Home |
Search |
Today's Posts |
|
#1
|
|||
|
|||
W2DU's Reflections III is now available from CQ Communications,Inc.
On May 24, 6:58*pm, Richard Clark wrote:
On Mon, 24 May 2010 07:06:44 -0700 (PDT), Keith Dysart wrote: I have often wondered if the manufacturer's tuning procedures had anything to do with maximizing output power transfer, or were they, in fact, optimizing some other aspect. This resolves quickly in measurement - no need to wonder unless it offers some secondary benefit of not measuring things. * An alternative is to simply examine conventional design considerations. *One can add to Plate current by throwing a lot of power into the grid. *More plate current yields more output power results, but grid lifetime plumments. One can do innumerable things to force an artificial outcome that strains to prove a distorted logic. *Examining a suite of sources, in initial conditions that are average for their application quickly reveals a common design paradigm. ****** The fundamental answer to your question is the manufacturer ultimately designs for market domination, or maximum investment return (the two don't necessarily converge). *Thus the marketplace gives us a spectrum of choice and the norm of the distribution reveals cautious design that has its eye on a value exchange expressed in money. *THAT is the only optimization you can expect = in an honest barter, you get what you pay for. 73's Richard Clark, KB7QHC You have gone to a bit higher level than I intended with my question and I agree with you conclusions at that level. But my question was more basic. When designing the filter for a PA, among other things, one uses the desired load to be applied to the tube and the disired load impedance to be supported and selects filter components to perform the desired transformation. When operating the radio, the operator has meters that measure some values, some knobs that control some component values and a procedure for adjusting these knobs. It is not at all obvious what exactly the result of performing the procedure is. Does it result in the same load being applied to the tube that was computed by the designer? There are some hints that the procedure will result in the load applied to the tube being real, but beyond that, what exactly are the circuit conditions that result? ....Keith |
#2
|
|||
|
|||
W2DU's Reflections III is now available from CQ Communications, Inc.
On Mon, 24 May 2010 16:23:26 -0700 (PDT), Keith Dysart
wrote: It is not at all obvious what exactly the result of performing the procedure is. Does it result in the same load being applied to the tube that was computed by the designer? Hi Keith, By and large, Yes. There are some hints that the procedure will result in the load applied to the tube being real, but beyond that, what exactly are the circuit conditions that result? I am a little lost on that. The load applied is the load applied (sorry for the Zen). If you mean that the load is transformed by tuning to a real R for the Plate to see, then, yes, that is operative. However, that is not the end of it. That R is seen as the loss of a now-poorer Q for the Plate tank. This is the distinction between loaded and unloaded Q. The Plate tank Q expressed in terms of loaded Q, to be effective, is quite low in comparison to its unloaded value. This value of loaded Q is roughly between 10 and 20 where the components in isolation (unloaded) could easily achieve 10 to 30 times that. The term "loaded" includes BOTH the plate and the applied load (whatever is presented to the antenna connection). The only time the unloaded Q of the Plate tank is at peak value is when it is sitting in isolation from the chassis, circuitry, and even mounts - which means it is not very useful in that configuration, except as a trophy. Many silver plate their tanks as trophies (because this rarely results in better operation). Now, let's return to my statement about what Q is "effective" AND that it measures out at roughly 10 to 20. This is straight out of Terman if you need a citation. As for explanation (also found in Terman), you have to consider that the Plate tank is the gate-keeper (as well as transformer of Z) of power. If you have too high a Q, the power is not getting THROUGH the tank as it must, and necessarily it remains in the tank (as energy, albeit). Consider further that ALL resonant circuits can be cast from series circuits to parallel circuits or parallel to series (a fact lost on some inventors of antennas). To describe the Plate tank in series terms as I do, then the plate resistance and load resistance combine in series through a simple circular path through ground. There are parallel tank designs where the resistances combine in parallel. The net result is the same insofar as Q is concerned. Consult Terman if that is confusing. No doubt others will either more clearly cite him, or add to the confusion. 73's Richard Clark, KB7QHC |
#3
|
|||
|
|||
W2DU's Reflections III is now available from CQ Communications,Inc.
