Insights
Ultrasound Staggered PRF: Preserve the Pulse Intervals Behind the Velocity Estimate
Why staggered Doppler processing needs actual pulse intervals, with an exact phase-aliasing example, noise trade-offs and a proposed OEM acquisition acceptance test.
ALEVA — Orange County, California, USA
Engineering question: When a Doppler acquisition alternates its pulse intervals, what must the acquisition and processing interface preserve so that a velocity estimate remains interpretable?
A single PRF field is sufficient only when it describes the time spacing actually used by the estimator. Alternating intervals, interleaved transmit angles and missing acquisitions can make that assumption false. The integration task is to preserve the relationship between each complex sample, its transmit event and the interval used for phase estimation.
Why staggered timing deserves an explicit interface
Posada and colleagues investigated staggered dual- and triple-PRF ultrafast color Doppler in a peer-reviewed 2016 journal paper. Their work establishes that changing the intervals between transmissions can provide information for resolving velocity aliasing. It is foundational research, not a newly released technique. [1]
A 2025 IEEE IUS conference paper by Bonciani and colleagues describes a real-time implementation using interleaved dual-PRF sequences on a platform combining FPGAs, DSPs and an embedded GPU. Its public abstract reports flow-phantom experiments with a linear array. That is relevant implementation evidence, but it does not establish performance for another probe, timing schedule or product. This article does not claim that ALEVA uses the reported implementation. [2]
A reproducible phase example
The following calculation is an ideal educational model, not a scanner measurement or a reproduction of either paper's experimental results. Assume a noiseless complex signal s(t) = exp(j 2π fD t), a constant Doppler frequency fD = 6,000 Hz, and alternating intervals T1 = 100 microseconds and T2 = 150 microseconds. Use the phase convention arg(s(t + T) s(t)*), wrapped to [−π, π).
The 100-microsecond pair advances by 0.6 cycles. After wrapping, its phase is −0.8π radians; dividing that phase by 2πT1 gives −4,000 Hz. The 150-microsecond pair advances by 0.9 cycles, wraps to −0.2π radians, and gives approximately −666.7 Hz. Both individual estimates are aliased, despite describing the same positive 6,000-Hz signal.
In this ideal constant-frequency model, the wrapped difference between those phases is 0.6π radians. Dividing it by 2π(T2 − T1) recovers 6,000 Hz. That difference measurement repeats every 1/(T2 − T1) = 20,000 Hz, giving a centered ambiguity interval of [−10,000, 10,000) Hz. This interval is a mathematical property of this chosen example, not a guaranteed usable velocity range.
Replacing the two intervals with their 125-microsecond average loses the distinction needed for this calculation. For example, averaging the two unit phase vectors and dividing the resulting phase by 2π times that average interval yields −2,000 Hz here. An average pulse rate therefore cannot substitute for the interval labels used by a staggered estimator.
The extension has a noise cost
Let the two measured phases have errors ε1 and ε2. Away from wrapping boundaries, the difference-based frequency error in this model is (ε2 − ε1)/[2π(T2 − T1)]. A smaller interval difference increases the ideal ambiguity interval while also increasing sensitivity to a given phase-difference error. Correlated errors, finite ensembles and wrap mistakes require explicit evaluation; the formula does not predict their probability.
A practical implementation may combine an ambiguity-resolving estimate with a more precise estimate from longer intervals. It should be evaluated on wrong-branch decisions and rejected estimates as well as average error. A visually smooth color map can conceal incorrect ambiguity selection, especially if smoothing is applied before results are inspected.
Preserve events, not just a nominal rate
For an OEM integration specification, we propose recording an event identifier, an acquisition-time reference, transmit angle or sequence role, interval class, and validity information alongside each ensemble. The estimator should document which event pairs it uses and the elapsed time for each pair. These are proposed interface requirements, not a claim of a universal standard.
If a sample is lost, the elapsed time across the gap is longer than the ordinary adjacent interval. Silently compacting the remaining samples changes the estimator's timing model. Likewise, equal spacing between all transmit events does not imply equal spacing between repeated acquisitions at one angle. A timestamp taken when a host buffer arrives is not a substitute for the relevant acquisition-event timing.
The schedule also remains constrained by echo return time. In a simplified pulse-echo model with sound speed 1,540 m/s, a 100-microsecond interval corresponds to 77 mm of one-way depth before the next transmission. This is only cT/2: transmit duration, receiver recovery, guard time and the actual sequence reduce the available margin. Staggered processing does not remove the need to assess range ambiguity.
An acceptance test that exposes the trade-off
Begin with synthetic complex tones generated from the actual event-time list. Sweep both frequency signs, wrap boundaries and the proposed extended range. Then introduce controlled phase noise, acceleration, missing events and incorrect interval labels. Keep the interval-aware output, ordinary single-interval estimates, validity mask and ambiguity decisions available for comparison.
Next test the complete acquisition path with an appropriate moving or flow phantom and an independent reference. Repeat at different depths and relevant sequence settings. Report valid-estimate fraction, branch-error frequency, bias and spread separately, with the reference uncertainty. Pass limits should follow the intended engineering use and measured uncertainty; this example supplies no universal acceptance threshold.
For ultrasound OEM development, the deliverable is a traceable timing specification plus evidence that acquisition, transport and processing preserve it. This complements Doppler wall-filter packet testing: correct pulse intervals do not by themselves establish clutter rejection, flow sensitivity or clinical performance.
Sources and scope
Posada D. et al. Staggered Multiple-PRF Ultrafast Color Doppler. IEEE Transactions on Medical Imaging, 35(6), 1510–1521, June 2016; online publication January 18, 2016. Peer-reviewed journal article. DOI: 10.1109/TMI.2016.2518638. Author-hosted paper.
Bonciani G., Guidi F., Giangrossi C., Varray F., Garcia D., Boni E., Ramalli A. Real-Time High Frame Rate Color Doppler Imaging with Staggered Pulse Repetition Frequency. 2025 IEEE International Ultrasonics Symposium, September 15–18, 2025, pp. 1–4. Published conference paper, not a journal article or preprint; discussion here is limited to the public abstract and publication record. DOI: 10.1109/IUS62464.2025.11201738. University of Florence record.
The calculations and proposed acceptance workflow above are original educational analysis. They do not establish product specifications, clinical effectiveness or regulatory status.