my characterization, not his, due to the frequency with which his name comes up), NIST has two purposes: one is to improve measurement science; the other is to set standards. They have an active MEMS program going back to when their interest lay in using MEMS devices to improve their measurement equipment. Over time, the MEMS activity moved from that into a standard-setting mode. The test structure project made it clear that a critical ingredient was missing from the process: industry consensus.

They have since taken on a roadmapping effort and have put out recommendations for testing standards, also done in conjunction with MIG. But the consensus-building process has also moved the area of focus out: before you can standardize a test or test process, you have to agree on what needs to be tested. And before you can agree on what needs to be tested, you have to agree on what customers care about and what should be put on the datasheet. So Michael Gaitan has noted that he’s very encouraged by the efforts to pin down the datasheet contents. It establishes consensus and sets a starting point from which, given continued consensus, further work can proceed.

Meanwhile, similar to their reliability project, iNEMI is also initiating a testing project, also with a looming February 1 deadline* for public input into the Statement of Work. While this might sound duplicative, Mr. Arnold reinforces iNEMI’s insistence that their work be complementary to other efforts, so they will explicitly be focusing on things not being worked by others.

Improved communication

Less evolved is a completely different effort to improve how a sensor communicates with a host (or anything else in the system). Right now, sensors typically offer I²C (found more frequently because it requires only two pins) and/or SPI, with separate interrupt signals. The mode of operation either has the host controlling everything through the bus outright, or the sensor using the interrupt line to alert the host, after which the host responds using the bus.

This creates several challenges. First, the number of pins proliferates. Even with a single I²C bus, you must manage multiple interrupt signals, both finding a place to send them (different processors have different interrupt-handling capabilities) and routing them on the board.

Then there’s performance. According to Lattice’s Satwant Singh, I²C, for example, handles only 400 kb/s, driving designers to create multiple I²C busses, exacerbating signal congestion. Power is the next problem: driving a 0 on an I²C bus can use more power than the sensor itself consumes. Add to this the fact that both I²C and SPI have limited protocol standardization, creating incompatibilities, and you have a situation where a number of players are considering trying to standardize a new higher-bandwidth protocol.

This is being managed through MIPI, and a number of existing standards are being considered for outright use or adaptation, including the RF Front End (RFFE), SLIMbus, and Battery Interface (BIF) standards. The focus is on keeping gate count low, with an eye on latency and bandwidth. The interrupt signal is also being considered for removal – that and any other necessary control signaling could be handled in-band over the higher-speed protocol.

This is still at an early stage – a so-called “Birds of a Feather” discussion. An outline is being put together for proposal to the MIPI board of directors; the expected timeline is a current topic of discussion.

Where folks fear to tread

If there’s one MEMS standards topic that raises hackles, it’s the thought of standardizing actual processes. MEMS is all about novel ways of building tricky structures, and no one is willing to subject a core process to a standard.

The good news is, no one is requesting that. But there have been suggestions – and, at this point, they’re only suggestions – that there might be some elements of processing that can be standardized. NIST’s Mr. Gaitan noted an analogy from AMFitzgerald’s Alissa Fitzgerald (who recently tooks steps in this direction via collaboration with Silex) likening the situation to a machine shop: you can build anything in such a shop that you can envision.** But you are still subject to standards: drill bit sizes, threads, and other such aspects.

The thinking is that there are such pieces in MEMS processing that could be harmonized, making it easier for devices to be manufactured in different fabs or by different foundries. So the topic has been broached – gently – but, as far as I’m aware, it’s only talk for now.

Unaccounted for

When I was rummaging through the Internet trying to find productive leads on this topic, I ended up a few places that were completely unreferenced in all of my other discussions. The Internet being the collector of all things relevant and ir-, it’s hard to decide how to disposition these things. Are they prior incarnations of projects that got renamed and then handed to other organizations and then renamed again? Were they failed projects? Are they alive and well in a different, parallel universe? I’m not sure, but here is a sampling of what I’ve stumbled across. I’m sure that, with a couple more months of work, I could sort through all of this, but I’m succumbing to the will of the publishing deadline as an excuse for saying “Enough!”

