0 00:00:00,000 --> 00:00:30,000 Dear viewer, these subtitles were generated by a machine via the service Trint and therefore are (very) buggy. If you are capable, please help us to create good quality subtitles: https://c3subtitles.de/talk/913 Thanks! 1 00:00:15,140 --> 00:00:17,239 For the first time today, we're going 2 00:00:17,240 --> 00:00:19,789 to visualize electromagnetic radiation 3 00:00:19,790 --> 00:00:22,639 in the range of the Wi-Fi frequencies, 4 00:00:22,640 --> 00:00:24,829 we have Friedemann Rhinehart from 5 00:00:24,830 --> 00:00:26,609 the Munich Technical Institute. 6 00:00:26,610 --> 00:00:28,669 He has these 7 00:00:28,670 --> 00:00:30,019 junior research project. 8 00:00:30,020 --> 00:00:31,789 And with one of his students, he pursued 9 00:00:31,790 --> 00:00:34,009 this project. The student is Philip Hall 10 00:00:34,010 --> 00:00:36,289 and the stage is yours. 11 00:00:36,290 --> 00:00:37,579 Thanks. 12 00:00:37,580 --> 00:00:38,580 Well. 13 00:00:43,620 --> 00:00:46,169 So good morning and 14 00:00:46,170 --> 00:00:48,479 thanks for having me here on this 15 00:00:48,480 --> 00:00:50,579 Congress, which I enjoyed very 16 00:00:50,580 --> 00:00:52,149 much so far. 17 00:00:52,150 --> 00:00:54,269 So in the talk that 18 00:00:54,270 --> 00:00:56,429 I would like to give, I would 19 00:00:56,430 --> 00:00:58,499 like to look at 20 00:00:58,500 --> 00:01:00,629 wireless signals from a very 21 00:01:00,630 --> 00:01:02,969 different perspective. 22 00:01:02,970 --> 00:01:05,159 When we think of 23 00:01:05,160 --> 00:01:07,439 wireless signals, we usually think 24 00:01:07,440 --> 00:01:10,139 of data packets, 25 00:01:10,140 --> 00:01:13,379 messages, run times, 26 00:01:13,380 --> 00:01:14,910 protocols and so on. 27 00:01:16,050 --> 00:01:18,089 But we can look at them from a very 28 00:01:18,090 --> 00:01:20,369 different point of view, from 29 00:01:20,370 --> 00:01:22,679 the point of view that 30 00:01:22,680 --> 00:01:24,420 it's actually just light. 31 00:01:25,660 --> 00:01:27,749 This is kind of a bold statement, 32 00:01:27,750 --> 00:01:29,609 but it's actually a fairly exact 33 00:01:29,610 --> 00:01:31,169 statement. So from a physicist's 34 00:01:31,170 --> 00:01:33,659 perspective, a wireless signal 35 00:01:33,660 --> 00:01:35,849 is virtually the same as the beam 36 00:01:35,850 --> 00:01:37,679 of a laser or the light coming from the 37 00:01:37,680 --> 00:01:38,729 sun. 38 00:01:38,730 --> 00:01:40,919 It's electromagnetic radiation, which 39 00:01:40,920 --> 00:01:42,779 is described by Maxwell's equations and 40 00:01:42,780 --> 00:01:43,769 so on. 41 00:01:43,770 --> 00:01:45,869 So it's an oscillating electric 42 00:01:45,870 --> 00:01:46,900 and magnetic field. 43 00:01:48,180 --> 00:01:50,279 The only difference to visible 44 00:01:50,280 --> 00:01:52,409 light that we can see with our eyes 45 00:01:52,410 --> 00:01:54,869 is that the wavelength of these signals 46 00:01:54,870 --> 00:01:56,159 is much longer. 47 00:01:56,160 --> 00:01:58,379 So while the visible light that 48 00:01:58,380 --> 00:02:00,809 makes up the visible world 49 00:02:00,810 --> 00:02:02,669 has a wavelength on the range in the 50 00:02:02,670 --> 00:02:05,099 range of a micrometer, the 51 00:02:05,100 --> 00:02:07,109 wireless signals that we typically employ 52 00:02:07,110 --> 00:02:09,659 in our devices have a 53 00:02:09,660 --> 00:02:11,549 wavelength in the range of a few 54 00:02:11,550 --> 00:02:13,619 centimeters to a few tens of 55 00:02:13,620 --> 00:02:14,939 centimeters. 56 00:02:14,940 --> 00:02:16,949 And as a second very important 57 00:02:16,950 --> 00:02:18,870 difference, we cannot see them. 58 00:02:20,110 --> 00:02:21,110 So. 59 00:02:23,420 --> 00:02:24,979 But from a physicists perspective, this 60 00:02:24,980 --> 00:02:27,559 is very interesting because 61 00:02:27,560 --> 00:02:30,739 that has somewhat surprising implication, 62 00:02:30,740 --> 00:02:33,139 it has the implication that 63 00:02:33,140 --> 00:02:35,329 our modern world is actually brightly 64 00:02:35,330 --> 00:02:37,759 lit by all the wireless devices 65 00:02:37,760 --> 00:02:39,709 that we use every day. 66 00:02:39,710 --> 00:02:42,619 So this whole room as we sit here, 67 00:02:42,620 --> 00:02:46,249 it's actually glowing in various colors. 68 00:02:46,250 --> 00:02:48,559 Only we can't see it because it's glowing 69 00:02:48,560 --> 00:02:50,719 somewhere in the microwave radio 70 00:02:50,720 --> 00:02:51,720 domain. 71 00:02:52,790 --> 00:02:55,789 And this also implies that 72 00:02:55,790 --> 00:02:57,949 there is kind of a potential big 73 00:02:57,950 --> 00:02:59,899 security leak associated with these 74 00:02:59,900 --> 00:03:01,999 wireless devices in the 75 00:03:02,000 --> 00:03:05,029 sense that any device that you use, 76 00:03:05,030 --> 00:03:08,179 no matter how well you encrypt your data, 77 00:03:08,180 --> 00:03:10,429 is going to transmit a 78 00:03:10,430 --> 00:03:12,559 full three dimensional picture of 79 00:03:12,560 --> 00:03:15,199 the world to 80 00:03:15,200 --> 00:03:18,049 your environment no matter what you do. 81 00:03:18,050 --> 00:03:20,359 And so 82 00:03:20,360 --> 00:03:22,579 when we realized that 83 00:03:22,580 --> 00:03:24,679 we got intrigued and we 84 00:03:24,680 --> 00:03:26,839 wondered whether there would be a way 85 00:03:26,840 --> 00:03:29,869 to actually record that radiation 86 00:03:29,870 --> 00:03:32,659 and to reveal that picture 87 00:03:32,660 --> 00:03:35,599 to, in a way, built a camera 88 00:03:35,600 --> 00:03:37,759 that would be sensitive to this kind 89 00:03:37,760 --> 00:03:39,889 of radiation and that would enable 90 00:03:39,890 --> 00:03:42,049 us to see the world in the way 91 00:03:42,050 --> 00:03:44,269 that we would see it if our eyes 92 00:03:44,270 --> 00:03:46,399 were sensitive in this 93 00:03:46,400 --> 00:03:48,169 frequency domain. 94 00:03:48,170 --> 00:03:51,049 And so to cut a long story short, 95 00:03:51,050 --> 00:03:52,789 we managed to do that. 96 00:03:52,790 --> 00:03:54,859 And this is the result 97 00:03:54,860 --> 00:03:56,209 that we got. 98 00:03:56,210 --> 00:03:59,089 This is a picture of our lab. 99 00:03:59,090 --> 00:04:01,249 And the colorful thing that you 100 00:04:01,250 --> 00:04:03,529 see is wireless 101 00:04:03,530 --> 00:04:05,659 radiation coming from a router on 102 00:04:05,660 --> 00:04:07,969 the back side of the room 103 00:04:07,970 --> 00:04:09,589 that we visualize with our method. 104 00:04:09,590 --> 00:04:11,749 And we can actually see it propagate 105 00:04:11,750 --> 00:04:12,739 in space. 106 00:04:12,740 --> 00:04:15,169 And we can even see that objects 107 00:04:15,170 --> 00:04:17,539 that we place in the lab cast a shadow 108 00:04:17,540 --> 00:04:18,749 in that beam. 109 00:04:18,750 --> 00:04:20,838 And in what follows, 110 00:04:20,839 --> 00:04:23,239 I would like to walk you through 111 00:04:23,240 --> 00:04:26,119 how we did that and 112 00:04:26,120 --> 00:04:28,819 what it might be used for, for. 113 00:04:28,820 --> 00:04:30,949 And and 114 00:04:30,950 --> 00:04:33,139 I also would like to discuss whether it's 115 00:04:33,140 --> 00:04:35,660 actually a serious security concern. 116 00:04:37,130 --> 00:04:39,319 But before I do that, I would like 117 00:04:39,320 --> 00:04:42,079 to start from a somewhat 118 00:04:42,080 --> 00:04:44,479 broader overview and tell you a bit 119 00:04:44,480 --> 00:04:46,699 more about what has 120 00:04:46,700 --> 00:04:49,369 happened in the past few years 121 00:04:49,370 --> 00:04:50,299 in this field. 122 00:04:50,300 --> 00:04:52,519 So it turns out that hacking 123 00:04:52,520 --> 00:04:54,769 wireless signals in the sense 124 00:04:54,770 --> 00:04:57,259 that thinking in the sense of thinking 125 00:04:57,260 --> 00:04:59,599 about other ways, other useful 126 00:04:59,600 --> 00:05:01,849 data that we can extract from 127 00:05:01,850 --> 00:05:04,189 that signals that are omnipresent 128 00:05:04,190 --> 00:05:06,259 in our modern world, it 129 00:05:06,260 --> 00:05:08,329 has become quite a vibrant area of 130 00:05:08,330 --> 00:05:10,729 research in the past five years. 131 00:05:10,730 --> 00:05:12,799 And we haven't been the only ones 132 00:05:12,800 --> 00:05:14,959 that realized that this is 133 00:05:14,960 --> 00:05:16,549 actually a very interesting thing to look 134 00:05:16,550 --> 00:05:18,829 at. So there have been many other labs 135 00:05:18,830 --> 00:05:20,989 that have pursued many different 136 00:05:20,990 --> 00:05:23,239 directions to look 137 00:05:23,240 --> 00:05:25,399 at Wi-Fi radiation from a different point 138 00:05:25,400 --> 00:05:27,469 of view. And maybe the 139 00:05:27,470 --> 00:05:29,539 most famous result is this. 140 00:05:29,540 --> 00:05:31,609 This is a result from a group 141 00:05:31,610 --> 00:05:33,829 at MIT in 142 00:05:33,830 --> 00:05:35,060 Boston, Massachusetts, 143 00:05:36,320 --> 00:05:38,599 and they actually 144 00:05:38,600 --> 00:05:41,029 have Wi-Fi routers in the way that 145 00:05:41,030 --> 00:05:43,549 they exploit the fact 146 00:05:43,550 --> 00:05:44,870 that modern routers 147 00:05:45,920 --> 00:05:47,989 have an array of antennas 148 00:05:47,990 --> 00:05:50,839 and they use that to actually 149 00:05:50,840 --> 00:05:53,119 send beams in different directions 150 00:05:53,120 --> 00:05:55,579 and to focus the beams on the devices 151 00:05:55,580 --> 00:05:57,649 that they want to interact with. 152 00:05:57,650 --> 00:05:59,899 So modern routers, they can actually send 153 00:05:59,900 --> 00:06:02,509 two different signals to 154 00:06:02,510 --> 00:06:04,459 cell phones on the right and on the left 155 00:06:04,460 --> 00:06:05,460 end of the room. 156 00:06:06,320 --> 00:06:08,149 And you can check that. 157 00:06:08,150 --> 00:06:10,309 You can use that to scan a beam 158 00:06:10,310 --> 00:06:12,439 and two dimensions and 159 00:06:12,440 --> 00:06:14,209 look at the reflections. 160 00:06:14,210 --> 00:06:16,369 So in a way to build a 161 00:06:16,370 --> 00:06:18,829 phased array radar, if you wish. 162 00:06:18,830 --> 00:06:21,079 And this is the kind of picture 163 00:06:21,080 --> 00:06:23,659 that you get with that. So you can scan 164 00:06:23,660 --> 00:06:25,009 a beam through the room. 165 00:06:25,010 --> 00:06:26,669 You can look at the reflections and you 166 00:06:26,670 --> 00:06:28,429 get a fairly detailed picture of the 167 00:06:28,430 --> 00:06:31,099 world. You can even vaguely resolve 168 00:06:31,100 --> 00:06:32,100 human beings. 169 00:06:33,870 --> 00:06:36,269 You can also just 170 00:06:36,270 --> 00:06:38,489 pursue a simpler approach, you can only 171 00:06:38,490 --> 00:06:40,949 map attenuation, 172 00:06:40,950 --> 00:06:43,259 so you take a Wi-Fi 173 00:06:43,260 --> 00:06:45,359 router and the Wi-Fi receiver 174 00:06:45,360 --> 00:06:47,729 and you map the attenuation 175 00:06:47,730 --> 00:06:50,069 of the signal between the two of them 176 00:06:50,070 --> 00:06:52,259 and you can then mount them 177 00:06:52,260 --> 00:06:54,449 on a movable platform. 