Ctrl-P to Print
Self-HEALing mixed-signal Integrated Circuits (HEALICs)
Program Manager and PoC: Dr. Sanjay Raman
Frequently Asked Questions
- Q: The BAA implies that designs done in CMOS technology nodes larger than 45 nm will be looked upon less favorably than those done at 45 nm and below. Can you please clarify? Are designs limited to CMOS or to Silicon-only processes?
A: This BAA does not specify any particular technology or technology node. The technology node is not a metric. Note, however, that ‘Relevance to the DARPA Mission’ is an evaluation criterion for proposals. Additionally, no semiconductor technology is off-limits for this BAA; recall, however, that there is a focus on mixed-signal (analog and/or RF + digital) designs. Also, proposers should be cognizant of the discussion of Program Phases starting on page 9 of the BAA document.
- Q: Must a demonstration design be manufactured? Or is yield and performance prediction ok?
A: Fabricated chips are required for evaluation against the program metrics.
- Q: Can we extrapolate yield based on shared fabrication runs?
A: Performance yield results from Multi-Project Wafers are acceptable as long as the extrapolated results are credible. A statistically significant number of fabricated die must be measured.
- Q: We would like to propose a demonstration that evolves an existing Mixed-Signal SoC design by designing into the next technology node. Is this acceptable?
A: It is advisable that the technology node for the Baseline design and for the Phase I + II designs should be kept the same in order to make fair comparisons; however, if circumstances require you to scale the design into a different technology node during the course of the program, a credible and justifiable mapping of performance metrics must be presented between the designs completed at different nodes.
- Q: It’s not explicitly mentioned in the BAA to push mixed-signal chip performance. Should our designs push state-of-the-art, or just prove the self-healing concept?
A: The intent of this program is not specifically to push the state of the art of any particular Mixed-Signal component/microsystem, such as ADCs or RF synthesizers; the intent is to dramatically improve the performance yield of existing designs through the use of self-healing. However, it is also desirable that the proposed performance metrics be sufficiently aggressive to show that performance metrics that could only be achieved in a small number of baseline chips can be achieved for a large number of chips with the application of self-healing technology.
- Q: Should self-healing occur all the time or can it be done during idle time in a particular application?
A: “Idle-time” or “off-line” self-healing may be acceptable; however, it is expected that a set of metrics for the performance of the self-healing circuitry (e.g. settling time) be defined in the proposal. If “idle-time” or “off-line” self-healing strategies are used, the time that the circuit is taken offline should be factored into the evaluation of these metrics.
- Q: What is the total Award amount for the DARPA/HEALICs
program?
A: The size of awards is determined in the course of the source-selection process. Proposers should propose a budget that is appropriate to their proposed statement of work.
- Q: How many proposals will be granted?
A: The number of awards to be made is determined in the course of the source-selection process. DARPA may select none, one, or several proposals for funding.
- Q: Can a particular
investigator be a participant on more than one HEALICs proposal?
A: Yes; there is no restriction on the number of teams on which a particular investigator can participate. However, DARPA could potentially select multiple teams for awards, in which case such an investigator would be expected to meet their commitments to each of the efforts they are participating in. If a particular investigator is spread too thin, this could impact the capabilities of the teams involved to accomplish their statements of work.
Other Technical Questions:
- Q: Are we limited to using Trusted Foundry or to US-based fabrication facilities?
A: No. On a related note, proposers are advised not to base their proposed efforts on the availability of “free” fabrication, unless such fabrication is specified in the proposal as a cost-sharing mechanism by one of the proposal team members.
- Q: Is design automation part of this BAA? Can an EDA vendor bid for a Phase I contract without teaming up with a design group?
A: Design automation may be an acceptable approach for some aspects of the program. However, fabricated chips are required for evaluation against the program metrics, so proposers should structure their teams appropriately.
- Q: HEALICs is a mixed-signal program. Does it mean RF front-end (e.g. LNA, PA) circuits are excluded?
A: No, they are not excluded. However, proposers should be cognizant of the discussion of Program Phases starting on page 9 of the BAA document.
- Q: What if our original design submission includes an inadvertent design error that is fixed in the process of redesign for self-healing? How can our performance yield be consistently evaluated?
A: In this case, the original baseline design would have to be re-evaluated for its performance and yield in order to make a fair comparison with the new self-healing design.
