Where Industry and Security Intersect
What's New | Sitemap | Search
OPEN SESSION: Wednesday, January 28th
Agenda Item Presentations/Discussions:
Approximately 45 people were in attendance.
Opening: Raj Shea of NASA Ames Research Center welcomed the ISTAC to the NASA Ames Conference Center.
Industry Telecomm Proposals for 2009: Harvey Trop of JDSU reported on the current status of two telecomm technologies (high speed data transmission and optical switching) that he had first discussed at the ISTAC meeting in April 2007. Both of these technologies are the subject of industry-initiated proposals for possible consideration at Wassenaar in 2009.
Regarding high-speed data transmission, Harvey explained that 40G systems are still emerging but that the cost differential between 40G and 10G has been reduced. The key technologies needed for 40G are high-speed drivers, pre-amplifiers and modulators (at low data rates, it is possible to directly modulate the laser but high data rates use a CW laser and external modulation). Harvey believes that 100G is still relevant from an export-control standpoint, but the situation is less clear for 40G. Some key 40G technologies reside only in fabs that are currently within Wassenaar member states and it is likely to be some time before 40G VCSELs will be available. On the other hand, a variety of 40G components is commercially available including from Huawei (China). It was also pointed out that despite the technical note defining where to measure the “total digital transfer rate” there is still ambiguity and that clarification of the Note would be beneficial.
Regarding Optical Switching, there are two broad classes of technology: opto-mechanical (e.g., mirrors or MEMS) and waveguide-based (change the refractive index of a material, such as by changing the electric field or temperature). Harvey suggests that the critical control may be switching speed, which is typically in the range of 1-100 ms for telecom but in the ns-us range for applications of military concern. This boundary maps neatly to the two technologies, with opto-mechanical generally slower than 1 ms and waveguide-based generally faster. Accordingly, he suggests that 5E1c3 might be modified to add frequency switching time as a control parameter, with a control threshold of 1 ms.
Joe Young asked several questions:
What power levels are generated by lasers used in telecomm, and what distance do signals at those power levels propagate?
Answer: The power levels are typically 0.1-20 mW, but they are not direct-drive. Rather, after the signal is created it is passed through a modulator and an amplifier; the output of the amplifier typically travels 100 km before a repeater is needed.
Compare lasers for telecomm with lasers for cutting?
Answer: Lasers for telecomm are single-mode, usually with wavelengths in the range 820-980 nm (often 850 nm for short-range telecomm) and power of about 1 mW, whereas lasers for cutting are multimode with power typically 10 W.
What is the risk of diversion of telecomm lasers?
Answer: Little or no risk, due to the low power.
Foreign Availability: Roz Thomsen spoke on work that the Alliance for Network Security is doing to gather data on foreign availability of routers and operating systems. Data was presented showing that there is significant production capability for routers that exceed the criteria in 740.17(b)(2) in China. Further work is needed to determine whether the Chinese products meet the comparable quality/comparable quantity requirement for foreign availability under the EAR. However, China (principally Huawei) is active in industry forums and in seeking patents. Indeed, it was noted that Huawei is claimed to be the fourth most-active filer of patent applications in 2007. In addition, Huawei has R&D facilities in several US cities and considerable sales in the U.S.
Regarding operating systems, the key barrier to US exporters is the control on open cryptographic interfaces, and there is significant Chinese development in that product space, principally Red Flag Linux and also Symbian (the leading operating system for mobile phones in China). Interestingly, Red Flag Linux is beginning to gain some market share in converged devices (smart phones, PDAs, etc) and some western companies are using or supporting Red Flag. Finally, it was noted that encryption algorithms are increasingly available in chip form (one Chinese supplier is HiSilicon), making them easier to use and ever-more ubiquitous. No immediate action is needed by the ISTAC as this work is still in progress. The ISTAC will consider further action when and if requested.
Foreign Microprocessors: Raph Martelles provided an overview of foreign microprocessors, with emphasis on those exceeding the technology control level of 3A2. Key players are: Taiwan (VIA Nano, a 64-bit x86 compatible chip), Korea (Samsung ATM Core, a 16- or 32-bit embedded processor), Japan (Renesas, 16- and 32-bit cores for embedded applications; and Toshiba TX4939, which is believed to be below the 3E002 threshold), India (MosChop MCS8140, a general-purpose network microprocessor), Russia (Elbrus E2K, believed to be a defunct), and China (Godson, also known as DragonChip or Loongson). The general conclusion is China is a serious player in this technology space, with intent to have independent microprocessor capability, but other countries are ARM licensees. As this presentation was a technology overview and status report, no specific follow-up action was needed.