178 00:06:54,450 --> 00:06:56,819 So as a fancy example, 179 00:06:56,820 --> 00:06:59,129 for instance, a drone so you can 180 00:06:59,130 --> 00:07:01,549 fly these two šemeta 181 00:07:01,550 --> 00:07:04,109 and receive a drones around a building 182 00:07:04,110 --> 00:07:06,749 map, the attenuation on 183 00:07:06,750 --> 00:07:09,359 every line of sight that you 184 00:07:09,360 --> 00:07:10,709 covered during that flight 185 00:07:11,910 --> 00:07:14,099 to tomography on that data. 186 00:07:14,100 --> 00:07:16,319 And by doing that, you can 187 00:07:16,320 --> 00:07:19,139 actually extract three dimensional 188 00:07:19,140 --> 00:07:21,479 views of what has been in between 189 00:07:21,480 --> 00:07:22,480 these drones. 190 00:07:24,870 --> 00:07:27,239 So there has been a lot of activity 191 00:07:27,240 --> 00:07:28,499 in that field. 192 00:07:28,500 --> 00:07:30,959 But still, this previous 193 00:07:30,960 --> 00:07:33,479 work did not yet fully 194 00:07:35,670 --> 00:07:37,439 answer this question that we got 195 00:07:37,440 --> 00:07:39,629 interested in, like what would the world 196 00:07:39,630 --> 00:07:41,369 look with wi fi eyes? 197 00:07:41,370 --> 00:07:43,469 And if you 198 00:07:43,470 --> 00:07:45,569 look at that question, the state 199 00:07:45,570 --> 00:07:47,789 of the art had been pretty much 200 00:07:47,790 --> 00:07:49,679 that we had something like the compound 201 00:07:49,680 --> 00:07:51,389 eye of an insect. 202 00:07:51,390 --> 00:07:53,699 There have been many ways to 203 00:07:53,700 --> 00:07:56,009 tamper with wireless radiations 204 00:07:56,010 --> 00:07:59,039 in in the sense that we 205 00:07:59,040 --> 00:08:00,929 look at different directions, either by 206 00:08:00,930 --> 00:08:02,819 directing a beam into some direction or 207 00:08:02,820 --> 00:08:04,919 by looking at attenuation 208 00:08:04,920 --> 00:08:06,989 in a certain direction and 209 00:08:06,990 --> 00:08:09,359 do that for multiple angles to 210 00:08:09,360 --> 00:08:11,619 get a picture of the world. 211 00:08:11,620 --> 00:08:13,769 And so we got interested in whether we 212 00:08:13,770 --> 00:08:15,899 could actually get a full, three 213 00:08:15,900 --> 00:08:18,209 dimensional view of the full 214 00:08:18,210 --> 00:08:20,489 wavefront and all the light propagating 215 00:08:20,490 --> 00:08:22,889 in space, pretty much in the same way 216 00:08:22,890 --> 00:08:24,749 as we can do it with visible light. 217 00:08:26,530 --> 00:08:28,689 So we thought about that 218 00:08:28,690 --> 00:08:30,819 and it turned out that you can do it and 219 00:08:30,820 --> 00:08:32,918 you can do it in a surprisingly simple 220 00:08:32,919 --> 00:08:35,109 way. So there is a technique 221 00:08:35,110 --> 00:08:38,168 that essentially solves the problem 222 00:08:38,169 --> 00:08:39,788 and it's holography. 223 00:08:39,789 --> 00:08:42,339 So holography holograms 224 00:08:42,340 --> 00:08:44,469 is something that you probably all have 225 00:08:44,470 --> 00:08:46,659 seen. It's these amazing 226 00:08:46,660 --> 00:08:48,759 pictures that appear three 227 00:08:48,760 --> 00:08:50,949 dimensional if you look at them and that 228 00:08:50,950 --> 00:08:53,109 you can tilt and you can actually look 229 00:08:53,110 --> 00:08:55,389 at objects from different angles. 230 00:08:55,390 --> 00:08:57,459 They do not look like a photograph. 231 00:08:57,460 --> 00:08:59,769 They look much more like 232 00:08:59,770 --> 00:09:01,899 a window into a virtual world 233 00:09:01,900 --> 00:09:04,089 that has been frozen by taking this 234 00:09:04,090 --> 00:09:05,559 hologram. 235 00:09:05,560 --> 00:09:07,959 And you can make them actually 236 00:09:07,960 --> 00:09:10,029 in a fairly simple way or you don't 237 00:09:10,030 --> 00:09:11,589 even need a lens. 238 00:09:11,590 --> 00:09:13,479 All you need to do to record such a 239 00:09:13,480 --> 00:09:15,669 hologram is illuminate 240 00:09:15,670 --> 00:09:18,099 your object with what we call coherent 241 00:09:18,100 --> 00:09:20,409 light and get to that in a second 242 00:09:20,410 --> 00:09:22,749 and record it on 243 00:09:22,750 --> 00:09:24,639 a face sensitive camera. 244 00:09:26,050 --> 00:09:28,689 So these are very technical terms. 245 00:09:28,690 --> 00:09:30,909 Let me try to explain that in a 246 00:09:30,910 --> 00:09:32,499 more visual way. 247 00:09:32,500 --> 00:09:35,079 So in a way, a hologram 248 00:09:35,080 --> 00:09:37,030 is a photograph to zero. 249 00:09:38,920 --> 00:09:41,469 It captures more information than 250 00:09:41,470 --> 00:09:43,629 a two dimensional photograph, which 251 00:09:43,630 --> 00:09:44,890 we usually take, 252 00:09:46,360 --> 00:09:48,549 which where information is restricted 253 00:09:48,550 --> 00:09:51,189 in the sense that the usual photograph 254 00:09:51,190 --> 00:09:53,469 only captures intensity 255 00:09:53,470 --> 00:09:55,539 of light so 256 00:09:55,540 --> 00:09:58,359 it can record in a two dimensional plane 257 00:09:58,360 --> 00:10:00,429 where there was much light and 258 00:10:00,430 --> 00:10:02,169 where there was very little light. 259 00:10:02,170 --> 00:10:04,329 And the resulting photo will be very 260 00:10:04,330 --> 00:10:05,529 bright and very dark. 261 00:10:05,530 --> 00:10:07,779 But it's going to be restricted to a two 262 00:10:07,780 --> 00:10:09,519 dimensional plane. It will appear as a 263 00:10:09,520 --> 00:10:11,889 flat piece of paper. 264 00:10:11,890 --> 00:10:12,890 If you look at it, 265 00:10:14,030 --> 00:10:15,639 a hologram can do more. 266 00:10:15,640 --> 00:10:18,249 So a hologram is a face coherent 267 00:10:18,250 --> 00:10:19,389 recording. 268 00:10:19,390 --> 00:10:21,459 And that, in a visual way means 269 00:10:21,460 --> 00:10:23,769 that not only records 270 00:10:23,770 --> 00:10:26,169 brightness of the light, it also 271 00:10:26,170 --> 00:10:28,449 records the direction of 272 00:10:28,450 --> 00:10:31,059 the light beams hitting this hologram 273 00:10:31,060 --> 00:10:32,739 during exposure. 274 00:10:32,740 --> 00:10:34,899 So in a hologram, in a way, you 275 00:10:34,900 --> 00:10:37,419 freeze all the light rates that enter 276 00:10:37,420 --> 00:10:39,159 this specific plane. 277 00:10:39,160 --> 00:10:41,379 And if you develop 278 00:10:41,380 --> 00:10:43,869 that hologram, you in a way revive 279 00:10:43,870 --> 00:10:45,939 all these light beams again. 280 00:10:45,940 --> 00:10:48,369 And so when we look at the hologram, 281 00:10:48,370 --> 00:10:50,529 we really look at the very same 282 00:10:50,530 --> 00:10:53,109 light rays at the very same light field 283 00:10:53,110 --> 00:10:54,609 that came from the object during 284 00:10:54,610 --> 00:10:55,599 exposure. 285 00:10:55,600 --> 00:10:57,609 And that in particular implies that we 286 00:10:57,610 --> 00:10:59,709 can look at it from various angles and 287 00:10:59,710 --> 00:11:02,229 we can have a three dimensional 288 00:11:02,230 --> 00:11:03,230 view of it. 289 00:11:04,840 --> 00:11:07,239 So how is 290 00:11:07,240 --> 00:11:08,589 that done? 291 00:11:08,590 --> 00:11:10,929 So this is done by 292 00:11:10,930 --> 00:11:13,479 a phase coherent recording 293 00:11:13,480 --> 00:11:15,729 and that is this this 294 00:11:15,730 --> 00:11:17,649 technical term essentially captures this 295 00:11:17,650 --> 00:11:19,929 idea of recording 296 00:11:19,930 --> 00:11:22,689 direction along 297 00:11:22,690 --> 00:11:24,399 with intensity. 298 00:11:24,400 --> 00:11:27,789 And to see how that works. 299 00:11:27,790 --> 00:11:29,979 I have to well, get back to the point 300 00:11:29,980 --> 00:11:32,109 that light actually 301 00:11:32,110 --> 00:11:34,989 is electromagnetic waves 302 00:11:34,990 --> 00:11:37,119 and waves, pretty much like waves 303 00:11:37,120 --> 00:11:39,429 on a lake, have valleys 304 00:11:39,430 --> 00:11:40,779 and hills. 305 00:11:40,780 --> 00:11:42,999 So areas where electric field is very 306 00:11:43,000 --> 00:11:44,979 strong and areas where electric field is 307 00:11:44,980 --> 00:11:47,169 very weak and these 308 00:11:47,170 --> 00:11:49,539 valleys and peaks propagate in space. 309 00:11:50,960 --> 00:11:53,989 So if you throw a stone into a lake 310 00:11:53,990 --> 00:11:56,089 and you watch the resulting 311 00:11:56,090 --> 00:11:58,819 waves, you will notice 312 00:11:58,820 --> 00:12:01,639 that these waves always 313 00:12:01,640 --> 00:12:03,889 move perpendicular to both 314 00:12:03,890 --> 00:12:05,059 the peaks and valleys. 315 00:12:06,410 --> 00:12:08,509 So the direction of 316 00:12:08,510 --> 00:12:11,419 the wave is in a way encoded 317 00:12:11,420 --> 00:12:13,339 in the difference between these peaks and 318 00:12:13,340 --> 00:12:14,689 valleys. 319 00:12:14,690 --> 00:12:16,819 So if you can take a photograph 320 00:12:16,820 --> 00:12:18,889 where you also 321 00:12:18,890 --> 00:12:21,079 register whether the peak came first 322 00:12:21,080 --> 00:12:23,239 or the valley came first, you 323 00:12:23,240 --> 00:12:26,179 also register direction 324 00:12:26,180 --> 00:12:29,119 and that's 325 00:12:29,120 --> 00:12:31,399 what it takes. So it takes a light source 326 00:12:31,400 --> 00:12:33,469 where these peaks and valleys are very 327 00:12:33,470 --> 00:12:34,369 well defined. 328 00:12:34,370 --> 00:12:36,109 This is what we call a coherent light 329 00:12:36,110 --> 00:12:38,479 source. And it takes a very sensitive 330 00:12:38,480 --> 00:12:40,759 camera, a camera that can 331 00:12:40,760 --> 00:12:42,919 record whether the peaks of the valley 332 00:12:42,920 --> 00:12:43,920 came first. 333 00:12:45,410 --> 00:12:47,779 Now, it turns out that this is a 334 00:12:47,780 --> 00:12:49,609 fairly difficult thing to do in the 335 00:12:49,610 --> 00:12:51,889 optical domain, people 336 00:12:51,890 --> 00:12:54,859 managed to do it, but it was 337 00:12:54,860 --> 00:12:56,629 such a big breakthrough that it actually 338 00:12:56,630 --> 00:12:58,879 got awarded a Nobel Prize 339 00:12:58,880 --> 00:12:59,880 40 years ago. 340 00:13:02,030 --> 00:13:04,279 Fortunately, it's actually 341 00:13:04,280 --> 00:13:06,679 much simpler in the 342 00:13:06,680 --> 00:13:08,749 radio frequency domain 343 00:13:08,750 --> 00:13:10,939 because the frequency here 344 00:13:10,940 --> 00:13:13,639 is much lower than for visible light 345 00:13:13,640 --> 00:13:16,099 and so we can actually record 346 00:13:16,100 --> 00:13:18,289 every peak and every value 347 00:13:18,290 --> 00:13:20,689 simply by registering 348 00:13:20,690 --> 00:13:23,299 the wave with a good oscilloscope 349 00:13:24,470 --> 00:13:25,470 and. 350 00:13:27,370 --> 00:13:28,899 That, in particular means that 351 00:13:30,130 --> 00:13:32,649 it should be possible to do holography 352 00:13:32,650 --> 00:13:34,959 of this wireless radiation, so to come 353 00:13:34,960 --> 00:13:37,269 up with an experimental setup 354 00:13:37,270 --> 00:13:39,489 where we take a picture 355 00:13:39,490 --> 00:13:42,069 in some two dimensional plane of space, 356 00:13:42,070 --> 00:13:44,799 we register the brightness 357 00:13:44,800 --> 00:13:46,600 and phase of the 358 00:13:47,740 --> 00:13:50,289 microwave light at every point in space, 359 00:13:50,290 --> 00:13:52,899 and we can then later on use that to 360 00:13:52,900 --> 00:13:56,039 reconstruct a virtual view of the world. 