pith. sign in

arxiv: 2606.04281 · v1 · pith:JFREMK6Ynew · submitted 2026-06-02 · 💻 cs.ET

Functional Interface Blocks for Neuromorphic Hardware: A Junction-Centered Framework

Wellington Avelino , Yann Beillard , Fabien Allibart , Dominique Drouin , Gilberto Medeiros-Ribeiro This is my paper

Pith reviewed 2026-06-28 07:02 UTC · model grok-4.3

classification 💻 cs.ET
keywords neuromorphic hardwareheterogeneous systemsinterface frameworkjunction conditionsfunctional blocksmemristive devicescurrent conveyor
0
0 comments X

The pith

Assigning drive and sense roles at junctions and grouping them into functional blocks solves compatibility issues in heterogeneous neuromorphic hardware.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes a junction-centered framework that describes connections between dissimilar neuromorphic devices by assigning drive or sense roles at each junction. These roles are then collected into a small set of canonical functional interface blocks that replace direct coupling. Direct coupling fails because the resulting load-line conditions place devices in unintended operating regimes. The blocks therefore supply a reusable layer that makes the overall system function correctly regardless of the individual device characteristics. The approach is shown in hardware by realizing the blocks with current conveyors and testing them in a circuit that pairs a memristive synapse with a unijunction-transistor neuron.

Core claim

Inter-device connections are described through assigned drive/sense roles and organized into canonical functional interface blocks; a CCII-based composite circuit realizes these primitives, and the resulting framework is validated by correct operation of a Pavlovian-conditioning demonstrator that combines a memristive synapse with a UJT post-neuron.

What carries the argument

The junction-centered interface framework that assigns drive/sense roles at each connection point and assembles them into reusable functional interface blocks.

If this is right

  • Designers can compose systems from different device technologies without custom interface circuitry for every pair.
  • System behavior can be analyzed from the local role assignments rather than from exhaustive simulation of every device characteristic.
  • A small library of blocks realized with current conveyors supplies a practical hardware layer for many device combinations.
  • The same role-assignment method can be applied to larger networks once the primitive blocks are verified.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same junction-role method could be used to match devices whose dynamics differ in time scale rather than in steady-state characteristics.
  • Software that automatically selects and wires the blocks could turn the framework into a design tool for mixed neuromorphic circuits.
  • Adding blocks that enforce timing constraints would extend the approach to systems where phase or frequency mismatch is the dominant problem.

Load-bearing premise

Defining drive and sense roles at junctions and collecting those roles into standard blocks will overcome the operating-regime mismatches that arise when devices with dissimilar dynamics are connected directly.

What would settle it

A test circuit built with the proposed blocks in which the devices still fail to reach the intended operating points because of unaccounted load-line effects.

Figures

Figures reproduced from arXiv: 2606.04281 by Dominique Drouin, Fabien Allibart, Gilberto Medeiros-Ribeiro, Wellington Avelino, Yann Beillard.

Figure 1
Figure 1. Figure 1: From neural-network abstraction to conditioned physical interaction enabled by the FIB. (a) Neural￾network representation, where computation is described at the functional level (neurons and weighted synapses). (b) Physical realization of the same network in a heterogeneous crossbar-like hardware substrate, where memristive synapses (R12-R42) connect input neurons (n1-n4) to a postsynaptic node, resulting … view at source ↗
Figure 2
Figure 2. Figure 2: Junction-to-FIB placement guide for a crossbar-based SNN. (a) Inference/read mode: External input variables xi(t) are conditioned at the front-end junctions J1-J4, passed through the pre-neuron stage, and converted into pre-synaptic row-driving signals applied at J4-J8. The crossbar performs the weighted summation, and the resulting column current is accumulated as I(t) at the column junctions J9-J12, wher… view at source ↗
Figure 3
Figure 3. Figure 3: Taxonomy of core FIB primitives. The interface blocks are organized by the electrical variable sensed at the input (rows) and the variable driven at the output (columns). Each cell depicts the corresponding primitive—VCVS, VCCS, CCVS, and CCCS—highlighting the key impedance requirements that enable functional decoupling: high input impedance for voltage sensing and low input impedance for current sensing, … view at source ↗
Figure 4
Figure 4. Figure 4: CCII-based FIB as a practical realization of the interface taxonomy. (a) Ideal second-generation current conveyor and its canonical interpretation: terminal Y is a high-impedance voltage input, terminal X is a low-impedance node with Vx ≈ Vy, and terminal Z is a high-impedance current output with Iz = ±Ix, corresponding to a composite VCVS-like (Y→X) and CCCS-like (X→Z) behavior. (b) Mode-dependent use of … view at source ↗
Figure 5
Figure 5. Figure 5: Experimental demonstration of Pavlovian conditioning with a memristive synapse and a UJT-based post￾neuron enabled by a CCII interface. (a) Simplified test setup showing the memristor synapse, the CCII-based interface (FIB), the bipolar write-back pulse driver used for plasticity, and the post-neuron implemented with a unijunction transistor (UJT). (b) Oscilloscope capture of a representative event, showin… view at source ↗
Figure 6
Figure 6. Figure 6: Pavlovian-conditioning validation protocol. Experimental results showing the four-phase Pavlovian proto￾col illustrated by spike rasters for the pre-neurons (Food/US and Bell/CS) and the post-neuron response (Salivation). The evolution of the memristive synapse during the training and validation protocol is summarized in Fig.7, which reports the resistance trajectory across epochs as an occurrence-coded ma… view at source ↗
Figure 7
Figure 7. Figure 7: Pavlovian-conditioning validation protocol. Experimental results showing the four-phase Pavlovian proto￾col illustrated by spike rasters for the pre-neurons (Food/US and Bell/CS) and the post-neuron response (Salivation). Beyond the average trend, the occurrence-coded representation provides a statistical view of the protocol repeatability: high-occurrence regions indicate that multiple runs reached simila… view at source ↗
read the original abstract

