mi17050592.pdf - 第31页
Micromachines 2026 , 17 , 592 31 of 34 data imputation techniques [ 44 ] must be developed to maintain alignment robustness against real-time pr obe drift and intermittent signal degradation. T ransitioning these methodo…
Micromachines 2026, 17, 592
30 of 34
5.6. Statistical Validity: Data Profiles Versus Averages
Data profile analysis (Figures 9 and 10) confirms the above mentioned findings [
18
].
For instance, while Var-Grad exhibits a misleadingly low average step count (26.1), its
data profile reveals a premature plateau at a 90.3% success rate, indicating fundamental
algorithmic limitations rather than mere parameter mistuning. Conversely, the steep profile
slopes of the Fix-Grad and Local BO algorithms demonstrate rapid, highly consistent
convergence. From a practical manufacturing perspective, these profiles explicitly quantify
worst-case performance and failure rates, directly informing throughput modeling by
accounting for severe manual intervention overhead.
5.7. Comparison to State-of-the-Art
Contextualized against recent automated alignment benchmarks that report multi-
second per-die execution times [
10
,
11
], the evaluated methods demonstrate highly com-
petitive, sub-second performance. Operating entirely in an open-loop configuration, the
Fix-Grad algorithm achieves a rapid 1.44 a.u. alignment time alongside 95.8% reliability
and 99.4% accuracy. Furthermore, experimental validation of the Local BO method yields
a 100% success rate and a 3–4
×
speedup (99.87 a.u.) over exhaustive reference scans.
Crucially, these capabilities are realized without closed-loop position feedback, relying
instead on computationally efficient algorithmic optimization, where even GP fitting runs
in real-time on embedded systems and robust feedforward hysteresis compensation [
14
,
33
].
This performance underscores a critical innovation: deploying sophisticated algo-
rithms to effectively compensate for simplified hardware mechanics. By eliminating the
necessity for bulky external positioning equipment, this approach enables the cost-effective
integration of optical alignment capabilities directly into the probe card. For high-volume
PIC manufacturing [
3
,
20
], this integrated architecture provides a profound advantage. It
inherently facilitates simultaneous multi-channel parallel testing, delivering multiplica-
tive throughput enhancements that scale significantly beyond standard per-die speed
improvements [5,12].
5.8. Limitations and Future Work
Several avenues warrant further investigation to advance automated wafer-level
photonic testing. Future research will explore data-driven control architectures to further
enhance system throughput. Specifically, replacing analytical hysteresis models with
Long Short-Term Memory (LSTM) networks is expected to yield higher accuracy by better
capturing deep, history-dependent piezoelectric behaviors. Concurrently, Reinforcement
Learning (RL) agents offer the potential to adaptively counteract complex mechanical wear
and drift over the lifespan of the probe card, outperforming static algorithms for dynamic
fine-stage optimization.
Additionally, expanding these methods to multi-fiber array alignments will require
strategies to balance peak loopback transmission with channel uniformity and crosstalk
limits across the array. Because the FAU moves as a rigid block, optimizing a single
channel can induce misalignment on peripheral channels due to planar tilt. Utilizing
the differential Y1/Y2 piezoelectric actuators for precise rotational tilt compensation will
be essential in navigating these multi-channel compromises. Finally, standardizing chip
layouts, such as fixing input/output port offsets and reference waveguide lengths across
different PIC designs, would significantly narrow the spatial bounding required for search
algorithms, accelerating the cross-scan initialization phase for both Bayesian and Gradient
methodologies. To mitigate the initialization challenges observed in Bayesian methods,
historical wafer data can also be leveraged to generate predictive spatial probability maps,
drastically reducing initial search areas. Furthermore, online learning and advanced sensor
https://doi.org/10.3390/mi17050592
Micromachines 2026, 17, 592
31 of 34
data imputation techniques [
44
] must be developed to maintain alignment robustness
against real-time probe drift and intermittent signal degradation.
