Information Systems Technical Advisory Committee
(ISTAC)

Date: July 23-24, 2014

Time: July 23 from 9:05 AM to 4:30 PM
July 24 from 9:00 AM to 12:30 PM

Location: July 23 in Room 3884, HCHB
July 24 in Room 3884, HCHB

Agenda Item Presentations/Discussions:

PUBLIC SESSION (July 23, 2014)

The meeting opened at 9:05 AM. Approximately 27 people were in attendance.

Opening: Jonathan Wise opened the meeting with introductions and request for comments from the public. There were no public comments.

Working Group Reports: Jonathan Wise reported on the activities of the Category 3A working group, Jeff Rogers reported on Category 3B, Henry Brandt 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: This group held a teleconference to begin planning Wassenaar proposals for 2015; these will be summarized in a separate open-session presentation. Mark Renfeld raised a possible ambiguity in several definitions in EAR part 772; this will be explained in a separate open-session presentation.

Cat 3B: This group has been coordinating with Semi to provide data and support for SEMI's Foreign Availability petition regarding 3B1c plasma dry-etch equipment. In parallel, this group is considering proposals for relaxation to 3B1c in the event that the Foreign Availability petition is unsuccessful.

Cat 4: This group has held several teleconferences. They are working on a revision to the "APP Practitioner's Guide," including addition of new/relevant examples and updates to address regulatory changes. The group is also working on Wassenaar proposals for 2015, which are expected to focus mostly on relaxation to APP thresholds.

Cat 5p1: This group has had no recent activity.

Cat 5p2: This group has had several conversations with Aaron Amundson regarding the OAM decontrol proposal, to refine the language of the proposal and do contingency planning.

Definition Issues in Cat 3: Jonathan Wise gave a presentation written by Mark Renfeld of Hewlett-Packard (Mark was attending the meeting by phone).

-Wassenaar uses double quotes to indicate terms that have a global definition; such terms are not defined locally.

-The EAR almost always follows Wassenaar practice, and uses double quotes to indicate terms that have a global definition (in part772). This presentation gives two examples where the EAR does not follow Wassenaar practice.

-In part 772, the (global, double-quote) definition of "monolithic integrated circuit" refers to "circuit element" (double quotes) and to 'chip' (single quotes). One would expect that because "circuit element" has double quotes, it would have a global definition in 772 but in fact it has only a local definition. One would also expect that because 'chip' has single quotes it would have a local definition but in fact it has neither local nor global definition. In Wassenaar, "circuit element" in double quotes has a global definition, but 'chip' in single quotes has neither local nor global definition.

-In part 772, the (global, double-quote) definition of "substrate" refers to "discrete components" (double quotes). One would expect that because "discrete components" has double quotes, it would have a global definition in 772 but in fact it has only a local definition. Further, that local definition refers to double-quoted "circuit element" which suggests the existence of a global definition but as described above "circuit element" is only defined locally within "monolithic integrated circuit". In Wassenaar, both "discrete component" and "circuit element" in double quotes have global definitions.

-It appears that 'chip' is not intended to have a local definition because it is used colloquially and is qualified: "...a so-called 'chip'"

-Although there is little or no evidence that the EAR definitions are causing confusion, there is an opportunity for clarification.

-Actions: The Cat 3A Working Group will write a letter to BIS about this and will copy the RPTAC.

Old Business/Open Business: Jonathan Wise summarized the status of old/open business items.

-From April 2014, PRBA, the Rechargeable Battery Association, represented by George Kerchner raised a concern that 3A1e1b may soon control many lithium-ion cells in common commercial use. Subsequent to the April ISTAC meeting, Jonathan Wise held two teleconferences with George to provide suggestions and guidance on how to develop a proposal for submission to BIS, and ISTAC member Erik Oliver has been assigned as a mentor. We expect that PRBA will submit a proposal before the nominal November 1, 2014 deadline.

-From April 2014, ViaSat raised a concern that the proposed 3A611.c would capture some MMICs that are in widespread civil use. ISTAC member Doug Carlson was assigned to lead a subgroup to develop an ISTAC position/statement on this matter. In the interim, however, on July 1, 2014, BIS published a Notice of Inquiry (79FR37548), seeking public comments on this same issue; comments are due on September 2, 2014. Doug will continue to lead the ISTAC subgroup, but there is now the possibility that instead of an ISTAC position/statement, responses will be in form of separate industry comments to the Notice of Inquiry.