Hello Keith,
Thank you for your response. I’m starting my answer to your statements by first quoting from one of your posts: “Try as I might, I have not been able to derive a mechanism to explain the observations in Reflections. But the explanations offered in Reflections require large chunks of linear circuit theory to be discarded, so this does not seem to be an appropriate path.” That you have been unable to derive a mechanism that explains the action in an RF power amplifier is evidence that you do not understand it. So let’s examine the action that follows an appropriate path that does not require any linear circuitry to be discarded. Further evidence that you do not understand it is that you used a bench power supply to describe the action, which you state has an infinite source resistance when the load exceeds 50 ohm, and zero source resistance when the load is less than 50 ohms. Unfortunately, this power supply in no way resembles an RF power amplifier, either in components or action. We’ll begin by stipulating that the ‘filter’ is a pi-network tank circuit, having a tuning capacitor at the input and a loading- adjustment capacitor at the output. We’ll also stipulate that the plate voltage and the grid bias are set to provide the desired conditions at the input of the tank circuit, which means that the desired grid voltage is that which results in the desired conduction time for the applied plate voltage. The result provides a dynamic resistance RL, which is determined by the average plate voltage VPavg and the average plate current IPavg appearing at the terminals leading to the input of the tank circuit. In other words, RL = VPavg/IPavg. To permit delivery of all available power to be delivered by the dynamic resistance RL, we want the input impedance appearing at the input of the tank circuit to be equal to RL. We’ll now go to the output of the tank circuit. We’ll assume the load to be the input of a transmission line on which there are reflections. The result is that the input to the line contains a real component R and a reactance jX. The output terminals of the tank circuit are the two terminals of the output-loading capacitor. When the line is connected to the output terminals of the tank circuit the reactance appearing at the line input is reflected into the tank circuit. This reactance is then cancelled by the tuning capacitor at the input of the tank circuit, resulting in a resonant tank circuit. We now need to adjust the output-loading capacitor to apply the correct voltage across the input of the transmission line so that the real component R appearing at the line input is reflected into the tank circuit such that the resistance RL appears at the input of the tank circuit, thus allowing all the available power to enter the tank circuit. In other words, adjusting the loading capacitor to deliver all the available power into the line also makes the output resistance of the tank circuit equal to the real component R appearing at the line input. With any other value of output resistance of the source, all the available power would not be delivered to the line. A corollary to that condition follows from the Maximum Power Transfer Theorem that for a given output resistance of the source (the tank circuit), if the load resistance is either increased for decreased from the value of the source resistance, the delivery of power will decrease. This condition also accurately describes the condition for the conjugate match. Keep in mind that the input impedance of the line is complex, or reactive, but the reactance of the correctly-adjusted tuning capacitor has introduced the correct amount of the opposite reactance to cancel the reactance appearing at the line input. Thus the line input impedance is R + jX and the output impedance of the source is R – jX, providing the conjugate match. You stated in one of your posts that the phase of the reflected wave in relation to that of the source wave results in a non-linear condition. This is totally untrue. The tuning action of the input capacitor in the tank circuit that cancels the line reactance caused by the reflection on the line in no way introduces any non-linearity in the circuit, and the condition in the vicinity of the output of the tank circuit is totally linear. Thus, circuit theorems that require linearity to be valid are completely valid when used with the RF power amplifier as described above. This applies to all RF power amplifiers, Class A, AB, B and C. I hope my comments above assist in understanding the action that occurs in RF power amplifiers. Walt Maxwell, W@DU |
#4
|
|||
|
|||
W2DU's Reflections III is now available from CQ Communications,Inc.
