Bunch Oscillation Recorder (BOR) Overview

• Bunch Oscillation Recorder (BOR) is a diagnostic system that records beam motion turn-by-turn and bunch-by-bunch, providing detailed insights into beam dynamics.
• BOR systems facilitate modal analysis, transient studies, and feedback optimization by measuring beam positions, phases, and charge variations with high resolution.
• Implementations range from RFSoC-based modules to oscilloscope and BPM setups, each offering real-time data for instability characterization and machine protection.

Searching arXiv for recent and foundational BOR-related papers to ground the article. Searching for “Bunch Oscillation Recorder RFSoC SuperKEKB” and related diagnostics. Bunch Oscillation Recorder (BOR) denotes a bunch-by-bunch, turn-by-turn diagnostic that records the motion of individual bunches versus turn and enables modal analysis, transient studies, and feedback optimization. In modern circular accelerators it is often implemented as an operational mode of a digital multi-bunch feedback system, but functionally equivalent BORs have also been realized with single-BPM digitizers, oscilloscope-based BPM acquisition, wall-current-monitor systems, and RFSoC-based beam-position monitors (Lonza et al., 2016). Across these implementations, the defining output is a record such as $x_b(n)$, $y_b(n)$, longitudinal phase $\phi_b(n)$, or bunch charge $Q_b(n)$, indexed by bunch $b$ and turn $n$, from which tunes, growth and damping rates, coupled-bunch modes, beam-beam coherent modes, and fast loss transients can be reconstructed (Nomaru et al., 2024).

1. Physical basis and measured observables

In coupled-bunch dynamics, the bunch centroid is modeled as a damped oscillator with instability growth and feedback damping terms. The cited formulation writes

$x(t) + 2(D-G)\,\dot{x}(t) + \Omega^2 x(t) = 0,$

with approximate solution

$x(t) = k\, e^{-(D-G)t}\, \sin(\Omega t + \phi),$

and, with feedback,

$$x(t) + 2(D - G + D_{fb})\,\dot{x}(t) + \Omega^2 x(t) = 0.$$

For stability, $D-G+D_{fb}>0$ is required (Lonza et al., 2016). A BOR therefore measures the variables needed to reconstruct $y_b(n)$0, $y_b(n)$1, $y_b(n)$2, and the oscillation frequency $y_b(n)$3.

For $y_b(n)$4 equally spaced bunches, the coupled-bunch modes are characterized by the bunch-to-bunch phase advance

$y_b(n)$5

with spectral components at

$y_b(n)$6

In practice, each mode appears as a pair of sidebands around every RF harmonic (Lonza et al., 2016). This is why a BOR must record bunch-resolved motion over enough turns to support both FFTs along the turn index and modal decomposition along the bunch index.

The measured quantity is not unique. In transverse systems it is the bunch position, often reconstructed from BPM difference signals. In longitudinal systems it can be beam synchronous phase, bunch length, or coupled-bunch instability amplitude extracted from wall-current-monitor waveforms (Steimel et al., 2012). In SuperKEKB, the BOR records horizontal position $y_b(n)$7, vertical position $y_b(n)$8, and bunch charge $y_b(n)$9 for all filled buckets preceding a beam abort (Nomaru et al., 18 Jul 2025). In the Hefei Light Source oscilloscope-based system, the recorded coordinates are $\phi_b(n)$0, $\phi_b(n)$1, and $\phi_b(n)$2 for 45 bunches over 2266 turns, with $\phi_b(n)$3 representing longitudinal phase or arrival time (Lu et al., 2021).

A recurring point in the literature is that BOR is best understood as a diagnostic function rather than a single hardware archetype. This suggests that the defining criterion is not the front-end technology but the availability of bunch-resolved, turn-resolved records from which oscillatory beam dynamics can be inferred.

2. Signal formation, pickups, and synchronization

BOR implementations typically begin with BPM or wall-current-monitor pickups. For a four-button BPM, the standard wideband combinations are

$\phi_b(n)$4

where $\phi_b(n)$5 are button signals (Lonza et al., 2016). In SuperKEKB, a vertical BOR uses two opposite buttons $\phi_b(n)$6 and $\phi_b(n)$7, with the vertical position reconstructed by the difference-over-sum formula

$\phi_b(n)$8

The analog chain there uses trombone delay equalization, 600 MHz low-pass filters, a 180° hybrid, and amplification of the $\phi_b(n)$9 path before RFSoC digitization (Nomaru et al., 2024).