Transitioning these methodologies to high-volume manufacturing demands broader
experimental validation across diverse foundries, device architectures, and multi-fiber
arrays. In such high-dimensional optimization landscapes, the algorithmic hierarchy may
shift favorably toward Bayesian surrogate models [
5
,
11
]. Crucially, these sub-second
alignment strategies must be seamlessly integrated into double-sided test cells to meet
the aggressive throughput demands of CPO production flows [
1
,
3
]. Finally, coupling
operational alignment metrics with digital twin frameworks and transformer-based sensor
placement [
45
,
46
] will enable robust structural health monitoring, facilitating predictive
maintenance and extending probe card lifespans in demanding industrial environments.
6. Conclusions
This paper presented a comprehensive evaluation of alignment optimization tech-
niques for wafer-level photonic integrated circuit testing, utilizing the Technoprobe Eclipse
Dynamic probe card system. Through robust experimental validation and extensive simu-
lations spanning diverse operating regimes, incorporating varying actuation speeds and
hysteresis conditions, this study systematically benchmarked seven alignment algorithms.
The results unequivocally establish the Fix-Grad algorithm as the optimal solution for
high-volume manufacturing. Across all evaluated conditions, this approach demonstrated
exceptional robustness, consistently delivering a 95.8% reliability rate and 99.4% align-
ment accuracy.
Transitioning to high-speed actuator drive electronics is essential for production via-
bility, providing a 14.6–21.9
×
throughput improvement over low-speed operation (Table 3).
In the ideal high-speed, hysteresis-compensated regime, the Fix-Grad algorithm achieved a
rapid alignment time of 1.44 a.u. per die. While uncompensated hysteresis approximately
doubles this execution time, algorithmic mitigation strategies successfully preserve final
measurement quality without requiring complex closed-loop positioning feedback. Alter-
native methods presented specific trade-offs—the Var-Grad technique offered the fastest
alignment (1.18 a.u.) but suffered a higher failure rate of 9.7%, rendering it suitable only for
non-critical screening. Furthermore, despite high sample efficiency, BO approaches exhib-
ited excessive computational overhead and heightened sensitivity to hysteresis, limiting
their viability for strict production environments.
From a production perspective, data profile analysis confirms that the high reliability
of the Fix-Grad method limits manual operator intervention to just 4.2% of tested dies. For a
standard 300 mm wafer containing 200 dies, this integrated architecture facilitates complete
optical characterization in approximately 288 to 664 a.u., depending on the extent of active
hysteresis compensation. By achieving highly reliable optical alignments entirely through
open-loop control, this system matches or exceeds the performance metrics reported in
recent literature [
10
,
11
]. Crucially, embedding this capability directly into the probe card
eliminates the need for bulky external positioning stages, significantly reducing overall
equipment footprint and enabling the parallel multi-channel testing required for emerging
CPO architectures [1,3].
Future research will focus on advancing data-driven control strategies to further
enhance system throughput and absolute accuracy. Key initiatives include implementing
Long Short-Term Memory networks to replace analytical models for history-dependent
hysteresis compensation, alongside deploying Reinforcement Learning agents to adaptively
counteract complex mechanical irregularities. Additionally, leveraging historical wafer
data to generate predictive spatial heatmaps will heavily optimize search initialization
parameters. Expanding these methodologies to accommodate simultaneous multi-fiber
https://doi.org/10.3390/mi17050592
Micromachines 2026, 17, 592
32 of 34
array alignments and fully integrating the resulting automated flows into advanced test
cells will be essential to meet the escalating manufacturing demands of next-generation
silicon photonics.
Author Contributions: Conceptualization, R.V., M.M. and S.M.; methodology, M.B., A.G. and R.V.;
software (algorithm writing and simulation), M.B.; investigation (research and experiment data
recording), M.B. and A.G.; resources (hardware design and build), A.G. and R.V.; writing—original
draft preparation, M.B.; writing—review and editing, M.B., A.G., M.M. and S.M.; supervision, M.M.
and S.M.; project administration, A.G. and R.V. All authors have read and agreed to the published
version of the manuscript.
Funding: Funded by the European Union (Grant Agreement No. 101072491). Views and opinions
expressed are however those of the author(s) only and do not necessarily reflect those of the European
Union or the European Research Executive Agency. Neither the European Union nor the granting
authority can be held responsible for them.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author.
