|
|
Thread Tools | Display Modes |
#1
|
|||
|
|||
Still lower noise radio astronomy (was: low-noise amplifiers for radio astronomy )
"John (Liberty) Bell" wrote in message
... George Dishman wrote: "John (Liberty) Bell" wrote in message ... wrote: "John (Liberty) Bell" wrote in message ... wrote: snip The DC input resistance of the HEMT is high but I'm not sure of the resistive component of the reactance for a microwave signal. Is there a resistance in series with the capacitance that is unmeasurable at DC? Good point. The capacitance is presumably from the gate to the source, and from the gate to the drain. So that resistance/impedance should be that of any source resistor/impedance and drain resistor/impedance in parallel. It's more complex but you need to look at the equivalent circuit I mentioned below for an example. However, you have the idea. This is starting to sound more like home territory to me, where a low impedance source is simply plugged direct into a high input impedance amplifier, with no resultant significant losses to worry about. Not quite. If you do that at RF, the impedance change acts like a mirror and reflects all the input power back to the source. I work in HF and the key is that we must _always_ match the impedance. OK. It seems that this is usually achieved by placing an additional load from the input to ground, before the L and C we were discussing. I don't like the idea of "wasting" that extra potential power, but if it is necessary, then it has to be done. That's one way. I'm more used to bipolars where the input impedance is intrinsically defined by the junction and operating point. However, this does bring me to an additional possible question on impedance matching which is probably easiest discussed abstractly in the context of operational amplifiers. Here the gain is defined as the feedback impedance divided by the source impedance (which is the whole point of op-amps) and the inverting input acts as a 'virtual earth'. Do you know exactly how this would affect the already discussed input impedance matching, when the source is fed into that inverting input via said compensating inductor? For an op-amp, the input is fed to the virtual earth via a resistor. That would define the input impedance. Any stray capacitance could be nulled by an input inductor leaving a pure resistance at the frequency of interest. (I am ignoring for the moment that an op-amp has both inverting and non inverting input transistors which will obviously increase the transistor noise) Sure but note that noise can be reduced by using two transistors in parallel. More components isn't always bad. After the LNA where the signal has been boosted by the gain, it is of less concern but the key to low noise is matching at the input. Yes, I did note, much earlier, that unless I could achieve adequate impedance matching at the input, there would be a problem with the design. However, it does seem from Steve Willner's comments that we should now be discussing direct impedance matching between the detector and the amp, as opposed to via a 50 ohm transmission line. Sure, but what is the source impedance of the pickup in the guide/horn/diplexer/whatever? If it is a standard component, it will have been designed to have a common impedance (50 or 75 most likely) but whatever it is the LNA will be matched to it. Again I suggest you look at this: http://www.submm.caltech.edu/cso/rec...ers/ballna.pdf Unfortunately, the computer I have borrowed for this response does not have a pdf reader, so the above will have to suffice until I get back to my own office. OK, my earlier comments will make more sense when you see this. However, are you getting distracted? Impedance matching is always considered in RF design and as you can see from that paper a lot of work goes into optimising the input network for the desired compromise (badwidth, noise, cost etc.) while your idea was aimed at reducing noise below what can be achieved by conventional methods. Wouldn't it be simpler to assume that the LNA designers have achieved the best matching for the lowest noise and then see what improvement your method can achieve treating the LNA as a black box? George |
#2
|
|||
|
|||
Still lower noise radio astronomy (was: low-noise amplifiers for radio astronomy )
George Dishman wrote:
Sure, but what is the source impedance of the pickup in the guide/horn/diplexer/whatever? If it is a standard component, it will have been designed to have a common impedance (50 or 75 most likely) but whatever it is the LNA will be matched to it. Given the stated estimated budget for upgrading the DSN, I doubt that those engineers would have necessarily been constrained to use standard components. Similarly, by combining your comments on the simplicity of the required sensor, and Steve Willner's comments on state-of-the-ast radio telescopes, and the relative insignificance of cryogenic costs for the preamp, I doubt that we are thus constrained either. This leads me to ask: What is so great about a 50 ohm 'standard', anyway? I suspect that this could just be historical. It was a convenient figure for driving and being driven by pre-WW2 valve amplifiers. Consequently I still cannot understand why a wire sensor at the focus of a modern telescope, or a knob in a waveguide, needs to be designed with a 50 ohm resistance. For that matter, I don't currently see why a TV arial needs to be 50 ohms either, if boosted by a preamp before being fed into coax. The title under current discussion is, after all "Still lower noise radio astronomy" not "the best we can manage using standard off-the-shelf components" Again I suggest you look at this: http://www.submm.caltech.edu/cso/rec...ers/ballna.pdf This is certainly a highly pertinent paper, for LNAs in the cm to mm range, and the transistor schematic is also helpful. (I am still waiting for my copy of Merian Pospieszalski's paper, which I hope will be more revealing on magnitudes for the various elements of that schematic). On first reading, however, I don't really understand the purpose of the Lange couplers, why we want to split the signal into 2 components with a 90 degree phase shift anyway, or why there are 4 of these couplers to the 2 transistors between input & output in the 1991 design. I note from this first reading, however: 1) that this amp does indeed appear to use standard components for I/O (and 50 ohm coax between transistors) 2) the inductors are also conventional