`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`APPLE INC.,
`Petitioner,
`
`v.
`
`MASIMO CORPORATION,
`Patent Owner.
`
`IPR2022-01299
`U.S. Patent 7,761,127
`
`DECLARATION OF MOHAMED DIAB
`
`
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`
`
`1.
`
`I, Mohamed Diab, am making this declaration at the request of Patent
`
`Owner Masimo Corporation (“Masimo”) in the matter of the Inter Partes Review
`
`No. IPR2022-01299 of U.S. Patent No. 7,761,127 (“the ’127 patent”). I understand
`
`that Apple submitted the ’127 patent as Exhibit 1001 in these proceedings. The ’127
`
`patent describes and claims the invention that came out of our development of a
`
`sensor capable of noninvasively measuring carboxyhemoglobin. I understand that
`
`this declaration is being submitted in this proceeding as Exhibit 2102.
`
`2.
`
`I gave deposition and hearing testimony in an ITC Investigation in
`
`which Masimo asserts the ’127 patent and other patents against Apple. My
`
`testimony in this declaration is similar to my testimony in the ITC Investigation.
`
`3.
`
`I started as an engineer at Masimo. My current position at Masimo is
`
`fellow scientist. I started working at Masimo in 1989 and have worked there ever
`
`since.
`
`4.
`
`In 1986, I graduated from Cal State Fullerton with a Bachelor of
`
`Science degree in electrical engineering with an emphasis on computer engineering.
`
`5.
`
`In the 1990s, I and the other engineers at Masimo were working on our
`
`first pulse oximeter. I was involved in the hardware design, the sensor design, and
`
`the algorithm design. The algorithm takes the signal from the sensor and calculates
`
`pulse rate, oxygen saturation, and other parameters.
`
`-1-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`6.
`A pulse oximeter is a device that noninvasively measures physiological
`
`parameters in a patient’s blood by transmitting light into a tissue site (such as a
`
`finger) and measuring the light after it has passed through the tissue. Figure 1 of the
`
`’127 patent depicts a pulse oximeter with a sensor attached to a patient’s finger.
`
`7.
`
`In a typical pulse oximeter, the sensor that attaches to a patient’s finger
`
`contains: (1) two light sources, generally light-emitting diodes (LEDs), and (2) a
`
`light detector (generally a photodetector). Top and bottom views of a representative
`
`
`
`Masimo rainbow® sensor are shown below.
`
`
`
`-2-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`
`
`
`
`
`Top view
`
`Bottom view
`
`8.
`
`Oxygen saturation (“SpO2”) is a parameter measured noninvasively by
`
`pulse oximeters. For an oxygen-saturation measurement, the LEDs typically
`
`transmit red and infrared light into the patient’s finger. Some of the transmitted light
`
`is absorbed by the tissue and pulsating blood flow in the finger. Bright red
`
`oxygenated blood absorbs light differently than blue-green tinted deoxygenated
`
`blood. The light detector measures the intensity of the light (i.e., amplitude) from
`
`both wavelengths after it passes through the tissue. The ratio of the amplitude of the
`
`measured pulsating intensity of the light detected at the red wavelength compared to
`
`light detected at the infrared wavelength indicates the level of oxygen saturation.
`
`Pulse oximeters typically have a calibration curve or lookup table that correlates
`
`ratios of red to infrared wavelengths to SpO2 values. This curve reflects empirical
`
`data of ratios to invasively measured oxygen saturation values determined with a
`
`blood draw. Accurate measurements of saturation require the calibration curve to
`
`correspond to the actual LED wavelengths during operation of the sensor.
`
`-3-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`9. Masimo has become a leading innovator in pulse oximeters that
`
`measure oxygen saturation. I and other engineers at Masimo are inventors on
`
`hundreds of patents for oxygen-saturation measurement using pulse oximeters. For
`
`example, we were the first to develop pulse oximeters that could accurately measure
`
`oxygen saturation while a patient is moving.
`
`10. Masimo’s pulse-oximetry algorithms were already extremely accurate
`
`in measuring oxygen saturation before the ’127 patent invention.
