Case 6:12-cv-00799-JRG Document 120-4 Filed 03/06/14 Page 1 of 8 PageID #: 3444
`
`Exhibit D
`
`
`
`
`
`

`
`Case 6:12-cv-00799-JRG Document 120-4 Filed 03/06/14 Page 2 of 8 PageID #: 3445
`
`C.A. No. 6:12-cv-00799-LED
`
`JURY TRIAL DEMANDED
`
`IN THE UNITED STATES DISTRICT COURT
`FOR THE EASTERN DISTRICT OF TEXAS
`TYLER DIVISION
`
`§
`INVENSYS SYSTEMS, INC.,
`§
`
`
`§
`
`Plaintiff,
`§
`v.
`
`§
`
`
`§
`EMERSON ELECTRIC CO. and
`§
`MICRO MOTION INC., USA,
`§
`
`
`§
`
`Defendants.
`§
`and
`
`§
`
`
`§
`MICRO MOTION INC., USA,
`§
`
`
`§
`
`Counterclaim-Plaintiff,
`§
`v.
`
`§
`
`
`§
`INVENSYS SYSTEMS, INC.,
`§
`
`
`
`Counterclaim-Defendant. §
`
`
`
`TECHNICAL TUTORIAL SCRIPT FOR COUNTERCLAIMANTS’ PATENTS
`
`This portion of Invensys’ tutorial will cover the background of the technology that is the
`
`subject of the patents asserted by Micro Motion and Emerson in this case, specifically U.S. Pat.
`Nos. 5,555,190 and 6,505,131.
`
`
`The technology of these patents also relates to Coriolis flowmeters in general.
`
`
`
`
`
`However, these patents are more specifically related to the measurement function, rather
`than the drive function, of the control and measurement system. They disclose various signal
`processing techniques for processing the flowtube sensor signals in order to determine, for
`example, the mass flow rate through the flowtubes.
`
`
`
`As explained previously, the sensor signals represent information about the position of
`the flowtubes as they vibrate under the drive signal. For ease of illustration, we usually show
`these signals in their ideal form as sine waves.
`
`
`
`However, in practice, the sensor signals will generally be corrupted by noise arising from
`a variety of sources in the environment.
`
`1
`
`
`

`
`Case 6:12-cv-00799-JRG Document 120-4 Filed 03/06/14 Page 3 of 8 PageID #: 3446
`
`
`
`These noise components present in the corrupted sensor signal are undesirable because
`
`they lead to measurement inaccuracies. As you can see, in the presence of noise, the phase shift
`between these two signals cannot be accurately determined. Signal processing techniques are
`used to enhance, or remove, the noise from these noisy signals so that useful information can be
`recovered.
`
`
`
`For Coriolis flowmeter sensor signals, this is possible because the true signal component
`of the sensor signal is at a single well-defined frequency, which is the oscillation frequency of
`the flowtubes. However, the contributing noise components that corrupt the sensor signal are
`present at a wide band of frequencies. This is an important point, because even though the
`additive effect of all the noise, across all frequencies, can be quite significant in aggregate, the
`noise component within the narrow range of frequencies where the true signal resides is actually
`quite small relative to the true signal from the sensor.
`
`
`
`We can use different colors to represent different frequencies as an analogy to illustrate
`this point. Suppose our useful signal is represented by the letter “a” and this information is at a
`frequency represented by the red color. Now, imagine that we have noise at many other
`frequencies added to this signal, with the result that the “a” in our signal becomes obscured. This
`broadband noise is illustrated by this multi-colored graphic, representing many different
`frequencies.
`
`
`
`We will refer to this combination of the original “a” with the colorful noise added as a
`noisy signal. As you can see, the addition of noise sources at all these colors has obscured the
`underlying signal of interest, the letter “a.” It becomes more difficult to see.
`
`
`
`Of course, our eyes are able to distinguish different colors, so we can still see the solid
`red “a” faintly in the background. However, if we imagine this was a black and white image,
`then it would be much more difficult in this presentation to see that there is a signal hiding in the
`graphic.
`
`
`
`This is the precise concept that is important for the signal processing. If we know that the
`signal is supposed to be at a certain frequency, such as red in this illustration, then by filtering
`the red out from all the other colors at each pixel, we are able to separate the useful information
`from the noise.
`
`
`
`Collectively, all the unwanted noisy colors drown out the useful information. However, if
`we separate out and isolate the frequency of interest, the noise level within that frequency is
`relatively small, compared to the signal level.
`
`
`
`
`
`U.S. Patent No. 5,555,190, or the ‘190 Patent, talks about processing noisy Coriolis
`flowmeter signals using adaptive notch filters implemented by digital signal processors.
`
`The ‘190 patent discloses two embodiments that take the digitized signals from the
`sensors and generate enhanced versions of the signals for subsequent processing.
`
`
`2
`
`
`

