Wideband overlapped polyphase channelizer for kinetic inductance detector readout: split-path architecture and experimental characterization

arXiv:2608.24459 · astro-ph.IM, eess.SP · Submitted 2026-08-25 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Astrophysics. The day in astrophysics: observations, models and what the data forces.

Vera: I'm Vera.

Jocelyn: And I'm Jocelyn, here with Subrahmanyan.

Vera: We're looking at the paper: "Wideband overlapped polyphase channelizer for kinetic inductance detector readout".

Paper summary: Vera: This paper introduces a high isolation wideband channelizer for microwave kinetic inductance detector readouts that is designed to mitigate spectral leakage and scalloping loss through a fifty percent overlapping polyphase filter bank <ref:2608.24459#pg1,designed to mitigate spectral leakage and scalloping loss through a 50>. The thesis is that this design, implemented on the Xilinx Zynq UltraScale+ RFSoC using custom High Level Synthesis with time critical output serializer logic in VHDL, processes a four point zero nine six gigasamples per second input stream with a per-per channel measurement.

Jocelyn: They claim they're achieving this by using a split path architecture where the delayed and non delayed polyphase branches compute concurrently to keep the bulk of the channelizer at two hundred fifty six megahertz while only one five hundred twelve megahertz subdomain is confined to the BRAM to FFT section.

Subrahmanyan: The key claim they are making is that this architectural parallelism enables a deep sixteen tap prototype filter which doubles the filter depth compared to eight tap filters common in high speed systems, and that this substantially improves channel isolation.

Vera: They also highlight solving two implementation hazards not documented in prior art: a pipeline stage skew affecting state toggles in HLS generated pipelines, and the selective state preservation policy needed to make the channelizer restartable without losing frame alignment.

Jocelyn: The experimental characterization they performed involves a systematic one hundred tone batch simulation of the channel response, which shows that within zero point one five decibels down to negative twenty eight decibels, both overlapping channels have full gain across their common band.

Subrahmanyan: Furthermore, they report that leakage into every measured channel is less than negative ninety four decibels relative power or lower from the tone is less than one point eight seven five megahertz or more, apart from images attributed to converter interleaving <ref:2608.24459#pg3>.

Vera: And in terms of the noise, the signal to noise ratio per sample lies within one point seven decibels of the limit set by the data sheet noise density of the converter for every measured channel.

Jocelyn: The final result is that an emulated detector pulse is recovered with a decay time of fourteen point six five microseconds, which has a standard uncertainty of zero point zero four microseconds, and this measurement is within zero point seven percent of the expected value.

Subrahmanyan: So, to summarize the core argument for this paper: they've designed an overlapped polyphase filter bank that significantly reduces scalloping loss to below zero point one decibel across the full two megahertz channel.

Conclusion: Vera: So, looking at the paper "Wideband overlapped polyphase channelizer for kinetic inductance detector readout: split-path architecture and experimental characterization" and what we just talked about, this work shows a practical way to build a high isolation channelizer on hardware.

Jocelyn: It really is about taking the idea of parallel processing over raw clock speed to get a deeper filter design than you'd see in aggressively clocked IP based designs for comparable bandwidth.

Subrahmanyan: By computing the delayed and non delayed polyphase branches in parallel and reordering at the frame level rather than the sample level, this design preserves a pipeline initiation interval of one while keeping the bulk of the channelizer at a robust two hundred fifty six megahertz.

Vera: That timing budget that they freed up is then invested in a sixteen tap Blackman prototype filter, achieving a stopband roughly fifteen decibels deeper and correspondingly higher inter channel isolation.

Jocelyn: It demonstrates that for dense arrays, you can trade architectural parallelism against clock speed to obtain a deeper, higher isolation channelizer than aggressively clocked IP based designs of comparable bandwidth.

Subrahmanyan: The fact that the parent readout system has been operated cryogenically on a real MKID array, identifying individual resonators in total darkness really solidifies this as an experimental benchmark for wideband frequency division multiplexed readouts.