`
`11.
`
`In about 2001, we started a project at Masimo to work on noninvasively
`
`measuring carboxyhemoglobin and other parameters within the hemoglobin species.
`
`The parameters within the hemoglobin species include oxyhemoglobin (blood
`
`oxygen saturation), carboxyhemoglobin, methemoglobin, and total hemoglobin.
`
`These parameters are much more difficult to measure noninvasively than the
`
`parameters traditionally measured by pulse oximetry.
`
`12. When carbon monoxide binds with hemoglobin, it displaces the oxygen
`
`and will not let the oxygen bind with hemoglobin for many hours to come. Thus, it
`
`turns the hemoglobin into a dysfunctional hemoglobin, causing carbon monoxide
`
`poisoning.
`
`13. Firefighters are exposed to carbon monoxide and, thus, can suffer from
`
`carbon monoxide poisoning. People that are exposed to furnaces or similar
`
`appliances lacking good combustion can also get carbon monoxide poisoning.
`
`-4-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`14. Carbon monoxide poisoning is very hard to diagnose because it looks
`
`like a flu. When people with high carbon monoxide poisoning go to the hospital,
`
`they look bright red and look like they have a flu.
`
`15.
`
`Thus, the noninvasive measurement of carboxyhemoglobin was a
`
`sought-after measurement to allow for early detection and treatment of carbon
`
`monoxide poisoning. Before our invention of the ’127 patent, no company had been
`
`able to noninvasively measure carboxyhemoglobin. To this day, no other company
`
`offers a competitive product.
`
`16. When Masimo started researching noninvasive measurement of
`
`carboxyhemoglobin, we looked into how carboxyhemoglobin interacts with light
`
`propagating through the tissue. I spent quite a bit of time looking into the theory
`
`behind it. Then, we conducted computer simulations trying to understand whether
`
`we could build a device that could measure carboxyhemoglobin in the tissue
`
`noninvasively, with reasonable accuracy that is relevant to the field. These
`
`simulations took about a year.
`
`17.
`
`The simulation results indicated that we could noninvasively measure
`
`carboxyhemoglobin. We also found that we could measure methemoglobin and total
`
`hemoglobin as well.
`
`18. After I had worked on carboxyhemoglobin measurement for about two
`
`years, Marcelo Lamego and others joined me on the rainbow® team. The name
`
`-5-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`rainbow® came from the number of colors, or wavelengths, of light emitted by the
`
`LEDs needed in the sensor. Typically, Masimo’s pulse oximeters that measure
`
`oxygen saturation use two LEDs with two wavelengths of light. Because the
`
`carboxyhemoglobin sensors used eight or more wavelengths, we called them
`
`rainbow® sensors.
`
`19.
`
`Exhibit 2131 is a true and correct copy of one of my research folders
`
`related to carboxyhemoglobin measurement. The tab on the first page includes
`
`“HbCO” in the label. “Hb” stands for hemoglobin and “CO” is carbon monoxide.
`
`20.
`
`The first page is the title page of a progress report we wrote in July
`
`2003. The report is titled, “RAINBOW PROJECT: Feasibility Study for
`
`Noninvasive Level Detection of Carbon Monoxide in Blood.” Mr. Lamego and I
`
`wrote the project report with input from Ammar Al-Ali, as shown in the footer of
`
`each page after the title page.
`
`21.
`
`In the report, we described a study we had done on 22 heavy smokers.
`
`The study shows that our measurements at that time had an error of about plus or
`
`minus six percent compared to the referee. We used about four sensors in the study.
`
`Each of the sensors had nine LEDs.
`
`22.
`
`The error of about plus or minus six percent was a step in the right
`
`direction and showed promise. However, the measurement was not accurate enough
`
`for a medical device.
`
`-6-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`23.
`
`
`
`
`
`
`
`
`
`24. While considering new sensor designs, I decided to look at the data
`
`itself and take a deeper look at what was causing the error.
`
`25.