The front-end may be heterodyne or direct-sampling. In digital feedback systems, the BPM signal is often down-converted to baseband by heterodyne demodulation, producing a pulse train with one pulse per bunch per turn; pulse amplitude is proportional to position or phase error times bunch charge (Lonza et al., 2016). In other systems the electrodes are digitized directly. The Hefei Light Source setup connects four BPM electrodes directly to a Siglent SDS6204H12.Pro oscilloscope with 12 bit resolution, 10 GSa/s sampling rate, and 2 GHz bandwidth, then reconstructs $Q_b(n)$0, $Q_b(n)$1, $Q_b(n)$2, and bunch charge in software (Lu et al., 2021). The Fermilab SBD system uses resistive wall current monitors with bandwidth $Q_b(n)$3 GHz and a 2 GHz-bandwidth, 5 GS/s oscilloscope to measure longitudinal bunch behavior (Thurman-Keup et al., 2011).

Precise timing is a universal BOR requirement. In the SSRF transverse feedback processor, the ADC runs at 500 MS/s, synchronous with the bunch rate, and a clocking scheme based on LMK04803 PLLs and SY89295 delay lines provides 10 ps step delay adjustment; the required timing precision is better than 20 ps (Zhan et al., 2019). In Tevatron coherent-mode detectors, the digitizer samples at 8 GS/s and the revolution-marker jitter is corrected offline using per-segment time stamps and spline interpolation, because any mismatch between BPM plate equalization and trigger timing converts longitudinal timing error into apparent transverse motion (Stancari et al., 2013). In the KARA synchronized diagnostics, timing alignment across KAPTURE, EOSD/KALYPSO, and a fast-gated camera is established with a hardware synchronization scheme and verified with an RF phase step that launches a synchrotron oscillation observed by all detectors at the same turn (Kehrer et al., 2017).

3. Architectures and representative implementations

Within digital multi-bunch feedback systems, the basic BOR-capable architecture consists of detector, ADC, de-multiplexer, per-bunch filters, re-multiplexer, DAC, power stage, and kicker. The same chain that stabilizes the beam can record bunch-by-bunch motion by tapping ADC data streams into memory and exporting them for offline analysis (Lonza et al., 2016). Section 6 of that work explicitly identifies integrated diagnostic tools such as ADC data recording, on-the-fly filter parameter modification, and injection of external digital samples, which are BOR capabilities in everything but name.

Representative implementations illustrate how broad the BOR concept has become.

System	Recorded quantity	Notable capability
Digital feedback diagnostic mode (Lonza et al., 2016)	$Q_b(n)$4, $Q_b(n)$5	ADC recording, transient generation, modal analysis
SSRF transverse processor (Zhan et al., 2019)	bunch-by-bunch transverse position	500 MS/s ADC, 256 Mega-samples DDR3, triggered capture
HLS oscilloscope system (Lu et al., 2021)	$Q_b(n)$6, $Q_b(n)$7, $Q_b(n)$8 of 45 bunches	2266 turns per shot, about 5 um in $Q_b(n)$9, about 0.5 ps in $b$0
Tevatron single-BPM system (Stancari et al., 2013)	transverse coherent oscillations	8 GS/s, response time of a few seconds, $b$1 tune resolution, 60 nm sensitivity
SuperKEKB RFSoC BOR (Nomaru et al., 2024)	bunch-by-bunch transverse position	100-turn pre-abort buffer, position resolution of 0.03 mm
SuperKEKB SBL BOR analysis (Nomaru et al., 18 Jul 2025)	$b$2, $b$3, $b$4 for all bunches	abort-triggered 100-turn ring buffer, sector-resolved loss analysis

The RFSoC BOR at SuperKEKB uses an AMD/Xilinx Zynq UltraScale+ RFSoC ZCU111, ADC sampling at $b$5 GS/s $b$6 MHz, live buffers of about $b$7, and fault buffers of about $b$8 with decimation by 8; the design goal is portable multi-location deployment rather than single-station permanence (Nomaru et al., 2024). The older VME-BOR at SuperKEKB had approximately $b$9 resolution and showed that SBLs are accompanied by substantial transverse oscillations, but one device per ring could not localize the origin of the disturbance (Nomaru et al., 2024).

Oscilloscope-based and single-BPM systems show the same functional structure in different hardware. The Hefei system is explicitly described as functionally a BOR implemented with a high-speed digital oscilloscope rather than dedicated bunch-by-bunch electronics (Lu et al., 2021). The Tevatron system is described as effectively a BOR implemented with a single high-$n$0 BPM, a 10-bit 8 GS/s digitizer, and offline Fourier analysis (Stancari et al., 2013).

4. Analysis methods and data products

BOR analysis usually begins with turn-by-turn series $n$1, $n$2, $n$3, or $n$4. The standard workflow is: per-bunch FFT over turns, identification of peaks near betatron or synchrotron tunes, and, when many bunches are filled, discrete Fourier transform over bunch index to obtain coupled-bunch mode amplitudes $n$5 (Lonza et al., 2016). Growth or damping rates are then extracted by fitting modal amplitudes or bunch amplitudes to exponentials in time.