Instituto de Astrof༺ sica de Canarias · Universidad de La Laguna

astro-ph.IM, eess.SP

Submitted: 2026-08-25

Updated: 2026-11-17

Comments: 18 pages, 8 figures, 3 tables. Submitted to Measurement (Elsevier). v2: revised and retitled version that adds the experimental characterization on the board (v1, submitted to JATIS, was titled "A high-isolation wideband channelizer for MKID readouts: a custom HLS implementation on RFSoC"). Extends Proc. SPIE 14156, 141561I (doi:10.1117/12.3103240)

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 92/100

The gist: A high-performance channelizer for Microwave Kinetic Inductance Detectors (MKIDs), designed to mitigate spectral leakage and scalloping loss through a 50% overlapping polyphase filter bank (PFB)…

Key concepts

Polyphase Filter Bank (PFB)
A filter bank that uses overlapping filters to improve sampling efficiency and reduce artifacts like scalloping loss. By using a 50% overlap, it allows consecutive data frames to share samples, effectively removing the blind spots where tones might be attenuated in standard critical sampling.
Split-Path Architecture
A design strategy where delayed and non-delayed polyphase branches are computed concurrently. This parallelism allows the bulk of the channelizer to run robustly at a lower clock speed (256 MHz) while confining only a small section to a higher frequency domain (512 MHz), enabling deeper filter prototypes.
Scalloping Loss
A problem in critically sampled filters where tones falling between two FFT bins are severely attenuated, causing signal loss. The overlapped PFB is specifically designed to overcome this by oversampling the channelizer in time, ensuring that even tones falling between bins are properly captured and recovered.
16-tap Blackman Prototype Filter
A deep filter design used in the channelizer that is twice as deep as common 8-tap filters. This increased depth provides a sharper transition band and a much deeper stop-band, which is crucial for suppressing crosstalk between adjacent channels in dense detector arrays.

Terminology

Summary

A high-performance channelizer for Microwave Kinetic Inductance Detectors (MKIDs), designed to mitigate spectral leakage and scalloping loss through a 50% overlapping polyphase filter bank (PFB), is implemented on a Xilinx Zynq UltraScale+ RFSoC using custom High-Level Synthesis (HLS) with time-critical output serializer logic in VHDL, resulting in a high-isolation benchmark for wideband frequency-division multiplexed readouts

System Architecture and Implementation

The design is implemented on a Xilinx Zynq UltraScale+ RFSoC ZCU111 using a custom design, primarily in Vitis HighLevel Synthesis (HLS), with the time-critical output serializer and glue logic in VHDL The split-path architecture computes the delayed and non-delayed polyphase branches concurrently, so that the bulk of the channelizer operates at a robust 256 MHz while a single 512 MHz subdomain is confined to the BRAM-to-FFT section This architectural parallelism enables a deep 16-tap prototype filter that doubles the filter depth of comparable high-speed systems and substantially improves channel isolation The functional readout system in operation today runs its entire processing chain in a single 256 MHz clock domain, crossing from the converter clock through an asynchronous FIFO

Polyphase Filter Bank Design

The channelizer uses a 50% overlapping polyphase filter bank (PFB) to mitigate scalloping loss A critically sampled PFB suffers from scalloping loss, where a tone falling between two FFT bins is attenuated by as much as 3.9 dB The overlapping PFB removes this blind spot by oversampling the channelizer in time by a factor of two—the so-called 2/1 overlap—so that consecutive FFT frames share 50% of their input samples This reordering requires reordering the polyphase outputs before the FFT according to the cyclic permutation R(n) = (n · D) mod M

Performance and Isolation Gains

The 16-tap Blackman prototype filter, which is twice the depth of 8-tap filters common in high-speed RFSoC channelizers, provides a sharper transition band and a substantially deeper stop-band This deeper stop-band means the 16-tap filter reaches the stop-band roughly 15 dB deeper at the first adjacent channel compared to an 8-tap design, which is precisely what is required to suppress crosstalk in a dense MKID array In simulation, the overlapped channelizer recovers the ∼3.9 dB scalloping loss of a critically sampled channelizer to below 0. p. 0

Where the paper says it

The 2/1 overlapped channel response is validated in simulation—recovering the ∼3. 9 dB scalloping loss of a critically sampled channelizer to below 0.

