Enthalpy

Intel reveals the Knights Landing, a new and much improved Xeon Phi meant for supercomputing. https://software.intel.com/en-us/articles/what-disclosures-has-intel-made-about-knights-landing https://en.wikipedia.org/wiki/Xeon_Phi Its 144 vector processors mul-acc 8 double floats each in one cycle at 1.3GHz, so one huge chip offers 3TFlops. Better: this edition accesses directly <=384GB of Dram to remove one bottleneck and ease programming. The Knights Landing smashes Gpu on double floats, consumes less and is much easier to program, so it's a clear choice for supercomputers. Since the architecture I propose deduces its performance from the Knights Landing, the number of mul-acc is the same and the consumption too, except that I assumed a maybe suboptimum 1GHz clock. Though, my architecture is obviously waaaaay better - but then, why? ----- Computer size One Knights Landing and its six Dram modules occupy 150mm*150mm*45mm, or 880mm3 rack volume per mul-acc. The boards I described on 09 November 2015, putting 2 chips in 24mm*20mm*20mm but with 8 nodes per chip, occupy 600mm3 per node. That's equivalent, unless one accepts to keep the network density and put more chips, which my design and the much easier cooling allow. Both options add a cabled network, which is more bulky in my design because its throughput is much bigger. ----- Chip production cost Again as many mul-acc for the same computing power, but my design needs no L2, no L3 nor "near memory", and it spreads the mul-acc over smaller chips. Scalar processors need more sequencers but fewer instructions make these simpler. The snappy Dram makes hyperthreading unnecessary, and the explicit macro-op instructions I described here on 01 and 02 January 2016 can replace out-of-order execution if desired. More silicon founders can make the smaller chips, possibly for cheaper than the huge Knights Landing. I explained here on Nov 12 and Dec 27, 2015 how to keep the Finfet and Dram chips separate. ----- Memory throughput At algorithms needing 2 double floats per operation, the Avx512 demands 128 bytes per cycle: for each core of the Knights Landing, that's 166GB/s, which only the registers and L1 can provide. The L2 is 4x, the near Dram 48x, the far Dram 260x too slow for simple programming. When a bigger problem lets an algorithm make more computations per data, the solution is to decompose the problem into varied subchunks that fit in the successive caches. Here is a simple example for matrix multiplication; some algorithms accept less memory throughput, but others like the Fourier transform demand more, and at least this one can be planned and it accesses consecutive data (one matrix is transposed in advance). The units are doublefloats (1W = 8B = 64b) and their mult-acc (fma). I suppose the algorithm makes few writes and loads complete submatrices at once - partial loads would be even more complicated. The program can't sweep through both source matrices. It must load 1024*1024 chunks in the near Dram, 64*64 in the L2, 16*16 in the L1, 8*8 in the registers, to make successive uses of a bigger chunk. That's a set of loops for each cache size. The compiler may manage the registers but not the L1, L2, near Dram. Each cache performance has a small margin in speed or chunk size, but it's eaten by Hyperthreading, and one same executable isn't optimum on the Knights Landing's predecessor and successor. Add mentally the vector instructions and alignment. In addition, every supercomputer is loosely multitask through a slower network. Mess. While feasible for matrix multiplication (imagine an Fft or a database), the Russian dolls data chunks make the program difficult to write, read, port, and more prone to bugs. Ancient computer designers wanted instead the same throughput at every cache level and called this the Golden Rule which my proposals follow. My architecture blanks out the Knights Landing here. ----- Network A 3TFlops Knights Landing chip has two Pci-E 3.0 16x buses. Each carries 16GB/s in and out. Imagine a 2D matrix of 5782 Knights Landing: through any network equator, the chip sends and receives 16GB/s. It computes 750 mul-acc in the time it receives one doublefloat. The transport of 1024*1024 chunks is 19x slower than their bidimensional Fourier processing. Well, the chip could that at most, but planned supercomputers foresee hyperdull fibre hypertori. The chips and networks I propose carry 2GB/s from and to each 2GFlops node in each dimension. A node computes 4 mul-acc while receiving one doublefloat. Even with complex 2*32 mul-acc, 4*4 chunks occupy the execution units with a 2D Fourier transform. If slashing the network by 4, 256*256 chunks suffice. Clean win here too. Marc Schaefer, aka Enthalpy