`
`Case 6:12-cv-00799-JRG Document 120-4 Filed 03/06/14 Page 4 of 8 PageID #: 3447
`
`Before we explain what a digital adaptive notch filter is, and how the ‘190 patent uses it
`
`for enhancing Coriolis flowmeter signals, let’s first talk about a simple digital notch filter. A
`digital notch filter is used to filter out, or block, a small range of frequency components from
`passing through the filter. However, any frequency components outside of that filter range can
`pass through. So it is a device that can distinguish between different frequencies, just like how
`our eyes can distinguish between colors.
`
`
`
`Now imagine a digital signal with many different frequency components, which we will
`represent here in purple, red, orange, green, and blue, using our color analogy. If we design a
`digital notch filter that blocks only the “red components”, then the output will be the same
`original signal, only with all the red colors removed.
`
`
`Let’s see what happens if we use this notch filter on the noisy red letter “a” signal.
`
`
`
`
`
`As you can see, the noisy signal comes into the input of the notch filter, carrying all the
`different colors. As the noisy signal hits the filter, all the colors are allowed to pass through,
`except for the red colors, including our signal, that are blocked.
`
`
`
`As you can see, the output of a digital notch filter does not actually provide the desired
`signal. Instead, the notch filter has allowed all of the noise components to pass through, but
`blocked, or removed, the desirable portion of the signal. Therefore, this is only the first step in
`the process for getting an enhanced version of the signal.
`
`
`
`The ‘190 patent teaches that to recover the useful information using digital notch filters,
`we must perform a second step, which is to subtract this filtered output from the original signal.
`By performing this filtering and then subtracting, the noise components can be removed from the
`noisy signal.
`
`
`There are two additional characteristics of a digital notch filter that are important as
`
`contemplated in the ‘190 patent for use in Coriolis flowmeters. Together, they determine the
`specific frequencies that will be blocked by the filter; these are adaptive in the ‘190 patent.
`
`
`
`The first characteristic relates to the width of the filter, meaning how wide a range of
`frequencies it can block. Here is an example of a wide notch filter blocking not just one color,
`but three: orange, red, and green.
`
`
`
`By comparison, the notch filter we saw previously is a much narrower filter, since it only
`blocks one color.
`
`
`
`These two examples illustrate the concept of width, which characterizes the range of
`frequencies that the notch can block. Note that both of these notch filters will block the red
`frequencies, but the wider notch blocks additional frequencies surrounding the red frequency.
`
`
`
`This leads us to the second characteristic, referred to as the center frequency. As the
`name suggests, this frequency corresponds to the center of the range of frequencies that will be
`
`3
`
`
`