`
`I noticed something peculiar about one of the four sensors we were
`
`using. Specifically, one sensor stood out as having a bias. When a sensor has a bias,
`
`that means the error is not random. Rather, the error is systematic. It is a hint that
`
`the sensor has properties that sets it apart from the rest.
`
`26.
`
`Initially, I thought something was wrong with the sensor itself.
`
`However, after careful examination of the sensor I found there was nothing wrong
`
`with how the sensor was built.
`
`27.
`
`I dug deeper, spending a couple of weeks looking in detail into the data.
`
`I found out that the sensor had a very sensitive LED whose wavelength was about
`
`one and a half nanometers off of the other three LEDs. We had carefully picked the
`
`LEDs to have about the same wavelength as each other, but this one LED was off
`
`by one and a half nanometers. To put things into perspective, the sensitive
`
`wavelength is 630 nm and this particular sensor had a wavelength of 631.5 nm.
`
`-7-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`28.
`I did a calculation and showed that, given the light absorption curve of
`
`carboxyhemoglobin, the LED being off by one and a half nanometers explained the
`
`bias in the sensor. We then started thinking about ways we could compensate for
`
`the LED wavelength shift.
`
`29. One cause of LED wavelength shift is the LED heating up. While the
`
`eye may see an LED as having a single color or wavelength, such as red, the LED
`
`wavelength has an emission spectrum that looks like a bell curve when measured
`
`using a device called a spectrophotometer. The emission spectrum has a peak and
`
`can be characterized by a weighted average called a centroid.
`
`30.
`
`If an LED heats up, which typically happens when an LED turns on,
`
`the peak, as well as the centroid, shifts to longer wavelengths. A mere increase of
`
`10 degrees Celsius in the LED temperature, which happens regularly when an LED
`
`is turned on, can cause a 2 nm shift in the wavelength. If this shift is not corrected
`
`it will render a carboxyhemoglobin measurement useless. We needed a technique
`
`that can calculate or adjust for that wavelength shift. We considered many
`
`techniques.
`
`31. We had previously designed a wavelength detector. Exhibit 2026 is a
`
`true and correct copy of an earlier Masimo patent (U.S. Patent No. 5,758,644)
`
`describing measuring LED wavelengths in real time without using a temperature
`
`sensor for the measurement, as shown in the patent figures below:
`
`-8-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`
`EX2026, Figs. 7, 9B. One embodiment described in that patent uses two
`
`photodiodes, a beam splitter, and an integrating optical sphere to determine the
`
`operating wavelengths of LEDs. Id., 20:9-28. However, this system was too
`
`unwieldy to implement in a small device such as the rainbow® sensors.
`
`32. We also thought we needed to know the junction temperature of the
`
`LED to calculate the wavelength. The junction temperature of the LED is the
`
`temperature at a location in the middle of the LED called the p-n junction. The LED
`
`has two different materials called p and n, and the junction between them is the p-n
`
`junction. If the junction temperature of an LED can be measured, the LED’s
`
`operating wavelength can be calculated. We initially thought that measuring the
`
`LED junction temperature to calculate the LED operating wavelength was
`
`conceptually the simplest and most straightforward solution to wavelength shift.
`
`Because each LED’s temperature changes based on complex interactions and is
`
`different from the temperatures of the other LEDs, we also thought we would need
`
`a separate thermistor for each LED to measure each LED’s junction temperature.