and of sufficiently few turns that they can: a) be wound by hand more accurately and cheaply than buying 'off the shelf' components b) utilise wire of whatever thickness you want, to vary their DC resistance (at the trade-off of varying their capacitance). This leads me to ask: at what frequency is it desirable/necessary to switch from such an inductor to a 'squiggle' on the pcb? 3) The discrete capacitors too look completely conventional. However, are you getting distracted? Impedance matching is always considered in RF design and as you can see from that paper a lot of work goes into optimising the input network for the desired compromise (badwidth, noise, cost etc.) while your idea was aimed at reducing noise below what can be achieved by conventional methods. Wouldn't it be simpler to assume that the LNA designers have achieved the best matching for the lowest noise and then see what improvement your method can achieve treating the LNA as a black box? That is precisely what I did at the beginning. This led me to conclude that achieving a 20dB improvement in s/n ratio for the amp was easier than falling off a log, and significantly larger improvements were possible with more care. However, as discussed before, that too is a question of diminishing returns. Why spend ages building and tweaking a couple of amplifiers to noise performances levels that nobody actually needs, when you can more easily and consistently produce a more economical device that still 'blows the socks off' the competition. Since then I have, I admit, been 'looking for problems' on several fronts. (Forewarned is forearmed.) However, I don't believe that line of enquiry is completely fruitless, if it holds out a possibility of also reducing external noise, and making the resultant LNA more easily and consistently manufacturable. Differences may arise because I am approaching this subject as a physicist, whereas you are approaching it as an RF engineer. Nevertheless, I believe a fusion of both approaches is required, to achieve the optimum system solution. John Bell (Change John to Liberty to bypass anti-spam email filter) |
#3
|
|||
|
|||
Still lower noise radio astronomy (was: low-noise amplifiers for radio astronomy )
John (Liberty) Bell wrote:
George Dishman wrote: Sure, but what is the source impedance of the pickup in the guide/horn/diplexer/whatever? If it is a standard component, it will have been designed to have a common impedance (50 or 75 most likely) but whatever it is the LNA will be matched to it. Given the stated estimated budget for upgrading the DSN, I doubt that those engineers would have necessarily been constrained to use standard components. I was not suggesting they were, read my comment again. As an aside, getting the budget means cutting costs to the minimum, however if that can be achieved by non-standard parts then certainly that would be considered. Usually "specials" put the price up significantly though. Similarly, by combining your comments on the simplicity of the required sensor, and Steve Willner's comments on state-of-the-ast radio telescopes, and the relative insignificance of cryogenic costs for the preamp, I doubt that we are thus constrained either. This leads me to ask: What is so great about a 50 ohm 'standard', anyway? I suspect that this could just be historical. It was a convenient figure for driving and being driven by pre-WW2 valve amplifiers. Consequently I still cannot understand why a wire sensor at the focus of a modern telescope, or a knob in a waveguide, needs to be designed with a 50 ohm resistance. For that matter, I don't currently see why a TV arial needs to be 50 ohms either, if boosted by a preamp before being fed into coax. TV was usually 75 ohm again for historical reasons, but that is not important, what matters is that the equipment and cables are matched. If you put a 50 ohm cable between a 75 ohm antenna and a 75 ohm input TV or vice versa then each mismatch causes a reflection and you see a ghost in the background of the picture due to the signal reflected first from the TV and then back down from the antenna. That is annoying for the viewer but would be disasterous for accurate measurements at the DSN. The title under current discussion is, after all "Still lower noise radio astronomy" not "the best we can manage using standard off-the-shelf components" Which is why I said "IF it is a standard component, ... but WHATEVER it is ..." Again I suggest you look at this: http://www.submm.caltech.edu/cso/rec...ers/ballna.pdf This is certainly a highly pertinent paper, for LNAs in the cm to mm range, and the transistor schematic is also helpful. (I am still waiting for my copy of Merian Pospieszalski's paper, which I hope will be more revealing on magnitudes for the various elements of that schematic). On first reading, however, I don't really understand the purpose of the Lange couplers, why we want to split the signal into 2 components with a 90 degree phase shift anyway, or why there are 4 of these couplers to the 2 transistors between input & output in the 1991 design. I am not familiar with the configuration but the article appears to say the benefit is they give low return loss over an octave of bandwidth. Consider that in the context of the S-band we were discussing where the bandwidth is less than 1.5% of the center. I note from this first reading, however: 1) that this amp does indeed appear to use standard components for I/O (and 50 ohm coax between transistors) The alternative is to get a manufacturer to set up a production line to make special connectors. That would be _very_ expensive, and the users would ahve to buy special matching connectors. A bit of different width track on the PCB matches the preferred standard connector. It would even be easy to have sevearl PCB designs with slightly different track width matching but the same layout for the circuitry to allow various connectors as customer options. 2) the inductors are also conventional and of sufficiently few turns that they can: a) be wound by hand more accurately and cheaply than buying 'off the shelf' components Hand winding is less accurate and much more expensive than off-the-shelf, perhaps $100 per hour for a skilled coil winder compared to ~$0.2 for off-the-shelf to a guaranteed specification. However, some parameters may require an air-spaced coil but you would need to look in more detail to find why they went that way. We use a lot of such coils but only in transmitters where voltages of many kV can appear if the output isn't properly matched and PCB parts would fry. b) utilise wire of whatever thickness you want, to vary their DC resistance (at the trade-off of varying their capacitance). This leads me to ask: at what frequency is it desirable/necessary to switch from such an inductor to a 'squiggle' on the pcb? It depends on the space, reliability and cost tradeoffs. 