Conflicts of Interest: Authors Mehdi Bejani, Alessia Galli, Riccardo Vettori, and Marco Mauri were
employed by the company Technoprobe S.p.A. The authors declare that this research was conducted
as part of a collaboration between Politecnico di Milano and Technoprobe S.p.A. within the EU-
funded project (Grant Agreement No. 101072491). The remaining authors declare that the research
was conducted in the absence of any commercial or financial relationships that could be construed as
a potential conflict of interest.
References
1.
Thomson, D.; Zilkie, A.; Bowers, J.E.; Komljenovic, T.; Reed, G.T.; Vivien, L.; Marris-Morini, D.; Cassan, E.; Virot, L.; Fédéli, J.-M.;
et al. Roadmap on silicon photonics. J. Opt. 2016, 18, 073003. [CrossRef]
2.
Maeda, Y.; Miura, T.; Matsuo, S.; Fukuda, H. Accurate Fiber Alignment using Silicon Photodiode on Grating Coupler for
Wafer-Level Testing. In Proceedings of the 24th OptoElectronics and Communications Conference (OECC) and 2019 International
Conference on Photonics in Switching and Computing (PSC), Fukuoka, Japan, 7–11 July 2019; pp. 1–3. [CrossRef]
3. Couny, F. Seamlessly move PIC testing from lab to fab. PIC Mag. 2022, 2, 34–36.
4.
De Coster, J.; De Heyn, P.; Pantouvaki, M.; Snyder, B.; Chen, H.; Marinissen, E.J.; Absil, P.; Van Campenhout, J.; Bolt, B. Test-station
for flexible semi-automatic wafer-level silicon photonics testing. In Proceedings of the 21th IEEE European Test Symposium
(ETS), Amsterdam, The Netherlands, 23–27 May 2016; pp. 1–6. [CrossRef]
5.
Petrini, M.; Baldi, R.; Seyfried, M.; Morichetti, F.; Melloni, A. Probe-card based automatic testing of photonic integrated filters. In
Proceedings: Optica Advanced Photonics Congress 2022; Technical Digest Series (Optica Publishing Group, 2022); Paper IM4C.2;
Optica Publishing Group: Washington, DC, USA, 2022. [CrossRef]
6.
Gnausch, T.; Grundmann, A.; Juhasz, T.; Kaden, T.; Büttner, R.; von Freyhold, T. Novel Opto-Electronical Probe Card for
Wafer-Level PIC Testing. In Proceedings of the 2019 Optical Fiber Communications Conference and Exhibition (OFC), San Diego,
CA, USA, 3–7 March 2019; pp. 1–3.
7.
Vettori, R.; Galli, A. A comprehensive study on improving probe card transmission lines for effective high-frequency wafer-level
testing. In Proceedings of the 2024 IEEE European Test Symposium (ETS), The Hague, The Netherlands, 20–24 May 2024.
8.
Vettori, R.; Galli, A. Eclipse Dynamic Probe Card: A Novel Approach for Wafer-Level Photonic Testing with Automated Fiber
Array Unit Alignment. In Proceedings of the 2025 IEEE International Test Conference (ITC), San Diego, CA, USA, 20–26
September 2025; pp. 446–449. [CrossRef]
9.
Bejani, M.; Appello, D.; Mauri, M.; Todaro, S.; Mariani, S. AI-Assisted Framework for Real-Time Monitoring and Management of
Probe Cards in Electrical Wafer Sort Applications. In Proceedings of the 2025 IEEE European Test Symposium (ETS), Tallinn,
Estonia, 26–30 May 2025; pp. 1–4. [CrossRef]
10.
Yuan, Q.; Peczek, A.; Frankel, J.; Rishavy, D.; Mai, C.; Christenson, E.; Pratap, D.; Zimmermann, L. Fully Automated Wafer-Level
Edge Coupling Measurement System for Silicon Photonics Integrated Circuits. IEEE Trans. Semicond. Manuf. 2025, 38, 168–177.
[CrossRef]
https://doi.org/10.3390/mi17050592