- Q: Is the integration of over 1 million transistors required for success in Phase II?
A: There is no hard metric for the number of transistors. The transistor counts enumerated in the discussion of the program phases are anticipated levels of complexity, but are not strictly required.
- Q: Robust control is optimal, but not required. How about neuroscience approaches?
A: If the proposed approach is able to meet the goals of the self-healing mixed-signal IC program, then it may be of interest.
- Q: Does the control circuitry have to be implemented in hardware, or is software control acceptable? In other words, can I take advantage of an on-chip microprocessor to perform self-healing algorithms?
A: Self-healing in software could be acceptable; however, it is expected that all of the self-healing circuitry will be fabricated on-chip.
- Q: If radical new approaches are taken to “heal” mixed-signal ICs, how sensible is it to compare the yield between the new HEALICs and the conventional counterparts?
A: The performance yield metrics are defined in the BAA.
- Q: Would MTO prefer analog healing rather than digital healing?
A: The BAA does not specify a preference between digital and analog healing methodologies.
- Q: Say I have a general approach that applies equally to ADC and RF circuits. How should I propose? Should I propose 2 designs? Or just pick one?
A: You may propose as many circuits as you like; however, your schedule and budget should be commensurate with the statement of work that you propose.
- Q: Should the self-healing be application independent to the best degree possible?
A: The self-healing methodologies should be as application independent as possible. For example, if the target mixed-signal core is a self-healing frequency synthesizer, it would be desirable for the self-healing approach to be generally applicable whether the synthesizer were to be used for radar or for communications applications.
- Q: Is there a preferred frequency range of operation?
A: No.
- Q: The BAA touted process portability. Are you ruling out CMOS + MEMS unique process combinations?
A: We are not ruling out CMOS + MEMS or any other unique process combinations. However, proposers are cautioned that the fundamental goal of the program is the development of self-healing IC design approaches.
- Q: This program is focused on improving yield for process variability, but how much will proposed techniques that directly address environment or ageing be rated or supported?
A: The proposals will be scored against the BAA-defined evaluation criteria.
- Q: Do we need to design two separate chips, one with self-healing and one without?
A: The baseline design would be without self-healing. If the question is whether two versions of the design need to be fabricated in Phase I (and II) of the program, not necessarily.
- Q: Will the final Phase 1 demonstration require controllers (ADC, FPGA, etc.) on the SoC?
A: The proposed self-healing control system should be on the chip.
- Q: Is the power overhead average from power on to power off? (i.e. can the peak power overhead be higher than 10%/5%?
A: It is up to the proposer to specify the modes of operation of their design, and to assign the appropriate power consumption metrics to those modes. Note that the power consumption metric is the percentage of power consumption above that of the baseline design.
- Q: Is the overhead metric (5%/10%) based on what is actually built, or what the full chip would be? Does the area overhead include multifunctional circuits that contribute to function and healing?
A: The overhead metrics apply to the fabricated self-healing designs in Phase I and Phase II.
- Q: The BAA pre-supposes a solution to variability using self-healing. Technologies that lower the base-line circuit variability could be used such that self-healing would have less correction required. Is this of interest for this BAA?
A: The goal of this program is the development of self-healing circuit design techniques and algorithms. While techniques such as “design-for-yield”/layout optimization may be able to partially meet the program metrics, these are not techniques of primary interest.
- Q: Are the MARCO/FCRP research results available to non-MARCO members?
A: You are encouraged to contact the FCRP directly to gain access to their publications and research results (http://fcrp.src.org/member/about/about_centers.asp?bhjs=1). In addition, there may be potential teaming opportunities with FCRP investigators.
- Q: How do you want to differentiate the HEALICs program objectives from what was presented as going on in the GSRC and C2S2 during the proposer’s day?
A: The FCRP information provided during the workshop was for background information and to highlight potentials for teaming with FCRP institutions. See the BAA for the goals and objectives of the HEALICs program.
- Q: Would VDD scaling, adaptive clock stretching to improve error resiliency be of interest?
A: If the approach can be applied to mixed signal SoC designs, then it may be of interest.
- Q: Can you clarify if there is any specific interest in supporting self healing IC research for space-based mixed signal SoCs under the HEALICs BAA?