3) The discrete capacitors too look completely conventional. Off course. Setting up a production line for a single design would be vastly expensive and I can't see what possible benefit special capacitors would provide. The PCB substrate would also be a standard material, you just select the most suitable. However, are you getting distracted? Impedance matching is always considered in RF design and as you can see from that paper a lot of work goes into optimising the input network for the desired compromise (badwidth, noise, cost etc.) while your idea was aimed at reducing noise below what can be achieved by conventional methods. Wouldn't it be simpler to assume that the LNA designers have achieved the best matching for the lowest noise and then see what improvement your method can achieve treating the LNA as a black box? That is precisely what I did at the beginning. This led me to conclude that achieving a 20dB improvement in s/n ratio for the amp was easier than falling off a log, and significantly larger improvements were possible with more care. As I remember, you hinted you had a method but said nothing about what it could achieve because of concerns over patenting. However, as discussed before, that too is a question of diminishing returns. Why spend ages building and tweaking a couple of amplifiers to noise performances levels that nobody actually needs, when you can more easily and consistently produce a more economical device that still 'blows the socks off' the competition. If you can improve noise by 20dB over current levels then it would be of huge value Since then I have, I admit, been 'looking for problems' on several fronts. (Forewarned is forearmed.) It is a good policy but impedance matching isn't an area that is neglected, it is probably one of the biggest drivers throughout the design process and something you can take for granted will have alread been optimised before you apply your new technique. However, I don't believe that line of enquiry is completely fruitless, if it holds out a possibility of also reducing external noise, and making the resultant LNA more easily and consistently manufacturable. Use of standard components available to a far tighter spec than any special goes a long way to achieving that. The noise reduction benefit would be your selling point. Differences may arise because I am approaching this subject as a physicist, whereas you are approaching it as an RF engineer. Nevertheless, I believe a fusion of both approaches is required, to achieve the optimum system solution. My own degree is in physics and several of our RF designers also followed that path. George |
#5
|
|||
|
|||
Still lower noise radio astronomy (was: low-noise amplifiers for radio astronomy )
"John (Liberty) Bell" wrote in message
... wrote: John (Liberty) Bell wrote: George Dishman wrote: Sure, but what is the source impedance of the pickup in the guide/horn/diplexer/whatever? If it is a standard component, it will have been designed to have a common impedance (50 or 75 most likely) but whatever it is the LNA will be matched to it. Given the stated estimated budget for upgrading the DSN, I doubt that those engineers would have necessarily been constrained to use standard components. I was not suggesting they were, read my comment again. In that case, read MY comment again on posting 6 of the original discussion "low-noise amplifiers for radio astronomy (was: Ranging and Pioneer)". I was the one who first emphasised the necessity of impedance matching the source and the amp, in this discussion. That's because I was taking it for granted, though I did mention matching cables around the same time. It really doesn't matter who first mentioned it, we both understand that impedance matching in RF electronics is an essential consideration. Bringing it up is like telling a professional gardener that watering the plants is a good idea ;-) snip I note from this first reading, however: 1) that this amp does indeed appear to use standard components for I/O (and 50 ohm coax between transistors) The alternative is to get a manufacturer to set up a production line to make special connectors. That would be _very_ expensive, and the users would ahve to buy special matching connectors. I probably agree with you for the amp output, but not for the input, since this can be just a wire placed at the focus of the telescope, as suggested by the combined comments of you and Steve Willner. That wire could itself be supplied as part of the amp which projects from the box through a tiny bit of insulated sleeving. And how would you ensure the impedance of the wire in the hole was matched to the input of the amplifier? If you look at the photograph in the article, you can see connectors on the right hand side. These have gold plated contacts, the insulator will be PTFE and of a controlled dielectric constant, and the cross section will have been modelled to ensure the impedance does not vary through the connector. I would expect to be able to find figures for measured insertion loss on the manufacturers website provided by a third-party test house. That would be still cheaper again, but, as I have already mentioned, price is not the central issue here. Exactly, so we buy connectors that we can trust instead of "doing it ourselves". 