PaperlayerView exact revision on arXiv

Source-linked AI summary

Frequency-Modulated Piezoelectric Haptic Display

Boyuan Liang, Lingfeng Sun, Masayoshi Tomizuka

arXiv:2609.30626v1cs.RO

TL;DR

Independently driving amplitude-controlled pixels can burden scaling toward large-area wearable haptic displays. The paper uses vibration frequency to encode intensity and develops a shared-source architecture for piezoelectric pixels. Volunteers distinguished the tested patterns and intensity levels, supporting feasibility within the investigated operating range.

  • Problem

    For wearable haptic displays, independently controllable pixels must be realized while minimizing high-power driving hardware that scales with pixel count.

    Source: §I · p. 1

  • Method

    The piezoelectric SSFM architecture shares a high-voltage amplifier across pixels while compact switches independently control their vibration frequencies.

    Source: §I · p. 1, §VII · p. 8

  • Results

    Participants reliably distinguished the tested patterns and intensity levels, providing evidence for FM-based distributed haptic rendering.

    Source: §I · p. 2

  • Takeaways & Limitations

    The results demonstrate feasibility of FM-based distributed haptic rendering and suggest a potential pathway toward more compact driving architectures for future wearable displays.

    Source: §I · p. 2

    The experiments were proof-of-concept tests of coarse stimulus levels; finer frequency, spatial, and temporal increments are needed to determine rendering resolution and perceptual bandwidth.

    Source: §V · p. 7

Abstract

from arXiv · show

We present a frequency-modulated (FM) haptic display based on piezoelectric vibrating actuators. Existing haptic displays commonly encode haptic intensity through the deformation amplitude of individual haptic pixels. Although amplitude-modulated (AM) approaches have enabled compact haptic pixels, independently controlling the deformation amplitude of a large number of pixels can require increasingly complex and bulky driving systems, posing challenges for scaling toward high-density, large-area wearable displays. To address this scaling challenge, we investigate an FM design principle in which haptic intensity is encoded through vibration frequency. We further develop a \textit{Shared-Source Frequency Modulation} (SSFM) structure in which multiple haptic pixels are powered by a common power amplifier while their vibration spectrum are controlled individually, reducing the need for independent high-power amplification at each pixel. A proof-of-concept piezoelectric haptic display was built and evaluated on rendering spatial and temporal haptic patterns through volunteer tests. The results show that participants reliably distinguished spatial and temporal patterns encoded using FM principles within the investigated operating range. These findings demonstrate the feasibility of FM-based distributed haptic rendering and suggest a potential pathway toward more compact driving architectures for future high-density, large-area wearable haptic displays.

I. INTRODUCTION

The paper addresses the scaling burden of independently driven haptic pixels by proposing frequency-modulated intensity control and a shared-source architecture. In proof-of-concept tests, participants distinguished the tested patterns and intensity levels.

I. INTRODUCTION

  • As independently controlled pixel arrays grow, amplitude-based driving can require more amplification, routing, or energy-delivery resources near the actuation surface.This scaling burden matters for wearables, where near-body components must remain compact, lightweight, and mechanically unobtrusive.Source: §I · p. 1
  • Frequency-dependent vibrotactile sensitivity motivates using vibration frequency, rather than mechanical amplitude alone, to encode distinguishable tactile intensity levels.The cited sensitivity peak is around 200 Hz, where displacement thresholds can reach sub-micrometer levels.Source: §I · p. 1
  • SSFM lets multiple piezoelectric pixels share a high-voltage amplifier while compact analog switches independently control each pixel’s vibration frequency.The architecture places the high-power amplification stage off-skin, leaving switching circuitry and interconnections near the pixels.Source: §I · pp. 1–2
  • Participants reliably distinguished the tested spatial-temporal patterns and FM-encoded intensity levels, providing evidence for FM-based distributed haptic rendering.The prototype and human-participant experiments evaluated recognition of patterns and discrimination of intensity levels.Source: §I · pp. 1–2

A. High-Resolution Haptic Displays

High-resolution haptic displays need both fine spatial detail and rapid temporal response, while scaling also depends on the driving resources required by individual pixels.

A. High-Resolution Haptic Displays

  • Temporal continuity may require refresh intervals below approximately 50 ms, while resolving fine tactile patterns can require spatial resolutions of a few millimeters.These are requirements suggested by perceptual studies, not measurements of the present system.Source: A. High-Resolution Haptic Displays · p. 2
  • Mechanical, pneumatic, thermal, electromagnetic, electrostatic, piezoelectric, and combustion-based approaches have advanced pixel size, response speed, and force output.Millimeter-scale pixels and rapid actuation have each been demonstrated across these actuation technologies.Source: A. High-Resolution Haptic Displays · p. 2
  • Amplitude modulation gives direct control of each actuator’s mechanical output, but large independently controlled arrays may require replicated amplification or other routing resources.Miniaturizing actuators alone does not necessarily reduce the hardware distributed across a large-area interface.Source: A. High-Resolution Haptic Displays · p. 2

B. Shared Actuation Resources for Large-Scale Displays

Shared actuation resources offer an alternative to replicating high-power hardware at every haptic pixel. The paper motivates an electrically driven approach compatible with wearable electronics.