-From April 2014 (closed session), an issue was raised that a strict reading of 3A001b4 (controls on microwave solid-state amplifiers) reveals that the frequency range and the output power are not linked. That is, the output power is not explicitly required to be within the associated frequency range. This leads to the unexpected outcome that a downconverter that would not be controlled on the basis of its output power at low frequency might still be controlled on the basis of its high-frequency input. However, this issue is believed to be resolved by December 2013 Wassenaar list changes, which adopt the now-standard linguistic construction of "within the frequency range", in this case "peak saturated output power within the [specified] frequency range". Author's note: The December 2013 Wassenaar changes were published by BIS on August 4, 2014 (79FR45288).

-From November 2013, Ben Flowe raised a question pertaining to the scope of control of 5A001b6 (voice coding at rates <2400 bit/sec). The ISTAC interpretation was that this entry applied only to items that can perform voice coding such that the output is intelligible to a human listener, and does not apply to applications such as speech-to-text conversion (e.g., dictation), speaker identification (e.g., identification of who is speaking on a multi-party call) or to speaker state recognition (e.g., emotional state). At that time, it was decided that the ISTAC Cat 5p1 working group would submit this matter to BIS in the form of an advisory opinion. In an advisory opinion issues on May 7, 2014, BIS concurred with this interpretation. There was a discussion of whether the ISTAC would request that this advisory opinion be published on the BIS web page containing advisory opinions, but it was decided instead to release the advisory to interested parties.

-From early 2012, the FPGA industry submitted a Wassenaar proposal to replace the problematic parameter "system gate count" with "aggregate one-way peak serial transceiver data rate." This proposal was adopted as part of the Dec 2012 Wassenaar list changes, and subsequently implemented in the EAR. The industry has requested that this parameter change be propagated to the 3A991.d AT controls, so that the problematic parameter would be removed there as well. Author's note: This change was implemented in the December 2013 Wassenaar regulation published by BIS on August 4, 2014 (79FR45288).

3A2a6 Instrumentation Data Recorders: Tom Hill of Tektronix spoke on several issues pertaining to 3A2a6 instrumentation data recorders. Key points were:

-The criteria for determination of what is an "instrumentation data recorder" are ambiguous. One interpretation is that determination could be based on the use to which the instrument is put (e.g., for recording data from instrumentation, but presumably not for recording internet downloads, television broadcasts, home movies and the like). An alternate interpretation is that determination could be based on the name/description of the instrument (e.g., it is described in the datasheet as being an instrumentation data recorder).

-Two examples were given: 1) a USB data logger that can store 32k points per channel at a rate of one data point per second, and 2) equipment marketed as a data recorder, having one or two channels each specified at 160 MB/s data rate and up to 2 Gbit/s aggregate, and having 20 TB of magnetic storage of 30 TB of solid-state storage. It was suggested that the USB devices is not a data recorder but that the other equipment likely is.

-The current 3A2a6 control language refers specifically to magnetic storage, with the consequence that recorders having solid-state memory (storage) are automatically excluded from control. Consideration could be given to modernizing 3A2a6 to include all types of storage, but this poses the risk of the undesired side effect of capturing consumer electronics that have USB ports.

-USB2.0 has data rate 60 MB/s or 480 Gbit/s aggregate, which is below the 3A2a6 thresholds of 100 MB/sec and 1 Gbit/sec aggregate. By contract, USB3.0 has data rate 625 MB/sec and 5 Gbit/sec aggregate, both of which clearly exceed the 3A2a6 thresholds.

-Because USB3.0 is ubiquitous in and so greatly exceeds the 3A2a6 thresholds, this may be an argument for deletion of entry 3A2a6.

-Actions: The Cat 3A Working Group will review this and will likely write a proposal to delete entry 3A2a6.

Ideas for 2014 Wassenaar Proposals: Jonathan Wise presented an overview of ideas for possible instrumentation-related Cat 3/4/5 Wassenaar proposals for 2014.

-3A1a5a: Seek broad relaxation for ADCs, to account for technology advances since the last broad update in 2010 (although the threshold for 3A1a5a1 was relaxed in 2013). Additionally, evaluate the possibility of adding another control parameter (possibly latency) to better differentiate ADCs of national security concern.

-3A1b4: Simple grammatical clarification to replace the phrase "microwave assemblies/modules containing microwave solid state amplifiers" with the phrase "amplifier modules/assemblies" to narrow the scope of control to items whose principal functionality is microwave amplifier. Alternately, move the text of Note 3 ("3A1b4 includes transmit/receive modules and transmit modules") to the header, approximately as "microwave sold-state amplifiers, transmit/receive modules and transmit modules".