A: DARPA does not specify interest in any particular design area. Note, however, that ‘Relevance to the DARPA Mission’ is an evaluation criteria for proposals.
Other Contracting Questions
- Q: Can the lead institution be a university or must it be a company?
A: The lead institution can be a university. If a Federally Funded Research and Development Center (FFRDC) or Government Laboratory is proposed as team member (either in prime or subcontractor role) the proposal must provide documentation to establish that such organizations are eligible to propose against a Government solicitation and support that they are providing unique capabilities not otherwise available in private industry (See BAA Section III “Eligibility Information”).
- Q: What feedback will be received from the abstract submission?
A: The feedback at a minimum will be to ‘Encourage’ or ‘Discourage’ submission of a full proposal. Additionally, all ‘Discourage’ letters will provide technical feedback regarding DARPA’s recommendation/rationale. ‘Encourage’ letters will provide technical feedback to the maximum extent that is deemed practical and/or which is in the Government’s best interest.
- Q: Can foreign partners, such as university professors, be co-investigators or subcontractors to a US investigator?
A: There are no restrictions on the participation of non-US citizens or permanent residents as investigators or co-investigators . There are also no specific restrictions on the participation of foreign institutions, but potential performers must be cognizant of ITAR and export control restrictions to the extent they may apply.
- Q: For consortiums or collaborations, do we need to arrange the group now and submit only one abstract before July 3rd, or do we submit an abstract with the piece of the solution we are proposing and then organize a consortium before the proposal deadline?
A: It is advisable to fully identify your team for the abstract submission. If not, this is OK, although this may affect DARPA’s recommendation, but the final team would need to be in place when the final proposal is submitted and fully detailed therein. For consortiums, as applicable when proposing an OTA TIA award instrument, the details of the consortium structure will be negotiated subsequent to selection for an award, but the capabilities and roles of the consortium members should be clearly discussed in the proposal.
- Q: What is the minimum size of the team that DARPA will consider funding?
A: The minimum size that can be proposed is a single investigator. However, the proposers should carefully consider the BAA requirements when making teaming decisions.
- Q: Is it necessary to have industrial partners?
A: No.
- Q: Will ideas in our abstract be visible to
competitors? For example: additional metrics which add
real value to product use, not ones that are just
academically interesting.
A: Abstracts are handled as source selection information and, as such, are not visible to anyone outside of the DARPA program and source selection teams.
- Q: How (if at all) is the IP on the baseline design protected?
A: It is assumed that proposers will identify their baseline designs as proprietary information. The Government is prohibited from disclosing proprietary information to 3rd parties except as may be permitted per the negotiated terms and conditions of any resulting award instrument. Within a given performer team, IP protection issues must be worked out between team members prior to proposal submission. See also BAA Section VI (B)(2) “Intellectual Property” for further details regarding proposer assertion requirements for all data and/or software that will be delivered to the Government with less than Unlimited Rights. Failure to properly respond to this BAA requirement may render a proposal non-responsive.
- Q: Please comment on ownership/use by DARPA of new IP that will evolve from this work.
A: In general, DARPA negotiates data and patent rights (IP) terms and conditions in accordance with FAR and DFARS Part 27. Guidance is similar for grants, cooperative agreements and Technology Investment Agreements (TIAs) which follow the DoD Grant and Agreement Regulations. In any case, the Government does not seek to take ownership of the resulting IP. If the resulting IP is funded solely by the federal government then typically the Government seeks a license allowing for government and/or 3rd party use of the resulting IP (e.g., Unlimited Rights). For that IP which is considered co-funded (government and contractor funded), the Government typically seeks a license allowing for government and/or 3rd party use for government purposes only (e.g., Government Purpose Rights). See also BAA Section VI (B)(2) “Intellectual Property” for further details.
- Q: Do you need a different account for each BAA teaming website?
A: Yes.
- Q: In section 2 (Proposal abstract format) the BAA states: “… the total length should not exceed 15 pages, excluding cover page and official transmittal letter. …
” and further in the same paragraph:“ No formal transmittal letter is required.”
We are somewhat confused in use of the two terms: “official transmittal letter” and “formal transmittal letter” Are the two the same? If so, our understanding is that no transmittal letter (either formal or official) is to be sent with the proposal abstract.
A: A transmittal letter is not required for the abstract submission. A transmittal letter IS required for a full proposal submissions.