It was in the same context of fleeing France that I observed the over-brilliant satellite which I believe had night lights: http://www.scienceforums.net/topic/78485-what-is-this-moving-light-in-the-sky/#entry901753

Hi imatfaal, thanks for your interest! Bb3 to Eb6 is the written range. The soprillo (not especially mine) sounds a seventh higher: Ab4-Eb7. Written C7 isn't neither standard range for a clarinet: not every player, mouthpiece, reed achieve it, and the fingerings depend on each instrument; standard practice goes to written G6 if memory serves. The smaller Eb clarinet sounds an Bb6 from the written G6. But above all, this is a saxophone. It has the sound quality, intonation, fingerings of a saxophone. While I certainly agree about the unique ppp to fff on the clarinet, I also consider that instruments are not equivalent. The same piece played on different instruments sounds differently, and even more remarkably, composers who play different instruments don't create the same pieces, jazz musicians even more so.

The "ferroelectric materials approaching their Curie point" are ceramic capacitors of type II and III. The proportion of BaTiO3 versus SrTiO3 adjusts the Curie point for maximum capacity around operation temperature. Then it depends on if you want the maximum capacity only (type III), or also pay attention to the temperature drift, dielectric polarization, dielectric losses, linearity (type II) - knowing that only type I are really good on these last aspects. So this is old past. Recent past are supercapacitors a well as niobium electrolytic capacitors (cheaper than tantalum). What the future is, I don't know.

No, it doesn't. Some properties like the momentum or the position are not quantized for free particles. Entanglement leaves uncertainty in the relation between the particles, which can't be arbitrarily accurate. This has been measured experimentally and compared with Heisenberg's relations.

Hey friends, have you already heard a soprillo? The rare instrument is a saxophone, but a higher-pitched one: in Bb, one octave over the soprano, one fifth over the sopranino. Made only by Benedikt Eppelsheim, a nice guy in Munich http://www.eppelsheim.com/en/instruments/soprillo/ (have a look at his other odd instruments too...) Any high-pitched wind instrument is difficult, and this one is very high-pitched, but Nigel Wood plays it nicely http://soprillo.com/soprillogy.php he recorded a Cd of inspired and diverse jazz, "Soprillogy", that's probably the debut of the instrument's reknown. Track samples there!

23Mm/s=c/13: this needs less antimatter, but the next star takes 2/3 century then. Dust: why worry? There is occasionally some on low-Earth orbit but not farther in our Solar system. Do you expect any in the interstellar medium? Then, whether beryllium or brick doesn't change anything - except that I'd avoid a brittle metal like beryllium as an armour. And I'd prefer redundancy (as combat aeroplanes have) and automatic leak tightening (works at tyres, possible source of inspiration). Pions and so on: why neglect photons? Gammas are "difficult" to reflect but they give substantial thrust. Even if the ones emitted forward are absorbed instead of reflected, they still give half of their ideal thrust. (Little) some serious science has been done (at Cern and elsewhere) about antimatter propulsion and is available on arXiv.

Well, there isn't much more to say with the little information provided... If 1T is enough, or at most 2T, use magnetic cores. Apparently you've decided not to. With magnetic cores, use permanent magnets. For some reason you don't want to. The next reasonable choice is superconductivity - again, you eliminate it. What remains is capacitive discharge, but figures will probably rule it out, especially the energy. Most desires are impossible in magnetic design. At some point, after checking a few figures like power and energy, one very often has to rethink the wishes. "Focus" a magnetostatic field is impossible under any circumstance, and multi-Tesla induction uses to be extremely impractical. If the goal is to induce a voltage, try many short pulses instead of one long if possible. For transcranial magnetic stimulation it would change everything: http://www.scienceforums.net/topic/70203-transcranial-magnetic-stimulation/ including reduced forces on the coils, which can have a better shape then. ---------- 1m3: these were 3500µF 350V, all in parallel, nicely connected by screws in copper bars. They had cost a shiny penny, sure.

