Analysis and Opinions Relating to US Patents 7184723 and 7327803

I have been retained by Mike Farmwald of SkyMoon Ventures, LLP of Santa Clara, CA to review US Patents 7,184,723 and 7,327,803 and express my opinions in regards to the technical feasibility of the inventions described therein. Both patents are authored by David F. Sorrels, Gregory F. Rawlins, and Michael W. Rawlins, and are assigned to ParkerVision, Inc., of Jacksonville, FL. The priority date on both patents is October 22, 2004 through provisional application number 60/620,972. The validity of the claims included in those patents is a matter of law, and I draw no specific conclusions in that regard. However, I do provide my opinions on the novelty of the inventions from a technical standpoint. In this regard, I saw no need to construe the meaning of claim terms outside their plain and ordinary meaning as used by those skilled in the art. The level of discussion assumes some general background in radio frequency (RF) electronics, though my intent is to make it accessible to a larger audience. My opinions rest solely on my review of the patents, a review or prior knowledge of some of the cited references, and my knowledge gained through my studies and experiences in RF power amplifier design and linearization over the past 20 years. I reserve the right to modify my opinion when presented materials of which I am unaware at the time of this writing.

Overview and Background

The ‘723 and ‘803 share a common specification, which I understand is shared by other patents issued to ParkerVision. In general, the purported invention described in the specification is an out-phasing technique that uses multiple power amplifiers (PAs) which are combined to achieve improved efficiency for signals with non-constant envelopes. Figure 5 from the common specification is shown below. It shows a bank of four vector modulators driving individual PAs which are summed together. The essence of the invention described relies on a well know technique in PA design called “out-phasing.” In summary, out-phasing is a process whereby a non-constant envelope, or amplitude modulated (AM) signal is decomposed into a set of constant amplitude, phase modulated (PM) signals which are individually amplified and recombined to restore the original envelope. The improvement in efficiency over conventional linear amplification schemes (class-A, class-AB, etc.) is afforded by the use of highly efficient, nonlinear PA modes (class-C/E/F, etc.) that may be used on PM signals without distorting the phase information. Thus, the phase information is preserved in each signal as the carrier amplitude is increased.

The ‘723 patent focuses on out-phasing techniques. Various architectures exist to achieve out-phasing. They differ mainly in the way the decomposed signals are recombined after being amplified. The most well known are “linear amplification using nonlinear components” (LINC), and that developed by Chireix. The LINC architecture combines the decomposed signals using a combiner that isolates the separate PAs. Thus, the signals presented by one PA to the combiner are not seen by the other PAs. In that way, the load of each PA does not vary. A number of methods to achieve this isolation using microwave circuits have been developed over the past half century. However, the problem with such isolation is that the combiner must be “balanced” in such a way that any power that is cancelled (e.g. two signals of equal amplitude and opposite phase) must necessarily be absorbed in a resistive isolation load. This power is not transmitted, and therefore wasted, drastically reducing the efficiency of the overall transmitter system when operating at power levels below the maximum. Various methods have been proposed to “recycle” this lost power. However, I know of no commercially successful LINC implementations in use in modern commercial wireless systems.

Unlike LINC, the Chireix technique does not isolate one PA from another. Instead, each PA presents an equivalent impedance to the others by summing the currents. The impedances combine in parallel as seen by any one PA, thereby shifting its “load line.” A load line represents a relationship between the current through a transistor and the voltage across it. For any device operating at some bias point, an optimum load line exists for which the maximum power may be extracted out of the device. If the impedance presented to a device is suboptimal, less power is transmitted to the load. In this way the amplitude modulation imposed on the original signal may be restored. The advantage of the Chireix technique over LINC is that power is not wasted in shifting the load line. Hence the efficiency of the transmitter is preserved at low output signal levels. Chireix PAs have been commercially attempted by at least one company known by me, Icefyre, Inc., which was acquired some years ago. However, the use of this technique is not widespread due to a number of practical issues involved in its implementation. One such limitation is the “load-pulling” effect caused by the out-phased currents. This effect creates an apparent reactive load that can cause instabilities and signal distortion. The references cited in both patents show that the inventors had knowledge of the original disclosures of LINC (see D.C. Cox), and Chireix methods, as well as many subsequent studies in these areas. My analysis of the purported invention disclosed in the claims of the ‘723 patent will focus on the power combining methods claimed or discussed as preferred embodiments.

The ‘803 patent also discusses dynamic biasing of the PA to optimize efficiency. Such techniques are well known among those skilled in the art. The concept behind dynamic biasing involves reducing the DC power when low RF power levels are required. Many techniques have been proposed over the years, and many of which are cited in the ‘803 references. Some practical  issues arise in implementing bias control. These involve limitations in the bandwidth used to control the voltage or current, and reductions in overall efficiency due to the power consumed by the control circuits. However, it is my understanding that dynamic biasing has been implemented in some manner in a number of handset transmitters that are in mass production today. My analysis of the ‘803 patent will focus on the combination of out-phasing and dynamic biasing.