Note that some of these may or may not look specifically like MEMS standards (see, for instance the DIN ones), but they came up specifically in MEMS-related contexts when I first found them.

• IEEE 1451. Yes, IEEE. And not P anything: it’s beyond that. This is a “Standard for a Smart Transducer Interface for Sensors and Actuators.” There are eight components, 1451.0-1451.7. It includes the definition of a “transducer electronic datasheet” or “TEDS.” Seriously: this never once came up spontaneously in any conversation I had with anyone. O_o…

  When I asked, Mr. Gaitan confirmed that NIST has been very involved with the project, although through a different NIST lab and individual. I will follow up on this for future coverage; this seems to be a complex standard. To be fair, it does appear to apply to all sensors and actuators, not just MEMS.

• Numerous projects under IEC TC 47/SC 47F, covering MEMS. The Japan Micromachine Center tracks this activity as well. They include, at various stages of development,
  • IEC 62047-11: Test method for linear thermal expansion coefficients of MEMS materials
  • IEC 62047-15: Test method of bonding strength between PDMS and glass
  • IEC 62047-16: Test methods for determining residual stresses of MEMS films; wafer curvature and cantilever beam deflection methods
  • IEC 62047-17: Bulge test method for measuring mechanical properties of thin films
  • IEC 62047-18: Bend testing methods of thin film materials
  • IEC 62047-19: Electronic compasses
  • IEC 62047-20: Test method for Poisson’s ratio of thin film MEMS materials
  • IEC 62047-22: Electromechanical tensile test method for conductive thin films on flexible substrates
  • PNW 47F-139: General rules for the assessment of micro-geometrical parameters
  • PNW 47F-140: Silicon-based MEMS fabrication technology – Basic regulation for layout design
  • PNW 47F-141: Silicon-based MEMS fabrication technology – Measurement method of cutting and pull-press strength of micro bonding area

I am attempting to contact the US National Committee Member (I’m intimidated already), and I will follow up in the future on this as well.

• DIN 1495-3: Sintered metal plain bearings subject to specific requirements for use in small-power and fractional horse-power electric motors – Part 3: Requirements and testing. (DIN refers to the Deutsches Institut für Normung, or German Standards Institute.)
   
• DIN 32561: Fertigungsmittel für Mikrosysteme – Werkstückträger – Anschlussmaße und Toleranzen. Oh, sorry, that would be “Production equipment for microsystems – Tray – Dimensions and tolerances.” Future work includes production equipment for microsystems, classification of microsystem components, terminology, and the interface between equipment and “end-effectors” (the business end of robotic tools/fingers/whatever).
   
• ASTM E08.05.03 – several thin-film related standards (E 2244, 2245, and 2256). This is older work (the visible activity is vintage 2003), so it may have been worked into other things given that this comes under the auspices of NIST.

Boiling it down

To summarize, then:

• Several low-level metrology standards exist.
• A few microfluidics standards exist.
• A Smart Sensor standard exists.
• A datasheet standard/agreement is in the works.
• Test and reliability standardization efforts are building.
• A sensor communication standard is being considered.

My head hurts.

*According to iNEMI’s Jim Arnold, this date may be extended if needed due to the fact that the holidays chewed up much of the input period.

**Yes, yes, I know. In drafting class we all learned about things that were easy to draw but could never actually be built. So let’s limit our thinking here to buildable things.

UPDATE NOTE: The machine shop analogy, originally attributed to Mr. Gaitan, has been corrected to attribute it to Alissa Fitzgerald.

More info, where available:

IEC SC 47F

IEEE 1451 (via NIST)

iNEMI reliability project

iNEMI testing project

Japan Micromachine Center (JMMC)

SEMI MEMS standards