The canonical BOR data products include bunch-by-bunch spectra, bunch-by-bunch tune measurements, 3D transient plots, and mode evolution maps. The digital-feedback literature shows explicit examples of a bunch-by-bunch spectrum, local excitation with tune extraction from the FFT of one bunch, a 3D transient plot of amplitude versus bunch index and turn, and multi-bunch spectrum versus time during feedback-off and feedback-on transients (Lonza et al., 2016). The Hefei system similarly produces 2D and 3D plots of $n$6, $n$7, and $n$8 versus bunch and turn, and histograms of bunch-by-bunch tune distributions (Lu et al., 2021).

Several BOR implementations rely on more specialized spectral processing. In the Tevatron coherent-mode detector, each slice within a bunch is Fourier transformed after application of a Slepian window of rank 2, then spectra from signal slices are divided by spectra from background slices to suppress narrowband lines caused by the time-interleaved ADCs. With up to 62,500 turns, the resulting frequency resolution is $n$9 in fractional tune, and oscillation amplitudes of 60 nm are detectable (Stancari et al., 2013). The same hardware and method were used to measure coherent beam-beam mode spectra and compare them with beam-beam models (Stancari et al., 2011).

Longitudinal BORs add model-based reconstruction. In the Fermilab Booster, a segmented-memory oscilloscope and a mountain range trigger capture 4 $x(t) + 2(D-G)\,\dot{x}(t) + \Omega^2 x(t) = 0,$0 sweeps at 5 MS/s through the 33 ms acceleration ramp; FFT-based processing with sinc deconvolution yields beam synchronous phase, bunch length, and coupled bunch instability amplitudes (Steimel et al., 2012). In the Tevatron and Main Injector SBD system, wall-current-monitor waveforms are dispersion-corrected by an FIR filter derived from the inverse of the cable impulse-response matrix, then used to compute bunch intensity, mean phase, rms width, skew, kurtosis, and, through longitudinal Hamiltonian reconstruction, longitudinal emittance and momentum spread (Thurman-Keup et al., 2011).

A broader extension of BOR methodology appears in KARA. There, KAPTURE samples Schottky-diode CSR signals synchronously at 499.71 MHz, one sample per bucket per turn, producing a 4-channel single-shot THz spectrometer capable of recording 500 million spectra per second and streaming readout (Steinmann et al., 2017). Combined with EOSD and fast-gated imaging under hardware synchronization, this yields a distributed, synchronized, multi-parameter BOR-like system for longitudinal phase-space studies (Kehrer et al., 2017).

5. Scientific and operational uses

The most direct use of a BOR is feedback optimization and instability characterization. Digital multi-bunch feedback systems use BOR-style recordings to create controlled growth and damping transients, measure instability growth rate $x(t) + 2(D-G)\,\dot{x}(t) + \Omega^2 x(t) = 0,$1, feedback damping rate $x(t) + 2(D-G)\,\dot{x}(t) + \Omega^2 x(t) = 0,$2, and tune shifts under different FIR or IIR settings, then adjust loop gain and phase accordingly (Lonza et al., 2016). The SSRF processor was designed for this dual role: precise timing, deep memory, and bunch-by-bunch FIR processing support both suppression of transverse instabilities and offline analysis of recorded ADC data (Zhan et al., 2019).

BORs are also central to tune and coherent-mode diagnostics. In the Tevatron, bunch-by-bunch coherent-mode detectors based on a single high-$x(t) + 2(D-G)\,\dot{x}(t) + \Omega^2 x(t) = 0,$3 BPM, 8 GS/s sampling, and weak band-limited noise excitation complement Schottky systems as diagnostic tools for tunes, tune spreads, beam-beam effects, and electron-lens studies (Stancari et al., 2011). These systems measured coherent mode spectra whose evolution reflected betatron tune settings, beam-beam parameter, and collision pattern (Stancari et al., 2013).

Longitudinal BORs support machine optimization across fast acceleration cycles and multi-ring transfer chains. The Fermilab Booster system records beam synchronous phase, bunch length, and coupled-bunch mode amplitudes from 400 MeV to 8 GeV in 33 ms, while the Tevatron/Main Injector SBD system is used for quality control of injection, coalescing optimization, timing diagnostics, longitudinal emittance tracking, and studies of longitudinal mode-1 instability and beam loading (Steimel et al., 2012). These are BOR functions in the longitudinal plane even when the hardware is framed as “longitudinal beam properties” or “sampled bunch display.”

