Simultaneous Multi-Site PPG Measurement: Techniques, Applications & Pulse Transit Time

Synchronization Requirements and Methods

Timing Accuracy Requirements

The required synchronization accuracy depends on the target application and the PTT values being measured. For PTT-based blood pressure estimation, PTT changes by only a few milliseconds over clinically relevant blood pressure changes, so timing errors of even a millisecond between sites translate into meaningful blood pressure error.

For aortic PWV estimation using finger-to-toe PTT, the typical PWV change of 1 m/s corresponds to a PTT change of approximately 20-40 ms over the 1.5 m path length. This relaxes the synchronization requirement to approximately 2-5 ms for clinically meaningful PWV resolution.

Hardware Synchronization

The most reliable synchronization approach uses a shared hardware clock or trigger signal between PPG AFE channels. Modern multi-channel AFEs such as the MAX86141 (2 channels) and ADPD4101 (8 input channels) provide inherent synchronization between channels because all channels share a single clock and sampling sequencer. When both PPG sensors connect to the same AFE, synchronization is exact to within the ADC conversion time (typically 1-10 us).

For sensors connected to separate AFEs on separate devices (e.g., a wrist device and an ear clip), hardware synchronization requires a wired trigger line or wireless trigger with deterministic latency. A GPIO trigger pulse from one device to the other can achieve synchronization better than 1 us when properly implemented.

Software Synchronization

When hardware connections between devices are impractical, software-based time synchronization is necessary. Several approaches are used:

Bluetooth timestamp synchronization: BLE time synchronization achieves typical accuracy of 1-5 ms, which is marginal for PTT applications on short arterial paths but adequate for finger-to-toe measurements. The precision depends heavily on the BLE stack implementation, connection interval, and whether the devices support the BLE Isochronous Channels feature introduced in Bluetooth 5.2.

Cross-correlation-based alignment: Rather than relying on absolute time synchronization, the PPG signals from both sites can be aligned using cross-correlation to find the optimal time delay. This approach is robust to clock drift and eliminates the need for precise time synchronization, but it assumes a specific model for the expected waveform delay (typically that the two waveforms have similar morphology, which is only approximately true between central and peripheral sites).

Network time protocols: For laboratory setups using wired connections, IEEE 1588 Precision Time Protocol (PTP) can synchronize devices to within 1 us over Ethernet. This is the standard approach for research-grade multi-site PPG systems.

Pulse Transit Time Estimation Algorithms

Feature-Based PTT Estimation

The traditional approach to PTT estimation identifies characteristic features (fiducial points) on each PPG waveform and calculates the time difference between corresponding features at the two sites.

Foot of the pulse wave: The pulse foot, defined as the point of maximum acceleration at the onset of the systolic upstroke, is the most commonly used fiducial point. It corresponds most closely to the actual arrival of the pressure wave front. Detection methods include the intersecting tangent method (drawing tangent lines on the diastolic trough and systolic upstroke and finding their intersection) and the second derivative maximum (the point of maximum positive acceleration in the waveform).

Peak of the pulse wave: The systolic peak is the easiest feature to detect but is influenced by wave reflection timing and local vascular compliance, making it a less pure measure of pulse wave arrival time. Peak-based PTT includes both the transit time and the local pulse wave buildup time, introducing a site-dependent offset.

Maximum slope (dP/dt max): The point of maximum rate of rise on the systolic upstroke. This feature is less sensitive to wave reflection effects than the peak and more robust to baseline wander than the foot. It represents a compromise between the theoretical purity of the foot and the detection reliability of the peak. For deeper analysis of PPG waveform derivatives, see our second derivative analysis guide.

Waveform-Based PTT Estimation

Instead of extracting discrete features, waveform-based approaches estimate PTT from the overall shape relationship between the two PPG signals.