Flash memory chips improve quickly. One single chip interface can toggle at 333MHz over 8 bits width, even at 128Gbit=16GByte size http://www.micron.com/products/nand-flash/mlc-nand and Usb sticks transfer >200MB/s. That is, 8 or 16 chips (this can begin at 256GByte) deliver a throughput like 2 or 4GByte/s that neither Usb 3.0 nor Sata/6000 can handle. A recent wide Pci-E (32GByte/s for x16 v4.0) can still carry the data but it won't cope forever. It's time to define a parallel interface for Flash storage. Maybe the disks must be modules plugged on the mobo like Dram modules are, or if possible have a wide cable to the mobo. Fibre optics won't help much. It looks like the bus must connect to the Cpu directly Maybe Flash chips on the video card would be useful to load textures faster, but the OS must be aware of it. If using the Avx256 properly, this machine at 2.2GHz would bring 422GFlops but the Knights Landing 3TFlops, doing the job in 1/3 day if the soft fits the hard. Your colleagues look like a customer group for the new toy. If I sometimes wanted to scoff I'd add: "and on how recent your software is".

In a variant, each amplifier stage can have several digital outputs, so there are fewer amplifier stages, each with a gain less small. The example diagram would have as many differential triplets, (unrepresented) flip-flops and gates for the same number of logic outputs, but grouped on fewer amplifier stages. Darlingtons would minimize interactions. Varied BiasN at differential triplets of the same amplifier stage makes them sensitive to varied thresholds. Marc Schaefer, aka Enthalpy

My 7T experiment : sure. It was a part of a project in a club, and anyway, we did it the same way as everyone. 1m3 electrolytic capacitors charged to 350V, their rated voltage. Thick copper bars between them, connected at their middles as a 3D tree. One 2.5kA thyristor (could be a Gto now) used at 20kA, its rated nonrepetitive peak current for 10ms half-wave. Normal thick wires but hold by good staples on the table - or they fly away. Normal connecting blocks, but double and triple-check they're tightened firmly - or they volatilize and you're deaf for a day. Fun: the cables are cold, and 10ms later they're lukewarm. I made the coil with copper foil, as broad as the coil, winded as a spiral in alternance with a plastic film. On a PVC tube kernel and hold in a hole in thick plywood. No cooling for 10ms, but afterwards the coil is warmer. It's the standard way to make strong magnets, SmCo and NdFeB. These need more than 2.3T so a magnetic core isn't an option, and since teslas in air demand megawatts, but the magnets react quickly, everyone uses capacitive discharge. Just ask at a company that makes magnets - possibly one that makes servomotors. ---------- For longer duration, the decent option is superconductivity. Some 8T are a difficult realistic target, and 20T an exotic achievement. A few people, maybe three research groups worldwide, produce >20T over many seconds. This needs copper because superconductors get resistice before, plus water to cool it, and many MW (optionally from a homopolar machine or similar). It also demands a very strong construction. A stronger induction has been achieved by capacitive discharge - 50T then, do I remember 200T now? The coil is extremely reinforced (understand: fibres alternate with copper) but it deforms and survives only a few shots. A compulsator could supposedly replace the capacitors and its construction be fused with the coil. The flux compressors, invented by Sakharov, achieve supposedly a bigger induction using explosives, metal and an initial field, but I don't know the figures. They serve in weapons as EMP sources.

It's worse than that. You claim to achieve 0.3c - which is indeed the kind of speed necessary to reach the nearest stars within a few years - and this speed demands to annihilate ~1/4 of the initial vessel's mass, needing 1/8 the initial mass in antimatter. Of course, the emitted energy must be fully directed opposed to the desired force, so it needs gamma rays mirrors. Then, you want to brake at destination, needing again 1/8 the mass, so the antimatter is 1/4 the vessel's start mass. Though, I approve the choice of antimatter as the only energy dense enough per kg to reach 0.3c - it's just that Mankind has no means to store nor poduce significant amounts, which isn't bad news considering its danger.