B. Shared Actuation Resources for Large-Scale Displays

  • A shared laser source has been steered among photoabsorbing elements to produce localized thermal actuation in arrays containing thousands of tactile elements.The approach reduces the need for dedicated high-power actuation hardware at individual pixels.Source: B. Shared Actuation Resources for… · p. 2
  • An electrohydraulic approach uses shared electrical resources and electrostatic forces to squeeze liquid and create localized surface deformation.Source: B. Shared Actuation Resources for… · p. 2
  • Optical and fluidic approaches suit form factors that integrate their energy-transmission mechanisms, while wearable electronics motivate exploring an electrically driven alternative.Thin flexible conductors and compact integrated switches are described as compatible with wearable electronic fabrication.Source: B. Shared Actuation Resources for… · p. 2

III. SHARED-SOURCE FREQUENCY MODULATION

The section outlines the SSFM structure and its operation, then describes dynamic pattern composition and waveform analyses.

III. SHARED-SOURCE FREQUENCY MODULATION

  • The section describes the SSFM structure and how it realizes independently controlled vibration frequencies at individual pixels.Source: §III · p. 2
  • It explains how dynamic haptic patterns are composed.Source: §III · p. 2
  • It presents spectral analysis of the output waveform and a first-order estimate of its thermal behavior.Source: §III · p. 2

A. Single-Pixel Frequency Modulation

Each pixel converts a shared base waveform into a selected nominal frequency by gating whole base periods. The resulting waveform retains higher-frequency spectral content, while the analysis also bounds switching and conduction losses.

A. Single-Pixel Frequency Modulation

  • A switch gates the shared base waveform for n −1 periods and passes one period to produce nominal frequency f0/n; n →∞ leaves the pixel static.Switching occurs only when Vb(t) < δ, limiting voltage discontinuities.Source: A. Single-Pixel Frequency Modulation · pp. 2–3
  • The nominal-frequency component remains dominant, while residual content near the base frequency persists and higher-order harmonics decay as m−3.The fundamental amplitude approaches V0/n for large n; retained high-frequency content may help keep low-nominal-frequency vibrations perceptible.Source: Figure 1 · p. 3, A. Single-Pixel Frequency Modulation · p. 4
  • The first-order thermal analysis accounts for switching-event dissipation and closed-channel ohmic loss, with switching frequency highest at n = 2.The switching contribution scales with δ², and the sinusoidal base waveform avoids full-supply-voltage switching; a numerical example reports over an order-of-magnitude difference from constant supply.Source: A. Single-Pixel Frequency Modulation · p. 4

B. Dynamic Pattern Generation

Dynamic patterns are generated by assigning each pixel an intensity sequence across fixed refresh intervals. Different per-pixel sequences can form moving tactile textures, including sweeping lines.

B. Dynamic Pattern Generation

  • Each pixel’s dynamic pattern is specified by an intensity sequence over fixed refresh intervals.The sequence spans P intervals of duration Δt.Source: B. Dynamic Pattern Generation · p. 4
  • Different intensity sequences across pixels render dynamic textures such as sweeping lines, moving points, and translating shapes.A left-to-right sweep is composed from the intensity sequences of five pixels.Source: B. Dynamic Pattern Generation · p. 4

IV. IMPLEMENTATION DETAILS

The proof-of-concept uses a 16-by-6 piezoelectric array with off-skin waveform generation and on-skin analog switching. Its implementation combines synchronized switch updates with a shared amplifier and estimates substantially lower dissipation than a constant-voltage supply.

IV. IMPLEMENTATION DETAILS

  • The prototype contains 16 rows and 6 columns of PowerHap 1204 piezoelectric actuators, mounted on resin holders with aluminum supports to limit bending under hand pressure.Each actuator measures 12 × 4 mm and provides approximately 20 µm deformation under a 50 V drive.Source: §IV · p. 4, Figure 2 · p. 5
  • Seven assigned intensity levels map to nominal frequencies from 200 Hz to 5 Hz, plus a static state, using a 200 Hz base waveform.The listed levels are n = 1, 2, 4, 10, 20, 40, ∞, corresponding to 200, 100, 50, 20, 10, 5 Hz, and static.Source: §IV · p. 4
  • The host sends each 96-actuator frame to the ESP32 every 50 ms, which updates analog switches over SPI when the waveform crosses the trigger threshold.The ESP32 must transfer all 96 switch states within the available threshold window.Source: §IV · pp. 4–5
  • A 4 MHz SPI clock transfers the 96 switch states within 30 µs, while daisy-chained 16-channel switches avoid adding off-skin-to-on-skin wires as pixels increase.The prototype uses HV2601 switches; the figure shows serial PC-to-ESP32 communication and SPI-controlled analog switch channels.Source: §IV · p. 5, Figure 2 · p. 5
  • Approximately 360 mW total dissipation is predicted for the switched driver, versus approximately 12 W for a constant 50 V supply.The paper notes that 12 W could risk actuator damage and unsafe skin-contact temperature; the sinusoidal base waveform is preferred.Source: §IV · p. 5