On May 25, 4:20 pm, walt wrote:
Hello Keith, Thank you for your response. I’m starting my answer to your statements by first quoting from one of your posts: “Try as I might, I have not been able to derive a mechanism to explain the observations in Reflections. But the explanations offered in Reflections require large chunks of linear circuit theory to be discarded, so this does not seem to be an appropriate path.” That you have been unable to derive a mechanism that explains the action in an RF power amplifier is evidence that you do not understand it. So let’s examine the action that follows an appropriate path that does not require any linear circuitry to be discarded. Further evidence that you do not understand it is that you used a bench power supply to describe the action, which you state has an infinite source resistance when the load exceeds 50 ohm, and zero source resistance when the load is less than 50 ohms. Unfortunately, this power supply in no way resembles an RF power amplifier, either in components or action. “No way” is a bit strong. The RF PA is constructed from a constant voltage source (the power supply) and a constant current controller (the tube), both aspects present in the bench supply example previously offered. A tube is often modelled as an ideal variable constant current source but unlike an ideal source, which can produce whatever voltage is needed to drive the current, the current produced by the tube is limited by the power supply voltage. Thus, assertions of linear behaviour need to be tempered by ensuring that such voltage limits are not exceeded. We’ll begin by stipulating that the ‘filter’ is a pi-network tank circuit, having a tuning capacitor at the input and a loading- adjustment capacitor at the output. We’ll also stipulate that the plate voltage and the grid bias are set to provide the desired conditions at the input of the tank circuit, which means that the desired grid voltage is that which results in the desired conduction time for the applied plate voltage. The result provides a dynamic resistance RL, which is determined by the average plate voltage VPavg and the average plate current IPavg appearing at the terminals leading to the input of the tank circuit. In other words, RL = VPavg/IPavg. To permit delivery of all available power to be delivered by the dynamic resistance RL, we want the input impedance appearing at the input of the tank circuit to be equal to RL. Most references use Vpeak and Ipeak, though they are usually related to average values with constants of proportionality so the computed RL will be the same. None-the-less, I prefer peak values since it ties better to the choices made in the design. The power that can be controlled by a control device (be it a switch, tube or transistor) is related to device limitations. So, for example, the maximum power that can be controlled by a 250V 1A switch is 250W. This occurs with a supply voltage of 250V and a load of 250 ohms. Increasing the supply voltage exceeds the switch capability as does reducing the load resistance. If the supply voltage is less than 250V then the maximum power occurs with a load that causes 1A to flow and is now a limit based on circuit choices and device capabilities. Note that these power limits have nothing to do with maximum power transfer in a linear circuit. Similarly in a tube circuit, the maximum power is limited by the supply voltage and the tube drive level (which sets the current that will flow in the tube). Maximum controlled power is then V*I and occurs with a load resistance of V/I. Increasing the load resistance reduces the power because there is insufficient voltage from the supply to drive more current through the load and reducing the load resistance reduces the power because less voltage is impressed across the load. Note that neither of these effects is related to the maximum power transfer in a linear circuit. We’ll now go to the output of the tank circuit. We’ll assume the load to be the input of a transmission line on which there are reflections. The result is that the input to the line contains a real component R and a reactance jX. The output terminals of the tank circuit are the two terminals of the output-loading capacitor. When the line is connected to the output terminals of the tank circuit the reactance appearing at the line input is reflected into the tank circuit. This reactance is then cancelled by the tuning capacitor at the input of the tank circuit, resulting in a resonant tank circuit. We now need to adjust the output-loading capacitor to apply the correct voltage across the input of the transmission line so that the real component R appearing at the line input is reflected into the tank circuit such that the resistance RL appears at the input of the tank circuit, thus allowing all the available power to enter the tank circuit. In other words, adjusting the loading capacitor to deliver all the available power into the line also makes the output resistance of the tank circuit equal to the real component R appearing at the line input. With any other value of output resistance of the source, all the available power would not be delivered to the line. A corollary to that condition follows from the Maximum Power Transfer Theorem that for a given output resistance of the source (the tank circuit), if the load resistance is either increased for decreased from the value of the source resistance, the delivery of power will decrease. This condition also accurately describes the condition for the conjugate match. While a conjugate match does result in a situation where altering the load will reduce the power transfer, it is not true that any situation where altering the load reduces the power transfer is also a conjugate match. The two examples above (bench power supply, tube in a circuit) amply demonstrate this. Keep in mind that the input impedance of the line is complex, or reactive, but the reactance of the correctly-adjusted tuning capacitor has introduced the correct amount of the opposite reactance to cancel the reactance appearing at the line input. Thus the line input impedance is R + jX and the output impedance of the source is R – jX, providing the conjugate match. This is quite in error, unless, by happenstance, RL is equal to Rp (plus the other contributors to source impedance). You stated in one of your posts that the phase of the reflected wave in relation to that of the source wave results in a non-linear condition. This is totally untrue. The tuning action of the input capacitor in the tank circuit that cancels the line reactance caused by the reflection on the line in no way introduces any non-linearity in the circuit, and the condition in the vicinity of the output of the tank circuit is totally linear. Thus, circuit theorems that require linearity to be valid are completely valid when used with the RF power amplifier as described above. This applies to all RF power amplifiers, Class A, AB, B and C. For any circuit with a conduction angle of less than 360 degrees, my simulations indicate otherwise. The reflection coefficient experienced by the reflected wave when it arrives at the amplifier output varies with the phase of the reflected wave. Since the reflection coefficient is a function of source impedance and line impedance, and the line impedance is not changing, this means that the source impedance is changing with the phase of the reflected wave. This is not a behaviour that is consistent with a linear circuit. Given the non-linearities in a circuit with a conduction angle of less than 360 degrees, this should not be a surprise. More, it would be a surprise if such a circuit did behave as a linear circuit. I hope my comments above assist in understanding the action that occurs in RF power amplifiers. Thank you. They have indeed helped clarify my thinking. We are still left with the puzzle of why the observations documented in Reflections report a reduction in power transfer when the load is changed in either direction. It seems unlikely that RL is, by happenstance, equal to Rp, which would be one explanation. It seems plausible that it is related to the behaviours associated with the examples I provided above, but I can not articulate a mechanism that satisfies. …Keith |
#5
|
|||
|
|||
W2DU's Reflections III is now available from CQ Communications, Inc.