`
`Case 6:12-cv-00799-JRG Document 120-4 Filed 03/06/14 Page 5 of 8 PageID #: 3448
`
`blocked by the notch filter. It marks the center of the range of targeted frequencies that will be
`blocked, so the notch is always most effective at blocking the center frequency.
`
`
`
`Here is an illustration of a notch filter that has a different center frequency.
`
`
`
`
`
`Recall our red letter “a” masked by the multi-colored noise dots.
`
`Suppose we implemented a notch filter at the incorrect center frequency for this signal,
`such as the orange notch from the previous slide.
`
`
`
`Here, all the orange get blocked, but our red signal passes through along with all the
`noise at the other colors. As you can see, if we implement a filter using the wrong center
`frequency, we will not be able to separate the signal from the noise. When we perform the
`subtraction, we will subtract out the signal as well, leaving us with just the orange noise.
`
`
`
`Therefore, it is important that the notch filter is correctly centered around the flowtube
`vibration frequency. Unfortunately, the flowtube vibration frequency is not known in advance,
`and it will also vary during operation as the density of the fluid flowing through the tubes change
`as well.
`
`
`
`If we implement a wide notch filter, then we increase our chances that the correct
`flowtube vibration frequency lies within this wider range.
`
`
`
`However, a wider notch is less effective at separating out the unwanted noise from the
`relevant signal. As illustrated, all the green and orange noise components are still tied up with the
`signal. A wide notch does not provide adequate separation compared to a narrow notch to give
`the desired enhancement.
`
`The ‘190 patent combines elements of both characteristics by using “adaptive” notch
`
`filters. The ‘190 digital notch filters are “adaptive” because they are able to change the width
`and/or the center frequency over time.
`
`
`
`In a first disclosed embodiment, only one adaptive digital notch filter is used for each
`sensor signal. The notch starts out wide, which ensures blocking of the correct flowtube
`frequency, while providing marginal improvement to the signal. Even though the correct
`frequency may be unknown, the range of possible frequencies is known, given the geometry of
`the flowtubes. The ‘190 patent teaches to adapt the notch filter width by making it narrower and
`
`4
`
`
`
`
`
`
`As shown, this notch filter is a narrow notch that blocks only the orange frequency.
`
`By specifying the center frequency, as well as its width, we will know how the notch
`filter will process an incoming signal.
`
`
`
`We will now explain in the next slides what the ‘190 patent has to say about these notch
`filters relative to Coriolis flowmeters, and why the ‘190 patent uses adaptive filters.
`
`

`
`Case 6:12-cv-00799-JRG Document 120-4 Filed 03/06/14 Page 6 of 8 PageID #: 3449
`
`narrower, and also to adapt the center frequency, to eventually block only the correct frequency
`matching the flowtube vibrations.
`
`
`
`The noise passing through the adaptive notch filter is subtracted from the original signal
`as explained previously.
`
`
`
`In a second disclosed embodiment, two adaptive digital notch filters are used for each
`sensor signal, one followed by the other. This is shown in figure 13 of the ‘190 patent. Unlike the
`first embodiment, these notch filters can only adapt the center frequency, but the notch widths
`are fixed. This embodiment uses a first adaptive filter that has a wide notch, followed by a
`second adaptive filter with a narrow notch. Of course, in this embodiment of the ‘190 patent,
`they still need to adapt the center frequency of both notch filters, in order to block the correct
`frequency. However, since the second notch is already a narrow notch, once it locks on to the
`correct center frequency, the filtering performance will match that in the first embodiment
`without needing to adapt the width of the notches.
`
`
`Once the noisy components have been subtracted from the sensor signals, the ‘190 patent
`
`teaches to determine the phase value for each sensor signal, and to determine the phase
`difference between the two.
`
`
`
`This figure in the ‘190 patent also shows the next stage in signal processing, where the
`mass flow is determined.
`
`
`
`Micro Motion’s other asserted patent, U.S. Patent No. 6,505,131, or the ‘131 patent,
`relates to a complementary aspect of the digital signal processing on the measurement side for
`Coriolis flowmeters.
`
`
`
`In Micro Motion’s own ‘190 patent that we just looked at, one problem it did not address
`was how to make the adaptive signal processing components work across different Coriolis
`flowtube geometries.
`
`
`
`Different flowtube designs in general produce widely varying ranges of vibration
`frequencies, and this full range of frequencies across designs is beyond what the flow meter
`electronics can handle. For example, larger bent tube flowmeters typically operate at lower
`frequencies, whereas straight tube flowmeters operate in a much higher frequency range. As a
`practical matter, what this means is that if they implemented the adaptive notch filters of the ‘190
`patent as the measurement electronics for one particular flowtube design, the same electronics
`may not be operable for a different flowtube design. They would need to implement new filters
`and build out new electronics for different flowtube designs.
`
`
`
`This is the problem that the ‘131 patent attempts to overcome. The ‘131 patent talks
`about a digital signal processor that is adaptable across different flowmeter designs, permitting a
`wider range of operating frequencies.
`
`
`
`
`5
`
`
`

