The pulse tone power (perceived as loudness) changed with saturation level for all pulse oximeters, as shown in Figure 2 and Table 4. Power increased as saturation decreased from 100%, peaking at 65% to 96% saturation depending on the pulse oximeter. In every pulse oximeter except the GE Solar, powers at low saturations (below thresholds of 83% to 12%, depending on the pulse oximeter) were less than that at 100% saturation. Peak powers ranged from 0.8 to 13.3 dB higher and nadir powers ranged from 3.0 to 13.8 dB lower than the powers at 100% saturation (a power difference of 10 dB is experienced as about a 2-fold change in loudness).
We have presented a comprehensive analysis of commercial pulse oximeter tones over the full range of saturation, from 0 to 100%. Our principal findings are as follows. First, the pitch of pulse tones may not decrease when saturation decreases below a certain level in some pulse oximeters. Second, the algorithms for mapping saturation to pitch are different among various brands of pulse oximeter and may even be different in pulse oximeters of the same model or similar models. Third, pitch steps occur with every change in saturation, or at regular saturation intervals, or at variable saturation intervals, depending on the pulse oximeter. Fourth, most pulse oximeters use pitch steps of constant magnitude (i.e., fixed frequency difference), but many pulse oximeters use pitch steps of constant proportion (i.e., fixed frequency ratio). Fifth, pulse tones generally become louder at modest levels of desaturation and quieter at deep levels of desaturation.
Comparison with Prior Studies
Our pulse tone frequency content results are directly comparable with those from 2 other studies. Santamore and Cleaver9 analyzed the fundamental frequencies at saturations >80% from 4 pulse oximeter models, 3 of which were the same or similar to ones we studied. Our results and those of Santamore and Cleaver for the Nellcor N-100 were within 20 Hz of each other. Similarly, our results for the GE Datex-Ohmeda CAM (a compact version of the AS/3) and the Philips monitors (formerly Agilent) were within 20 Hz of the results of Santamore and Cleaver for the Datex-Ohmeda AS/3 and Agilent ACMS M1177A, respectively.
In contrast, our results differed from those reported by Chandra et al.,10 who analyzed the frequency content at saturations >85% of 5 pulse oximeter models, 3 of which were the same or similar to ones we studied. They reported frequencies that were consistently 20 to 85 Hz higher than ours when they compared the Datex AS/3 with the Datex CAM, the Ohmeda Biox 3700 with the Ohmeda Biox 3740, and the Hewlett-Packard M116a with Philips monitors. We cannot explain why the results of Chandra et al. were slightly different from ours, particularly because they did their Fourier analysis using a Hanning algorithm with a window size of 1024 samples, which should yield equivalent but less-precise results. Chandra et al. also reported only cursory information regarding the timbre, or harmonic content, of the pulse tones. Their result of 3 harmonics for the Hewlett-Packard monitor is consistent with the Philips spectrogram in our Figure 1, which shows 3 harmonics at high saturations but up to 5 major harmonics at lower saturations. They reported 12 harmonics for the Ohmeda Biox 3700, but only 6 primary harmonics are visible on our spectrogram of the Ohmeda Biox 3740 in Figure 1. In summary, a comparison of our findings with those from previous studies provides some evidence that manufacturers tend to use the same frequency ranges in their different model pulse oximeters, even when the manufacturer changes ownership (e.g., Hewlett-Packard became Agilent and then Philips).
Challenges for the Design of Pulse Oximeter Tones
There are 2 goals in designing a pulse oximeter sonification that are at cross-purposes to one another. One goal is to have pitch changes at each step that are large enough to be easily perceived by the user. A competing goal is that saturation values from 0 to 100 need to be conveyed using a limited range of sound frequencies.
Our results show that manufacturers respond to these competing goals in different ways. An early design, implemented in the Nellcor N-100 pulse oximeter, had a fundamental frequency of 650 Hz at 100% saturation with a decrease of 5 Hz at every 1% decrease in saturation. Although this sonification was clinically useful, Schulte and Block7 later tested peoples’ abilities to detect changes when saturation values decreased incrementally from 100% to 90%, and they demonstrated that a 5-Hz step change was too small to be consistently heard.
Later designs have used 1 or more of 4 different techniques to improve perceivability of saturation changes while staying within a limited frequency range: (1) increasing the frequency range, (2) using ratio step sizes, (3) not changing pitch at every step change in saturation, and (4) including a lower saturation cutoff.