`
`-9-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`33. Many factors contribute to these complex interactions. Typically,
`
`about 90% of an LED’s energy is converted into heat. Accordingly, placing multiple
`
`LEDs in a typical finger-based sensor with a very small footprint can raise each
`
`LED’s junction temperature and the temperature at the base of the LED by tens of
`
`degrees Celsius. By contrast, a typical electronic chip has a large body area and
`
`many legs connected to the circuit board, which helps dissipate the heat generated
`
`inside the chip. By way of example, a small 8-pin electronic chip has a footprint of
`
`approximately 6 x 10 millimeters, which is 1,500 times larger than the footprint of
`
`an LED used in a rainbow® sensor. In addition, LEDs typically have different drive
`
`currents that may change based on differences in light absorption in the patient’s
`
`finger being measured. Accordingly, the amount of heat generated inside each LED
`
`is not constant but changes from patient to patient. Further, the LEDs need to be
`
`located near each other to minimize errors induced by varying photon mean path
`
`length, which adds complexity because heat interactions between the LEDs affect
`
`each of the LED’s temperatures. In addition, the LEDs need to be encapsulated to
`
`isolate them from contact with the human body, and the encapsulation layer’s poor
`
`heat conductivity limits heat dissipation. The rapid pulsing on and off of the LEDs
`
`also causes rapid fluctuation in LED temperatures. Finally, ambient temperature and
`
`skin temperature, which are not constant, affect the LED temperatures. Further, the
`
`appropriate design of a substrate to have a thermal mass capable of stabilizing and
`
`-10-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`normalizing a bulk temperature will differ depending on the amount of heat injected
`
`into the substrate by the LEDs and the arrangement of the LEDs, which may differ
`
`significantly based on how many LEDs are needed and how much heat each LED
`
`generates.
`
`34. The net result of these and other factors is that each LED temperature
`
`is dynamic and has complex interactions with other LEDs based on their drive
`
`currents. Those complex interactions and the LEDs having different temperatures
`
`from each other led us to initially believe we would need a separate thermistor for
`
`each LED to measure each LED’s junction temperature.
`
`35. However, we found that it was impractical to measure the LED junction
`
`temperature. It is very difficult to locate, at the p-n junction of an LED, a
`
`temperature sensor that can measure the LED junction temperature while a sensor in
`
`a very small device is in operation. Because each LED has a different junction
`
`temperature depending on the p-n junction type and the amount of energy pumped
`
`into the LED, it would have been even more complicated and impractical to use
`
`multiple temperature sensors to measure each LED junction temperature of the
`
`multiple LEDs of the rainbow® sensors. Furthermore, in a sensor, the LED junction
`
`temperature is not fixed. Rather, the LED junction temperature is dynamic. Our
`
`tests showed that the LED junction temperature is influenced by the ambient as well
`
`as the patient body temperatures. This implies that each LED’s wavelength is also
`
`-11-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`dynamic and must be measured in real time in order to correct for errors in the final
`
`physiologic parameter measurement.
`
`36. Therefore, we explored many different techniques. Eventually, we
`
`came up with a hypothesis that maybe a single thermistor could measure temperature
`
`of a thermal mass to which the LEDs are thermally bonded, and that we could
`
`calibrate to each LED the measured temperature could be used to estimate the
`
`operating wavelengths of each of the LEDs. I am not aware of this hypothesis being
`
`disclosed or suggested by anyone before we tried it. Also, this hypothesis was
`
`contrary to the intuitive belief, which we initially held, that measuring the LED
`
`junction temperature or an ambient temperature near the LEDs would provide the
`
`most accurate estimate of LED junction temperature for correcting wavelength shift.
`
`37. We could not just look up, in a technical data sheet or patent, a
`
`previously designed circuit board with a thermal mass that would work for our
`
`purpose. We also could not just buy an off-the-shelf circuit board with a thermal
`
`mass that would work for our purpose. Circuit boards are specially designed for a
`
`specific purpose within a specific product. I am not aware of any circuit boards
`
`before the ones we developed for the rainbow® sensors that were designed to have
`
`a thermal mass whose temperature could be used to reliably estimate the operating
`
`wavelengths of multiple LEDs. Eventually, however, we were able to design a
`
`thermal mass and measure the temperature of the thermal mass to reliably estimate
`
`-12-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`the operating wavelengths of the LEDs. We initially formed the thermal mass from
`
`four internal copper layers of the substrate. The copper layers were thermally
`
`connected to the LEDs and a temperature sensor. The temperature sensor measured
`
`the temperature of the thermal mass. We used the measured temperature to estimate
`
`the LED wavelengths. When we used this technique with a properly designed
`
`thermal mass, the error dropped dramatically.
`
`38.