On Wed, 26 May 2010 04:50:30 -0700 (PDT), Keith Dysart
wrote: a conjugate match does result in a situation where altering the load will reduce the power transfer .... We are still left with the puzzle of why the observations documented in Reflections report a reduction in power transfer when the load is changed in either direction. Hi Keith, Stripping away everything that you offer as objections to what is not in Walt's premise (I cannot vouch for his attempts to explain the universality of it), your statements come into conflict. If you offer you find a puzzle about measurements, then that is simply researched at the bench instead of in expansive wanderings in myriad qualifications. Do you have documented measurements under initial conditions identical to Walt's that run counter to Walt's quantitative results? I suspect not, or we would be talking about competing bench results instead. This would be a more productive and genuine debate seeking explanation for what you describe as the "puzzle." Barring quantitative evidence, anything that continues this rag-chew is a simple example of "modeling is doomed to succeed." 73's Richard Clark, KB7QHC |
#6
|
|||
|
|||
W2DU's Reflections III is now available from CQ Communications,Inc.
On May 26, 11:50*am, Richard Clark wrote:
On Wed, 26 May 2010 04:50:30 -0700 (PDT), Keith Dysart wrote: a conjugate match does result in a situation where altering the load will reduce the power transfer ... We are still left with the puzzle of why the observations documented in Reflections report a reduction in power transfer when the load is changed in either direction. Hi Keith, Stripping away everything that you offer as objections to what is not in Walt's premise (I cannot vouch for his attempts to explain the universality of it), your statements come into conflict. If you offer you find a puzzle about measurements, then that is simply researched at the bench instead of in expansive wanderings in myriad qualifications. *Do you have documented measurements under initial conditions identical to Walt's that run counter to Walt's quantitative results? I suspect not, or we would be talking about competing bench results instead. *This would be a more productive and genuine debate seeking explanation for what you describe as the "puzzle." * Barring quantitative evidence, anything that continues this rag-chew is a simple example of "modeling is doomed to succeed." 73's Richard Clark, KB7QHC Hi Keith, Sorry, OM, but you still misunderstand various aspects of RF power amp operation. First, the power supply is not the limiting factor concerning plate current. The grid drive is what determines the plate current, and thus the output power. Second, the tank circuit is an energy storage device that isolates the non-linear input from the linear output. That the output is linear is because the voltage and current are in phase at the output of the tank circuit. The effect of the energy storage of the tank results in the tank becoming the source of the energy appearing at the output. Third, the action of plate resistance Rp occurs only in the formation of RL, and has no further effect on any action downstream of the input of the tank circuit. Thus, it has no bearing on the development of the conjugate match that occurs at the junction of the tank output and the input of the transmission line. Fourth, as I said earlier, the the action of the bench power supply that you presented in no way models the action of the RF power amplifier. Furthermore, you are incorrect when you say that when varying the load in either direction causing the power deliver to decrease there is no conjugate match. In saying what you did violates the theorem of Maximum Transfer of Power. Fifth, as I stated earlier, when the reactance appearing at the input of the load (the transmission line with reflections) is canceled by the opposite reactance introduced by the pi-network tuning capacitor, the output impedance of the source (the tank circuit) is the conjugate of the line-input impedance. If you cannot accept this as fact you have a problem. Sixth, your understanding of the effect of the reflected wave on the source wave is flawed. The non-linearity of the plate current when the conduction time is less than 360° has no relation to the action downstream of the input to the tank circuit, because from that point on the voltage current relationship is linear. If you cannot accept this as fact you have still another problem. Seventh, your belief that because there is a conjugate match at the output of the tank there must be a conjugate match at the input of the tank is also not true. The effect of the energy storage in the tank isolates the non-linearity af the input from the linear operation at the output, permitting a conjugate match at the output, while not allowing it to occur at the input. These seven comments are born out (proven) by the results of many measurements I made using laboratory grade instruments, HP and General Radio. If you check my record as a professional electrical engineer regarding the measurements I've made that led to successful hardware flying on various Earth-orbiting platforms, you must accept the validity of the measurements I made on RF power amplifiers that prove my position. As I said earlier, no one but you has considered my position on this subject incorrect. Therefore, if you cannot agree with the comments I made above, but still consider my statements in Reflections flawed, then there is no point in my making any further comments. I hope someday you'll finally understand what's really happening within the RF amplifier. Walt Maxwell, W2DU |