EDRS is a new relay satellite in geosynchronous orbit that receives data by laser from Earth-observation satellites (typically on low orbit hence seeing the ground stations shortly) and relays it by radio to ground stations http://www.bbc.com/news/science-environment-35446894 I suggested here on Jan 10 and Jan 15, 2016 to transmit several beacons to separate ground stations and encode data on the polarization as well http://www.scienceforums.net/topic/76627-solar-thermal-rocket/page-3#entry900362 and this works from 36,000km distance 100 times better than from the Moon to transmit much more than EDRS' 180MB/s. Several primary sources targeting different ground stations can share one mirror. Lasers pumped directly by sunlight as I suggested here on May 24, 2014 may be more efficient http://www.scienceforums.net/topic/76627-solar-thermal-rocket/page-2#entry806581 Marc Schaefer, aka Enthalpy

Hello everybody! This shall be an analog-to-digital converter (Adc) for radiofrequencies. Up to now they're "Flash" converters or similar, whose dynamic range doesn't fulfil radiofrequency desires, because their complexity and the offset and 1/F noise limit them, especially if using the fastest components. My proposal instead uses several stages that are Ac-coupled to remove the offset and 1/F noise, where every stage contributes the digital output and has a modest gain. It is similar to a "logarithmic amplifier" (not the one with an op amp, but the amplifier-detector formerly used in radars and spectrum analyzers) like the LT5538 http://www.linear.com/docs/26333block diagram on page 7 Logarithmic amplifiers cumulate an absolute value of the local amplitude over all stages. As opposed, my converter outputs "strongly negative / weak / strongly positive" at each stage. Consequently, the converter's output is a mu-law if all stages have equal gains https://en.wikipedia.org/wiki/%CE%9C-law_algorithm which fits a wide dynamic range. Some logic determines which amplifier stage is the first to make the signal stronger than the threshold; together with the sign, this is the raw output data. More stages with less individual gain (like 2, or 1.25, or even less) give a finer conversion. Software downstream can represent the mu-law data in a linear way, for instance as float numbers, and optionally adjust the values in accordance to the identified thresholds to improve the conversion linearity. No sample-and-hold is needed. The amplifier chain is extremely fast; the synchronizing flip-flops limit the speed rather. The clock can be dispatched to follow the amplifier chain's delays, as in Analog Devices' old patent for logarithmic amplifiers. Some in-package logic can pack the data to reduce the outside wiring. In this example (two stages only, without the flip-flops nor logic), bipolars ease the stage comparators, but many diagrams are possible, with fast Fet too: The differential triplet in the example has the Bias' high enough that the current flows through the central transistor when the local signal is weak; a signal strong enough at this stage would make one base positive enough to lower one output depending on the signal's sign. Only stability limits the dynamic range. Differential operation helps, as the SO41 showed, and proper supply wiring in the chip too. Grounding the positive supply should improve a bit, and separate supply regulators per stage as well. If breadboarding, beware the Bfr90 and others demand their emitter grounded. Marc Schaefer, aka Enthalpy