361 00:13:56,040 --> 00:13:58,119 And this is pretty much how we did 362 00:13:58,120 --> 00:14:00,849 it. So this is our experimental setup. 363 00:14:00,850 --> 00:14:02,979 We explicitly 364 00:14:02,980 --> 00:14:05,559 wanted to capture light from arbitrary 365 00:14:05,560 --> 00:14:07,959 devices. So the 366 00:14:07,960 --> 00:14:10,239 light sources that we used always were 367 00:14:10,240 --> 00:14:12,399 commercial off the shelf 368 00:14:12,400 --> 00:14:14,589 Wi-Fi routers that we 369 00:14:14,590 --> 00:14:16,629 asked to download a big video from 370 00:14:16,630 --> 00:14:18,579 YouTube to generate a lot of traffic and 371 00:14:18,580 --> 00:14:20,559 data that we could image. 372 00:14:20,560 --> 00:14:22,659 But we make no assumption whatsoever on 373 00:14:22,660 --> 00:14:24,279 the signal that they actually emit. 374 00:14:24,280 --> 00:14:25,749 It could be anything. It could even be 375 00:14:25,750 --> 00:14:26,750 encrypted. 376 00:14:27,950 --> 00:14:30,199 We sometimes play some objects 377 00:14:30,200 --> 00:14:32,179 in the beam to get a more interesting 378 00:14:32,180 --> 00:14:34,639 picture, and we finally 379 00:14:34,640 --> 00:14:37,429 record this 380 00:14:37,430 --> 00:14:39,649 beam of microwave light 381 00:14:39,650 --> 00:14:42,259 in a scanning aperture approach. 382 00:14:42,260 --> 00:14:44,479 So unfortunately, it's very, very 383 00:14:44,480 --> 00:14:46,969 expensive to to build microwave 384 00:14:46,970 --> 00:14:49,099 cameras. You would need a huge array of 385 00:14:49,100 --> 00:14:51,379 antennas. So in a first 386 00:14:51,380 --> 00:14:53,089 step, it's easier to just take one 387 00:14:53,090 --> 00:14:55,249 antenna and scan it across every pixel 388 00:14:55,250 --> 00:14:56,250 that you want to emit. 389 00:14:57,650 --> 00:14:59,449 This is what we did. 390 00:14:59,450 --> 00:15:01,639 So we have an antenna that we mounted 391 00:15:01,640 --> 00:15:03,829 on a scanning platform and 392 00:15:03,830 --> 00:15:06,589 we reference that signal to 393 00:15:06,590 --> 00:15:08,720 the signal of a stationary antenna. 394 00:15:10,350 --> 00:15:12,839 This is what it looked like in real life, 395 00:15:12,840 --> 00:15:14,999 so it's you see, it's not 396 00:15:15,000 --> 00:15:17,189 very fancy, it's actually it's probably 397 00:15:17,190 --> 00:15:18,929 the cheapest experiment I've ever been 398 00:15:18,930 --> 00:15:20,489 involved in. 399 00:15:20,490 --> 00:15:23,069 And we used plywood 400 00:15:23,070 --> 00:15:25,259 and fishtank and tools like 401 00:15:25,260 --> 00:15:27,569 that to do it essentially 402 00:15:27,570 --> 00:15:29,549 to solve this problem of scanning the 403 00:15:29,550 --> 00:15:31,799 antenna through our lab. 404 00:15:31,800 --> 00:15:33,929 But at some point we were quite well 405 00:15:33,930 --> 00:15:35,489 and so we could record these pictures. 406 00:15:36,600 --> 00:15:38,789 So the signal that we get from 407 00:15:38,790 --> 00:15:41,129 the set up is 408 00:15:41,130 --> 00:15:43,949 a wave, a wave that we can record 409 00:15:43,950 --> 00:15:46,109 with an oscilloscope and 410 00:15:46,110 --> 00:15:48,539 that we actually record twice, once 411 00:15:48,540 --> 00:15:50,639 coming from the scanning antenna. 412 00:15:50,640 --> 00:15:52,409 This is what is going to make up the 413 00:15:52,410 --> 00:15:54,749 image data later on and once 414 00:15:54,750 --> 00:15:56,190 from this reference antenna. 415 00:15:57,930 --> 00:15:58,930 So. 416 00:16:00,390 --> 00:16:02,369 We make no assumption whatsoever on the 417 00:16:02,370 --> 00:16:04,439 signal that is transmitted, 418 00:16:04,440 --> 00:16:07,319 so we cannot rely on knowing 419 00:16:07,320 --> 00:16:08,789 what kind of bits and bytes are 420 00:16:08,790 --> 00:16:11,129 transmitted here, but it turns 421 00:16:11,130 --> 00:16:13,349 out that if you Fourier transform 422 00:16:13,350 --> 00:16:15,119 these signals and you divide them by each 423 00:16:15,120 --> 00:16:17,369 other, you can, in a way, 424 00:16:17,370 --> 00:16:20,369 normalize that signal to that reference 425 00:16:20,370 --> 00:16:22,469 and get 426 00:16:22,470 --> 00:16:24,749 a virtual second wave 427 00:16:24,750 --> 00:16:26,819 where the 428 00:16:26,820 --> 00:16:28,889 bit pattern actually cancels. 429 00:16:28,890 --> 00:16:31,079 And you only record 430 00:16:31,080 --> 00:16:33,259 the, um, the 431 00:16:33,260 --> 00:16:36,599 the face delay and the attenuation 432 00:16:36,600 --> 00:16:38,579 of that wave as it travels to the 433 00:16:38,580 --> 00:16:40,559 scanning antenna. 434 00:16:40,560 --> 00:16:42,839 So for every pixel we record a set 435 00:16:42,840 --> 00:16:44,999 of data, like on the upper left, 436 00:16:45,000 --> 00:16:47,159 we do some processing to 437 00:16:47,160 --> 00:16:49,439 remove the bit pattern and only 438 00:16:49,440 --> 00:16:51,599 get the coherent wave. 439 00:16:51,600 --> 00:16:53,699 And then we end up with something on 440 00:16:53,700 --> 00:16:55,919 the like on the 441 00:16:55,920 --> 00:16:57,299 lower part of this plot. 442 00:16:57,300 --> 00:16:59,609 So we do that for every pixel, 443 00:16:59,610 --> 00:17:01,049 for every pixel we get. 444 00:17:02,130 --> 00:17:04,439 The brightness, the amplitude 445 00:17:04,440 --> 00:17:07,618 and the face, the direction of the beam, 446 00:17:07,619 --> 00:17:09,509 we actually get that for a wide range of 447 00:17:09,510 --> 00:17:11,759 frequencies because Wi-Fi 448 00:17:11,760 --> 00:17:12,760 is 449 00:17:13,950 --> 00:17:15,149 multi frequency scheme. 450 00:17:15,150 --> 00:17:16,589 So you can have different channels. 451 00:17:17,619 --> 00:17:19,719 And we end up with a dataset like below. 452 00:17:20,890 --> 00:17:22,358 So this is a hologram. 453 00:17:24,040 --> 00:17:26,318 It's unfortunately not something 454 00:17:26,319 --> 00:17:28,749 that you would easily be able to 455 00:17:28,750 --> 00:17:31,449 make sense of just by looking at it. 456 00:17:31,450 --> 00:17:32,450 And so 457 00:17:34,300 --> 00:17:36,519 we we since we don't have eyes for this 458 00:17:36,520 --> 00:17:38,349 radiation, we cannot see the radiation by 459 00:17:38,350 --> 00:17:40,119 itself. And even if we record it like 460 00:17:40,120 --> 00:17:41,979 that, the information is not very 461 00:17:41,980 --> 00:17:42,980 meaningful. 462 00:17:43,900 --> 00:17:45,969 But fortunately, there 463 00:17:45,970 --> 00:17:47,799 exists a number of reconstruction 464 00:17:47,800 --> 00:17:49,089 algorithms. 465 00:17:49,090 --> 00:17:50,889 So this problem has been solved and 466 00:17:50,890 --> 00:17:53,889 coherent optics where people 467 00:17:53,890 --> 00:17:54,999 actually can. 468 00:17:56,010 --> 00:17:58,679 Render three dimensional views of 469 00:17:58,680 --> 00:18:01,019 what a hologram would look like simply 470 00:18:01,020 --> 00:18:03,299 by knowing the pattern 471 00:18:03,300 --> 00:18:05,659 that you produce on a photographic plate, 472 00:18:05,660 --> 00:18:08,459 so by scanning a photographic plate and 473 00:18:08,460 --> 00:18:10,139 feeding that in the right algorithm, you 474 00:18:10,140 --> 00:18:11,759 can get a three dimensional view of the 475 00:18:11,760 --> 00:18:14,759 hologram without ever 476 00:18:14,760 --> 00:18:16,979 looking at it in real light. 477 00:18:16,980 --> 00:18:19,439 And this you can also do with 478 00:18:19,440 --> 00:18:21,419 microwave radiation. 479 00:18:21,420 --> 00:18:23,579 So from that hologram, 480 00:18:23,580 --> 00:18:25,649 we can reconstruct three 481 00:18:25,650 --> 00:18:26,650 dimensional views. 482 00:18:27,780 --> 00:18:29,939 We did that and this is a 483 00:18:29,940 --> 00:18:32,099 while. This is the test setup that in 484 00:18:32,100 --> 00:18:33,569 the end worked best. 485 00:18:33,570 --> 00:18:35,729 So we have a commercial Wi-Fi router 486 00:18:35,730 --> 00:18:37,529 sitting on the back of the lab. 487 00:18:37,530 --> 00:18:38,819 It's labeled embitter. 488 00:18:38,820 --> 00:18:40,169 In the upper picture. 489 00:18:40,170 --> 00:18:42,389 We play some absorbing objects in 490 00:18:42,390 --> 00:18:44,699 the beam that we made from 491 00:18:44,700 --> 00:18:46,079 aluminum tin foil. 492 00:18:47,100 --> 00:18:49,169 And then we record a hologram in a 493 00:18:49,170 --> 00:18:50,939 plane that you cannot see here. 494 00:18:50,940 --> 00:18:53,069 That is essentially the front 495 00:18:53,070 --> 00:18:54,959 plane of that image. 496 00:18:54,960 --> 00:18:57,179 We feed the data into a reconstruction 497 00:18:57,180 --> 00:18:58,299 algorithm. 498 00:18:58,300 --> 00:19:00,539 And in that way, we can now 499 00:19:00,540 --> 00:19:02,789 render a three dimensional view, 500 00:19:02,790 --> 00:19:04,859 as it would appear if we had 501 00:19:04,860 --> 00:19:06,899 eyes for Wi-Fi radiation and we looked 502 00:19:06,900 --> 00:19:09,089 through that plane of the 503 00:19:09,090 --> 00:19:10,090 above picture. 504 00:19:11,340 --> 00:19:13,709 So we can actually, 505 00:19:13,710 --> 00:19:16,259 since it's a three dimensional approach, 506 00:19:16,260 --> 00:19:18,329 we can actually not 507 00:19:18,330 --> 00:19:20,039 only reconstruct this view, we can 508 00:19:20,040 --> 00:19:22,739 actually focus on different planes. 509 00:19:22,740 --> 00:19:24,989 And so this is what we do here. 510 00:19:24,990 --> 00:19:27,089 So obviously, a first interesting thing 511 00:19:27,090 --> 00:19:29,609 to do is to focus back 512 00:19:29,610 --> 00:19:32,429 into the emitter plane. 513 00:19:32,430 --> 00:19:34,529 If you focus into the plane of the 514 00:19:34,530 --> 00:19:35,530 Wi-Fi router, 515 00:19:36,930 --> 00:19:39,239 you actually nicely see a bright 516 00:19:39,240 --> 00:19:42,149 spot at one point in the image, 517 00:19:42,150 --> 00:19:44,249 which is the image 518 00:19:44,250 --> 00:19:46,049 of this glowing light bulb in the 519 00:19:46,050 --> 00:19:47,050 microwave domain. 520 00:19:48,570 --> 00:19:50,399 It turns out, interestingly, that the 521 00:19:50,400 --> 00:19:52,529 picture is not very pretty if you just 522 00:19:52,530 --> 00:19:55,589 do it like that, because 523 00:19:55,590 --> 00:19:58,169 if you only do it with one frequency, 524 00:19:58,170 --> 00:20:00,239 with one wavelength of light, you 525 00:20:00,240 --> 00:20:02,129 become very sensitive to something that 526 00:20:02,130 --> 00:20:03,749 is known as speckles. 