-3A2c5: Consider adding RTBW as a control parameter, parallel to the structure of 3A2c4.

-3A2c4/3A2c5: Clarify the control language for RTBW to capture instruments that might mischievously be decontrolled by addition of infrequent gaps.

-3A2d: Consider separation of arbitrary waveform generators from signal generators. One possible approach would be to control arbitrary waveform generators on the basis of sample rate and bit resolution, but this approach is unlikely to be straightforward.

-3A2e: Clarify the new Technical Note (from Dec 2012 Wassenaar) regarding Nonlinear Vector Network Analyzers. The intent is not to change the scope of control but to clarify the Note to assure that all readers, worldwide, will have the same understanding and interpretation.

-3B1a1: Replace the problematic phrase "capable of" (which potentially captures may tools that were neither designed nor intended to be used for "other than silicon"). One possible approach is to substitute "designed or optimized".

-3D3: Delete this entry, as the subject software is believed not to exist and unlikely to exist in the future.

-3E2c: This control on certain DSP technology currently refers to "fixed-point" calculations. Due to technology advances in the field, this proposal would seek relaxation by limiting this control to floating point calculations.

-Cat 3/Cat 5p1 "Frequency Switching" Definition: This definition was originally "to within 100 Hz of final frequency", but that was thought to be too tight and was changed to "to with +/- 0.05% of final frequency", but this is proving to be too loose. This proposal is to seek an intermediate value, tentatively 1 ppm.

-4A3/4D1/4E1: Update the APP thresholds to account for continuing advances in HPC technology and performance levels, or delete these entries. It is expected that this will become a recurring annual request. Dave Robertson commented that the last major issue for these ECCNs was aggregation, and wonders whether it is time to look for a new discriminator (control parameter) instead. Henry Brandt explained that the largest HPC systems do not aggregate. Jim Ramsbotham added that the basis of export control is that the performance curve has a knee. Dave responded that if there is no knee in the curve then there is nothing left to control.

-4A3e: Notionally move this control from Cat 4 to Cat 3, because this instrumentation is more commonly thought of as electronic test & measurement equipment than as computer equipment. Modification of the control parameters might also be considered (e.g., alignment with changes to 3A1a5a such as to add latency as a control parameter).

-5A1d: This proposal would exclude from control phase array antennas for consumer products (e.g., WiGig) by limiting the scope of control to phased array antennas for radar. Frank Quick suggested that if these antennas are limited only to radar, then they would no longer belong in Cat 5p1, and should perhaps move to Cat 7. It was also suggested that the antenna cannot exist independently from the driving electronics, the implication of which is that consideration should be given to rewriting the control to address that.

-Cat 5p2 Encryption Bit-Strength Thresholds: This proposal would seek, once again, to relax the bit-strength thresholds in Cat 5p2. The proposed new values of the thresholds are still under evaluation.

-Action: The several Working Groups will continue to review their respective proposals, with intent to submit drafts to BIS by November 1, 2014.

ADC Proposal for 2015 Wassenaar: David Robertson of Analog Devices outlined a proposal for updates to and relaxations of the 3A1a5a ADC controls. Key points were:

-ADCs are a $2B annual market, which is dominated by US companies. High-speed ADCs are a $250M-300M market but are the fastest-growing segment. US exports to major markers in China are affected by US export license processes (e.g., as compared to other Wassenaar members), and US dominance is threatened by emerging suppliers in China, who are aided by Chinese government subsidies. The US ADC industry also must obtain many technology licenses, both for its foreign design centers and for foreign nationals working in the US.

-The technology drivers (market trends) for ADC development are: 1) Broadband systems, to move increasingly large amounts of data. Microwave point-to-point systems now use 500 MHz and 1 GHz BW channels (in the upper frequency bands), and 5G cellular will have significant bandwidth requirements. 2) Multi-band/multi-mode systems such as software-defined radios and instruments, and the emergence of "cognitive" wideband systems; 3) FPGAs and ASICs at and beyond the 28 nm lithography node can process data at multiple GHz. In summary, RF data converters are becoming mainstream.