Three of the manufacturers whose pulse oximeters we studied (GE Datex-Ohmeda, Philips, and Masimo) used the first technique, listed previously, of expanding the frequency range in their pulse oximeters. Doing so allows the use of larger step sizes within the set frequency range. Four of the manufacturers whose pulse oximeters we studied (Philips, GE Healthcare, GE/Marquette, and Masimo) used uniform ratio step sizes, the second of the techniques listed previously, in some or all their pulse oximeters. Uniform ratio step sizes translate to larger steps in hertz at high frequencies and smaller steps at low frequencies, which preserve approximately equal pitch differences for human auditory perception. However, the resulting average step size is smaller than it would be with fixed step sizes, allowing larger perceivable steps to fit within a set frequency range. Four manufacturers (GE Datex-Ohmeda, GE Healthcare, GE/Marquette, and Masimo) adopted the third of the techniques listed previously of only changing the pitch at specific step changes in saturation in some or all their pulse oximeters. Doing so decreases the number of audible steps, which allows a larger pitch step size within a set frequency range, but it also decreases the clinical utility of the sonification, in that clinicians are not alerted to a change in saturation until it has changed by >1%. In some implementations, the clinician is not alerted until the saturation changes by >5%. Four manufacturers (GE Healthcare, GE/Marquette, GE Datex-Ohmeda, and Ohmeda) used the fourth of the techniques listed previously of introducing a lower saturation level cutoff to the sonification. Doing so is another way to limit the number of steps and thus increase pitch step size within a set frequency range. But introducing a lower saturation level cutoff also decreases clinical utility of the sonification because the clinician stops receiving auditory feedback when the saturation decreases below a certain level, in one implementation when saturation decreases <70%.
From a clinical perspective, the first 2 techniques for increasing perceivability (increasing the frequency range and using ratio step sizes) have no drawback, but the last 2 techniques (not changing pitch at every saturation change and using a lower saturation cutoff) do have potential clinical drawbacks. Surprisingly, in the 30 years that pulse oximeters have been in use, there appears to be no report that some pulse oximeters use a lower saturation cutoff level below which there are no further pitch changes. Clinicians may easily assume that pulse oximeter pitch changes whenever saturation changes and may provide clinical care during a crisis based on the assumption that the saturation is stable if the pitch does not change. During a desaturation event, clinicians will focus on improving ventilation and oxygen delivery until the pulse oximeter saturation stops decreasing and starts increasing. Any delay in the sonification of current saturation values potentially degrades this feedback loop. For instance, a clinician may mistakenly assume that patient oxygenation is not deteriorating further once the pitch reaches the lower saturation cutoff or there may be a delay in realizing that therapies are beginning to improve oxygenation when the pitch does not change until saturation changes by >5%. It is well known that pulse oximeters are less accurate at lower saturations, but their direction of change information is reliable at all saturation levels.
Linear Versus Logarithmic Step Sizes
Our study reveals that the pulse tone step sizes in current commercial pulse oximeters are based on the difference between tones frequencies (which we and others refer to as a “linear” relationship) or based on the ratio of tone frequencies (which we and others refer to as a “logarithmic” relationship). In designing a pulse oximeter sonification, one goal might be to code a percentage decrease in saturation as that same percentage decrease in perceived pitch.
Morris and Mohacsi8 found that experienced anesthesiologists could not accurately assess the percentage difference in saturation when listening to pulse tones at 2 different saturation levels from a Datex AS/3 pulse oximeter. The pulse tones were at saturation levels between 70% and 100%, and the authors reported that the fundamental frequencies were between 640 and 1000 Hz. Their subjects, who used a Datex AS/3 pulse oximeter at least 1 day per week, consistently underestimated the percentage difference in saturation, perceiving saturation differences of −25% to +25% as being between −5% to +5%. The Datex AS/3 uses a linear mapping of hemoglobin saturation to pulse tone frequency, with a step change of 20 Hz for every 2% change in saturation. In a subsequent study, Brown et al.12 compared anesthesiologists’ abilities to judge absolute saturation values, direction of change in saturation between 2 values, and magnitude of change in saturation between 2 values when listening to pulse oximeter tones with linear versus logarithmic relationships between saturation and pitch. They found that judgments in all 3 categories were somewhat better with logarithmic mapping and suggested that a logarithmic mapping should become a standard for all pulse oximeter sonifications.