527 00:20:03,750 --> 00:20:06,299 So this microwave light bounces 528 00:20:06,300 --> 00:20:08,039 back and forth between all the walls in 529 00:20:08,040 --> 00:20:10,109 the lab, and these waves 530 00:20:10,110 --> 00:20:11,849 interfere with each other in a very 531 00:20:11,850 --> 00:20:13,949 erratic way. So we end 532 00:20:13,950 --> 00:20:16,679 up with something that looks like 533 00:20:16,680 --> 00:20:18,869 clouds and 534 00:20:18,870 --> 00:20:20,939 we can interestingly enhance that quite 535 00:20:20,940 --> 00:20:23,609 a bit if we repeat the experiment for 536 00:20:23,610 --> 00:20:25,069 different frequencies. 537 00:20:25,070 --> 00:20:27,449 So four different frequency bands within 538 00:20:27,450 --> 00:20:29,879 a Wi-Fi channel and we superimpose 539 00:20:29,880 --> 00:20:31,049 the images. 540 00:20:31,050 --> 00:20:33,209 We get a much clearer picture where 541 00:20:33,210 --> 00:20:35,129 it is very clear that the router is the 542 00:20:35,130 --> 00:20:36,889 brightest spot. 543 00:20:36,890 --> 00:20:38,209 This is actually something that is very 544 00:20:38,210 --> 00:20:40,609 difficult to do with real light people 545 00:20:40,610 --> 00:20:42,769 struggling to make that work 546 00:20:42,770 --> 00:20:44,959 for microwave light, it's 547 00:20:44,960 --> 00:20:46,340 essentially straightforward. 548 00:20:47,720 --> 00:20:49,939 And we can then go on and focus 549 00:20:49,940 --> 00:20:52,369 in different planes, so, for instance, 550 00:20:52,370 --> 00:20:54,349 in the plane of this absorbing object in 551 00:20:54,350 --> 00:20:56,539 the beam, and if we do that, 552 00:20:56,540 --> 00:20:58,519 we nicely see that this so. 553 00:20:58,520 --> 00:21:00,709 So as we move move from 554 00:21:00,710 --> 00:21:02,839 the emitter to the object, we nicely see 555 00:21:02,840 --> 00:21:04,939 how the beam expands and 556 00:21:04,940 --> 00:21:06,979 that when once we get to the object 557 00:21:06,980 --> 00:21:09,139 plane, there is a big shadow appearing 558 00:21:09,140 --> 00:21:10,609 in the light wave. 559 00:21:12,140 --> 00:21:14,239 Nicely, you can actually 560 00:21:14,240 --> 00:21:16,519 see that if you focus slightly 561 00:21:16,520 --> 00:21:18,679 below, slightly below or slightly 562 00:21:18,680 --> 00:21:19,760 above that plane, 563 00:21:20,900 --> 00:21:22,249 the image blurs. 564 00:21:22,250 --> 00:21:24,229 So it really is a three dimensional 565 00:21:24,230 --> 00:21:26,359 image. It's like 566 00:21:26,360 --> 00:21:28,009 focusing with a photo camera. 567 00:21:28,010 --> 00:21:30,649 You can blur it if you for defocus 568 00:21:30,650 --> 00:21:31,650 into the wrong plane. 569 00:21:33,160 --> 00:21:34,160 So 570 00:21:35,290 --> 00:21:37,079 that's essentially the data that we took, 571 00:21:38,320 --> 00:21:40,479 there is potential for a lot more data 572 00:21:40,480 --> 00:21:41,769 processing. 573 00:21:41,770 --> 00:21:44,019 So it turns out that there 574 00:21:44,020 --> 00:21:46,449 are numerous, very powerful 575 00:21:46,450 --> 00:21:48,849 imaging schemes that work with coherent, 576 00:21:48,850 --> 00:21:51,549 visible light, so incoherent, 577 00:21:51,550 --> 00:21:53,679 visible light. You can take just plain 578 00:21:53,680 --> 00:21:55,719 dumb photographs as we do that most of 579 00:21:55,720 --> 00:21:56,649 the time. 580 00:21:56,650 --> 00:21:59,349 But you can also play clever tricks 581 00:21:59,350 --> 00:22:01,839 to enhance, for instance, absorbing 582 00:22:01,840 --> 00:22:03,429 objects in the light. 583 00:22:03,430 --> 00:22:05,829 It's called dark field microscopy. 584 00:22:05,830 --> 00:22:08,199 You usually do it with microscopy. 585 00:22:08,200 --> 00:22:10,269 You can image polarization 586 00:22:10,270 --> 00:22:13,539 of light or you could enhance 587 00:22:13,540 --> 00:22:14,540 only weakly 588 00:22:15,610 --> 00:22:16,779 refractive objects. 589 00:22:18,190 --> 00:22:20,379 And all these schemes we could actually 590 00:22:20,380 --> 00:22:22,599 emulate by just numerically 591 00:22:22,600 --> 00:22:24,879 post-processing the hologram data. 592 00:22:24,880 --> 00:22:26,859 So there's a lot of things to be looked 593 00:22:26,860 --> 00:22:29,169 at like 594 00:22:29,170 --> 00:22:31,119 what does the world look like if we 595 00:22:31,120 --> 00:22:33,519 enhance absorbing objects or 596 00:22:33,520 --> 00:22:35,739 do chairs have an index of 597 00:22:35,740 --> 00:22:37,839 refraction that is different 598 00:22:37,840 --> 00:22:40,269 from air and questions 599 00:22:40,270 --> 00:22:41,270 like that? 600 00:22:42,520 --> 00:22:44,589 So in a way, to 601 00:22:44,590 --> 00:22:46,420 summarize that, we 602 00:22:47,500 --> 00:22:49,539 managed to record pictures and we 603 00:22:49,540 --> 00:22:51,699 actually managed to establish 604 00:22:51,700 --> 00:22:54,099 kind of an analogy between wireless 605 00:22:54,100 --> 00:22:56,649 communication and coherent optics. 606 00:22:56,650 --> 00:22:59,319 So I made up a little dictionary 607 00:22:59,320 --> 00:23:01,329 what that how 608 00:23:02,680 --> 00:23:05,139 terms translate into each other. 609 00:23:05,140 --> 00:23:07,239 So wireless signals that 610 00:23:07,240 --> 00:23:09,999 we usually think of as packets of data, 611 00:23:10,000 --> 00:23:11,049 it's actually light. 612 00:23:12,490 --> 00:23:14,170 It can have different colors. 613 00:23:16,020 --> 00:23:18,269 We this is a degree of freedom 614 00:23:18,270 --> 00:23:20,369 that we haven't used yet, but we could 615 00:23:20,370 --> 00:23:21,720 do it, for instance, bye 616 00:23:23,100 --> 00:23:24,089 bye, bye bye. 617 00:23:24,090 --> 00:23:26,159 Splitting signals, according to their 618 00:23:26,160 --> 00:23:27,959 ID that they transmit. 619 00:23:27,960 --> 00:23:28,960 So in a way, 620 00:23:30,090 --> 00:23:32,759 once we look at the big Pappert pattern 621 00:23:32,760 --> 00:23:34,799 signals with different bit patterns in a 622 00:23:34,800 --> 00:23:37,329 way correspond to different colors. 623 00:23:37,330 --> 00:23:39,749 And so we could have multiple 624 00:23:39,750 --> 00:23:42,239 emitters in the picture, 625 00:23:42,240 --> 00:23:44,909 we could actually tell them apart and 626 00:23:44,910 --> 00:23:47,129 would get a multicolored picture of where 627 00:23:47,130 --> 00:23:49,379 every router in 628 00:23:49,380 --> 00:23:51,599 our room illuminates the room 629 00:23:51,600 --> 00:23:53,399 with a different color. 630 00:23:53,400 --> 00:23:55,139 This would be a very interesting thing to 631 00:23:55,140 --> 00:23:56,140 do in the future. 632 00:23:57,300 --> 00:23:58,979 It turns out that the pictures can be a 633 00:23:58,980 --> 00:24:00,989 lot better if you do white light 634 00:24:00,990 --> 00:24:02,639 holography, if you use more than one 635 00:24:02,640 --> 00:24:04,229 frequency band. 636 00:24:04,230 --> 00:24:06,089 This is something that has an analogy in 637 00:24:06,090 --> 00:24:08,699 the wireless domain where increasing 638 00:24:08,700 --> 00:24:10,859 the bandwidth usually improves 639 00:24:10,860 --> 00:24:12,420 performance of radar. 640 00:24:14,190 --> 00:24:15,869 If you want to learn something about the 641 00:24:15,870 --> 00:24:17,549 environment rather than the signal 642 00:24:17,550 --> 00:24:19,709 itself, you usually do time domain 643 00:24:19,710 --> 00:24:22,109 ranging in the classical wireless 644 00:24:22,110 --> 00:24:24,209 domain. But in optics, it's just 645 00:24:24,210 --> 00:24:26,309 imaging. So you form an image 646 00:24:26,310 --> 00:24:27,540 and you look at it. 647 00:24:29,070 --> 00:24:31,319 And what used to be the runtime delay 648 00:24:31,320 --> 00:24:33,569 of a signal is now 649 00:24:33,570 --> 00:24:35,910 facing direction of a light filled. 650 00:24:38,040 --> 00:24:40,799 So that was 651 00:24:40,800 --> 00:24:42,119 the work that we did. 652 00:24:42,120 --> 00:24:43,889 We were very happy that we made it work, 653 00:24:43,890 --> 00:24:46,139 we actually got it accepted in 654 00:24:46,140 --> 00:24:48,359 this journal and that was kind 655 00:24:48,360 --> 00:24:50,039 of interesting. It's a fairly prestigious 656 00:24:50,040 --> 00:24:51,509 journal in our field. 657 00:24:51,510 --> 00:24:53,609 And so when you submit a paper, 658 00:24:53,610 --> 00:24:56,399 you have to fill a questionnaire 659 00:24:56,400 --> 00:24:58,709 on why you believe this paper 660 00:24:58,710 --> 00:25:01,019 is worthy of being published there. 661 00:25:01,020 --> 00:25:03,119 And there are categories like, you 662 00:25:03,120 --> 00:25:05,279 know, it's a big advance of 663 00:25:05,280 --> 00:25:07,109 an established technique. 664 00:25:07,110 --> 00:25:09,089 It's an unexpected breakthrough. 665 00:25:09,090 --> 00:25:10,049 It's this and that. 666 00:25:10,050 --> 00:25:12,119 And we decided to opt for category number 667 00:25:12,120 --> 00:25:14,549 four, which is it is of singular 668 00:25:14,550 --> 00:25:16,200 appeal to all physicists. 669 00:25:18,300 --> 00:25:19,300 So. 670 00:25:25,020 --> 00:25:27,119 So so by that argument, we 671 00:25:27,120 --> 00:25:29,909 actually made it accept it there and 672 00:25:29,910 --> 00:25:32,309 it was met with a lot of interest, 673 00:25:32,310 --> 00:25:34,439 not only in science, but even 674 00:25:34,440 --> 00:25:36,509 more so in the in 675 00:25:36,510 --> 00:25:38,669 the real world of media 676 00:25:38,670 --> 00:25:40,019 and of companies. 677 00:25:40,020 --> 00:25:41,579 And so that's what I would like to look 678 00:25:41,580 --> 00:25:43,949 at in the second part of my talk. 679 00:25:43,950 --> 00:25:46,229 What can we actually do with that? 680 00:25:46,230 --> 00:25:47,759 Will it ever be useful? 681 00:25:49,110 --> 00:25:51,239 So if you think about that question, 682 00:25:51,240 --> 00:25:52,859 what could you use it for? 683 00:25:54,150 --> 00:25:55,799 There is a very obvious answer, 684 00:25:56,910 --> 00:25:59,039 if you don't have 685 00:25:59,040 --> 00:26:00,150 it right now, then 686 00:26:01,500 --> 00:26:04,349 you can actually read it on Russia today 687 00:26:04,350 --> 00:26:06,959 where they 688 00:26:06,960 --> 00:26:08,309 picked up our story. 689 00:26:08,310 --> 00:26:10,379 And the headline was that this could be 690 00:26:10,380 --> 00:26:12,359 used to map your home. 691 00:26:12,360 --> 00:26:14,609 So you may wonder whether I 692 00:26:14,610 --> 00:26:16,319 know some people say this is only fake 693 00:26:16,320 --> 00:26:17,819 news, but I can tell you that at least 694 00:26:17,820 --> 00:26:19,889 for this story, this is not fake news 695 00:26:19,890 --> 00:26:21,589 that's based on a on the real thing. 696 00:26:22,740 --> 00:26:25,319 So it's 697 00:26:25,320 --> 00:26:27,569 this idea. Can we use it to spy 698 00:26:27,570 --> 00:26:29,819 on other people or in 699 00:26:29,820 --> 00:26:32,249 a in a more optimistic fashion? 700 00:26:32,250 --> 00:26:34,319 Can we use it for security, enforcement 701 00:26:34,320 --> 00:26:36,269 and security applications? 