2) the inductors are also conventional and of sufficiently few turns that they can: a) be wound by hand more accurately and cheaply than buying 'off the shelf' components Hand winding is less accurate and much more expensive than off-the-shelf, perhaps $100 per hour for a skilled coil winder compared to ~$0.2 for off-the-shelf to a guaranteed specification. However, some parameters may require an air-spaced coil but you would need to look in more detail to find why they went that way. We use a lot of such coils but only in transmitters where voltages of many kV can appear if the output isn't properly matched and PCB parts would fry. The photo of the LNA shows 3 turn and 5 turn air spaced inductors. These are sufficiently easy and quick to wind that you could have many hundreds made per hour on legal minimum wage, once you have determined the number of required turns and diameter of each turn. However, I admit that my original experience in such an area was with low resistance wire wound power resistors. I found I could make dozens of tight spec., custom resistance, high wattage devices from a single coil of electric fire element, for the same cost as a single off-the-shelf resistor, which often got delivered days later, and up to 50% out of spec. Only problems we (a) you couldn't solder the ends for a production run (b) when the idea was submitted to a popular electronics magazine, they turned it into a four page feature article without crediting me, or paying me a bean. : ( You can use them reliably with screw connector blocks as long as they are made from high temperature plastic. I used that method to make 8 ohm 100W loads out of 1kW elements back in the 1970s when I designed disco amps. You had to couple pairs in reverse to cancel the inductance of course otherwise you tended to get a high power RF transmitter ;-) My current experience (this year) is with a 1kW HF transmitter where we have air-spaced coils, but they have to be held with rigid spacing between the turns to avoid microphony and of course you have to use silver plated wire because of the skin effect. Production test has to tunes certain coils by squeezing them, trial and error. If there was an alternative we would take it. snip That is precisely what I did at the beginning. This led me to conclude that achieving a 20dB improvement in s/n ratio for the amp was easier than falling off a log, and significantly larger improvements were possible with more care. As I remember, you hinted you had a method but said nothing about what it could achieve because of concerns over patenting. There is no problem with discussing what it CAN achieve, only HOW. (I have, by now, been down that route several times before, already.) If I remember correctly, nobody actually bothered to ask me what it could achieve, in quantitative terms. However, I have already answered that unasked question by implication, by describing noise Ts of resultant amps at an ambient (room) T. OK, I stand corrected, you did indeed give temperatures. If you can improve noise by 20dB over current levels then it would be of huge value I can, and more, if we are referring to amp, not system. Then you have a fortune at your fingertips. I am sceptical of course since predicting random noise is impossible in theory but there are many neat tricks that appear to defy theory so I will have to wait and see. No doubt you will make the headlines. My own degree is in physics and several of our RF designers also followed that path. OK Let me ask you one final question pertinent to our continuing apparent disagreement over ideal design impedances for a system, which I am now going to take to extremes: Nowhere have I said anything about "ideal design impedances", all I have said is that whatever the impedance, you must keep things matched to avoid power loss and reflections. The requirement is that the output of each stage should be matched to the input of the next. What I am saying is that this is standard practice of course hence is a distraction from the topic of your noise reduction scheme. If you place two identical metal poles on at tall building, and connect one to earth with a thick copper strip, and the other to earth via a 10 meg resistor, which will attract a lightning strike most? Both equally since the upward pilot is at low current. When struck, which will then carry most energy to earth? Wrong question, in our system we are trying to get the maximum power into our amplifier so the question is which design will extract most energy from the strike. Assuming the resistor is rated for the job, and taking typical values of 10MV and 10kA capacity for the strike and 1 ohm for the copper strap, I get 10MW (~10MV at 1A) dissipated in the 1M resistor and only 1MW (~10kA at 10kV) for the strap, the resistor is better. Using a 1k resistor would produce the best result, 5MV at 5kA giving 25MW. Note the 10MV at 10kA cloud output looks like a 1k source. Now, explode a nuclear warhead in the stratosphere above them. Which will now absorb and transfer the most electromagnetic (pulse) energy to earth? Looking forward, with anticipation, to your answer. I guess you are referring to EMP in which case the answer is probably neither. You have a vertically descending pulse with essentially dipole receiving antennas (ignoring any horizontal runs of wire across the roof) trying to work in "near-vertical incidence" mode. Dipoles have a null in the direction of their axis. George |
#6
|
|||
|
|||
Still lower noise radio astronomy (was: low-noise amplifiers for radio astronomy )
wrote:
"John (Liberty) Bell" wrote in message ... (Snip material long since agreed, on desirability of impedance matching.) And how would you ensure the impedance of the wire in the hole was matched to the input of the amplifier? Trial and error on pre-production prototype. Measure impedance after using small drill bit to create hole, use larger drill bit if indicated, then match amplifier to most sensible figure you can thus achieve, using standard drill bits, the given sheet thickness of the employed box, and the dielectric constant of your chosen insulator. The photo of the LNA shows 3 turn and 5 turn air spaced inductors. These are sufficiently easy and quick to wind that you could have many hundreds made per hour on legal minimum wage, once you have determined the number of required turns and diameter of each turn. However, I admit that my original experience in such an area was with low resistance wire wound power resistors. I found I could make dozens of tight spec., custom resistance, high wattage devices from a single coil of electric fire element, for the same cost as a single off-the-shelf resistor, which often got delivered days later, and up to 50% out of spec. Only problems we (a) you couldn't solder the ends for a production run (b) when the idea was submitted to a popular electronics magazine, they turned it into a four page feature article without crediting me, or paying me a bean. : ( You can use them reliably with screw connector blocks That is obvious. Screw connectors were explicitly covered in the published article. I used that method to make 8 ohm 100W loads out of 1kW elements