Pitch is “that attribute of auditory sensation in terms of which sounds may be ordered on a scale extending from high to low.”13 It is a perceptual phenomenon that is related to the frequency of sound: higher frequencies are perceived as higher pitch. Musical scales are certainly logarithmic, with each note in an octave having a frequency twice that of the same note in the octave below. However, musical scales coordinate the harmonic content of notes and are not necessarily applicable to the design of a pulse oximeter sonification. Importantly, psychoacoustic studies have shown that the relationship between pitch perception and frequency is neither linear nor logarithmic. The Mel scale is a perceptual scale of pitches judged by listeners to be equal in distance from one another.11,14 This scale reveals that a doubling in frequency >300 Hz is perceived as less than a doubling of pitch. Interestingly, no pulse oximeter sonification uses Mel scale mapping.
Loudness Variation at Different Saturation Levels
We are unaware of any previous report that pulse tone loudness changes over the range of a pulse oximeter’s saturation values. This finding could be an artifact of the recording system used for the present study, an unpublicized design feature of pulse oximeters, or an unrecognized imperfection of current pulse oximeter implementations. To exclude artifact because of the frequency response of our recording system, we plotted power versus fundamental frequency of all our pulse oximeters on the same graph (Fig. 3). If pulse oximeter power was constant over the range of saturation values and the frequency response of our recording system was not flat, then we would expect all pulse oximeters to have the same pattern of power changes as a function of frequency. Figure 3, however, shows that the patterns are different for each pulse oximeter brand, excluding artifact as an explanation for the finding. Conceivably, pulse oximeter designers intentionally vary the volume of pulse tones at different saturation values to compensate for human perception. From an acoustic perspective, sound pressure level in decibels needs to increase in low and high frequency ranges if a percept of constant loudness is to be maintained (so-called equiloudness curves).15 But the equal loudness level curve in Figure 3 indicates that designers do not intentionally vary the volume of pulse tones so that they are perceived as being equally loud as saturation decreases. This leads us to assume that the finding is an unrecognized consequence of current pulse oximeter implementations.
From a clinical perspective, it might be appropriate to intentionally increase perceived volume during desaturation to help draw clinicians’ attention and aid in auditory monitoring during treatment. However, in all but one pulse oximeter model that we tested, loudness decreased below baseline at deep saturation levels. This could have negative clinical implications, and one author has reported difficulty in hearing pulse tones when responding to deep desaturation events when saturation level is critically low. It is unclear whether the change in volume with saturation is by design or an unintended consequence of small speaker size. Either way, it is apparent that this aspect of current pulse oximeter performance has not been adequately specified or tested.
This study has several limitations. First, the Fourier transform window size was limited by the short duration of the pulse tones, yielding a frequency resolution of 5.4 Hz. Therefore, the fundamental frequencies reported in Table 2 may be slightly incorrect. Second, power measurements were made from numerical analysis of our recordings, not by using a calibrated sound level meter. Power measurements could therefore have been affected by the dynamic response of our microphone. If this were the reason for the variation in loudness with saturation, then measured power would be a function of pulse beep fundamental frequency. However, when results from all the pulse oximeters were charted as relative power versus fundamental frequency, there was no uniform effect of fundamental frequency, as shown in Figure 3. Moreover, preliminary measurements of the GE/Marquette Solar monitor with a calibrated sound level meter showed the same pattern of power changes as we found in our recordings. Third, recordings were taken with the microphone positioned adjacent to the pulse oximeter speaker and not from the position of a potential user. Because of this, some recordings contained audible contamination from internal monitor components, such as fans. Although these background sounds were included in the audio spectrum analysis, they were constant across each recording. Another consequence of taking sound recordings adjacent to the speakers is that loudness was not affected by room physics. This is important because users hear pulse oximeter sonifications in free-field, not through headphones, which may affect perceived volume as a function of saturation.
Although pulse oximeters from a number of large manufacturers were studied, many were not. There are hundreds of pulse oximeter models and many are low-cost instruments, such as the LifeBox, designed for use in low-income countries. The findings reported here cannot be directly extended to other models of pulse oximeters, but our open-source methods could be applied to evaluate the pulse tones from any pulse oximeter.
In summary, we found little uniformity in the implementation of pulse tone beeps across different pulse oximeter models and sometimes within the same or similar models. The only similarity between models was that pitch always increased with higher saturation. We found that each pulse oximeter used one or more of the 4 techniques to address the competing constraints of pitch change perception and frequency range. Finally, we found that pulse tone loudness was not uniform across the range of saturation values.