702 00:26:36,270 --> 00:26:38,699 And this is something that many 703 00:26:38,700 --> 00:26:40,170 media jumped on so 704 00:26:41,790 --> 00:26:43,619 many of the headlines that we thought 705 00:26:43,620 --> 00:26:44,969 were related to that. 706 00:26:44,970 --> 00:26:45,929 Can we use that? 707 00:26:45,930 --> 00:26:48,179 Can can my neighbor see me in my bathroom 708 00:26:48,180 --> 00:26:50,759 when I have my cell phone lying there 709 00:26:50,760 --> 00:26:52,589 and things like that? 710 00:26:52,590 --> 00:26:54,239 So it's a question to which we somehow 711 00:26:54,240 --> 00:26:55,240 need to respond. 712 00:26:56,310 --> 00:26:57,659 Would that be feasible? 713 00:26:57,660 --> 00:27:00,509 And at least in principle, it 714 00:27:00,510 --> 00:27:01,259 looks like that. 715 00:27:01,260 --> 00:27:02,759 So if you have a Wi-Fi router in your 716 00:27:02,760 --> 00:27:04,889 room and you walk around, 717 00:27:04,890 --> 00:27:06,809 then your buddy will scatter some 718 00:27:06,810 --> 00:27:08,939 reflections and you can pick 719 00:27:08,940 --> 00:27:11,009 up these reflections, run them 720 00:27:11,010 --> 00:27:13,109 through our algorithm and get 721 00:27:13,110 --> 00:27:14,309 an image of the world. 722 00:27:15,330 --> 00:27:17,459 And I'm somewhat skeptical whether 723 00:27:17,460 --> 00:27:19,649 this is a really serious use case, 724 00:27:19,650 --> 00:27:21,869 though, because, first of all, 725 00:27:21,870 --> 00:27:23,619 remember what it looked like. 726 00:27:23,620 --> 00:27:25,739 So it's a huge device and 727 00:27:25,740 --> 00:27:27,929 it's always going to remain a huge device 728 00:27:27,930 --> 00:27:30,269 because you need to have 729 00:27:30,270 --> 00:27:32,399 information on a very 730 00:27:32,400 --> 00:27:33,869 large area. 731 00:27:33,870 --> 00:27:35,579 This is what the whole scheme is based 732 00:27:35,580 --> 00:27:37,769 on. So even if you mounted 733 00:27:37,770 --> 00:27:39,509 that on a drone, you still would need to 734 00:27:39,510 --> 00:27:41,669 fly around 735 00:27:41,670 --> 00:27:44,579 many points around a building 736 00:27:44,580 --> 00:27:47,309 which would probably not go unnoticed. 737 00:27:48,750 --> 00:27:50,819 And maybe even more importantly, 738 00:27:50,820 --> 00:27:52,649 it's something that actually is already 739 00:27:52,650 --> 00:27:54,419 being done right now. 740 00:27:54,420 --> 00:27:56,699 So for 741 00:27:56,700 --> 00:27:58,709 many of these applications, you actually 742 00:27:58,710 --> 00:28:00,539 don't need imaging. 743 00:28:00,540 --> 00:28:02,699 Imaging is kind of a bonus, but 744 00:28:02,700 --> 00:28:05,099 it's a complicated and expensive bonus. 745 00:28:05,100 --> 00:28:07,199 And you can learn a lot about the world 746 00:28:07,200 --> 00:28:10,019 just by looking at the 747 00:28:10,020 --> 00:28:13,179 signal at one point in space. 748 00:28:13,180 --> 00:28:14,669 And this is something that is going 749 00:28:14,670 --> 00:28:17,039 commercial. So there are companies that 750 00:28:17,040 --> 00:28:19,199 sell systems where you 751 00:28:19,200 --> 00:28:21,299 essentially plug something like 752 00:28:21,300 --> 00:28:24,509 a second router into your room. 753 00:28:24,510 --> 00:28:26,729 This router is not emitting 754 00:28:26,730 --> 00:28:28,499 signals that are used for communication. 755 00:28:28,500 --> 00:28:30,719 It's actually looking at signals that 756 00:28:30,720 --> 00:28:33,179 are scattered from the environment, 757 00:28:33,180 --> 00:28:35,459 that it's analyzing them in a 758 00:28:35,460 --> 00:28:36,959 very detailed way. 759 00:28:36,960 --> 00:28:39,059 And by doing that, you can 760 00:28:39,060 --> 00:28:41,489 actually get a fairly good idea of 761 00:28:41,490 --> 00:28:42,899 whether there are people moving in the 762 00:28:42,900 --> 00:28:45,329 room. How many of them are there? 763 00:28:45,330 --> 00:28:47,099 Are they in the bathroom or in the 764 00:28:48,840 --> 00:28:51,089 kitchen or whatever else. 765 00:28:51,090 --> 00:28:53,099 So it's something where our probably is 766 00:28:53,100 --> 00:28:55,349 too complicated, which is good 767 00:28:55,350 --> 00:28:57,929 news because it's not a security concern. 768 00:28:57,930 --> 00:29:00,539 The bad news, if you wish, is that 769 00:29:00,540 --> 00:29:02,349 Wi-Fi is a very leaky thing. 770 00:29:02,350 --> 00:29:04,979 And this general idea 771 00:29:04,980 --> 00:29:07,079 that we can use the stray radiation to 772 00:29:07,080 --> 00:29:08,639 spy on other people, it's actually a 773 00:29:08,640 --> 00:29:10,009 valid security concern. 774 00:29:12,170 --> 00:29:14,989 OK, if we wanted to 775 00:29:14,990 --> 00:29:17,179 to to do it, if we wanted to 776 00:29:17,180 --> 00:29:19,189 sell that as a security application, I 777 00:29:19,190 --> 00:29:21,349 think we would run into a second very 778 00:29:21,350 --> 00:29:22,789 important problem. 779 00:29:22,790 --> 00:29:25,729 And the important problem here is that 780 00:29:25,730 --> 00:29:28,069 if you have an application 781 00:29:28,070 --> 00:29:30,139 where people actually are willing to buy 782 00:29:30,140 --> 00:29:31,819 expensive equipment to get a lot of 783 00:29:31,820 --> 00:29:33,979 information, then it's 784 00:29:33,980 --> 00:29:35,569 going to be some kind of dedicated 785 00:29:35,570 --> 00:29:36,679 device. 786 00:29:36,680 --> 00:29:38,899 And so it's 787 00:29:38,900 --> 00:29:40,669 not so much more expensive to actually 788 00:29:40,670 --> 00:29:43,009 combine that with a dedicated 789 00:29:43,010 --> 00:29:45,139 TaylorMade emitter that 790 00:29:45,140 --> 00:29:47,359 is emitting a specific signal. 791 00:29:47,360 --> 00:29:49,909 And once you can do that, 792 00:29:49,910 --> 00:29:51,619 you can play a trick, which is called 793 00:29:51,620 --> 00:29:53,269 Ultra Bend. 794 00:29:53,270 --> 00:29:55,879 And this is a very powerful trick 795 00:29:55,880 --> 00:29:58,189 that is spoiling the game 796 00:29:58,190 --> 00:30:00,229 for many of these applications for us. 797 00:30:00,230 --> 00:30:02,359 So I'm going to explain it in a little 798 00:30:02,360 --> 00:30:03,649 bit more. 799 00:30:03,650 --> 00:30:05,899 So the idea that you might 800 00:30:05,900 --> 00:30:08,059 want to use is that if you want 801 00:30:08,060 --> 00:30:10,789 to build something like an X-ray forums, 802 00:30:10,790 --> 00:30:13,309 you emit some signal from a dedicated 803 00:30:13,310 --> 00:30:14,239 emitter. 804 00:30:14,240 --> 00:30:16,759 You detect the reflection in a dedicated 805 00:30:16,760 --> 00:30:19,039 detector, you measure the runtime 806 00:30:19,040 --> 00:30:20,839 delay. And from that you infer whether 807 00:30:20,840 --> 00:30:22,999 there are people moving around and at 808 00:30:23,000 --> 00:30:24,090 which distance they are. 809 00:30:25,400 --> 00:30:27,589 So if you do that, the resolution 810 00:30:27,590 --> 00:30:30,859 that you get is related 811 00:30:30,860 --> 00:30:32,989 to the bitrate of your 812 00:30:32,990 --> 00:30:35,899 signal, if you which it's essentially 813 00:30:35,900 --> 00:30:38,209 the inverse of the bandwidth. 814 00:30:38,210 --> 00:30:40,489 So if these data packets are very 815 00:30:40,490 --> 00:30:42,829 broad, the resolution 816 00:30:42,830 --> 00:30:44,239 will drop. 817 00:30:44,240 --> 00:30:46,969 And from the point of view of radar, 818 00:30:46,970 --> 00:30:49,099 wi fi actually is not a very 819 00:30:49,100 --> 00:30:51,319 good signal because it has a very 820 00:30:51,320 --> 00:30:52,579 narrow bandwidth. 821 00:30:52,580 --> 00:30:54,799 So for two point four gigahertz Wi-Fi, 822 00:30:54,800 --> 00:30:56,629 it's actually only 20 megahertz. 823 00:30:56,630 --> 00:30:58,879 And this corresponds to 824 00:30:58,880 --> 00:31:01,279 a resolution of 10 meters. 825 00:31:01,280 --> 00:31:04,579 And if you build a dedicated emitter, 826 00:31:04,580 --> 00:31:06,979 obviously, you can drop 827 00:31:06,980 --> 00:31:08,989 that restriction. You can build an 828 00:31:08,990 --> 00:31:11,419 emitter with a much higher bitrate. 829 00:31:11,420 --> 00:31:12,709 And this is what people do. 830 00:31:12,710 --> 00:31:14,959 So you can push that 831 00:31:14,960 --> 00:31:17,119 up to a bit rate of gigahertz to 832 00:31:17,120 --> 00:31:19,249 tens of gigahertz and emitter. 833 00:31:19,250 --> 00:31:21,679 That is sending some garbage signal 834 00:31:21,680 --> 00:31:23,779 that is spread across 835 00:31:23,780 --> 00:31:26,359 all the frequencies up to 10 gigahertz. 836 00:31:26,360 --> 00:31:28,489 And by doing that, you actually get 837 00:31:28,490 --> 00:31:31,309 down to a centimeter resolution in radar. 838 00:31:31,310 --> 00:31:33,379 So if you invest into a dedicated 839 00:31:33,380 --> 00:31:35,359 device, this is a very powerful trick to 840 00:31:35,360 --> 00:31:37,669 play and this is actually being 841 00:31:37,670 --> 00:31:39,889 done. So there is I did some research 842 00:31:39,890 --> 00:31:41,839 on the Internet and there are devices for 843 00:31:41,840 --> 00:31:44,119 security forces that look 844 00:31:44,120 --> 00:31:46,189 a bit like these devices that you use 845 00:31:46,190 --> 00:31:48,229 to search for pipes before drilling into 846 00:31:48,230 --> 00:31:49,399 the wall. 847 00:31:49,400 --> 00:31:51,469 You place them on a wall. 848 00:31:51,470 --> 00:31:53,239 It takes a few seconds and it's going to 849 00:31:53,240 --> 00:31:55,399 tell you whether there are people in 850 00:31:55,400 --> 00:31:57,469 that room that you can't see and 851 00:31:57,470 --> 00:31:58,759 at which distance they are. 852 00:31:58,760 --> 00:32:00,469 So it gives you an image on the display, 853 00:32:00,470 --> 00:32:01,910 which looks pretty much like that. 854 00:32:03,700 --> 00:32:06,099 So to sum that up, 855 00:32:06,100 --> 00:32:08,229 security applications, 856 00:32:08,230 --> 00:32:10,029 well, are an interesting thing to think 857 00:32:10,030 --> 00:32:12,189 about, but it's probably not a really 858 00:32:12,190 --> 00:32:14,109 realistic use case for this scheme. 859 00:32:16,010 --> 00:32:17,989 Let's look at something else. 860 00:32:17,990 --> 00:32:19,939 Let's look at civil engineering, I 861 00:32:19,940 --> 00:32:21,379 already mentioned it. There are these 862 00:32:21,380 --> 00:32:23,629 devices that you use to search 863 00:32:23,630 --> 00:32:25,969 for water pipes before making a drill. 864 00:32:25,970 --> 00:32:28,129 And these devices, well, they 865 00:32:28,130 --> 00:32:29,130 work. But 866 00:32:30,200 --> 00:32:31,159 I did it myself. 867 00:32:31,160 --> 00:32:33,289 And it's it's not a very reliable 868 00:32:33,290 --> 00:32:34,579 thing to work with. 869 00:32:34,580 --> 00:32:37,559 So there is room for improvement. 870 00:32:37,560 --> 00:32:39,679 And it's interesting to ask whether 871 00:32:39,680 --> 00:32:42,439 we could fill a gap in that market. 