Cross-correlation: Computing the cross-correlation function between windowed segments of the two PPG signals and finding the lag at maximum correlation. This provides a robust PTT estimate that averages over the entire waveform rather than relying on a single feature. The temporal resolution is limited by the sampling rate (typically 0.01-0.04 ms at 25-100 Hz) but can be improved with parabolic interpolation around the correlation peak.

Phase difference in the frequency domain: Computing the phase of the cross-spectral density between the two signals at the cardiac fundamental frequency. This approach naturally provides sub-sample temporal resolution and is robust to noise, but assumes that the phase relationship is linear across the cardiac frequency band (which is approximately true for PTT but breaks down if the waveform morphology differs substantially between sites).

Gao et al. (2016) (doi:10.14814/phy2.12768) compared noninvasive PTT estimates against invasively measured PTT in swine over a wide blood pressure range. PTT between two PPG waveforms correlated well with all blood pressure levels (group average R² 0.81-0.85, not significantly different from invasive PTT), whereas ECG-to-PPG pulse arrival time correlated less well (best R² 0.70, for systolic BP) because it includes the pre-ejection period. This supports measuring true PTT between two PPG sites rather than using the ECG R-peak as the proximal marker.

Clinical Applications

Cuffless Blood Pressure Estimation

The primary clinical driver for multi-site PPG research is cuffless blood pressure monitoring. The inverse relationship between PTT and blood pressure is well established, though its strength varies across studies, individuals, and conditions (see the review by Mukkamala et al., 2015).

The current state of the art for PPG-PPG PTT blood pressure estimation achieves mean absolute errors of 5-10 mmHg for systolic and 4-7 mmHg for diastolic pressure in controlled laboratory studies. However, several challenges limit clinical adoption:

Calibration requirement: The PTT-BP relationship varies between individuals due to differences in arterial path length, vascular compliance, and autonomic tone. Individual calibration against a reference cuff measurement is necessary, typically using a linear or logarithmic model: BP = a * PTT + b or BP = a * ln(PTT) + b.

Calibration drift: The calibration parameters drift over time due to changes in vascular tone, hydration status, and long-term arterial remodeling. Recalibration is typically needed every 1-4 weeks, which limits the practical utility of continuous monitoring. For a broader discussion of cuffless blood pressure challenges, see our cuffless blood pressure guide.

Arterial Stiffness Assessment

Aortic PWV is an independent predictor of cardiovascular mortality and is recommended by ESH/ESC guidelines for cardiovascular risk stratification. Reference values for carotid-femoral PWV are well established: an expert consensus recommends 10 m/s as the cut-off for carotid-femoral PWV (Van Bortel et al., 2012; doi:10.1097/HJH.0b013e32834fa8b0).

Multi-site PPG provides a noninvasive, operator-independent alternative to tonometry-based PWV measurement. Finger-to-toe PPG PWV has been studied as a surrogate for carotid-femoral PWV, the gold standard. The main limitation is that the finger-to-toe path includes both elastic central arteries and stiff muscular peripheral arteries, making the measured PWV a composite value that does not directly correspond to aortic stiffness.

Peripheral Vascular Disease Screening

Bilateral multi-site PPG at symmetric body sites (left vs. right fingers, left vs. right toes) can detect vascular asymmetry indicative of peripheral arterial disease (PAD). Because left and right pulses are highly similar in healthy people (Allen and Murray, 2000), PPG-based detection of peripheral vascular disease is one of the clinical applications reviewed by Allen (2007) (doi:10.1088/0967-3334/28/3/R01). See our peripheral artery disease guide for diagnostic-accuracy evidence.

The advantage of PPG over the traditional ankle-brachial index (ABI) is that PPG can be performed as a continuous, unattended measurement rather than requiring a trained operator with a Doppler probe and sphygmomanometer. Multi-site PPG waveform analysis also provides additional information about the severity and location of stenosis through wave reflection timing analysis.