-For telecommunications infrastructure, the desired resolution is 12, 14 or 16 bits (high dynamic range allows reception of weak signals in the presence of strong signals, and enables multicarrier/software radio techniques). Increasingly, "software-defined radio" architectures are seeing to capture entire bands with a single ADC: 350 MSPS for capturing 120 MHz bands, and 2.5-2.7 GSPS for capturing 1 GHz cable bands.

-The history of Wassenaar changes for ADCs is as follows: 2004, control parameter was changed from conversion time to conversion rate. 2006, broad changes to control thresholds. 2011, broad changes to control thresholds, and updates/clarifications to the Notes. 2013, relaxation to control threshold only for 8-bit ADC.

-The general approach of this proposal is to bring the control thresholds within a factor of five of the commercial state-of-the art. For this proposal/analysis, "commercial state-of-the-art" means that devices are available from a single vendor (or at most two or three vendors); that special manufacturing (e.g., process, packaging, test), not yet widely available might be used; and that the devices may demand a premium in terms of high price (>$100) or high power usage (several watts). "Mainstream/mass-market" means that devices are available from multiple vendors (>3); that innovation reflects power reduction and improvements in secondary specifications; that pricing reflects competition and improvements on the technology learning curve (e.g., <$50 or <1 Watt power); and that the fab processes are "conventional" and are generally available in foundries and from subcontractors. "Commodity" means that it is implementable by any good team "schooled in the art"; that it is generally available as an IP block, such that the function is routinely embedded in ASICs and SOCs; and that the price is <$10.

-The proposal has three parts: 1) updates to the control thresholds; 2) revision to the groupings of bit resolution within the several subparagraphs, for consistency of grammatical structure; and 3) consideration of latency as an additional control parameter.

-The proposed updates to the control thresholds are as indicated in the following table:

-The proposed revisions to the grammatical structure are as follows: Currently, the groupings are 8 or 9 bit; 10 or 11 bit; 12 bit; 13 or 14 bit; >14 bit. The proposed new groupings are 8 or 9 bit (no change); 10 or 11 bit (no change), 12 or 13 bit (groups 13 bit with 12 rather than with 14); 14 or 15 bit (groups 15 bit with 15 rather than in the umbrella >14); and 16 bit or more (umbrella starts at 16 bit rather than 14).

-As a possible new control parameter, latency (the delay between when the sample is taken and when the complete digital output from that sample is available to the user) may serve to differentiate applications of national security concern (e.g. electronic warfare) from typical commercial applications (e.g. telecom). This may be especially relevant as ADC sample rates exceed 1 GSPS. Interestingly, and perhaps ironically, this approach goes back to the 2004 Wassenaar controls, which were based on conversion time (i.e., latency). Sources of latency include pipeline, interleaving, digital post-processing, and data interface. Special designs are typically required to achieve latency <50 ns.

-Possible control language for latency would use the "OR" construction, e.g. 3A1a5a1: A resolution of 8 bit or more, but less than 10 bit, with an output rate greater than 2 billion words per second OR a throughput latency of less than 50 ns.

-Action: Dave will develop this into a proposal and will submit it to BIS by November 1, 2014.

Data Converters and Instruments: Dave Robertson continued with a high-level overview of the relationship to and role of data converters (both ADC and DAC). Key points were:

-Wassenaar controls on ADC-based instruments are 3A1a5a (ADC ICs themselves), 3A2a5 (waveform digitizers), 3A2a6 (digital instrumentation data recorders), 3A2a7 (real-time oscilloscopes), 3A2c (signal analyzers), 3A2e (network analyzers, receive side), 3A2f (microwave test receivers) and 4A3e (digitzers for computers). Wassenaar controls on DAC-based instruments are 3A1a5b (DAC ICs themselves), 3A1a13 (direct digital synthesizers), 3A1b10 (oscillators), 3A1b11 (frequency synthesizer assemblies), 3A2d (frequency-synthesized signal generators) and 3A2e (network analyzers, transmit side).

-The block diagrams of ADC-based instruments and of DAC-based instruments are essentially the inverse of each other, both having signal conditioning, filtering, data conversion, signal processing, and a clock. Each block has certain critical technical functions that may be pertinent for export control. Instrument performance and fidelity is hugely dependent on clock jitter and phase noise.

-It was suggested that analysis of instrument block diagrams from the data-conversion perspective could be the basis of a formalism with which modify the Wassenaar controls to address software-defined instruments.

-Action: This presentation was mostly informational in nature, intended to stimulate thinking. To that end, no specific actions were necessary, but the 3A Working Group will revisit these ideas as necessary.

The open session was adjourned at 11:50 AM.