872 00:32:42,440 --> 00:32:44,959 So the way that would work 873 00:32:44,960 --> 00:32:46,220 would be that 874 00:32:47,270 --> 00:32:49,609 you rely on the radiation 875 00:32:49,610 --> 00:32:51,949 that is emitted from wireless 876 00:32:51,950 --> 00:32:54,469 devices in other rooms across 877 00:32:54,470 --> 00:32:55,900 the wall that you want to drill. 878 00:32:57,350 --> 00:33:00,229 And that will illuminate this wall 879 00:33:00,230 --> 00:33:02,539 with microwave light so 880 00:33:02,540 --> 00:33:04,759 that pipes and power lines in the wall 881 00:33:04,760 --> 00:33:07,399 would cast a shadow and then 882 00:33:07,400 --> 00:33:09,949 you would record a hologram across 883 00:33:09,950 --> 00:33:11,599 the wall that you want to drill, 884 00:33:11,600 --> 00:33:13,609 reconstruct that image and see every 885 00:33:13,610 --> 00:33:14,610 single pipe. 886 00:33:15,830 --> 00:33:17,899 Still, we have this challenge of 887 00:33:17,900 --> 00:33:19,099 technical complexity. 888 00:33:19,100 --> 00:33:20,779 But I think in this case it would be 889 00:33:20,780 --> 00:33:23,449 surmountable because 890 00:33:23,450 --> 00:33:25,669 we could actually scale up this 891 00:33:25,670 --> 00:33:28,009 system from a single 892 00:33:28,010 --> 00:33:29,989 antenna device that is scanning the plane 893 00:33:29,990 --> 00:33:32,599 to a one dimensional array of antennas, 894 00:33:32,600 --> 00:33:34,939 like a magic wand that would wave across 895 00:33:34,940 --> 00:33:35,839 the wall. 896 00:33:35,840 --> 00:33:37,339 And that would give you a picture on your 897 00:33:37,340 --> 00:33:39,229 smartphone of what's inside. 898 00:33:40,760 --> 00:33:42,889 So this is a fairly realistic 899 00:33:42,890 --> 00:33:45,139 thing, unfortunately, 900 00:33:45,140 --> 00:33:47,419 here, to, um, you 901 00:33:47,420 --> 00:33:49,549 have to compete with ultraright bent. 902 00:33:49,550 --> 00:33:51,649 So if you built a complicated 903 00:33:51,650 --> 00:33:53,419 magic wand, it's not so much more 904 00:33:53,420 --> 00:33:55,639 expensive to include an ultra wide 905 00:33:55,640 --> 00:33:58,069 bent emitter and to do very 906 00:33:58,070 --> 00:33:59,129 precise radar. 907 00:33:59,130 --> 00:34:00,859 And there are companies doing that you 908 00:34:00,860 --> 00:34:02,239 may have read about all about. 909 00:34:02,240 --> 00:34:04,339 It's a kind of a radar that you can 910 00:34:04,340 --> 00:34:06,379 plug on the back of your smartphone and 911 00:34:06,380 --> 00:34:08,509 you can X-ray your wall 912 00:34:08,510 --> 00:34:10,729 and it's going to do some signal 913 00:34:10,730 --> 00:34:12,289 processing and give you actually a 914 00:34:12,290 --> 00:34:14,429 picture of where the pipe is. 915 00:34:15,830 --> 00:34:18,589 So there would be strong competition. 916 00:34:18,590 --> 00:34:20,448 Still, our scheme would have some 917 00:34:20,449 --> 00:34:22,879 advantages so we could use radiation 918 00:34:22,880 --> 00:34:25,009 that is actually coming from things 919 00:34:25,010 --> 00:34:26,479 behind the wall. That could be an 920 00:34:26,480 --> 00:34:27,468 advantage. 921 00:34:27,469 --> 00:34:29,089 And so in principle, that could be 922 00:34:29,090 --> 00:34:30,499 something to think about. 923 00:34:30,500 --> 00:34:31,500 So it could work. 924 00:34:34,350 --> 00:34:36,839 There's one thing that we initially 925 00:34:36,840 --> 00:34:39,059 thought would be a good idea 926 00:34:39,060 --> 00:34:41,158 that is tracking emitters inside 927 00:34:41,159 --> 00:34:42,658 buildings. 928 00:34:42,659 --> 00:34:44,819 So this is actually 929 00:34:44,820 --> 00:34:47,428 a very important 930 00:34:47,429 --> 00:34:49,379 thing that is becoming more and more 931 00:34:49,380 --> 00:34:51,448 important as we move towards Internet 932 00:34:51,449 --> 00:34:53,849 of Things to actually 933 00:34:53,850 --> 00:34:55,979 tell where in the building, 934 00:34:55,980 --> 00:34:58,439 in three dimensional space, you have some 935 00:34:58,440 --> 00:35:00,209 RF tag. 936 00:35:00,210 --> 00:35:02,429 And our scheme in principle could 937 00:35:02,430 --> 00:35:05,669 do that. So we could think of installing 938 00:35:05,670 --> 00:35:07,859 a two dimensional, huge array of many, 939 00:35:07,860 --> 00:35:09,779 many antennas in the ceiling of this 940 00:35:09,780 --> 00:35:12,149 building, doing holography 941 00:35:12,150 --> 00:35:14,249 and then see actually the 942 00:35:14,250 --> 00:35:16,229 position of each and every smartphone and 943 00:35:16,230 --> 00:35:18,269 WiFi Briffa router moving around in real 944 00:35:18,270 --> 00:35:19,270 time. 945 00:35:20,080 --> 00:35:22,159 We didn't do the experiment, but we did 946 00:35:22,160 --> 00:35:24,039 a numerical simulation to see whether 947 00:35:24,040 --> 00:35:26,229 this could work and this 948 00:35:26,230 --> 00:35:27,189 is what you see here. 949 00:35:27,190 --> 00:35:29,319 So we built a 3-D model 950 00:35:29,320 --> 00:35:31,659 of a virtual storage hall 951 00:35:31,660 --> 00:35:34,029 with steel bars in the 952 00:35:34,030 --> 00:35:36,309 floors and with some steel 953 00:35:36,310 --> 00:35:38,439 shelves in one of the floor. 954 00:35:38,440 --> 00:35:40,689 And the image that we get 955 00:35:40,690 --> 00:35:41,799 is encouraging. 956 00:35:41,800 --> 00:35:43,959 So with that scheme, we probably would be 957 00:35:43,960 --> 00:35:46,149 able to track down embitters 958 00:35:46,150 --> 00:35:48,309 with centimeter scale precision 959 00:35:48,310 --> 00:35:50,409 at video rates and what 960 00:35:50,410 --> 00:35:52,689 is more, and we probably would even 961 00:35:52,690 --> 00:35:55,719 be able to get information about 962 00:35:55,720 --> 00:35:57,219 the objects in that building. 963 00:35:57,220 --> 00:35:59,289 So this is a movie of the simulation 964 00:35:59,290 --> 00:36:01,989 data and it's essentially 965 00:36:01,990 --> 00:36:04,449 the hologram reconstruction as we focus 966 00:36:04,450 --> 00:36:06,939 onto successively lower planes 967 00:36:06,940 --> 00:36:08,829 moving all the way from the top of the 968 00:36:08,830 --> 00:36:11,289 building to the ground floor. 969 00:36:11,290 --> 00:36:13,509 And so this is what we get as we move 970 00:36:13,510 --> 00:36:15,429 down and we nicely see these shelves and 971 00:36:15,430 --> 00:36:17,319 we nicely see these spires casting 972 00:36:17,320 --> 00:36:19,989 shadows in the beam before we converge 973 00:36:19,990 --> 00:36:22,149 on the emitter, which is appearing 974 00:36:22,150 --> 00:36:25,749 as a spot with centimeter resolution. 975 00:36:25,750 --> 00:36:26,750 So this could work, 976 00:36:27,910 --> 00:36:29,979 but obviously it would be 977 00:36:29,980 --> 00:36:32,349 a very expensive thing to do, so you 978 00:36:32,350 --> 00:36:35,049 would need a large array of antennas. 979 00:36:35,050 --> 00:36:37,179 It's probably not not 980 00:36:37,180 --> 00:36:38,579 something that you could do right away. 981 00:36:41,050 --> 00:36:43,419 But when we talk to companies, it turned 982 00:36:43,420 --> 00:36:44,949 out that there is a solution that is 983 00:36:44,950 --> 00:36:46,629 somewhere in between where it could get 984 00:36:46,630 --> 00:36:48,939 really interesting and 985 00:36:48,940 --> 00:36:51,909 it's in a way 986 00:36:51,910 --> 00:36:55,029 a reduced implementation of that scheme. 987 00:36:55,030 --> 00:36:57,579 So rather than going 988 00:36:57,580 --> 00:36:59,649 to a full two dimensional array 989 00:36:59,650 --> 00:37:01,779 of antennas, we could actually 990 00:37:01,780 --> 00:37:04,719 go to a one dimensional array 991 00:37:04,720 --> 00:37:07,119 like this magic wand, and 992 00:37:07,120 --> 00:37:09,459 that still would speed up acquisition 993 00:37:09,460 --> 00:37:11,649 by a factor of 100000 compared 994 00:37:11,650 --> 00:37:13,689 to our single antenna proof of principle 995 00:37:13,690 --> 00:37:15,189 implementation. 996 00:37:15,190 --> 00:37:17,349 So you could have you could actually 997 00:37:17,350 --> 00:37:19,809 take pictures of large 998 00:37:19,810 --> 00:37:22,449 scale, 10 meter scale structures 999 00:37:22,450 --> 00:37:24,699 and probably on a minute to our 1000 00:37:24,700 --> 00:37:27,209 scale, if you 1001 00:37:27,210 --> 00:37:29,229 can scanned this device around the 1002 00:37:29,230 --> 00:37:31,299 structure by, say, a 1003 00:37:31,300 --> 00:37:32,300 drone or a car. 1004 00:37:33,760 --> 00:37:36,009 And once you have that data, 1005 00:37:36,010 --> 00:37:37,929 you could actually look into the building 1006 00:37:37,930 --> 00:37:40,119 with wi fi eyes and understand how 1007 00:37:40,120 --> 00:37:41,889 radiation propagates in their. 1008 00:37:43,210 --> 00:37:45,279 And this could actually be 1009 00:37:45,280 --> 00:37:47,649 relevant for this whole field 1010 00:37:47,650 --> 00:37:48,760 of indoor tracking. 1011 00:37:49,970 --> 00:37:52,009 So maybe it would be a bit more specific 1012 00:37:52,010 --> 00:37:54,109 on that, this is actually a 1013 00:37:54,110 --> 00:37:56,299 very important unsolved challenge 1014 00:37:56,300 --> 00:37:57,769 at the moment. 1015 00:37:57,770 --> 00:37:59,869 So many companies get more and more 1016 00:37:59,870 --> 00:38:02,269 interested in locating 1017 00:38:02,270 --> 00:38:05,389 RF tags with centimeter scale precision 1018 00:38:05,390 --> 00:38:08,179 to track inventories or devices. 1019 00:38:09,600 --> 00:38:12,499 And it's a market that is 1020 00:38:12,500 --> 00:38:14,839 actually predicted to grow to a billion 1021 00:38:14,840 --> 00:38:16,460 dollar size in a few years. 1022 00:38:17,540 --> 00:38:19,969 However, at the press. 1023 00:38:19,970 --> 00:38:21,769 So, OK, so one very strange, 1024 00:38:21,770 --> 00:38:23,899 straightforward application where we 1025 00:38:23,900 --> 00:38:26,029 could enter here is that we could do 1026 00:38:26,030 --> 00:38:28,849 site survey for existing solutions. 1027 00:38:28,850 --> 00:38:30,979 So if you bought 1028 00:38:30,980 --> 00:38:33,079 some system for tracking and 1029 00:38:33,080 --> 00:38:35,209 you found that for 1030 00:38:35,210 --> 00:38:36,679 some reason it didn't work in your 1031 00:38:36,680 --> 00:38:38,989 factory hall, then you could come 1032 00:38:38,990 --> 00:38:40,969 with this one dimensional array of 1033 00:38:40,970 --> 00:38:43,069 antennas, get this full picture, and you 1034 00:38:43,070 --> 00:38:45,439 could see that there is actually a nasty 1035 00:38:45,440 --> 00:38:46,999 reflection on this metallic wall. 1036 00:38:47,000 --> 00:38:49,309 And if you replace that by some 1037 00:38:49,310 --> 00:38:51,289 concrete coating, it will make things 1038 00:38:51,290 --> 00:38:53,269 much better. This kind of thing. 1039 00:38:53,270 --> 00:38:55,399 So this could be actually 1040 00:38:55,400 --> 00:38:57,589 interesting to to get the full picture 1041 00:38:57,590 --> 00:38:59,569 only for that reason. 