#7
|
|||
|
|||
W2DU's Reflections III is now available from CQ Communications,Inc.
On May 26, 1:42 pm, walt wrote:
Hi Keith, Sorry, OM, but you still misunderstand various aspects of RF power amp operation. First, the power supply is not the limiting factor concerning plate current. The grid drive is what determines the plate current, and thus the output power. Of course the grid drive is one of the factors which controls the current flowing in the load. But the power supply is also one of the limiting factors. Reducing the power supply voltage below that which is necessary to cause the desired current to flow in the load will reduce the power output. Similarly, increasing the load resistance will eventually raise it to the point where the voltage is no longer adequate to cause the desired current to flow. Second, the tank circuit is an energy storage device that isolates the non-linear input from the linear output. That the output is linear is because the voltage and current are in phase at the output of the tank circuit. Can one not have a linear circuit where the current and voltage are not in phase? Also, if one loads any tank circuit with a resistance, the output current and voltage will be in phase and if it is loaded with a reactance, they won’t be. The effect of the energy storage of the tank results in the tank becoming the source of the energy appearing at the output. Yes, but that does not make the output independent of the input. Third, the action of plate resistance Rp occurs only in the formation of RL, and has no further effect on any action downstream of the input of the tank circuit. Thus, it has no bearing on the development of the conjugate match that occurs at the junction of the tank output and the input of the transmission line. I do not understand what is being said here. Fourth, as I said earlier, the the action of the bench power supply that you presented in no way models the action of the RF power amplifier. Furthermore, you are incorrect when you say that when varying the load in either direction causing the power deliver to decrease there is no conjugate match. In saying what you did violates the theorem of Maximum Transfer of Power. The definition I use for conjugate match is one where the source impedance is the complex conjugate of the load impedance. When this situation occurs between linear networks, maximum power is transferred between the networks. None-the-less, just because maximum power is being transferred between two networks does not mean they are complex conjugates of each other. This is demonstrated with the non-linear behaviour of the bench power supply example. Maximum power is transferred but the source and load impedance are not complex conjugates. Fifth, as I stated earlier, when the reactance appearing at the input of the load (the transmission line with reflections) is canceled by the opposite reactance introduced by the pi-network tuning capacitor, the output impedance of the source (the tank circuit) is the conjugate of the line-input impedance. If you cannot accept this as fact you have a problem. Perhaps I am not computing the impedances correctly. Let us see if I have done so for the following example. Consider a generator constructed of current source in parallel with a resistor, driving a PI network, connected to a load. generator filter load 6.945uH +-------+------- ----+---/\/\/\/---+---- ---+ | | | |1.398 | +---+ \ | | nF \ 3.75 | I | /8000 ----- ----- / 50 ohm MHz | | \ ----- ----- \ +---+ / |295.5 | / | | | pF | | +-------+------- ----+-------------+---- ---+ Looking into the input of the filter, then impedance is 1500 ohms. This is the load applied to the generator and is computed by applying the rules for series and parallel components to the 50 ohm load, and the two capacitors and inductor in the PI network. It is, I hope, generally accepted that the generator will have an output impedance of 8000 ohms. The output impedance of the filter is computed by applying the rules for series and parallel components to the 8000 ohm generator impedance and the 3 components in the filter. The result is 58.00 /_ 68.60 ohms. Note that the component values were taken from a PA design where the