Forum Discussion
Architectural feedback: <1ns Async Egress Gate + SHA3-256 XOR core on Agilex 7 (896 logic cells)
Hi everyone,
I’ve been developing a hardware-native cyber defense protocol (RQSD-CD v1.0 / Project Trident) designed to protect vital infrastructure (like undersea fiber backbones and 6G edge nodes) against physical sabotage and post-quantum interception.
The core has been synthesized with zero errors in Altera Quartus Prime Pro 26.1.1 targeting an Agilex 7 fabric, occupying 896 logic cells.
The architecture relies on three parallel hardware pillars:
1. Dynamic Ingress Failover: Continuous hardware-level health monitoring across 3 parallel channels with a 0us switch-over to backup paths.
2. Post-Quantum Cipher Core: A synchronous SHA3-256 bitwise XOR matrix utilizing hardware-native entropy, paired with an inverse Reversal Core for lossless decryption.
3. Asynchronous Egress Gate: An intelligent 3-step pulsed crash counter to filter out transient network drops. On the 3rd consecutive failure, it bypasses the system clock (rising_edge(clk)) to force the physical I/O pins into a High-Impedance ('Z') state in < 1ns.
My question to the senior FPGA architects here:
Since Prong 3 intentionally operates asynchronously outside the clock domain to achieve sub-nanosecond physical isolation, how would you best mitigate potential metastability issues on the downstream bus during that exact switching instant on an Agilex 7?
I’ve published the baseline HDL code (trident_core_base.vhd), the pulsed testbench, and the Questa RTL simulation waveforms on GitHub for architectural review (IP registered under Benelux i-DEPOT 158618):
https://github.com/avanessen81-design/RQSD-CD-Protocol-v1.0-Project-Trident
Any critical feedback on the timing constraints or async gate behavior is much appreciated!
2 Replies
- RQSD_Protocol
New Member
Thank you for this detailed architectural breakdown and the specific references to the Quartus Timing Analyzer guidelines (CDC-50011 and RDC). You make a very valid point regarding the difference between zero-delay RTL simulation behavior in Questa and actual silicon propagation delays on the Agilex 7 IOE. For the next revision of the hardware-ready Enterprise core, I will implement a proper two-stage synchronizer chain (without combinational logic in between) feeding the dedicated FAST_OUTPUT_ENABLE_REGISTER, and validate the exact post-fit pin-tristate delay and MTBF using the Quartus Timing Analyzer.Appreciate the exact guidance on making the asynchronous isolation path silicon-safe!
- ai_responder
Community Manager
There are three distinct technical issues here worth unpacking carefully.
The combinational / asynchronous OE path is the most dangerous element
What you describe in Prong 3 — logic that "bypasses rising_edge(clk)" to assert the output enable — is exactly the topology the Quartus Prime Design Assistant flags as rule CDC-50011 (Combinational Logic Before Synchronizer Chain). Incorporating logic between the registers in a synchronizer undermines its function of mitigating metastability risk when transferring signals across clock domains. The synchronizer's flip-flops are timed to capture and stabilize an asynchronous signal; introducing combinational logic can add unpredictable delays, complicate timing analysis, and lead to setup and hold violations — including glitches that, for asynchronous crossings, can cause the synchronizer to capture a value that is not real. 3.1.5. Debug CDC Example 2—Additional Logic in the Crossing — Quartus Prime Pro Edition User Guide Timing Analyzer
For a correct synchronizer, the destination of an asynchronous transfer must form a chain of two or more registers, with each register in the same clock domain as the destination of the transfer; no combinational logic is present between any of the registers in the chain; and no combinational logic is present on the path to the first register in the chain. 3.1.3.3. Identifying Synchronizer CDC Issues — Quartus Prime Pro Edition User Guide Timing Analyzer
Driving the IOE output-enable directly from a combinational crash-counter result violates all three conditions.
**The "< 1ns" tristate claim applies to RTL simulation, not to silicon**
I/O timing data from the data sheet is typically used prior to design to estimate the timing budget. The Quartus Prime Timing Analyzer provides more accurate and precise I/O timing data based on the specifics of the design after place-and-route is complete. I/O Timing — Agilex™ 7 FPGAs and SoCs Device Data Sheet M-Series In RTL simulation (Questa), a signal assignment to'Z'resolves in zero simulator time; that is a modelling artifact. On Agilex 7 silicon, the actual output-enable to pin-tristate delay is what the Timing Analyzer reports post-fit for your specific speed grade, IOE placement, and OE path length. No sub-nanosecond GPIO output-enable-to-HiZ spec is stated in Agilex 7 documentation for a combinationally driven OE path.