1 dB across the full 2 MHz channel

Design Challenges and Solutions

The design addresses two implementation hazards not documented in prior art: a pipeline-stage skew affecting state toggles in HLS, and the state-preservation requirements of a restartable design The fix for the pipeline-stage skew is to make the dependency explicit by grouping all frame-boundary toggles into a single unified update expression so that HLS sees one dependency chain and schedules the state transition atomically in one stage Furthermore, a selective soft reset policy is implemented where only the flag that gates the first output frame is cleared on stop, preserving state variables like write ptr and ping pong to maintain phase-locked alignment upon restart

Resource Efficiency and Trade-offs

The design prioritizes architectural parallelism over raw clock speed, keeping the bulk of the pipeline at a conservative 256 MHz while confining the only 512 MHz subdomain to a small BRAM-to-FFT section The dominant arithmetic cost is the pair of polyphase banks, requiring 512 DSP slices in total, which is 12% of the available resources on the device This trade-off results in a resource-efficient, high-isolation benchmark for wideband frequency-division multiplexed readouts The frame-level reordering preserves a pipeline initiation interval of one (II = 1) at the cost of a single state bit

Experimental Validation

The channel response was validated in simulation, recovering the ∼3.9 dB scalloping loss to below 0. p. 0

Where the paper says it

A high-isolation wideband channelizer for MKID readouts: a custom HLS implementation on RFSoC Alberto Hernandez Fern ´ andez ´ a,b,*, David D´ıaz Mart´ına , Jose Javier D ´ ´ıaz Garc´ıaa , Roger John Hoylanda , Luis Fernando Rodr´ıguez Ramosa , Diego Portero Rodr´ıgueza,c, Silvestre Rodr´ıguez Perez ´ b a Instituto de Astrof´ısica de Canarias, Electronics Department, V´ıa Lactea s/n, San Crist ´ obal de La Laguna, 38205 ´ Santa Cruz de Tenerife, Spain bUniversidad de La Laguna (ULL), Departamento de Ingenier´ıa Industrial, ESIT, Camino San Francisco de Paula 19, San Cristobal de La Laguna, 38200 Santa Cruz de Tenerife, Spain ´ cUniversidad de La Laguna (ULL), Departamento de Astrof´ısica, 38206 La Laguna, Tenerife, Spain Abstract. We present a high-performance channelizer for Microwave Kinetic Inductance Detectors (MKIDs), designed to mitigate spectral leakage and scalloping loss through a 50% overlapping polyphase filter bank (PFB).

1 dB across the full 2 MHz channel, and characterized systematically with a 100-tone batch frequency sweep The parent readout system has been operated cryogenically with an MKID array in an adiabatic demagnetization refrigerator (ADR), identifying individual resonators in total darkness This setup exercises the complete end-to-end RF–digital–RF loop on which the overlapped channelizer will operate once promoted to the deployed bitstream The result is a resource-efficient, high-isolation benchmark for wideband frequency-division multiplexed readouts

Comparison with Prior Art

The comparison with the UCSB design shows that the frame-level design doubles the arithmetic (two physical branches) but keeps all of the pipeline except the small BRAM-to-FFT subdomain at 256 MHz The frame-level design invests freed timing budget in filter depth, achieving a stopband roughly 15 dB deeper and correspondingly higher inter-channel isolation The comparison demonstrates that for dense arrays, the deep-filter design converts otherwise idle DSP resources into isolation Both approaches are complementary as they both preserve II = 1 and are constrained by the same fclk ·SSR = 2fs identity The work documents the design decisions and implementation hazards that make such a design synthesizable and operable on a single RFSoC

Conclusion

The central result is that architectural parallelism can be traded against clock speed to obtain a deeper, higher-isolation channelizer than aggressively clocked, IP-based designs of comparable bandwidth By computing the delayed and non-delayed polyphase branches in parallel and by reordering at the frame level rather than the sample level, the design preserves a pipeline initiation interval of one while keeping the bulk of the channelizer at a robust 256 MHz The timing budget thus freed is invested in a 16-tap Blackman prototype—twice the depth of standard high-speed implementations—yielding a stopband roughly 15 dB deeper and correspondingly higher inter-channel isolation The parent readout system has been operated cryogenically on a real MKID array, identifying individual resonators in total darkness Taken together, the work shows that a custom HLS approach—prioritizing parallelism over raw clock rate—can outperform standard IP-based methodologies for wideband FDM readout The design documents the design decisions and implementation hazards that make such a design synthesizable and operable on a single RFSoC