back in the 1970s when I designed disco amps. That sounds precisely right. That was when my idea was published, and dummy amplifier loading was the suggested primary application for the idea. I can even give you the magazine name, if you don't remember. You had to couple pairs in reverse to cancel the inductance of course otherwise you tended to get a high power RF transmitter ;-) And that was NOT covered in the published article, which suggests it was not obvious to the publisher. In fact, I don't think that it is necessarily true anyway, given that an audio pickup is only supposed to pick up the audio range, and a vinyl disc is definitely not modulated at radio frequencies, even if the stylus & cartridge could pick such modulation up, (which they can't). Sounds to me like your early amps were unstable, and subject to HF oscillation (which is not at all unusual). snip Production test has to tune certain coils by squeezing them, trial and error. If there was an alternative we would take it. Obviously. But this is what you have to do with prototypes. Once you know exactly what you want, via that process, you can churn the things out using the detailed knowledge you have thus gained, for the production run. wrote: John (Liberty) Bell wrote: That is precisely what I did at the beginning. This led me to conclude that achieving a 20dB improvement in s/n ratio for the amp was easier than falling off a log, and significantly larger improvements were possible with more care. As I remember, you hinted you had a method but said nothing about what it could achieve because of concerns over patenting. There is no problem with discussing what it CAN achieve, only HOW. (I have, by now, been down that route several times before, already.) If I remember correctly, nobody actually bothered to ask me what it could achieve, in quantitative terms. However, I have already answered that unasked question by implication, by describing noise Ts of resultant amps at an ambient (room) T. OK, I stand corrected, you did indeed give temperatures. If you can improve noise by 20dB over current levels then it would be of huge value I can, and more, if we are referring to amp, not system. Then you have a fortune at your fingertips. I will not take that advice too seriously until/unless I have a microwave engineer ready to do the 'obvious' twiddly RF detailing, and potential customers clamouring for the product. I am sceptical of course That is natural and normal, at this stage, for every genuinely inventive step that has ever been taken. This is why Confidential Sight Agreements are invariably agreed for much less than the invention is actually worth, to permit potential partners, developers, and angels, an opportunity for informed evaluation of the inventive proposal, prior to serious commitment. OK Let me ask you one final question, which I am now going to take to extremes: If you place two identical metal poles on at tall building, and connect one to earth with a thick copper strip, and the other to earth via a 10 meg resistor, which will attract a lightning strike most? Both equally since the upward pilot is at low current. Your response surprises me. You are claiming that all lightning conductors I remember seeing on buildings as a little boy were, in fact, a complete waste of time and money? When struck, which will then carry most energy to earth? Wrong question, in our system we are trying to get the maximum power into our amplifier so the question is which design will extract most energy from the strike. Not so. You are making unwarranted assumptions. If you want to second guess what I am considering at present, you would be closer to the mark if you equated the 10 meg resistor to the source impedance, not the amplifier input impedance. Assuming the resistor is rated for the job, and taking typical values of 10MV and 10kA capacity for the strike and 1 ohm for the copper strap, I get 10MW (~10MV at 1A) dissipated in the 1M resistor and only 1MW (~10kA at 10kV) for the strap, the resistor is better. Using a 1k resistor would produce the best result, 5MV at 5kA giving 25MW. Note the 10MV at 10kA cloud output looks like a 1k source. Wrong answer to the described question (which you chose to ignore). The question is: what is the resultant energy injected into the earth. Now, explode a nuclear warhead in the stratosphere above them. Which will now absorb and transfer the most electromagnetic (pulse) energy to earth? Looking forward, with anticipation, to your answer. I guess you are referring to EMP Yes, EMP is the acronym for ElectroMagnetic Pulse in which case the answer is probably neither. Sorry, wrong again, since your conclusion is that neither will be affected. EMP will fry everything in the neighbouring vicinity, including solid copper and solid silver wire. In order to take this question seriously, you thus need to assume that the explosion is sufficiently far away that such catastrophic damage does not occur. In which case, the following additional comments become pertinent. You have a vertically descending pulse Nonsense. You have assumed, without justification, not only that the warhead will explode exactly vertically, but also that it will explode exactly vertically above two different poles, in two different places, at the same time. That is impossible. with essentially dipole receiving antennas (ignoring any horizontal runs of wire across the roof) trying to work in "near-vertical incidence" mode. Dipoles have a null in the direction of their axis. Sorry, you have made yet another unwarranted assumtion. You have additionally assumed that both poles are mounted exactly vertically. My question said nothing about the orientation of the poles. In practice, by far the easiest way to put poles on a roof is to lay them flat on the slope of that roof, in whatever orientation is convenient. Please try answering the question again, this time a little more carefully. [Mod. note: I feel this thread may be becoming slightly acrimonious: please try to treat each other with courtesy, or alternatively take the parts of the discussion with little direct relevance to astrophysics research to private e-mail, where you can say what you like -- mjh] John Bell (Change John to Liberty to respond by email) |
#7
|
|||
|
|||
Still lower noise radio astronomy (was: low-noise amplifiers for radio astronomy )
|
#8
|
|||
|
|||
Still lower noise radio astronomy (was: low-noise amplifiers for radio astronomy )