I suggested sister crafts orbiting Uranus and Neptune: http://www.scienceforums.net/topic/76627-solar-thermal-rocket/#entry756556 whose mass were inaccurate but improved anyway by my better scenario http://www.scienceforums.net/topic/76627-solar-thermal-rocket/page-2#entry818683 leaving 2t at destination. Even at Neptune, 30.1AU=4.50Tm from the Sun, concentrators can direct the 1.51W/m2 sunlight on solar cells. Fifteen D=4.572m concentrators and 50% efficient multigap cells would produce skinny 176W electricity, but the unfoldable AstroMesh claims to be bigger http://www.northropgrumman.com/businessventures/astroaerospace/products/pages/astromesh.aspx up to D=25m for AM-1 and D=50m for AM-2, of which D=12m has flown. D=25m would provide 352W electricity and D=50m 1408W without plutonium. Their RF reflector is a mesh but I hope some metallized film would reflect sunlight. The manufacturer claims fuzzy ~0.3kg/m2 improving with size, or 150kg for D=25m. The fuzzy shape accuracy and maximum frequency include 26GHz. I wish the Astromesh would concentrate sunlight for my sunheat engine, but I suppose it's not accurate enough. Targeting solar cells is easier. Earth is close to the Sun as seen from Uranus and Neptune, so the same concentrator would double as an antenna, for instance with one secondary mirror made of mesh to redirect only the RF and some mechanical or electronics means to steer the RF or light independently. Modulating the phase and the polarization cumulates again the bandwidth, but with the wideband noise exceeding the signal from Neptune, Hadamard and Reed-Muller soft-decoding gains only as the Log of the band spreading. Marc Schaefer, aka Enthalpy

Collisions were a concern very early in space exploration, say before Pioneers were launched. These told not to worry. Meanwhile, spacecraft have passed through the rings of Saturn without noticing anything, because even these are essentially void. Intercontinental ballistic missiles are tracked by several ground-based radars. Norad has some in northern America, Russia has (or at least the Soviet Union had) one in Krasnoyarsk and elsewhere. The aim was to shelter the populations and start a retaliation strike. Meanwhile it may help trying to intercept the missile. Consequently, people launching a rocket or satellite tell it in advance to the world to avoid misunderstandings and undesireable consequences. Essentially, international organisms attribute a number to every new orbital object. There might be objects that were not declared, but very, very few. Said radars follow every object in low-Earth orbit down to a non-disclosed size that may be like 0.2m, even passive ones. A concern is that they can't detect smaller debris from past collisions. In geosynchronous orbit, they detect every object but bigger, possibly 0.5m. Then, you have just telescopes. Satellites are very visible when they get sunlight but the observer stays in the night and are an annoyance to astronomers. I saw once with naked eyes at ~800km distance a Progress (20m2?) following Mir, so a D=1m telescope instead of a D=1cm eye must see 20cm2 objects at that distance. Small amateur telescope see geosynchronous (36.000km) easily, bigger telescopes and special software see smaller ones. A Lidar must be better, especially if based in space.

It is known experimentally and explained that the drag of a submarine increases at the surface. Even at identical engine power, it gives a higher speed underwater. The extra drag at the surface results from the skull making waves, which doesn't relate with surface tension.

You get a square pulse from several capacitors bridged by inductors, in an LC ladder made of discrete elements that mimick a propagation line. That was used to power radar transmitters. More efficient, no need for synchronization nor software. More exotic material : if you have developed one that Mankind still ignores, fine. If not, forget the hope of finding one in a catalogue. The best induction at saturation is 2.3T for Fe-Co, and that's it - enough people have researched the topic that the outcome is known. Coil construction too is very well known, both with and without a permeable core. If the only goal is strong induction, it consists essentially in packing the conductor as much and as closely to the gap as possible, that is, as a fat ring. Then you have variants if the forces are destructive or if cooling is difficult. Well, again, magnetic design is fully specific to each problem. Once a bozo came to me who wanted to concentrate a magnetostatic field at a distance - is aim was probably a weapon to knock down people at a time the TMS (transcranial magnetic stimulation) began to work at contact distance. My general impression is that you're weaker on electromagntism than on general electronics, and that will be a very hard nut to crack. EM is difficult for real - university courses are only the very beginning of the art. Unfortunately, the more EM designs one has made, the better one realizes that most EM design goals are unrealistic. Have you decided what induction in what volume you want, what magnetic energy it means, and what copper losses it implies? The known optimum design, a fat ring, puts limits on them, which use to be a game stopper for hand-held designs. As an example, I achieved 7T in 50mm*50mm*50mm for 10ms, and this meant almost 10MW copper losses and 1m3 capacitors.