Future Directions

The immediate next step is to promote the overlapped channelizer from the cycle-accurate testbench into the deployed bitstream, replacing the classic non-overlapped PFB and adding auxiliary modules its architecture requires Beyond that, work continues on full multi-channel transport over the 10 GbE link to a network-attached storage backend Work also continues on real-time phase processing for singlephoton event detection and mechanical modifications to the cryostat to allow controlled optical illumination of the array through a fibre The modular, reconfigurable structure of the design is intended to let these enhancements be added block-by-block without redesigning the channelizer core The authors thank the Instrumentation Area of the Instituto de Astrof´ısica de Canarias (IAC) and the Universidad de La Laguna (ULL) for their support in the development of the eDAS The simulation data and analysis scripts that support the findings are not publicly available at this time, as the design is part of an ongoing instrument development at the Instituto de Astrof´ısica de Canarias The authors declare that there are no financial interests, commercial affiliations, or other potential conflicts of interest that could have influenced the objectivity of this research or the writing of this paper The code, data, and materials are obtained from the corresponding author upon reasonable request The authors thank an AI language assistant (Claude, Anthropic) for language editing and LaTeX formatting of the manuscript

References

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Architectural Parallelism for Higher Isolation: The system can achieve a deeper, higher-isolation channelizer than aggressively clocked, IP-based designs of comparable bandwidth by computing delayed and non-delayed polyphase branches in parallel at a robust 256 MHz while confining the 512 MHz section to a small BRAM-to-FFT area.

  2. Deep Filter Implementation: The system can achieve a stopband roughly 15 dB deeper than an 8-tap design by using a 16-tap Blackman prototype filter, which results in substantially improving inter-channel isolation.

  3. Overlapping Filter Response Correction: The system can recover signal loss, as the overlapped polyphase filter bank achieves flat coverage across the full channel, recovering the ~3.9 dB scalloping loss of a critically sampled channelizer to below 0.1 dB.

  4. Resource-Efficient High-Performance Design: The system can implement this high-isolation channelizer on a single RFSoC by using custom Vitis HLS and split-path topology, which is resource-efficient compared to systems that require aggressive clocking or vendor IP cores.

  5. Timing Constraint Management: The system maintains a pipeline initiation interval of one while operating the bulk of the channelizer at 256 MHz by employing a frame-level reordering scheme instead of sample-level interleaving, which is necessary to preserve II = 1 at this clock speed.

  6. Robustness Against Implementation Hazards: The system can be made synthesizable and operable by explicitly grouping all frame-boundary toggles into a single unified update expression to resolve the pipeline-stage skew that silently desynchronizes state toggles in HLS.

  7. Operational Reliability: The system can survive stop/restart cycles without losing alignment by employing a selective soft reset, which preserves state variables necessary for phase locking while clearing only the flag that gates the first output frame.

Abstract

Frequency-multiplexed readout of kinetic inductance detector arrays measures the complex transmission of thousands of superconducting resonators. The channelizer that separates them introduces systematic effects: a gain that depends on where each tone falls in the channel grid, leakage between channels and added noise. This article presents a 4096-channel polyphase channelizer with 2/1 overlap for a 4.096 GS/s input, implemented on a first-generation radio-frequency system-on-chip, and a procedure to characterize it on the board with two receive chains, one digitizing the excitation as a reference. A split-path architecture computes the two overlap states in parallel filter branches at 256 MHz and reorders them frame by frame, under three control rules; it keeps only the transform stage at 512 MHz and affords a prototype of 16 taps per polyphase path (65536 coefficients) at the multiplication rate of a published 8-tap bank. Each selected channel is exported as a complex series at 1 MS/s. The measured channel response follows the coefficient-table model within 0.15 dB down to-28 dB, with both overlapping channels at full gain across their common band. Leakage into every measured channel 1.875 MHz or more from the tone is-94 dBc or lower, apart from images attributed to converter interleaving. The signal-to-noise ratio per sample lies within 1.7 dB of the limit set by the data-sheet noise density of the converter, and an emulated detector pulse is recovered with a decay time of 14.65 μ s (standard uncertainty 0.04 μ s), within 0.7% of the expected value.

Sources

Related papers