John (Liberty) Bell wrote:
wrote: "John (Liberty) Bell" wrote in message ... OK Let me ask you one final question pertinent to our continuing apparent disagreement over ideal design impedances for a system, which I am now going to take to extremes: If you place two identical metal poles on at tall building, and connect one to earth with a thick copper strip, and the other to earth via a 10 meg resistor, which will attract a lightning strike most? When struck, which will then carry most energy to earth? Assuming the resistor is rated for the job, and taking typical values of 10MV and 10kA capacity for the strike and 1 ohm for the copper strap, I get 10MW (~10MV at 1A) dissipated in the 1M resistor and only 1MW (~10kA at 10kV) for the strap, the resistor is better. Using a 1k resistor would produce the best result, 5MV at 5kA giving 25MW. Note the 10MV at 10kA cloud output looks like a 1k source. The question was: "Which will then carry most energy to earth?", not "Which will dissipate most energy en-route?". Your answer says that a 1 ohm strap transmits 10kA , and a 1M resistor transmits 1A . Since power is proportional to current squared, the correct answer is that the strap will transmit 100,000,000 times as much power as a 1 M resistor. Yes, that's correct. However my understanding of an analogy with an electronic circuit would be that this would represent "ground bounce", a problem particularly in digital circuits where the inductance of the ground plane can cause spikes on the 0V pin of ICs. It would manifest as a small amount of signal on the ground plane near the source of the HEMT in an LNA and is one reason why choosing a ground point for a scope probe requires care when working at RF. (According to your own figures, impedance matching instead, using a 1 K resistor, would result in a 75% loss of transmitted power to the point under present scrutinisation.) The "matched" condition maximises power into the load, it is not intended to maximise a problematic side-effect. I apologise to both you and the moderator if I sounded a bit 'shirty' in my earlier more comprehensive response to your posting. However, you cannot afford to assume anything about what is inside the 'black box', or why I consequently ask the questions that I do. Similarly, I should make allowances for the fact that you don't actually know why I ask such questions. I will also apologise for not answering directly but in the past I have been accused of taking questions too literally as a way of avoiding a point. Drawing the analogy with the circuit structure suggested the style of my response. A literal answer would be that since the earth connection is hypothetically a zero resistance point, both methods would produce zero power which would have been an equally useless answer. I'll reply to your more detailed post later as I'm pushed for time at the moment. In the meantime try plotting a graph of SNR versus load resistance for a given ource resistance. Remember for a given badwidth, the noise power is independent of the resistance hence noise voltage varies as the sqrt of the resistance. best regards George |
#9
|
|||
|
|||
Still lower noise radio astronomy (was: low-noise amplifiers for radio astronomy )
"John (Liberty) Bell" wrote in message
... wrote: "John (Liberty) Bell" wrote in message ... (Snip material long since agreed, on desirability of impedance matching.) And how would you ensure the impedance of the wire in the hole was matched to the input of the amplifier? Trial and error on pre-production prototype. I meant how would you ensure it stayed stable once a unit was built but it ws really rhetorical. A company in the business of manufacturing radios doesn't want to worry about such details. In fact though my local part of the business has about 500 employees with over 100 in the engineering side of product development, we no longer have even a model shop. Measure impedance after using small drill bit to create hole, use larger drill bit if indicated, then match amplifier to most sensible figure you can thus achieve, using standard drill bits, the given sheet thickness of the employed box, and the dielectric constant of your chosen insulator. What insulator vbg? You want something to support the wire and it needs to be correctly shaped to provide that support without distorting the field. You can find the best shape by computer modelling but we are not a plastics moulding company. Far simpler just to buy an off-the-shelf part. The photo of the LNA shows 3 turn and 5 turn air spaced inductors. These are sufficiently easy and quick to wind that you could have many hundreds made per hour on legal minimum wage, once you have determined the number of required turns and diameter of each turn. However, I admit that my original experience in such an area was with low resistance wire wound power resistors. I found I could make dozens of tight spec., custom resistance, high wattage devices from a single coil of electric fire element, for the same cost as a single off-the-shelf resistor, which often got delivered days later, and up to 50% out of spec. Only problems we (a) you couldn't solder the ends for a production run (b) when the idea was submitted to a popular electronics magazine, they turned it into a four page feature article without crediting me, or paying me a bean. : ( You can use them reliably with screw connector blocks That is obvious. Screw connectors were explicitly covered in the published article. I used that method to make 8 ohm 100W loads out of 1kW elements back in the 1970s when I designed disco amps. That sounds precisely right. That was when my idea was published, and dummy amplifier loading was the suggested primary application for the idea. I can even give you the magazine name, if you don't remember. I don't think I ever saw it anywhere, it was just an obvious solution. You had to couple pairs in reverse to cancel the inductance of course otherwise you tended to get a high power RF transmitter ;-) And that was NOT covered in the published article, which suggests it was not obvious to the publisher. In fact, I don't think that it is necessarily true anyway, given that an audio pickup is only supposed to pick up the audio range, and a vinyl disc is definitely not modulated at radio frequencies, even if the stylus & cartridge could pick such modulation up, (which they can't). Sounds to me like your early amps were unstable, and subject to HF oscillation (which is not at all unusual). They were indeed initally, given an inductive load. Pairing the segments solved it while I developed the amp to the point where it didn't care. snip Production test has to tune certain coils by squeezing them, trial and error. If there was an alternative we would take it. Obviously. But this is what you have to do with prototypes. With air-spaced coils you have to do it in production too :-( This kit has been made for many years. Once you know exactly what you want, via that process, you can churn the things out using the detailed knowledge you have thus gained, for the production run. Trouble is they don't stay at the same value during the handling and soldering processes. It is OK for the less critical values but a pain in filters where to have manually adjust and then lock with some suitable varnish or proprietary compound. If you can improve noise by 20dB over current levels then it would be of huge value I can, and more, if we are referring to amp, not system. Then you have a fortune at your fingertips. I will not take that advice too seriously until/unless I have a microwave engineer ready to do the 'obvious' twiddly RF detailing, and potential customers clamouring for the product. That wasn't advice, just an acknowledgement of reality. My advice would be that your first step should be to produce a SPICE model (or whatever your favourite simulator uses) to find out if it works before building prototypes. Building shouldn't be too much of a problem but measuring the noise level is going to be _really_ difficult (A.K.A. expensive!!!). snip OK Let me ask you one final question, which I am now going to take to extremes: If you place two identical metal poles on at tall building, and connect one to earth with a thick copper strip, and the other to earth via a 10 meg resistor, which will attract a lightning strike most? Both equally since the upward pilot is at low current. Your response surprises me. You are claiming that all lightning conductors I remember seeing on buildings as a little boy were, in fact, a complete waste of time and money? No. As a strike comes down it spreads out in many threads like a river delta. The spike produces a high field at the tip inducing breakdown and the avalanche effect propagates that upwards. When the upwards 'pilot' makes contact with the downwards flow, the current increases ionising more air and the charge from the remaining downward branches is pulled back. Both initial flows are essentially electrostatic and at relatively low current so both your spikes will "attract a lightning strike" equally, both designs would work. Once the plasma path is complete, the current leaps to tens of kA and the resistor would in reality explode while the copper strap can take the current. You asked "which will attract a lightning strike most" and the _literal_ answer is that they will attract equally, survivability is a different question. When struck, which will then carry most energy to earth? Wrong question, in our system we are trying to get the maximum power into our amplifier so the question is which design will extract most energy from the strike. Not so. You are making unwarranted assumptions. If you want to second guess what I am considering at present, you would be closer to the mark if you equated the 10 meg resistor to the source impedance, not the amplifier input impedance. OK, then it is a difficult analogy to follow. Assuming the resistor is rated for the job, and taking typical values of 10MV and 10kA capacity for the strike and 1 ohm for the copper strap, I get 10MW (~10MV at 1A) dissipated in the 1M resistor and only 1MW (~10kA at 10kV) for the strap, the resistor is better. Using a 1k resistor would produce the best result, 5MV at 5kA giving 25MW. Note the 10MV at 10kA cloud output looks like a 1k source. Wrong answer to the described question (which you chose to ignore). The question is: what is the resultant energy injected into the earth. OK, answering literally, zero in both cases since the aerth point is rthe reference for measurements. Now, explode a nuclear warhead in the stratosphere above them. Which will now absorb and transfer the most electromagnetic (pulse) energy to earth? Looking forward, with anticipation, to your answer. I guess you are referring to EMP Yes, EMP is the acronym for ElectroMagnetic Pulse in which case the answer is probably neither. Sorry, wrong again, since your conclusion is that neither will be affected. EMP will fry everything in the neighbouring vicinity, including solid copper and solid silver wire. Only if a current is induced in the wire and the wire is not rated to handle the current. Since we are talking of lightning conductors, it is reasonable to assume they are rated to take direct strikes. In order to take this question seriously, you thus need to assume that the explosion is sufficiently far away that such catastrophic damage does not occur. In which case, the following additional comments become pertinent. You have a vertically descending pulse Nonsense. You have assumed, without justification, not only that the warhead will explode exactly vertically, but also that it will explode exactly vertically above two different poles, in two different places, at the same time. That is impossible. EMP is produced by the prompt gamma emission displacing the charge in the ionosphere. To create it you need an explosion substantially above that and generally if the spikes were withing a mile of each other there would be little difference in the unduction. with essentially dipole receiving antennas (ignoring any horizontal runs of wire across the roof) trying to work in "near-vertical incidence" mode. Dipoles have a null in the direction of their axis. Sorry, you have made yet another unwarranted assumtion. You have additionally assumed that both poles are mounted exactly vertically. That is necessary for a lightning conductor to work otherwise the sharp point doesn't spray the charge upwards towards the downcoming strike. The same effect is why the conductor is as high as possible, it needs to intercept the strike before any other upwards flow from trees, buildings, etc.. My question said nothing about the orientation of the poles. In practice, by far the easiest way to put poles on a roof is to lay them flat on the slope of that roof, in whatever orientation is convenient. Certainly easier but completely useless if you are trying to attract