The async reset / control deassertion problem (the real hazard)
Your crash counter, once it asserts the HiZ signal, eventually must deassert it. A reset domain crossing (RDC) occurs when a reset signal originates in one clock domain and then resets registers or logic in a different clock domain. Similar to clock domain crossings, RDCs require careful handling because the reset signal is asynchronous to the clock of the destination domain — a condition that can lead to any of the following issues if not managed properly: 3.2.1. RDC Timing Overview — Quartus Prime Pro Edition User Guide Timing Analyzer metastability — if you do not properly synchronize a reset signal, flip-flops can enter a metastable state if the reset deasserts near the clock edge of the destination domain, leading to uncertainty in the flip-flop output; 3.2.1. RDC Timing Overview — Quartus Prime Pro Edition User Guide Timing Analyzer glitches — asynchronous reset signals can cause glitches during assertion or deassertion that propagate through the system and cause logic to reset unexpectedly. To prevent glitches, drive asynchronous reset signals with a single flip-flop (no combinational logic). A reset synchronizer does not prevent asynchronous reset glitching — only adherence to the no-combinational-logic-in-between guideline can prevent glitching. 3.2.1. RDC Timing Overview — Quartus Prime Pro Edition User Guide Timing Analyzer
Recommended mitigation approach
For a topology with an asynchronous signal originating from an external source or a different clock domain, you must synchronize it to avoid metastability. Similar to signals that cross domain crossings, the synchronizer typically requires a series of flip-flops clocked by the destination clock domain. These flip-flops sequentially capture and stabilize the asynchronous signal, filtering out potential metastability. You then use the output of the synchronizer to drive the sequential logic elements within the destination clock domain. This approach ensures all elements within the same clock domain are correctly reset, maintaining the integrity of the system's behavior. 3.2.1. RDC Timing Overview — Quartus Prime Pro Edition User Guide Timing Analyzer
For the OE register specifically, once the synchronizer output is registered, use theFAST_OUTPUT_ENABLE_REGISTERassignment, which implements an output enable register in a cell that has a fast, direct connection to an I/O pin. If such a fast, direct connection is not available in the I/O cell hardware, the Fitter locks the output enable register in the LAB adjacent to the I/O cell it is feeding. Turning on the Fast Output Enable Register option can help maximize I/O timing performance, for example, by permitting fast clock-to-output times. FAST_OUTPUT_ENABLE_REGISTER — Quartus Prime Pro Edition Settings File Reference Manual
For metastability analysis on the synchronizer chain itself, metastability problems can occur when a signal transfers between circuitry in unrelated or asynchronous clock domains because the signal does not meet setup and hold time requirements. Circuit designers typically use a sequence of registers (a synchronization register chain or synchronizer) in the destination clock domain. The mean time between failures (MTBF) is an estimate of the average time between instances of failure due to metastability; the Timing Analyzer analyzes the potential for metastability and can calculate the MTBF for synchronization register chains. 1.1.6. Metastability Analysis — Quartus Prime Pro Edition User Guide Timing Analyzer You can set a minimum required MTBF globally or per-chain usingMIN_MTBF_REQUIREMENTwhich specifies the minimum acceptable MTBF, either globally or for a specific synchronizer chain. The MTBF value used will be 10 to the power of the setting value, in years. If the MTBF of a synchronizer chain is less than this value, it will be marked as a dangerous, asynchronous transfer in need of additional synchronization registers. MIN_MTBF_REQUIREMENT — Quartus Prime Pro Edition Settings File Reference Manual
Bottom line: replace the combinational crash-counter-to-OE path with a registered synchronizer chain feedingFAST_OUTPUT_ENABLE_REGISTER, constrain the async source withset_false_pathand an SDCset_max_skewon the OE bus, then run Report Asynchronous CDC and check for CDC-50011 violations before claiming any switching time spec.
Note: To continue using AI Responder, please start your next inquiry with ai_responderDisclaimer: AI-generated responses may contain errors or omissions. Please verify technical guidance with official Altera documentation before implementation.
Thank you very much.