desired load for the tube was 1500 ohms. And 8000 is not an unreasonable slope for the plate E-I curve of a tube. This has not resulted in a conjugate match. Sixth, your understanding of the effect of the reflected wave on the source wave is flawed. The non-linearity of the plate current when the conduction time is less than 360° has no relation to the action downstream of the input to the tank circuit, because from that point on the voltage current relationship is linear. If you cannot accept this as fact you have still another problem. It is, perhaps, this claim of isolation that is most strange. It seems quite at odds with the rules for connected networks. Seventh, your belief that because there is a conjugate match at the output of the tank there must be a conjugate match at the input of the tank is also not true. The effect of the energy storage in the tank isolates the non-linearity af the input from the linear operation at the output, permitting a conjugate match at the output, while not allowing it to occur at the input. It was my understanding that in a sequence of connected linear networks, if any connection exhibited a conjugate match, then they all were conjugately matched. Is this not correct? Are you saying that if a conjugate match is present between the line and the antenna, it might not be present between the transmitter and the line? These seven comments are born out (proven) by the results of many measurements I made using laboratory grade instruments, HP and General Radio. If you check my record as a professional electrical engineer regarding the measurements I've made that led to successful hardware flying on various Earth-orbiting platforms, you must accept the validity of the measurements I made on RF power amplifiers that prove my position. I quite believe your measurements. It is the conclusion that they prove a conjugate match that I find impossible to accept. Both because there are other situations that can lead to power behaviours that may appear similar to the power behaviour of a conjugate match and the method proposed for computing source impedance is quite at odds with linear theory. But the quality of the measurements suggest it is worthwhile to explore other explanations. ....Keith |
#8
|
|||
|
|||
W2DU's Reflections III is now available from CQ Communications,Inc.
On May 26, 8:53*pm, Keith Dysart wrote:
It was my understanding that in a sequence of connected linear networks, if any connection exhibited a conjugate match, then they all were conjugately matched. Is this not correct? The theorem requires linear *lossless* networks which do not exist in reality, i.e. networks containing only reactances. Therefore an *ideal* system-wide conjugate match cannot exist in reality just as a lossless transmission line cannot exist in reality. In low-loss systems, we can only achieve a system-wide near-conjugate match with an ideal conjugate match existing at one point, e.g. the Z0-match point where reflected energy flowing toward the source is eliminated. Are you saying that if a conjugate match is present between the line and the antenna, it might not be present between the transmitter and the line? Yes, speaking for me, in the real world, it is easy to prove that the system-wide impedance looking in one direction is not always exactly the conjugate of the impedance looking in the other direction. Thus the "maximum power transfer" assertion has to be modified to "maximum *available power* transfer". In the real world, ohmic and dielectric losses reduce the power available to be delivered to the load. It's easy to see. Let's say we have a completely flat 50 ohm system; 50 ohm source, 50 ohm coaxial feedline, and 50 ohm antenna. Now assume we install an antenna tuner between the source and the feedline that exhibits some series impedance and we adjust the tuner such that the source sees 50 ohms. At the output of the tuner looking toward the antenna, we will see 50 ohms. Looking back through the tuner toward the source, we will see the tuner impedance in series with 50 ohms. That proves it is not an *ideal* (lossless) conjugate match although it may be considered to be a near-conjugate match, as close as we can come in the real world. What I don't know is how close a real-world conjugate match has to be to an ideal lossless conjugate to be called a "conjugate match". A purist might argue that an ideal conjugate match cannot exist in reality. A realist might argue that if we are within 10% of an ideal conjugate match, then it is a real-world conjugate match, by definition. Note that I am not speaking for Walt here, just for myself. -- 73, Cecil, w5dxp.com |
Reply |
Thread Tools | Search this Thread |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
Transmission Line Reflections | Antenna | |||
Reflections on rrap | Policy | |||
Reflections on rrap | Policy | |||
Reflections on rrap | Antenna | |||
Reflections on rrap | Policy |