If the fusion of 1 atom = 4e-25kg 233U releases 200MeV = 3e-15J, the fuel itself can accelerate to 100km/s = 0.0004c only, and the complete reactor, spacecraft and load less than that. Nuclear fission and fusion are too weak to achieve a significant fraction of speedlight - the claim of 0.3c, needed to reach stars within a decent delay, is not plausible.

If 10ms duration were possible, capacitor discharge would achieve a higher induction. For a few seconds, only magntic poles and coils. Then both the material limits the induction (2.1T for pure Fe, at the worst places, so 1.8T in air would be an achievement, and 2.3T for Fe50-Co50) and the power. It's essentially a matter of gap length : zero length takes nearly zero watt, but the desired length would often demand MW. Permanent magnets would have been the solution of choice. I can't really imagine an application that excludes them, especially if you accept pole shoes, which have always some remanence and are polarized my Earth's field. Focussing a static magnetic field outside the poles is impossible. One can only bring the flux to the poles' faces and hope it won't spread too much. This is a difficult part of the design, often done with coils close to the gap or aroud the gap - but if only the induction counts and leaks are accepted, then conical poles work. I'm afraid more can't be said without knowing the project better, but designing a magnetic circuit is difficult, takes time to learn, and you'll fail the first 101 to 102 times.

This is how to bring a decommisioned WorldView from Earth orbit to Mars orbit. We couldn't transmit a complete map at full KH-11 resolution anyway. As opposed, the lighter WorldView 2 and 3 add eight colour bands to the panchromatic one, including true colours that are all-important for public support. http://www.satimagingcorp.com/satellite-sensors/worldview-2/ http://www.satpalda.com/product/worldview-2/ https://en.wikipedia.org/wiki/WorldView-2 WorldView-2 nears its end of life, WorldView-3 leaves some years more, and GeoEye is a similar candidate, all with image quality as you experience on Google Earth. The Spot series has a worse resolution, Helios maybe. WorldView-2 moves with 7469m/s at 773km altitude. I hope to put it at sun-synchronous 350km above Mars so 3384m/s there give the same line frequency. This improves the panchromatic resolution from 0.46m to 0.21m and the multispectral one from 1.85m to 0.84m but shrinks the swathe from 16.4km to 7.4km. I checked that a D=1m F=3m paraboloid is perfect within 1nm at 773km and 2.5nm at 350km once the focal plane is adjusted - but what happens to the probable Ritchey-Chrétien is parsecs beyond my skills. ---------- The transport combines chemical propulsion at escape and capture to exploit the Oberth effect with my sunheat engines as I described here on Jul 27, 2014. This table is only from Earth to Mars hence should be clearer: An Atlas V 511 or Delta IV M+(5,2) launches the 8100kg tug to 773km 98° - or an Ariane V, an H-II, maybe a Falcon 9 if building light. The tug grasps the 2300kg empty WorldView. From 10400kg, the transport leaves 4610kg on low Martian orbit, or 2310kg for the used tug. The engine is a throttled down RL-10, possibly without nozzle extension. Since 8kN thrust would suffice, it could be a 1bar pressure-fed or 70bar electrically pumped design. 10 sunheat engines take about 10 months at Earth and at Mars. A slingshot at our Moon could spare the chemical engine. Capture at Mars by the sunheat engine remains less efficient, but the craft is simpler. ---------- To the KH-11 design, the tug adds the RL-10, a small oxygen tank (toroid on the sketch) and is smaller. Scaled like the hydrogen volume, the tank and truss would weigh 670kg, ten sunheat engines 250kg, the RL-10 300kg, leaving 1t for the equipment bay. ---------- At Mars, two engines control the attitude and orbit, four supply electricity through solar cells of varied bandgaps, four make the already described transmission. The 4/7 weaker received field loses one bit, from 6 to 5 per polarisation and phase, transmitting mean 45MB/s including the eclipses. The 0.21m 11bit complete map transmits in 4.1 years uncompressed, and the 8 colour channels with 4*4 times less resolution in 2.0 years. Marc Schaefer, aka Enthalpy