`
`I was initially skeptical that a thermal mass could be properly designed
`
`so that the temperature of the thermal mass could be used to reliably estimate the
`
`operating wavelengths of multiple dissimilar LEDs. This technique had not been
`
`done before and had not been shown to work. Further, I knew that the measured
`
`temperature of the thermal mass would not be the LED junction temperature and
`
`would never match the LED junction temperature. I also knew that each LED would
`
`have a different junction temperature and a different wavelength variation with
`
`temperature. Therefore, a single temperature measurement could not possibly match
`
`all of the junction temperatures. In view of these issues, I thought it would be
`
`difficult, and maybe even impossible, to design a thermal mass that could produce a
`
`single temperature measurement that would correlate with all LED junction
`
`temperatures to allow estimating all LED operating wavelengths with sufficient
`
`accuracy. However, by calibrating each LED wavelength against the temperature of
`
`the thermal mass we were able to reliably estimate multiple LED wavelengths. I
`
`-13-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`was pleasantly surprised by how accurately we could predict the operating
`
`wavelength of each LED of the rainbow® sensor.
`
`39. The thermal mass does two things that allow its bulk temperature, as
`
`measured by the thermistor, to be meaningful for estimating LED wavelengths. It
`
`stabilizes the bulk temperature so that it does not fluctuate too quickly to be useful
`
`for estimating LED wavelengths. And it normalizes the temperatures at the base of
`
`the LEDs so that the variation in base temperatures is small enough that the single
`
`thermistor measurement can represent all of the LED temperatures and be used to
`
`accurately estimate LED wavelengths. The stabilization and normalization do not
`
`mean the measured bulk temperature does not change at all over time or that there is
`
`no temperature variation across the thermal mass. However, the temporal and spatial
`
`changes in temperature are small enough that the bulk temperature can be used to
`
`accurately estimate all the LED wavelengths. Both stabilization and normalization
`
`provide a bulk temperature that can be used to accurately estimate all the LED
`
`wavelengths.
`
`40. We did two things to confirm that we could reliably estimate the LED
`
`operating wavelengths using this technique. I conducted hundreds of simulations
`
`using a finite element partial differential equation solver software product called
`
`Flex PDE. This software product allowed me to virtually build and simulate sensors
`
`with specific properties and geometries, run a simulation on each sensor, and get
`
`-14-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`results that are meaningful and realistic without physically building the sensor itself
`
`to optimize performance.
`
`41. Exhibit 2103 is a true and correct copy of printed output from some of
`
`the simulations that I ran.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`.
`
`42.
`
` In conducting the simulations, I made a model of sensors with
`
`specifically defined thermal and geometric properties, simulated applying heat to the
`
`-15-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`sensor, such as by turning on one or both of the LEDs, and watched the results. The
`
`image above shows a steady-state temperature distribution for one of the simulated
`
`sensors.
`
`43.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`-16-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`
`
`
`
`
`44.
`
`I recently created a model similar to the design of a representative early
`
`rainbow® sensor using the same simulation software I used when developing the
`
`invention. Below is an output plot from a simulation of that model, with the top
`
`three curves showing the temperatures of three LEDs and the bottom curve labeled
`
`“d” showing the thermal mass temperature at the location of the thermistor on the
`
`other side of the substrate, both as a function of time.
`
`
`
`EX2135
`
`-17-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`Exhibit 2135 is a true and correct copy of the above output plot. The three LEDs
`
`have different temperatures because they have different drive currents. Exhibit 2136
`
`is a true and correct copy of a temperature-distribution map showing a steady-state
`
`temperature distribution of a cross-section of the simulated representative early
`
`rainbow® sensor.
`
`45. As shown, the bottom right corner of Exhibit 2135 includes a schematic
`
`showing the placement of 8 LEDs (shown as small squares) on the substrate.
`
`Starting at the left corner of the output plot of Exhibit 2135 shown above, when the
`
`LED is not turned on, the temperature of each of the LEDs and the thermal mass is
`
`295 Kelvin (the ambient temperature), which is equivalent to 71.33 degrees
`
`Fahrenheit. This simulation assumed the sensor is in a roughly room-temperature
`
`environment.