Heterogeneous neuromorphic hardware integrates devices with dissimilar electrical characteristics and dynamics, making functional compatibility at their interconnections a primary design challenge. Direct coupling alone is insufficient to ensure correct operation, because the load-line conditions established at each junction determine the effective operating regime. Here, we propose a junction-centered interface framework in which inter-device connections are described through assigned drive/sense roles and organized into canonical functional interface blocks. As a concrete hardware realization, a second-generation current conveyor (CCII)-based implementation is then adopted as a composite realization of these interface primitives. The framework is validated experimentally in a Pavlovian-conditioning demonstrator combining a memristive synapse with a unijunction-transistor (UJT) post-neuron. By linking local junction conditions to reusable interface functions, the proposed methodology provides a systematic basis for the design and analysis of heterogeneous neuromorphic systems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The paper proposes a junction-centered interface framework for heterogeneous neuromorphic hardware. Inter-device connections are described by assigning drive/sense roles at junctions and organizing them into canonical functional interface blocks; these are realized in hardware via second-generation current conveyors (CCII) and validated experimentally in a single Pavlovian-conditioning demonstrator that combines a memristive synapse with a unijunction-transistor (UJT) post-neuron.

Significance. If the framework can be shown to generalize, it would supply a reusable, junction-level methodology for ensuring functional compatibility between devices whose I-V characteristics and dynamics differ, addressing a recognized obstacle that direct coupling cannot reliably solve.

major comments (1)
  1. [Abstract and experimental validation] Abstract and experimental validation section: the central claim that the methodology 'provides a systematic basis for the design and analysis of heterogeneous neuromorphic systems' rests on the assertion that drive/sense assignment plus canonical blocks resolve compatibility for arbitrary dissimilar device dynamics. Only a single Pavlovian demonstrator (memristive synapse + UJT post-neuron realized with CCII primitives) is supplied; no second device pair, no enumeration of failure modes outside the example, and no general derivation mapping blocks to arbitrary I-V curves are given, leaving the scope of the claim untested.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed assessment and for identifying the need to clarify the intended scope of the framework's generality. We respond to the major comment below.

read point-by-point responses

  1. Referee: [Abstract and experimental validation] Abstract and experimental validation section: the central claim that the methodology 'provides a systematic basis for the design and analysis of heterogeneous neuromorphic systems' rests on the assertion that drive/sense assignment plus canonical blocks resolve compatibility for arbitrary dissimilar device dynamics. Only a single Pavlovian demonstrator (memristive synapse + UJT post-neuron realized with CCII primitives) is supplied; no second device pair, no enumeration of failure modes outside the example, and no general derivation mapping blocks to arbitrary I-V curves are given, leaving the scope of the claim untested.

    Authors: The framework begins from the observation that compatibility at each junction is governed by the drive/sense roles that set the local load-line conditions. These roles are abstracted into a fixed set of canonical functional interface blocks whose definitions are independent of any particular device's I-V curve or dynamics. The CCII implementation then supplies a concrete circuit-level realization of those blocks. The Pavlovian demonstrator confirms that the blocks operate correctly when applied to one heterogeneous pair. We agree that the experimental section contains only a single device pair and does not enumerate failure modes or supply an exhaustive I-V mapping; such additions would require further experimental work beyond the present scope. The claim in the abstract is therefore that the role-based construction itself constitutes a systematic methodology, not that every possible device pair has been exhaustively validated. We are willing to revise the abstract wording to make this distinction explicit. revision: partial