V. EXPERIMENTS

Six volunteer experiments tested whether the shared-source FM prototype could convey spatial, temporal, and intensity information. Results support feasibility at the tested levels, while leaving finer perceptual limits unresolved.

V. EXPERIMENTS

  • At least 90% of localization responses landed on the stimulated pixel or an immediate neighbor, although exact identification fell to 25%–55% at four proximal sites.Exact identification reached 100% for pixel 1 and 85% for pixel 2; the authors note that a tactile-acuity gradient is plausible but was not isolated.Source: §V · pp. 5–6, Figure 3 · p. 6
  • All five tested frequency pairs were judged correctly in every trial, showing that the prototype’s tested intensity levels were separable, not establishing a discrimination threshold.The frequency pairs differed by a factor of two or more.Source: §V · p. 6, Figure 3 · p. 6
  • Participants identified all six stimulated finger pairs with at least 90% accuracy, and three pairs reached 100%.Errors confused one finger with an immediately adjacent finger.Source: §V · pp. 6–7, Figure 3 · p. 6
  • Every trial correctly identified the direction of the cardinal and diagonal linear-motion patterns.The moving front used a two-level frequency-coded intensity gradient.Source: Figure 3 · p. 6, §V · p. 7
  • Rotational speed responses were 91%, 85%, and 94% correct at 16, 32, and 64 mm/s, respectively, with errors only between adjacent levels.The tested levels differed by successive factors of two, so the experiment did not establish the finest renderable speed difference.Source: Figure 3 · p. 6, §V · p. 7
  • The experiments support SSFM feasibility: participants distinguished frequency-coded intensity, finger combinations, and dynamic motion while sharing one amplification source.Performance suggests sufficient spatial and temporal independence for the tested patterns, but the stimulus levels were coarse.Source: §V · p. 7
  • The proof-of-concept tests used coarse stimulus levels and were not designed to characterize perceptual limits; finer frequency, spatial, and temporal increments remain to be tested.Eleven untrained participants took part across the six experiments.Source: §V · pp. 5, 7

VI. DISCUSSION: TOWARDS WEARABILITY

The discussion outlines a possible wearable implementation using flexible circuitry, compact switches, and smaller piezoelectric actuators. The proposed engineering path remains a projection, not a demonstrated device.

VI. DISCUSSION: TOWARDS WEARABILITY

  • Each HV2601 switch provides 16 channels in a 7×7 mm footprint, while commercially available multilayer piezoelectric actuators can be as small as 2×2 mm.The smaller actuators provide about 3 µm of travel, compared with 20 µm for the prototype’s PowerHap 1204 actuator.Source: §VI · p. 7
  • A flexible PCB could carry HV2601 switches and directly mounted millimeter-scale actuators, with a fixed set of conductors linking the on-skin layer to an off-skin amplifier.At the projected dimensions, switching circuitry would occupy roughly 3 mm2 per pixel.Source: §VI · pp. 7–8
  • These are design projections, not demonstrated results; a practical wearable device still needs to address skin coupling and the perceptual effects of reduced displacement.The authors leave implementation to future work.Source: §VI · p. 8

VII. CONCLUSION

The conclusion presents SSFM as a feasible way to share high-voltage amplification while independently setting pixel frequencies. Volunteer tests support FM rendering in the tested conditions, but the stimulus levels were coarse.

VII. CONCLUSION

  • SSFM uses one high-voltage amplifier for a common base waveform and compact analog switches to set each pixel’s nominal frequency independently.The architecture places the amplifier off-skin and reduces the on-skin driver to switching circuitry and interconnections.Source: §VII · p. 8
  • On a 96-pixel prototype, volunteers distinguished the tested spatial, temporal, and intensity patterns, supporting FM rendering with a shared amplifier.The conclusion reports at least 90% correct localization to the stimulated pixel or an immediate neighbor, error-free nominal frequency pairs and linear sweep directions, at least 90% accuracy for finger pairs, and 85%–94% accuracy for rotational speed.Source: §VII · p. 8
  • The stimulus levels were coarse; the authors identify measuring vibration-amplitude discrimination thresholds and realizing the wearable implementation as next steps.The conclusion also notes that the smaller-actuator wearable implementation remains future work.Source: §VII · p. 8
Loading 2609.30626v1…