1042 00:38:59,570 --> 00:39:01,639 And but it could also be 1043 00:39:01,640 --> 00:39:03,889 interesting as an R&D tool 1044 00:39:03,890 --> 00:39:06,109 to actually understand better what 1045 00:39:06,110 --> 00:39:07,849 these signals look like and how they 1046 00:39:07,850 --> 00:39:10,669 propagate to make these schemes better. 1047 00:39:10,670 --> 00:39:13,129 And there's work to do because 1048 00:39:13,130 --> 00:39:15,229 at the moment these 1049 00:39:15,230 --> 00:39:17,899 schemes are, well, very successful, 1050 00:39:17,900 --> 00:39:20,359 but still not fully satisfactory. 1051 00:39:20,360 --> 00:39:22,369 So people have tried many things. 1052 00:39:22,370 --> 00:39:24,619 You can move around with a camera 1053 00:39:24,620 --> 00:39:26,689 and scan barcodes, you can use 1054 00:39:26,690 --> 00:39:28,849 RFID, but then you 1055 00:39:28,850 --> 00:39:30,439 can only read them from a very close 1056 00:39:30,440 --> 00:39:32,869 distance. So it's not very convenient for 1057 00:39:32,870 --> 00:39:34,280 large scale factories. 1058 00:39:35,420 --> 00:39:37,729 You can use all kinds of beacons. 1059 00:39:37,730 --> 00:39:38,730 You can check 1060 00:39:40,040 --> 00:39:42,109 labels with ultrasound, Bluetooth, 1061 00:39:42,110 --> 00:39:43,219 Wi-Fi, whatever. 1062 00:39:44,600 --> 00:39:46,759 But with that, you run 1063 00:39:46,760 --> 00:39:48,589 into this bandwidth issue and you only 1064 00:39:48,590 --> 00:39:50,359 get meter scale resolution. 1065 00:39:50,360 --> 00:39:52,459 OK, here again, you 1066 00:39:52,460 --> 00:39:54,499 can play the ultra wideband trick so you 1067 00:39:54,500 --> 00:39:56,689 can have dedicated amateurs with 1068 00:39:56,690 --> 00:39:58,759 a very wide frequency band 1069 00:39:58,760 --> 00:40:00,529 and that will give you a centimeter scale 1070 00:40:00,530 --> 00:40:02,239 resolution. 1071 00:40:02,240 --> 00:40:04,729 And or this already is incredibly 1072 00:40:04,730 --> 00:40:07,009 successful today, but it comes 1073 00:40:07,010 --> 00:40:09,169 at a large price and it's 1074 00:40:09,170 --> 00:40:10,639 a physical price. 1075 00:40:10,640 --> 00:40:12,979 These chips are expensive, so 1076 00:40:14,780 --> 00:40:16,459 the tags will cost you something like 10 1077 00:40:16,460 --> 00:40:18,109 years or more. 1078 00:40:18,110 --> 00:40:20,389 And it's very hungry 1079 00:40:20,390 --> 00:40:22,729 in terms of power. So you need to 1080 00:40:22,730 --> 00:40:24,259 ship it with a battery and you need to 1081 00:40:24,260 --> 00:40:26,419 replace that battery every few months 1082 00:40:26,420 --> 00:40:28,099 or years. And this is not very 1083 00:40:28,100 --> 00:40:29,569 convenient. 1084 00:40:29,570 --> 00:40:31,789 So there is a quest out there 1085 00:40:31,790 --> 00:40:34,489 to make these schemes work with passive 1086 00:40:34,490 --> 00:40:36,679 tax to where we 1087 00:40:36,680 --> 00:40:38,749 actually rather than having an 1088 00:40:38,750 --> 00:40:41,089 active emitter on a label, 1089 00:40:41,090 --> 00:40:43,219 we just have some absorbing little 1090 00:40:43,220 --> 00:40:45,289 thing. And we illuminated by some 1091 00:40:45,290 --> 00:40:47,539 other source and we imaged the shadow 1092 00:40:47,540 --> 00:40:48,770 that it casts in the beam. 1093 00:40:50,690 --> 00:40:52,309 At the moment, this is not doable. 1094 00:40:54,120 --> 00:40:56,299 You in principle could solve 1095 00:40:56,300 --> 00:40:57,559 that problem by many ways. 1096 00:40:57,560 --> 00:40:59,659 You could buy more receivers, you 1097 00:40:59,660 --> 00:41:02,329 could do more signal processing, 1098 00:41:02,330 --> 00:41:04,429 whatever, but it's not clear 1099 00:41:04,430 --> 00:41:06,889 which route would be the most successful. 1100 00:41:06,890 --> 00:41:09,349 And so rather than trying them 1101 00:41:09,350 --> 00:41:11,659 by trial and error one by one. 1102 00:41:11,660 --> 00:41:14,389 One interesting idea could be 1103 00:41:14,390 --> 00:41:16,940 to actually record this entire wavefront. 1104 00:41:18,210 --> 00:41:20,970 Look at the picture, understand 1105 00:41:21,990 --> 00:41:24,029 how radiation propagates in typical 1106 00:41:24,030 --> 00:41:26,339 buildings and 1107 00:41:26,340 --> 00:41:28,529 what kind of what part of the signal 1108 00:41:28,530 --> 00:41:30,809 you actually need to see this shadow 1109 00:41:30,810 --> 00:41:32,009 best. 1110 00:41:32,010 --> 00:41:34,079 And here are the key 1111 00:41:34,080 --> 00:41:35,609 make believe could come in. 1112 00:41:35,610 --> 00:41:38,099 We could record this for WaveFront 1113 00:41:38,100 --> 00:41:40,319 and we could simulate any kind 1114 00:41:40,320 --> 00:41:42,479 of reduced scheme, like a scheme with 1115 00:41:42,480 --> 00:41:44,639 only a few antennas or looking at only 1116 00:41:44,640 --> 00:41:47,009 part of the frequency band 1117 00:41:47,010 --> 00:41:49,649 with our algorithms. 1118 00:41:49,650 --> 00:41:51,690 So summing that up, 1119 00:41:52,830 --> 00:41:54,239 it's an interesting technique. 1120 00:41:54,240 --> 00:41:56,489 I don't think it's a serious security 1121 00:41:56,490 --> 00:41:58,919 concern, but reduced implementations 1122 00:41:58,920 --> 00:42:00,119 actually might be one. 1123 00:42:01,350 --> 00:42:03,629 Civil engineering might be an application 1124 00:42:03,630 --> 00:42:06,299 where we could move into 1125 00:42:06,300 --> 00:42:08,399 indoor tracking probably is too 1126 00:42:08,400 --> 00:42:10,619 difficult, but R&D for 1127 00:42:10,620 --> 00:42:12,749 simpler tracking schemes could 1128 00:42:12,750 --> 00:42:14,369 be a very viable way to go. 1129 00:42:15,820 --> 00:42:17,949 With that, I'm at the end, I would 1130 00:42:17,950 --> 00:42:20,169 like to 1131 00:42:20,170 --> 00:42:22,029 acknowledge the person that did all the 1132 00:42:22,030 --> 00:42:24,249 work, Philip Hall here 1133 00:42:24,250 --> 00:42:26,439 actually was brave enough to to start 1134 00:42:26,440 --> 00:42:29,199 this project as a bachelor thesis 1135 00:42:29,200 --> 00:42:31,149 when at the time when everyone else was 1136 00:42:31,150 --> 00:42:33,459 frowning upon the idea and 1137 00:42:33,460 --> 00:42:35,409 he was very successful in making all this 1138 00:42:35,410 --> 00:42:36,429 work. 1139 00:42:36,430 --> 00:42:37,430 So. 1140 00:42:45,350 --> 00:42:47,449 So that's the end 1141 00:42:47,450 --> 00:42:49,549 of my talk, I would like to draw 1142 00:42:49,550 --> 00:42:51,649 your attention to the fact that 1143 00:42:51,650 --> 00:42:53,389 you can look at wireless signal from a 1144 00:42:53,390 --> 00:42:54,679 very different perspective. 1145 00:42:54,680 --> 00:42:56,779 It's just light and you can hack 1146 00:42:56,780 --> 00:42:59,359 them by just 1147 00:42:59,360 --> 00:43:01,669 using this stray radiation that 1148 00:43:01,670 --> 00:43:02,959 you get. 1149 00:43:02,960 --> 00:43:04,549 And if you do it right, you can actually 1150 00:43:04,550 --> 00:43:06,349 take real pictures. 1151 00:43:06,350 --> 00:43:08,719 This could be useful for maybe tracking, 1152 00:43:08,720 --> 00:43:10,789 maybe civil engineering, 1153 00:43:10,790 --> 00:43:13,099 probably most as an R&D tool 1154 00:43:13,100 --> 00:43:15,589 to implement to to, 1155 00:43:15,590 --> 00:43:18,109 um, develop future 1156 00:43:18,110 --> 00:43:19,519 wireless applications. 1157 00:43:19,520 --> 00:43:20,539 So thank you very much. 1158 00:43:32,030 --> 00:43:33,499 We have 15 minutes for Q&A. 1159 00:43:43,490 --> 00:43:44,490 And. 1160 00:43:49,040 --> 00:43:50,989 Hello, thank you for your time. 1161 00:43:50,990 --> 00:43:52,639 It was really interesting, I was 1162 00:43:52,640 --> 00:43:54,949 wondering if you could build Wi-Fi 1163 00:43:54,950 --> 00:43:55,950 lenses. 1164 00:43:57,890 --> 00:44:00,420 That's a good question. 1165 00:44:01,580 --> 00:44:03,739 So it 1166 00:44:03,740 --> 00:44:06,079 probably would be difficult 1167 00:44:06,080 --> 00:44:08,449 if it if it should be a physical lens, 1168 00:44:08,450 --> 00:44:10,579 because that so you not only 1169 00:44:10,580 --> 00:44:12,049 would need to find the right material, 1170 00:44:12,050 --> 00:44:14,329 which you probably could, you also 1171 00:44:14,330 --> 00:44:16,439 would need to make it larger to to, 1172 00:44:16,440 --> 00:44:18,860 uh, because the wavefront are large. 1173 00:44:20,030 --> 00:44:22,489 But again, 1174 00:44:22,490 --> 00:44:24,439 somewhat paradoxically, there are 1175 00:44:24,440 --> 00:44:26,389 actually much simpler schemes to do that. 1176 00:44:26,390 --> 00:44:28,909 So if you if 1177 00:44:28,910 --> 00:44:30,979 you take a train somewhere 1178 00:44:30,980 --> 00:44:33,559 in the country, you may see that nowadays 1179 00:44:33,560 --> 00:44:35,689 these satellite receivers, they are no 1180 00:44:35,690 --> 00:44:37,279 longer like a parabolic mirror, although 1181 00:44:37,280 --> 00:44:39,589 sometimes they tend to be flat, like 1182 00:44:39,590 --> 00:44:41,089 just a flat plate. 1183 00:44:41,090 --> 00:44:44,149 And this is in a way, due to a virtual 1184 00:44:44,150 --> 00:44:46,369 lens for this radiation 1185 00:44:46,370 --> 00:44:47,449 in that frequency band. 1186 00:44:47,450 --> 00:44:49,579 So it's a phased array where you 1187 00:44:49,580 --> 00:44:51,619 have many receivers and you delay the 1188 00:44:51,620 --> 00:44:53,329 phase electronically, which is 1189 00:44:53,330 --> 00:44:55,459 essentially what lens does in 1190 00:44:55,460 --> 00:44:56,899 the optical domain. 1191 00:44:56,900 --> 00:44:59,059 And in doing 1192 00:44:59,060 --> 00:45:01,309 that, you can simulate any kind of lens, 1193 00:45:01,310 --> 00:45:03,409 including a parabolic reflector looking 1194 00:45:03,410 --> 00:45:04,410 at the antenna. 1195 00:45:05,270 --> 00:45:06,409 So for our scheme, 1196 00:45:07,430 --> 00:45:09,529 it's actually even better we don't need 1197 00:45:09,530 --> 00:45:11,839 a lens because we can simulate any 1198 00:45:11,840 --> 00:45:13,969 kind of lens in the computer. 1199 00:45:13,970 --> 00:45:16,219 And this is something that we 1200 00:45:16,220 --> 00:45:18,769 actually thought about doing, like trying 1201 00:45:18,770 --> 00:45:20,959 different building different kinds of 1202 00:45:20,960 --> 00:45:23,299 virtual objective lenses to 1203 00:45:23,300 --> 00:45:25,759 look at different parts of the radiation. 1204 00:45:25,760 --> 00:45:28,309 But I think electronic solutions 1205 00:45:28,310 --> 00:45:30,529 will probably be the easier way to go 1206 00:45:30,530 --> 00:45:31,550 than a physical lens. 1207 00:45:33,360 --> 00:45:35,579 We have one question from the Internet. 1208 00:45:35,580 --> 00:45:36,629 What is your prognosis? 1209 00:45:36,630 --> 00:45:38,609 When can we expect a practical camera for 1210 00:45:38,610 --> 00:45:40,709 a different spectrum of ranges like 1211 00:45:40,710 --> 00:45:41,710 TV or radio? 