Circularity Check

0 steps flagged

No circularity: framework is an independent organizing proposal

full rationale

The paper introduces a junction-centered interface framework that assigns drive/sense roles and organizes connections into canonical functional interface blocks, then realizes them via CCII primitives and validates the approach with one Pavlovian-conditioning circuit. No equations, fitted parameters, or self-referential definitions appear in the provided text that would make any claimed result equivalent to its inputs by construction. The central claim is presented as a conceptual methodology rather than a derivation that loops back on itself, and no load-bearing self-citations or uniqueness theorems from the authors are invoked. The derivation chain is therefore self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, background axioms, or new physical entities are identifiable from the provided text. The framework itself introduces conceptual blocks whose independent evidence cannot be assessed here.

pith-pipeline@v0.9.1-grok · 5689 in / 1118 out tokens · 25777 ms · 2026-06-28T07:02:04.335721+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

31 extracted references

  1. [1]

    Mead C 2002Proceedings of the IEEE781629–1636

  2. [2]

    Indiveri Get al.2011Frontiers in Neuroscience573

  3. [3]

    Adhikari S Pet al.2012IEEE Transactions on Neural Networks and Learning Systems23 1426–1435

  4. [4]

    Rozenberg M J, Schneegans O and Stoliar P 2019Scientific Reports911123

  5. [5]

    Pickett M D, Medeiros-Ribeiro G and Williams R S 2013Nature Materials12114–117

  6. [6]

    Grollier Jet al.2020Nature Electronics3360–370

  7. [7]

    Sowmya Net al.2023International Journal of Computing and Digital Systems141–1

  8. [8]

    Brown T Det al.2024Nature633804–810

  9. [9]

    Markovi´ c D, Mizrahi A, Querlioz D and Grollier J 2020Nature Reviews Physics2499–510

  10. [10]

    Series B- Biological Sciences140177–183

    Hodgkin A L and Huxley A F 1952Proceedings of the Royal Society of London. Series B- Biological Sciences140177–183

  11. [11]

    Mahowald M and Douglas R 1991Nature354515–518

  12. [12]

    Del Valle J, Salev P, Tesler F, Vargas N M, Kalcheim Y, Wang P, Trastoy J, Lee M H, Kassabian G, Ram´ ırez J Get al.2019Nature569388–392

  13. [13]

    Chicca E, Stefanini F, Bartolozzi C and Indiveri G 2014Proceedings of the IEEE1021367–1388

  14. [14]

    Yang J J, Pickett M D, Li X, Ohlberg D A, Stewart D R and Williams R S 2008Nature nanotechnology3429–433

  15. [15]

    Wang Y Fet al.2015Scientific Reports510150

  16. [16]

    Garg Net al.2024 Versatile cmos analog lif neuron for memristor-integrated neuromorphic circuits2024 International Conference on Neuromorphic Systems (ICONS)(IEEE)

    2024

  17. [17]

    Huang Yet al.2024Nature Reviews Electrical Engineering1286–299

  18. [18]

    Koo R Het al.2025Device

  19. [19]

    Li Jet al.2023Nanoscale Horizons81456–1484

  20. [20]

    Sedra A and Smith K 1970IEEE Transactions on Circuit Theory17132–134

  21. [21]

    Lecerf G, Tomas J, Boyn S, Girod S, Mangalore A, Grollier J and Sa¨ ıghi S 2014 Silicon neuron dedicated to memristive spiking neural networks2014 IEEE International Symposium on Cir- cuits and Systems (ISCAS)(Melbourne, VIC, Australia: IEEE) pp 1568–1571

    2014

  22. [22]

    Sedra A Set al.2011Microelectronic Circuitsvol 4 (Oxford: Oxford University Press)

  23. [23]

    Wang Zet al.2020Nature Reviews Materials5173–195

  24. [24]

    Sarkar Tet al.2022Nature Electronics5774–783

  25. [25]

    Kim Yet al.2024Scientific Reports148356

  26. [26]

    Zamarre˜ no-Ramos Cet al.2011Frontiers in Neuroscience526 10 XXX XXXXXXXXXXJournalvv(yyyy) aaaaaa Wellington Avelinoet al

  27. [27]

    Alexander C K and Sadiku M N O 2021Fundamentals of Electric Circuits7th ed (McGraw-Hill)

  28. [28]

    Musil R, Prokop V and Prokop R 2009Annual Journal of Electronics73–76

  29. [29]

    Querlioz Det al.2013IEEE Transactions on Nanotechnology12288–295

  30. [30]

    El Mesoudy Aet al.2022Microelectronic Engineering255111706

  31. [31]

    Hebb D O 1949The Organization of Behavior: A Neuropsychological Theory(Wiley) Acknowledgments The authors acknowledge the support of 3iT (Interdisciplinary Institute for Technological Innova- tion) and the Department of Physics of the Universidade Federal de Minas Gerais (UFMG). The authors also thank the colleagues and collaborators who contributed throu...