Astronomers make from Earth images of Mars using radiotelescopes as radars, especially at Arecibo, so a first answer would be: Radius of Mars=3390km and satellite=5m (ratio 4.6e11) Distance of Mars=80Gm (nearest) hence satellite = 100Mm (ratio 823, acts as 4th power) pessimistic, because Mars images have several pixels, and satellites contain metal. Or: the D=0.5m F/A-15 radar sees the R=1700km Moon at 384Mm distance, so a D=50m radar sees a R=5m object at 66Mm. Better values exist at Near-Earth Object detectors, at Norad, and the like. Let's evaluate a tailored setup. Transmit 1MW at 30GHz from a D=70m antenna. Could be more. A 10m2 target at arbitrary 1Gm receives 250µW, wow, and reradiates them uniformly over 2pi srd (half space). A D=70m antenna receives 10-19W=-160dBm, wow again. The receiver is far enough from the transmitter that Earth shields it. Integrate with phase coherency for 1min: the signal energy is 6e-18J, the noise energy at 30K noise temperature is 2e-22J. So for S/N=14dB, you obtain 6Gm or 15* Earth-Moon. Or integrate with phase coherency (how?) for 10h. The signal energy from 1Gm is 3.6e-15J, so at S/N=14dB the range is 30GM or 0.2* Sun-Earth. Phase coherency for 1min and intercorrelation of two receiving antennas give a range between both. You can exaggerate a bit. Antenna fields cumulate more transmit and receive area and transmit more power. A corner reflector at the satellite would gain a lot. A Lidar is obviously better. It too measures distances and speeds. Some asteroids were already announced that would have been impossible to detect with a radar.

Helios was developed years ago, when blue Led weren't so common, and satellite designers are extremely conservative people. So the light source was more probably a xenon lamp, needing <10 times more electricity, or <36kWe from time to time - easy.

Distance is measured by the two-way propagation delay. On such distant objects an active transponder is needed, so all technological delays are first identified. Satellite designers don't want to carry an atomic clock if the payload needs none, and anything less accurate would be too inaccurate. Imagine a 10-11 clock: after 2 years of Earth-Venus-Venus-Earth flybies, it has drifted by 300µs or 2*50km, not good enough for an Earth flyby. Some craft are designed with an accurate transponder whose delay is well identified before launch. Others have relaxed needs. Also, satellite designers wanted to know the distance at a time (1950) when putting atomic clocks onboard was excluded, but radars were already routine.

Figures - the beginning of science. My white LED flashlight consumes 360mW and, put at the proper distance to illuminate D=10m, makes a spot quite visible under full moonlight. Cameras not so exceptional make perfect pictures under full moonlight conditions. I've seen some pictures made under starlight only, which is much more dificult, but these were noisy and the exposure time wouldn't fit the speed of a satellite. The LED's performance permits to illuminate even D=1km: that's more than is needed to check where one person is for instance. This takes 10,000 times more power than D=10m, or 3.6kW from time to time: easy. At diffraction limit, a lens would need D~2mm - it's rather the emission area of the LED that defines the spot. This permits many lenses, maybe one per LED. Or, since most observation satellites observe many columns at a time but one single row that sweeps as the satellite moves, the lighting as well can cover 10km width and 100m length - or even less to consume less. A different option is to reflect sunlight, when available, to the ground. D=2m, not very flat, suffice. Tilting the satellite when it takes high-resolution images, as is done commonly with spy satellites to reduce the apparent ground speed, needs to orient the reflector too. The white flashlight construction wouldn't be taken 1:1 at a satellite. It's a deep blue LED with green and red phosphors of varied divergence. LED specialized for each colour of the satellite's sensor seem a better fit - or semiconductor lasers, whose beam is easier to tailor. I don't believe I saw sunlight reflected by a satellite. There was a satellite at that place after the lighting went off, as I saw because the satellite was outside Earth's shadow, but the intense light was about as strong as an Iridium flash, needing a very flat illuminated area, and lasted much longer than an Iridium flash, which contradicts the flat area. For various reasons I suppose the craft was French, typically a Helios.