We have 3 recommendations for the design of future pulse oximeter sonifications on the basis of our findings and the clinical implications discussed previously. The first is to use a large frequency range and a logarithmic mapping so that pitch change is easy to detect at all steps. The second is to implement a pitch change at every 1% change in saturation, from 100% to 0%. The third is to increase the perceived loudness of pulse tones modestly as saturation decreases, perhaps in a stepwise fashion, and avoid pulse tones that are quieter than the 100% saturation tone.
In the meantime, anesthesia providers, and other clinicians who use variable-pitch pulse oximeter tones, should learn from this study that, in some current pulse oximeters, pulse tone pitch does not change with every transition in saturation value and that some pulse oximeters have a saturation threshold below which pulse tone pitch remains constant. They should also understand that pulse tones might become quieter and more difficult to hear during deep desaturation events. We urge clinicians to investigate whether the pulse oximeters that they use have these limitations, and if so to modify their reliance on pulse tones. One approach, during a deep desaturation event when vision is otherwise occupied, would be to have an assistant increase the volume if pulse tones become too quiet to hear or call out saturation values if they continue to change and the pitch does not.
Name: Robert G. Loeb, MD.
Contribution: This author helped design the study, conduct the study, collect the data, analyze the data, and prepare the manuscript.
Attestation: Robert G. Loeb approved the final manuscript, attests to the integrity of the original data and the analysis reported in this manuscript, and is the archival author.
Conflicts of Interest: Robert G. Loeb is on the Masimo Scientific Advisory Board and has received consulting fees for this.
Name: Birgit Brecknell, PhD.
Contribution: This author helped analyze the data and prepare the manuscript.
Attestation: Birgit Brecknell approved the final manuscript and attests to the integrity of the original data and the analysis reported in this manuscript.
Conflicts of Interest: None.
Name: Penelope M. Sanderson, PhD.
Contribution: This author helped analyze the data and prepare the manuscript.
Attestation: Penelope M. Sanderson approved the final manuscript.
Conflicts of Interest: None.
This manuscript was handled by: Maxime Cannesson, MD, PhD.
a Original recordings can be viewed online at: https://www.youtube.com/playlist?list=PLjZEF-1WCbVyTAY7egkB5w7ujzTZpDinn
b The MATLAB script, instructions, executable code, and sample data files are available online at: http://www.mathworks.com/matlabcentral/fileexchange/55001-pulse-oximetry-project
c Reconstructed videos can be viewed online at: https://www.youtube.com/playlist?list=PLjZEF-1WCbVy2c8494z7YzjDafut6AoiW
1. New W Jr, Corenman JE Pulse oximeter monitor. 1987 U.S. Patent 4,653,498;
3. Kramer G. Some organizing principles for representing data with sound. Auditory Display. 1994;18:185–221
4. Sanderson PM, Watson MO, Russell WJ. Advanced patient monitoring displays: tools for continuous informing. Anesth Analg. 2005;101:161–8
5. Craven RM, McIndoe AK. Continuous auditory monitoring—how much information do we register? Br J Anaesth. 1999;83:747–9
6. Sanderson PM, Watson MO, Russell WJ, Jenkins S, Liu D, Green N, Llewelyn K, Cole P, Shek V, Krupenia SS. Advanced auditory displays and head-mounted displays: advantages and disadvantages for monitoring by the distracted anesthesiologist. Anesth Analg. 2008;106:1787–97
7. Schulte GT, Block FE Jr. Can people hear the pitch change on a variable-pitch pulse oximeter? J Clin Monit. 1992;8:198–200
8. Morris RW, Mohacsi PJ. How well can anaesthetists discriminate pulse oximeter tones? Anaesth Intensive Care. 2005;33:497–500
9. Santamore DC, Cleaver TG. The sounds of saturation. J Clin Monit Comput. 2004;18:89–92
10. Chandra D, Tessler MJ, Usher J. Audio spectrum and sound pressure levels vary between pulse oximeters. Can J Anaesth. 2006;53:26–32
12. Brown Z, Edworthy J, Sneyd JR, Schlesinger J. A comparison of linear and logarithmic auditory tones in pulse oximeters. Appl Ergon. 2015;51:350–7
13. ANSI. Acoustical Terminology. ANSI S1.1-1994. 1994 New York American National Standards Institute
14. Rasch R, Plomp RDeutsch D. The perception of musical tones. In: The Psychology of Music. 1999 2nd ed. San Diego Academic Press:89–112
© 2016 International Anesthesia Research Society
15. ISO. Acoustics–Normal equal-loudness-level contours. ISO 226: 2003. 2003 Geneva, Switzerland International Organization for Standardization