`
`Case 6:12-cv-00799-JRG Document 120-4 Filed 03/06/14 Page 7 of 8 PageID #: 3450
`
`Suppose we have an existing measurement electronics module, for example, adaptive
`
`notch filters that are configured to work for bent tube flowmeters with typical vibration
`frequencies from about 20-200 Hertz, or 20-200 cycles per second.
`
`
`
`If we try and use these components with another bent tube flowmeter, so long as the
`possible range of flow tube vibrations coincide, there does not appear to be any problems.
`
`
`However, if we tried to use the same measurement electronics with a straight tube flow
`
`meter, then as designed, the electronics will not work correctly.
`
`
`
`This is because the vibration frequencies are typically at 300-800 Hz, which is higher
`than the range of frequencies where these electronics are configured for proper operation. So this
`won’t work, and the signals cannot be processed.
`
`
`In order to use the same back end signal processing components across different
`
`flowmeters possibly vibrating at a wide range of frequencies, the ‘131 patent describes a method
`that can translate or shift the frequency content of the sensor signals from its original frequency
`to a new frequency. This new frequency is within the range that the electronics can handle. To
`illustrate, we still have the straight flowtubes vibrating at the higher frequency as before.
`However, instead of sending these signals directly to the back end electronics, the signals first go
`through the frequency translation process.
`
`
`
`The result of the translation is a signal at a lower frequency that is now within the range
`of what the electronics can accept and process.
`
`
`
`The ability to shift the signal frequency to a new frequency, independent of the actual
`vibration frequency allows the same signal filtering and measurement electronics to be used for a
`wide variety of flowtube designs. With the convenience that the frequency of the translated
`signal can be arbitrarily chosen over a wide range, it is natural to translate the signals to the
`center frequency of the subsequent notch filters, such as those described in Micro Motion’s ‘190
`patent.
`
`
`In describing how to perform this translation, the ‘131 patent makes use of many
`
`mathematical concepts and complicated equations. There is one mathematical concept that is
`important to explain in this tutorial, and it is known as a dot product.
`
`
`
`A dot product is a special type of multiplication that involves two sequences of numbers.
`We will show the steps you need to take for calculating the dot product. Here’s an example.
`Suppose we have two sequences, we’ll call them sequence A and sequence B. Sequence A has 4
`numbers, all shown in red. Since these numbers can be any value, we will just refer to them as
`a1, a2, a3, and a4. Sequence B also has 4 numbers, shown in blue and we will label them as b1,
`b2, b3 and b4.
`
`
`
`The dot product of sequence A with sequence B is the sum of the products of the
`corresponding components. What that means is you take the first component of sequence A, that
`is a1, and then the first component of sequence B, that is b1, and you multiply them together.
`
`6
`
`
`

`
`Case 6:12-cv-00799-JRG Document 120-4 Filed 03/06/14 Page 8 of 8 PageID #: 3451
`
`Then you take the second components of each, so a2 and b2, and multiply them together,
`
`and so on. To finish calculating the dot product, you then add up all 4 numbers to arrive at the
`final answer. That sum is the dot product.
`
`
`
`You can of course, calculate dot products of much longer sequences as well. You just
`keep multiplying the corresponding components, and add them all up. However, one important
`observation here is that the dot product must be performed on two equal length sequences. In
`other words, the sequences must have the same number of components. Otherwise, one sequence
`will have extra components with no corresponding partner, so the dot product cannot be
`calculated.
`
`The other important observation is that the result of the dot product of two equal length
`
`sequences of numbers must be a single number, because you need to add together all the
`products of the corresponding components.
`
`
`
`
`
`
`EAST\72690749
`
`7
`
`
`This concludes Invensys’ tutorial regarding the patents at issue in this case.