Laser Interferometry: David George of Agilent presented an introduction to and technology overview of laser interferometry, which is a technique that permits measurement of displacements as small as the sub-nanometer range. Although traditionally considered as a Cat 2 issue (2.B.6 and 2.B.8), interferometry has significant use in lithography steppers and interferometry equipment specially designed for lithography might be controlled in 3.B.1.f. Such equipment is normally sold to lithography equipment manufacturers, who integrate it into their tools and often develop their own control software. Subsequent to the technology overview, Jonathan Wise summarized the existing export controls for interferometers and suggested three possible issues for consideration: 1) clarification of controls on systems vs components; 2) review of the values of the control thresholds; 3) consideration of whether to differentiate between single-axis optics and multi-axis optics. No specific suggestions were made and any work done on this would be coordinated with the MPETAC.
LVS for 3A002: Jonathan Wise raised the question of whether license exception LVS for ECCNs 3A002.c, 3A002.d, 3A002.e and 3A002.f is still needed. It is known that some low-value arbitrary waveform generators had been exported with LVS prior to implementation of the Dec 2006 Wassenaar regulation (published in the EAR Nov 5, 2007). With that regulation, however, those instruments were decontrolled to 3A992.a, thereby obviating the need for any license exceptions. The issue of concern now is that the availability of LVS may create an incentive to set an artificially low price for options or license keys that upgrade an otherwise-uncontrolled (3A992.a) instrument to 3A002 performance. For example, an instrument hardware platform configured for 3A002 functionality might cost $35k, and would require an export license for shipment to many destinations. But the same instrument hardware platform configured for 3A992 functionality might cost $32k and be exportable as NLR, and a $3k license key upgrading the platform to 3A002 performance might be sold as 3A002 (LVS). That is, instrument capability that would normally be IVL could potentially be exported under a license exception in certain situations. To eliminate this possibility, it may be appropriate to consider removing LVS eligibility from ECCNs 3A002.c through 3A002.f.
Working Group Reports: Jonathan Wise reported on the activities of the Category 3A working group, Jeff Rogers reported on Category 3B, Bill VanLoo reported on Category 4, Frank Quick reported on Category 5 Part 1, and Roz Thomsen reported on Category 5 Part 2. Key points from the Working Group reports were:
Cat 3A: Eight Wassenaar proposals were been submitted to BIS, and the originators of those proposals have been working as necessary with BIS to refine and develop them. Three major topics are coming into focus for 2009: High-speed data converters; digital instrument architectures including software-defined instruments and digitizers; and the implications of emerging telecomm standards at ~60 GHz
Cat 3B: This group has held multiple conference calls and continues to work closely with BIS to develop proposals for 3B1c (etch) and 3B1e (cluster tools). OTE has contacted the group about possibly doing a foreign availability study for etch equipment.
Cat 4: A total of four Cat 4 Wassenaar proposals have been submitted to BIS. Issues for possible consideration in 2009 are the distinction between “electronic assemblies” and “systems” and whether technology has evolved so as to necessitate clarification of this distinction; APP aggregation of coprocessors; “Cat 4 containment”, i.e., determining what is datacomm that properly belongs in Cat 4 and what is telecomm that properly belongs in Cat 5p1; and APP calculation for “difficult” processors.
Cat 5p1: This group will have formal kickoff in February 2009. The group is working with BIS to advance two proposals (5E1c1 data transmission rate) and 5E1c3 (optical switching).
Cat 5p2: This group is working on forensics tools, concurrent VoIP session, wireless & performance, and certified operating systems. More generally, the focus is not just encryption but security technology.
Silicon Processing Technology: Hans Stork of Applied Materials spoke on emerging trends for manufacturers of semiconductor fab equipment. As a consequence of the current economic slowdown, customers are purchasing less new equipment and instead are seeking to extend the capability and performance of existing equipment. To respond to this new customer behavior, it was suggested that the equipment manufacturers will develop and market processing technology. Several examples of new processing techniques that can-for certain well-defined repeating geometries-double the device density were presented. Nevertheless, only incremental improvements, not fundamental breakthroughs in manufacturing technology, are seen out to about the 10 nm process generation. Finally, it was also noted that the skill (know-how and experience) of the user can result in as much as a factor of two difference in results on a given tool. As this presentation was informational in nature, no follow-up activity is required.
HPC Clusters: Henry Brandt provided a technology and engineering overview of HPC clusters. The major reasons for the existence of HPC clusters are to leverage the low entry cost of commodity PCs and the ability to scale up as needed for workload. Broadly speaking, there are two approaches to building HPC clusters: Roll-Your-Own (“RYO”) and vendor-provided. The key point of this presentation is that assembling a reliable HPC cluster is more complex than is commonly believed and that the engineering and integration services that vendors can provide are extremely valuable. A particularly important factor to consider when assembling a cluster is the reliability of the individual components (motherboards, DIMMs, hard drives, etc), and vendors have the experience—often obtained through statistical testing—to know which components and which brands deliver the necessary MTBF. Another important factor is system operation: knowing which operating systems are most reliable and how often one should do checkpoint backup. As this presentation was a technology overview and status report, no specific follow-up action was needed.
The open session was adjourned at 4:00pm.