At light sources, BORs are routinely used for three-dimensional bunch diagnostics. The Hefei Light Source system extracts $x(t) + 2(D-G)\,\dot{x}(t) + \Omega^2 x(t) = 0,$4, $x(t) + 2(D-G)\,\dot{x}(t) + \Omega^2 x(t) = 0,$5, $x(t) + 2(D-G)\,\dot{x}(t) + \Omega^2 x(t) = 0,$6, and bunch charge from 500 $x(t) + 2(D-G)\,\dot{x}(t) + \Omega^2 x(t) = 0,$7 oscilloscope records, obtains the three-dimensional tune of each bunch in normal operation, and observes the spectrum peak of the transverse quadrupole oscillation when beam is incentived (Lu et al., 2021). The same work proposes extension to a “逐束团 6维(质心+尺寸)测量系统,” linking BOR practice to full bunch-by-bunch centroid-plus-size diagnostics.

In SuperKEKB, BORs have become machine-protection instruments. The RFSoC BOR records the last 100 turns before an abort, while a subsequent two-BOR analysis at Fuji-RFSoC and D5-RFSoC measured $x(t) + 2(D-G)\,\dot{x}(t) + \Omega^2 x(t) = 0,$8, $x(t) + 2(D-G)\,\dot{x}(t) + \Omega^2 x(t) = 0,$9, and $x(t) = k\, e^{-(D-G)t}\, \sin(\Omega t + \phi),$0 for typically 2346 of 5120 buckets and showed that sudden beam loss develops on a timescale of a few to tens of turns. Among 58 SBL events with both BORs active, 55 showed clear pressure bursts, and in all 58 events the sector-loss parameter $x(t) = k\, e^{-(D-G)t}\, \sin(\Omega t + \phi),$1 satisfied $x(t) = k\, e^{-(D-G)t}\, \sin(\Omega t + \phi),$2, indicating that most loss occurred in Section 1 and strongly implicating the D06 collimator region (Nomaru et al., 18 Jul 2025). The same study found that about 69% and 73% of events at Fuji-RFSoC, and about 81% and 83% at D5-RFSoC, had oscillation duration $x(t) = k\, e^{-(D-G)t}\, \sin(\Omega t + \phi),$3 turns in the horizontal and vertical planes, respectively (Nomaru et al., 18 Jul 2025).

6. Limitations, misconceptions, and development trajectory

BOR performance is limited by ADC resolution, analog noise, dynamic range, timing jitter, latency, and memory bandwidth. The digital-feedback literature notes that 8-bit ADCs may be marginal for very small oscillations, that large DC offsets can saturate ADCs, DACs, or amplifiers, and that clock jitter combined with non-flat pulse tops causes measurement noise and cross-talk (Lonza et al., 2016). The SSRF processor reports ENOB better than 9 bits from 100 kHz to 700 MHz and closed-loop attenuation better than 40 dB, but still identifies timing precision and ADC resolution as design constraints (Zhan et al., 2019). The SuperKEKB RFSoC BOR attains about $x(t) = k\, e^{-(D-G)t}\, \sin(\Omega t + \phi),$4 resolution, but the authors attribute the gap to the $x(t) = k\, e^{-(D-G)t}\, \sin(\Omega t + \phi),$5 level achieved elsewhere primarily to sub-optimal use of ADC input range and plan a custom high-gain mezzanine (Nomaru et al., 2024).

A common misconception is that a BOR is necessarily a standalone instrument. The literature repeatedly shows the opposite: in digital feedback systems it is often the same BPM–ADC–FPGA platform operated in a recording mode, with extra memories, trigger logic, and software (Lonza et al., 2016). Another misconception is that BOR data alone localize the source of an instability. SuperKEKB explicitly shows that a single recorder cannot distinguish a small kick near the recorder from a larger kick at a phase-degenerate location; multiple BORs distributed around the ring are required for localization (Nomaru et al., 2024).

BORs also complement rather than replace other diagnostics. Standard BPM systems average over many bunches or longer times and miss fast bunch-resolved dynamics; BLMs provide local loss signals but no bunch-by-bunch structure; vacuum gauges detect pressure bursts with millisecond response; Schottky monitors measure incoherent tune distributions; and beam-size or CSR diagnostics reveal dimensions of the longitudinal phase space that centroid-only BORs do not record (Nomaru et al., 18 Jul 2025). This suggests that the contemporary trend is toward synchronized, multi-parameter BOR ecosystems rather than isolated recorders.

The development trajectory follows accelerator electronics more broadly: DSP-based implementations in early bunch-by-bunch feedback systems, FPGA-based diagnostic platforms with internal memories and control-system integration, oscilloscope-based rapid-deployment systems, and recent RFSoC-based portable monitors (Lonza et al., 2016). In the same direction, KARA’s synchronized diagnostics and KAPTURE-based THz spectrometer extend BOR practice from centroid motion to continuous monitoring of CSR intensity, energy spread, longitudinal profile, and micro-bunching dynamics (Steinmann et al., 2017). The likely implication is that “BOR” now names a family of bunch-resolved transient-recording techniques whose common purpose is to make oscillatory beam dynamics directly observable on the turn-by-turn timescale.