Analysis of the Claimed Inventions

            The ‘723 patent contains 92 dependent claims, which incorporate either claim 1 or claim 41. Claim 1 is a methods claim which is mirrored by the apparatus claim 41. I believe that one of ordinary skill in the art would interpret claim 1 as comprising of a frequency reference signal that is manipulated by a set of control signals to produce a set of constant envelope, phase modulated signals. These signals are then independently amplified by transistors whose outputs are current summed at a single node. Because current summing affords no isolation between transistors, it is then clear that the claimed invention has a closer similarity to the Chireix technique than LINC. It is well known in the Chireix technique that any phase difference (other than 180°) between currents summed in the common output node creates the effect of a reactive load. Electrical engineers are taught in their first few years of undergraduate studies that reactive loads dissipate no real power, but only “apparent” power, which thereby reduces the overall efficiency of power transfer into the load. The Chireix technique overcomes this loss by connecting a compensating reactance to offset the reactance created by the out-phased currents. Claims 1 and 41 do not mention the use of such a compensating reactance, though dependent claim 9 mentions “harmonically shaping” the output signal, which implies a frequency dependent (i.e. reactive) load. In this way, the invention described by claim 9, which incorporates claim 1 might be interpreted by one of ordinary skill in the art as some sort of improved Chireix architecture. However, when looking for support in the 64 column common specification, I found no embodiments that described how such harmonic shaping compensates for the load-pulling effects of the out-phased currents. I did find a description of element 780, which is labeled a “pull-up impedance” in Figures 7A/B, and 8A/B/C/D. Figure 12/A/B, 13A/B show the same pull-up impedance (element 1288), along with an additional “impedance matching network (element 1290). Figures 17A/B and 18A/B also show the pull-up and matching networks as elements 1778 and 1780, respectively. Lastly, Figures 41 and 42 show a “Z_p” and a “Z_m” as elements 3610 and 3720, respectively, which I interpret as denoting the pull-up and matching impedances. When reviewing the preferred embodiments that described these elements, the pull-up impedance was described only as an impedance “coupled between a vector power amplifier and a power supply.” I believe that one of ordinary skill in the art would interpret this as simply a bias feed, with no particular intent to perform load-pulling compensation. Likewise, the “impedance matching network” is described as exactly that, which is “coupled to the output of a vector power amplifier 1200,” and described further as simply an ordinary matching circuit to achieve optimum power transfer of the PAs to the load. Upon my review of sections 3.4.5 on “Waveform distortion compensation” and section 3.6 “Harmonic Control,” I found no mention of load-pulling compensation, only discussions of generally known harmonic terminations to maximize fundamental signal power transfer. Hence, it was not apparent to me that the specification disclosed the best method for overcoming one of the major disadvantages of the Chireix architecture.

            The ‘803 patent contains one independent methods claim (claim 1), which is mirrored by a means plus function claim (claim 38), and an apparatus claim (claim 73). The other 110 claims are dependent on one of the aforementioned claims. The independent claims generally describe a method of controlling PA output power through dynamic biasing. One of ordinary skill in the art would not recognize this as being novel in and of itself. However, upon my first reading, I felt that the combination of dynamic biasing and the out-phasing techniques described by the common specification and the ‘723 patent might be novel. Dynamic biasing may be accomplished in three different ways: 1) Controlling the bias current in a transistor proportionally to the desired power, 2) controlling the supply voltage applied to a transistor proportionally to the desired power, and 3) a combination of (1) and (2). The ‘803 is specifically focused on implementation (1), as I have described above. The relevant sections of the specification are 3.5.2 “Output stage current control,” and 3.7 “Power control.” Upon reviewing those sections, I found reference to an “autobias control,” which is performed by elements 714 and 716 in Figure 7, and in a more detailed fashion in Figure 33. The operation of the autobias circuit is illustrated in detail in Figures 38-43. Both Figure 38 and 43 show bipolar transistors which are supplied a base current I_ref that controls the collector current through the transistor current gain b. Such circuits are well known among those skilled in the art. The only novelty would be in the formation of the current control transfer function, several examples of which are illustrated in Figure 41. It is clear from the description of the preferred embodiment that this transfer function is calculated from the baseband I/Q signals, though the relationship between this complex envelope signal and the actual control current applied to the I_ref terminal in Figure 38 and 43 is not discussed in any detail. Therefore, the best mode for this autobias current control is not fully disclosed.

            After a thorough review of the patent and reference materials, it is not clear to me that the independent claims disclose any inventions that were not already known to those of ordinary skill in the art. Furthermore, the specification did not provide sufficient support for overcoming implementation issues well known in these architectures. Though I did not exhaustively review the limitations included in the many dependent claims, I feel those inventions are of questionable value if the claims they depend on are fundamentally flawed from a technical point of view.