`
`46. As shown, upon simulating the turning on of the LED, both the LED
`
`temperatures and the thermal mass temperature begin to rise. The depicted
`
`undulation in the LED temperatures occurs because the LEDs are sequentially turned
`
`on and off, which is how pulse oximeters and other light-based physiological
`
`measurement systems work.
`
`47. As shown, the temperature of the thermal mass is stabilized and follows
`
`along with the temperatures of the LEDs, while maintaining a delta (difference in
`
`temperature) in between. The lack of undulation in the temperature of the thermal
`
`-18-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`mass indicates that the temperature is stabilized and does not experience rapid
`
`temperature changes. The maintenance of the delta between the thermal mass and
`
`the LEDs means that the thermal mass, as measured by the thermistor, does not
`
`actually match the temperatures of the LEDs. However, the thermistor-measured
`
`temperature of the thermal mass is correlated with the temperatures of the LEDs so
`
`that the temperature of the thermal mass can be used to reliably estimate the
`
`operating wavelengths of the LEDs. The thermal mass temperature provides a
`
`baseline temperature shared by the base of the LEDs, and from which all the LED
`
`junction temperatures can be estimated, so that the thermal mass temperature can be
`
`used to accurately estimate all the LED operating wavelengths. The thermal mass
`
`resists temperature change on a scale relevant to estimating the LED wavelengths.
`
`48. The actual rainbow® sensors have multiple LEDs but a single
`
`thermistor that measures the temperature of a single thermal mass. The temperatures
`
`of the LEDs differ from each other and from the temperature of the thermal mass.
`
`However, the single thermistor-measured temperature of the thermal mass correlates
`
`with the LED temperatures, and, thus, can be used to reliably estimate the operating
`
`wavelengths of the multiple LEDs.
`
`49. The temperature of the thermal mass, or the temperatures of the LEDs,
`
`do not remain constant. As the graph above shows, the LED temperatures naturally
`
`rise after the LEDs are turned on. Further, the thermal mass temperature naturally
`
`-19-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`follows all of the LED temperatures but is not equal to any of the LED temperatures
`
`as heat is distributed from the LEDs into the thermal mass. In addition, the thermal
`
`mass temperature is different from the LED temperatures and sits on top of the
`
`environment temperature, 295 Kelvin in the above simulation. And in the multi-
`
`LED rainbow® sensors, the temperatures of the LEDs are different from each other.
`
`Therefore, temperature stabilization in the ’127 patent is not about keeping the
`
`thermal mass and LED temperatures constant or equal to each other, but about
`
`maintaining a correlation between thermal mass and LED temperatures, so that the
`
`temperature of the thermal mass can be used to reliably estimate LED wavelengths.
`
`50. Exhibit 2137 is a true and correct copy of the temperature-distribution
`
`map below, which shows a steady-state temperature-distribution of the top surface
`
`of the simulated representative early rainbow® sensor substrate to which the base of
`
`the LEDs are bonded.
`
`-20-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`
`EX2137
`
`
`
`While the temperature-distribution map shows a temperature gradient across the
`
`substrate, the differences in base temperature at the location of the LEDs are small
`
`enough that the single thermistor measurement can be used to accurately estimate
`
`all LED wavelengths.
`
`51.
`
`I also recently created a model similar to the design of the current
`
`rainbow® sensors using the same simulation software I used when developing the
`
`invention. Exhibit 2138 is a true and correct copy of an output plot from a simulation
`
`-21-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`of that model, shown below with the top three curves showing the temperatures of
`
`three LEDs and the bottom curve labeled “d” showing the thermal mass temperature
`
`at the location of the thermistor, both as a function of time.
`
`EX2138
`
`
`
`As shown, the bottom right corner of Exhibit 2138 includes a schematic showing the
`
`placement of 8 LEDs (shown as small squares) on the substrate. The three LEDs
`
`have different temperatures because they have different drive currents or different
`
`locations on the substrate. Exhibit 2139 is a true and correct copy of a temperature-
`
`-22-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`distribution map showing a steady-state temperature distribution of a cross-section
`
`of the simulated representative current rainbow® sensor.