Practical Implementation Considerations

Multi-Channel Data Acquisition

For research-grade multi-site PPG systems, simultaneous acquisition of 2-8 PPG channels plus accelerometer and ECG reference channels is common. Key hardware considerations include:

Sampling rate matching: All PPG channels should be sampled at the same rate and ideally from the same clock domain. Mismatched sampling rates require resampling, which introduces interpolation error that directly affects PTT accuracy.

Channel crosstalk: When multiple PPG sensors are close together (e.g., multi-wavelength sensors on adjacent fingers), optical crosstalk between LED sources can corrupt the signals. Sequential LED driving (time-division multiplexing) eliminates optical crosstalk at the cost of a small timing offset between channels (typically 0.1-1 ms, which is negligible for most applications).

Ambient light rejection: Each measurement site may have different ambient light conditions, requiring independent ambient light cancellation for each channel. This is handled automatically by most multi-channel AFEs through correlated double sampling on each channel independently.

ECG-PPG vs. PPG-PPG Transit Time

An important distinction in the literature is between ECG-derived pulse arrival time (PAT) and PPG-PPG pulse transit time (PTT). PAT, measured as the delay between the ECG R-wave and a peripheral PPG feature, includes both the pre-ejection period (PEP, the isovolumetric contraction time of the left ventricle, typically 30-80 ms) and the actual pulse transit time. PEP varies with cardiac contractility, preload, and afterload, introducing a confounding variable that degrades the PAT-BP correlation.

PPG-PPG PTT, measured between two peripheral PPG sites, largely eliminates the PEP component because both sites are downstream of the aortic valve. This makes PPG-PPG PTT a purer measure of pulse wave velocity and, theoretically, a better surrogate for blood pressure. However, PPG-PPG PTT requires two wearable sensors rather than the single sensor plus ECG electrode combination used for PAT measurement.

Conclusion

Multi-site PPG measurement extends the clinical and research utility of photoplethysmography far beyond what is possible with a single sensor, enabling assessment of arterial stiffness, blood pressure tracking, and peripheral vascular disease screening. The critical technical challenges are sensor synchronization (sub-millisecond for PTT applications), robust fiducial point detection for accurate PTT estimation, and individual calibration for blood pressure applications. As wearable form factors mature and multi-device synchronization improves, multi-site PPG has the potential to bring continuous, cuffless hemodynamic monitoring from the research laboratory to routine clinical practice. For engineers building multi-site PPG systems, careful attention to the signal processing pipeline and power management is essential for practical implementation.

References

• Laurent, S., et al. (2006). Expert consensus document on arterial stiffness: methodological issues and clinical applications. European Heart Journal, 27(21), 2588-2605. https://doi.org/10.1093/eurheartj/ehl254
• Allen, J., & Murray, A. (2000). Similarity in bilateral photoplethysmographic peripheral pulse wave characteristics at the ears, thumbs and toes. Physiological Measurement, 21(3), 369-377. https://doi.org/10.1088/0967-3334/21/3/303
• Allen, J. (2007). Photoplethysmography and its application in clinical physiological measurement. Physiological Measurement, 28(3), R1-R39. https://doi.org/10.1088/0967-3334/28/3/R01
• Mukkamala, R., et al. (2015). Toward ubiquitous blood pressure monitoring via pulse transit time: theory and practice. IEEE Transactions on Biomedical Engineering, 62(8), 1879-1901. https://doi.org/10.1109/TBME.2015.2441951
• Gao, M., Olivier, N. B., & Mukkamala, R. (2016). Comparison of noninvasive pulse transit time estimates as markers of blood pressure using invasive pulse transit time measurements as a reference. Physiological Reports, 4(10), e12768. https://doi.org/10.14814/phy2.12768
• Van Bortel, L. M., et al. (2012). Expert consensus document on the measurement of aortic stiffness in daily practice using carotid-femoral pulse wave velocity. Journal of Hypertension, 30(3), 445-448. https://doi.org/10.1097/HJH.0b013e32834fa8b0