lightning. Please try answering the question again, this time a little more carefully. OK, that's done above. To summarise, both will attract equally but the resistor would in reality blow like a fuse while the copper could take the surge. [Mod. note: I feel this thread may be becoming slightly acrimonious: I hope that doesn't happen. I made certain assumptions in good faith and I have apologised to John if he got the incorrect impression that it was an argumentative device. That was not the intention, the answers I gave should have illustrated the benefit of matching which was the point in question. I have now answered directly though I suspect they will again not be what John was expecting and I can understand why John might feel somewhat exasperated as a result. Having had to design equipment to survive EMP, I probably know more about the subject than is optimal for the discussion. please try to treat each other with courtesy, or alternatively take the parts of the discussion with little direct relevance to astrophysics research to private e-mail, where you can say what you like -- mjh] It is certainly drifting off-topic and if John wishes to switch to email I am happy to continue that way. My address is at the top of this post. I don't bother anti-spamming it as it has been publicly available since before spam was invented :-( George |
#10
|
|||
|
|||
Still lower noise radio astronomy (was: low-noise amplifiers for radio astronomy )
wrote:
"John (Liberty) Bell" wrote in message ... wrote: "John (Liberty) Bell" wrote in message ... You want something to support the wire and it needs to be correctly shaped to provide that support without distorting the field. You can find the best shape by computer modelling but we are not a plastics moulding company. Far simpler just to buy an off-the-shelf part. That would be true if that standard part gave similar performance, which it presumably does in your designs. However we already know that the detector and connector contribute a significant amount of noise in their own right, in the arena of radio astronomy. That significance becomes much greater when your preamp is a further 20 dB quieter. Significantly improving s/n ratio before the preamp would obviously be worth doing if this could be achieved for the modest cost of getting a plastic moulding company to supply you with a run of insulating washers of appropriate dimensions. snip material going off topic. (I have already mentioned that this amp design permits fairly relaxed component toleraces, and so there is little point in discussing the need for tighter component tolerences in other products, here.) If you can improve noise by 20dB over current levels then it would be of huge value I can, and more, if we are referring to amp, not system. Then you have a fortune at your fingertips. I will not take that advice too seriously until/unless I have a microwave engineer ready to do the 'obvious' twiddly RF detailing, and potential customers clamouring for the product. That wasn't advice, just an acknowledgement of reality. My advice would be that your first step should be to produce a SPICE model (or whatever your favourite simulator uses) to find out if it works before building prototypes. As already indicated under Steve Willner's thread, I have not been involved in electronic design for about 25 years, except to the extent of recently applying my thus acquired knowledge to the problem of further reducing preamplifier noise (as a consequence of our discussion under Pioneer and Ranging). (Even when I was thus involved, this was more in a research as opposed to manufacturing environment). Consequently, I think I would be better advised to go into a partnership with a company that is already 'tooled up' for RF/microwave manufacturing. Building shouldn't be too much of a problem but measuring the noise level is going to be _really_ difficult (A.K.A. expensive!!!). That was equally true for prior art LNAs. In practice, testing that the idea works at particularly low frequencies is easy and inexpensive. By chaining amplifiers together (without an active input signal!) it does not take too long before you can display the noise floor of both the prior art device and the new device together on an oscilloscope, for comparison purposes. Viability at higher frequencies is simply a matter of extrapolation. snip additional material going off topic The only reason for my final question was to demonstrate that the whole subject of optimum impedance matching could change if your circuit allows you to treat the sensor as a current injector, not as a voltage source. However, it seems from your response that I may be the only one to have implimented a design solution with a current injector before (to the delight of the customer). Bearing this firmly in mind, perhaps my question could have been more appropriately phrased thus: If you terminate a TV arial (pointing towards the transmitter) with a 75 ohm resistor between its two connectors, how much current flows between those terminals. If you now connect a piece of copper wire between those two terminals, how much current flows then? More, or less? I don't bother anti-spamming it as it has been publicly available since before spam was invented :-( What we have done in such situations is set up an autoresponder on the original address, and then set up a somewhat different email address on the email server that only genuine respondents can get to. You can delete the autoresponder if and when you know there is negligible chance of genuine respondents still using the original address. (We can even supply you with a suitable email server, if your current host does not provide that flexibility.) John Bell (Change John to Liberty to bypass anti-spam email filter) |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
[sci.astro] General (Astronomy Frequently Asked Questions) (2/9) | [email protected] | Astronomy Misc | 0 | May 3rd 06 12:33 PM |
[sci.astro,sci.astro.seti] Welcome! - read this first | [email protected] | Astronomy Misc | 9 | February 2nd 06 01:37 AM |
SCIENTIFIC PANEL TO REVIEW ED CONRAD'S BRAIN.... | Ed Conrad | Astronomy Misc | 0 | December 25th 05 02:24 PM |
CORRUPT THEORY 'WINS' ANOTHER ROUND -- Evolution vs. Intelligent Design | Ed Conrad | Astronomy Misc | 2 | December 21st 05 08:55 PM |
[sci.astro] General (Astronomy Frequently Asked Questions) (2/9) | [email protected] | Astronomy Misc | 0 | October 6th 05 02:35 AM |