`
`52. As shown above in the plot of Exhibit 2138, starting at the left corner,
`
`when the LED is not turned on, the temperature of each of the LEDs and the thermal
`
`mass is 295 Kelvin (ambient temperature), which is equivalent to 71.33 degrees
`
`Fahrenheit. This simulation assumed the sensor is in a roughly room-temperature
`
`environment. This simulation shows that the thermal mass of the current rainbow®
`
`sensors achieves the same results as the thermal mass of the early rainbow® sensors,
`
`including that the thermal mass temperature is not equal to, but correlates with, all
`
`of the LED temperatures, so that the thermal mass temperature can be used to
`
`accurately estimate LED wavelengths. The thermal mass resists temperature change
`
`on a scale relevant to estimating LED wavelengths.
`
`53. Exhibit 2140 is a true and correct copy of the temperature-distribution
`
`map below, which shows a steady-state temperature-distribution of the top surface
`
`of the simulated representative current rainbow® sensor substrate to which the base
`
`of the LEDs are bonded.
`
`-23-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`
`EX2140
`
`
`
`While the temperature-distribution map shows a temperature gradient across the
`
`substrate, the differences in base temperature at the location of the LEDs are small
`
`enough that the single thermistor measurement can be used to accurately estimate
`
`all LED wavelengths.
`
`54.
`
`In parallel to the simulations we did during development of the
`
`invention, we also ran experiments on actual rainbow® sensors by using a
`
`spectrophotometer to detect, very accurately, the wavelength of each LED. By
`
`testing actual rainbow® sensors, we showed that, if we measured the temperature of
`
`-24-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`the thermal mass in the sensor and calibrated for each LED, we could estimate the
`
`LED operating wavelengths while the sensor was in operation typically within plus
`
`or minus 0.1 nanometer (one standard deviation).
`
`55. Specifically, we verified the accuracy of the wavelength-estimation
`
`algorithm in the rainbow® sensors by comparing actual wavelengths measured by a
`
`spectrophotometer with the wavelengths estimated by the algorithm, which relies on
`
`the thermal mass bulk temperature and LED drive current as inputs. For example,
`
`the spectrophotometer may detect an actual wavelength of 632.2 nm, while the
`
`wavelength-estimation algorithm calculates a wavelength of 632.1 nm, indicating an
`
`error of only 0.1 nm. Masimo created characterization stations, where it has tested
`
`and calibrated and continues to test and calibrate every single rainbow® sensor, both
`
`early and current, to ensure the accuracy of wavelength estimation in each sensor.
`
`These tests are done for each LED and particular drive current so the accuracy of
`
`wavelength estimation at different drive currents is verified for each LED. EX2128
`
`is a set of output plots from a characterization station. One such plot is shown below.
`
`-25-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`
`EX2128
`
`
`
`The circles on the plot correspond to actual measured and estimated wavelength
`
`values. The x-axis indicates the actual operating wavelength (centroid) of an LED
`
`as measured by a referee (a spectrophotometer). The y-axis indicates the operating
`
`wavelength (centroid) of an LED as estimated by the wavelength-estimation
`
`algorithm that uses the thermistor measurement of bulk temperature as an input. The
`
`diagonal is the identity line where ideally all measurement points should lie in order
`
`to exactly match the referee. As shown, the circles are very close to the diagonal
`
`line. This close correlation confirms that the bulk temperature closely correlates
`
`with, and allows reliable estimation of, LED operating wavelengths. Note that there
`
`-26-
`
`REDACTED
`
`MASIMO 2102
`Apple v. Masimo
`IPR2022-01299
`
`
`
`IPR2022-01299
`Apple Inc. v. Masimo Corp.
`are two clusters of data points in the graph, each cluster belonging to a different bulk
`
`temperature which is intentionally induced while calibrating the sensor. The plot
`
`shown above