1212 00:45:43,680 --> 00:45:46,049 So, um. 1213 00:45:46,050 --> 00:45:47,759 Well, I mean, technically, I don't see an 1214 00:45:47,760 --> 00:45:50,309 obstacle for transferring 1215 00:45:50,310 --> 00:45:52,739 this technique to other frequency domains 1216 00:45:52,740 --> 00:45:54,840 like radio or so. 1217 00:45:57,180 --> 00:45:59,279 I don't think we will see it in 1218 00:45:59,280 --> 00:46:01,109 the near future as a commercial device 1219 00:46:01,110 --> 00:46:03,839 because applications are so limited. 1220 00:46:03,840 --> 00:46:05,999 But maybe some lab will pick up the 1221 00:46:06,000 --> 00:46:07,569 idea and do it. 1222 00:46:07,570 --> 00:46:09,809 So it might be something that we might 1223 00:46:09,810 --> 00:46:11,219 read about in a few years. 1224 00:46:14,500 --> 00:46:15,500 Microphone one. 1225 00:46:16,330 --> 00:46:17,350 Thank you for the talk. 1226 00:46:18,760 --> 00:46:20,769 Do you know how much of this technology 1227 00:46:20,770 --> 00:46:22,959 is present in a typical quarter so 1228 00:46:22,960 --> 00:46:25,149 it could affect to gain some 1229 00:46:25,150 --> 00:46:26,439 physical information about its 1230 00:46:26,440 --> 00:46:27,440 surroundings? 1231 00:46:28,390 --> 00:46:30,609 That's yeah, this is an interesting 1232 00:46:30,610 --> 00:46:31,869 question. 1233 00:46:31,870 --> 00:46:33,729 So obviously, you couldn't use it for our 1234 00:46:33,730 --> 00:46:35,589 scheme because you would need to scan it. 1235 00:46:37,280 --> 00:46:39,499 You could probably, to some 1236 00:46:39,500 --> 00:46:41,629 extent, use it for these other schemes of 1237 00:46:41,630 --> 00:46:43,699 the other groups by using this memo 1238 00:46:43,700 --> 00:46:45,620 capability of scanning the beam. 1239 00:46:46,640 --> 00:46:48,769 I don't know how technically easy 1240 00:46:48,770 --> 00:46:50,269 it is to to do that. 1241 00:46:50,270 --> 00:46:53,059 You probably could do it in some way. 1242 00:46:53,060 --> 00:46:55,399 But the other groups 1243 00:46:55,400 --> 00:46:57,049 actually use development kits for 1244 00:46:57,050 --> 00:46:58,699 routers. So it doesn't seem to be very 1245 00:46:58,700 --> 00:47:00,859 straightforward to at least or 1246 00:47:00,860 --> 00:47:02,959 at least a few would think 1247 00:47:02,960 --> 00:47:05,059 about coming up with some project to 1248 00:47:05,060 --> 00:47:07,399 do it. It's probably easier to 1249 00:47:07,400 --> 00:47:09,469 buy a development kit for a 1250 00:47:09,470 --> 00:47:11,539 router which has a somewhat more 1251 00:47:11,540 --> 00:47:12,979 dedicated hardware than the router 1252 00:47:12,980 --> 00:47:14,449 itself. 1253 00:47:14,450 --> 00:47:16,889 And yeah, microphone 1254 00:47:16,890 --> 00:47:19,489 for, yes, 1255 00:47:19,490 --> 00:47:21,859 the technology that you showed, 1256 00:47:21,860 --> 00:47:23,899 it looks a bit like a static radar 1257 00:47:23,900 --> 00:47:26,059 system, but the power levels are 1258 00:47:26,060 --> 00:47:26,989 much lower here. 1259 00:47:26,990 --> 00:47:28,729 What's about the ratio between a 1260 00:47:28,730 --> 00:47:31,289 commercial radar system and actually 1261 00:47:31,290 --> 00:47:32,689 a great deal of power between a 1262 00:47:32,690 --> 00:47:34,909 commercial radar system and just 1263 00:47:34,910 --> 00:47:36,049 Wi-Fi? 1264 00:47:36,050 --> 00:47:38,179 Oh, I probably should be frank 1265 00:47:38,180 --> 00:47:39,859 in saying that. I don't know. 1266 00:47:39,860 --> 00:47:41,969 So my 1267 00:47:41,970 --> 00:47:44,269 take on it is that at least military 1268 00:47:44,270 --> 00:47:46,939 radar is employing huge powers like 1269 00:47:46,940 --> 00:47:48,709 Kilowatt maybe. 1270 00:47:48,710 --> 00:47:50,659 And we are definitely way below that 1271 00:47:50,660 --> 00:47:52,459 because we work with commercial devices 1272 00:47:52,460 --> 00:47:54,979 that are in the watt range 1273 00:47:54,980 --> 00:47:55,980 at most. 1274 00:47:57,250 --> 00:47:58,759 So that would be a factor of a thousand, 1275 00:47:58,760 --> 00:48:00,699 but there might be advanced radars which 1276 00:48:00,700 --> 00:48:02,250 actually use much less power. 1277 00:48:05,800 --> 00:48:06,800 Microphone three. 1278 00:48:07,570 --> 00:48:09,789 So how long did it take to take 1279 00:48:09,790 --> 00:48:11,829 one of the hologram pictures and how 1280 00:48:11,830 --> 00:48:13,179 large is the picture? 1281 00:48:13,180 --> 00:48:15,369 So it took us a 1282 00:48:15,370 --> 00:48:17,289 night. So we typically programed the 1283 00:48:17,290 --> 00:48:19,119 device when we left the lab and when we 1284 00:48:19,120 --> 00:48:20,979 came in the morning, if we were lucky, 1285 00:48:20,980 --> 00:48:21,980 the picture was there. 1286 00:48:24,480 --> 00:48:26,909 So it took something like 1287 00:48:26,910 --> 00:48:29,549 seconds of acquisition time per pixel, 1288 00:48:29,550 --> 00:48:31,659 and the pixels were something like 1289 00:48:31,660 --> 00:48:33,659 a few hundred by a few hundred pixels in 1290 00:48:33,660 --> 00:48:34,660 size. 1291 00:48:35,580 --> 00:48:37,949 So that took a night. 1292 00:48:37,950 --> 00:48:40,589 And this, of course, is a big drawback 1293 00:48:40,590 --> 00:48:42,239 as it stands right now. 1294 00:48:42,240 --> 00:48:44,489 But as I said, I think it would be very 1295 00:48:44,490 --> 00:48:46,499 easy to get better in that respect. 1296 00:48:46,500 --> 00:48:48,839 So even if we only scale 1297 00:48:48,840 --> 00:48:50,969 it in one dimension, you take an array 1298 00:48:50,970 --> 00:48:53,249 of antennas in one dimension, like 1299 00:48:53,250 --> 00:48:55,799 a magic wand with many, many antennas. 1300 00:48:55,800 --> 00:48:57,869 And you can it only in the second one, 1301 00:48:57,870 --> 00:48:59,939 potentially the third one, then you 1302 00:48:59,940 --> 00:49:02,129 could speed up acquisition by a factor of 1303 00:49:02,130 --> 00:49:03,809 100 2000. 1304 00:49:03,810 --> 00:49:05,789 So it would be in something like a minute 1305 00:49:05,790 --> 00:49:07,889 to arrange to get a full picture 1306 00:49:07,890 --> 00:49:08,890 of a building. 1307 00:49:11,300 --> 00:49:12,980 Right, microphone one. 1308 00:49:14,600 --> 00:49:17,389 So in that case, wouldn't 1309 00:49:17,390 --> 00:49:19,489 be performing help 1310 00:49:19,490 --> 00:49:21,589 because, well, it works 1311 00:49:21,590 --> 00:49:23,059 both ways. 1312 00:49:23,060 --> 00:49:24,849 And this way you could also. 1313 00:49:27,870 --> 00:49:30,299 Scan the signal, you eliminate 1314 00:49:30,300 --> 00:49:31,619 the space with 1315 00:49:32,700 --> 00:49:35,579 wouldn't that decrease the 1316 00:49:35,580 --> 00:49:38,339 critical dimensions you need to have 1317 00:49:38,340 --> 00:49:40,769 on the acquisition antenna 1318 00:49:40,770 --> 00:49:42,479 array a bit? 1319 00:49:42,480 --> 00:49:44,159 Yeah, this is a very good point. 1320 00:49:44,160 --> 00:49:46,409 And we actually thought about 1321 00:49:46,410 --> 00:49:48,539 that. So you could come up with kind of 1322 00:49:48,540 --> 00:49:50,729 a hybrid scheme where you do 1323 00:49:50,730 --> 00:49:52,979 scanning to cover some 1324 00:49:52,980 --> 00:49:55,079 space, but you take more than 1325 00:49:55,080 --> 00:49:57,149 one antenna, so you take a set 1326 00:49:57,150 --> 00:49:59,669 of antennas. That is like the inverse 1327 00:49:59,670 --> 00:50:01,829 of the memo area in a 1328 00:50:01,830 --> 00:50:02,729 commercial router. 1329 00:50:02,730 --> 00:50:04,829 So it could record 1330 00:50:04,830 --> 00:50:06,779 where the light is coming from, a long 1331 00:50:06,780 --> 00:50:09,659 haul two-dimensional set of directions, 1332 00:50:09,660 --> 00:50:11,759 and then you might be able 1333 00:50:11,760 --> 00:50:14,009 to actually accelerate 1334 00:50:14,010 --> 00:50:15,179 that scheme quite a bit. 1335 00:50:15,180 --> 00:50:17,629 And I think that if we wanted 1336 00:50:17,630 --> 00:50:19,709 to go for passive localization 1337 00:50:19,710 --> 00:50:21,749 and all these things, probably the 1338 00:50:21,750 --> 00:50:23,489 solution in the end would look like that 1339 00:50:23,490 --> 00:50:26,069 having a set of some enhanced 1340 00:50:26,070 --> 00:50:27,689 antennas with a good direction, 1341 00:50:27,690 --> 00:50:29,819 sensitivity and having a discrete set of 1342 00:50:29,820 --> 00:50:32,069 them at strategically placed 1343 00:50:32,070 --> 00:50:33,070 points in a room 1344 00:50:34,170 --> 00:50:36,119 that might be sufficient to get all the 1345 00:50:36,120 --> 00:50:38,099 information you need to at least see the 1346 00:50:38,100 --> 00:50:39,100 emitters. 1347 00:50:41,560 --> 00:50:44,019 I can think of a smaller 1348 00:50:44,020 --> 00:50:46,449 array with a lot of antennas 1349 00:50:46,450 --> 00:50:47,559 like eight by eight, 1350 00:50:49,380 --> 00:50:51,579 but this would basically end up being 1351 00:50:51,580 --> 00:50:53,320 a base array 1352 00:50:55,540 --> 00:50:58,179 radar, but on steroids. 1353 00:50:58,180 --> 00:50:59,219 Right, right. 1354 00:50:59,220 --> 00:51:01,779 So this is but this is a very realistic 1355 00:51:01,780 --> 00:51:02,769 thing to do. I believe. 1356 00:51:02,770 --> 00:51:04,899 So, um, still, if you 1357 00:51:04,900 --> 00:51:06,879 so if you ran holographic, you were 1358 00:51:06,880 --> 00:51:08,709 processing on that data, it would be a 1359 00:51:08,710 --> 00:51:11,019 very coarse resolution image because 1360 00:51:11,020 --> 00:51:13,179 it's only about eight pixels. 1361 00:51:13,180 --> 00:51:15,339 But if you combine that with scanning 1362 00:51:15,340 --> 00:51:17,409 it to some only some strategic 1363 00:51:17,410 --> 00:51:19,959 points, then that might be sufficient 1364 00:51:19,960 --> 00:51:22,359 to actually get good pictures 1365 00:51:22,360 --> 00:51:23,829 even without scanning a whole two 1366 00:51:23,830 --> 00:51:24,830 dimensional plane. 1367 00:51:26,120 --> 00:51:27,259 We have another question from the 1368 00:51:27,260 --> 00:51:29,389 Internet, aside from 1369 00:51:29,390 --> 00:51:31,009 the paper, did you publish source code to 1370 00:51:31,010 --> 00:51:32,479 hardware schematics to make it easier for 1371 00:51:32,480 --> 00:51:34,729 others to reproduce the results? 1372 00:51:34,730 --> 00:51:37,209 We didn't publish it, so 1373 00:51:37,210 --> 00:51:39,469 we wrote a very extensive 1374 00:51:39,470 --> 00:51:41,659 supplementary material where we describe 1375 00:51:41,660 --> 00:51:43,819 every well, every detail of how we 1376 00:51:43,820 --> 00:51:45,979 actually did the acquisition, including 1377 00:51:45,980 --> 00:51:47,749 references to all the hardware components 1378 00:51:47,750 --> 00:51:49,369 and plans. 1379 00:51:49,370 --> 00:51:52,069 But we already got requests from 1380 00:51:52,070 --> 00:51:54,199 hackers and students that wanted to 1381 00:51:54,200 --> 00:51:55,399 recreate that. 1382 00:51:55,400 --> 00:51:57,769 And of course, if people 1383 00:51:57,770 --> 00:51:59,389 just write us an e-mail, we will be very 1384 00:51:59,390 --> 00:52:01,579 willing to to give 1385 00:52:01,580 --> 00:52:03,739 away all the software and information. 1386 00:52:07,200 --> 00:52:09,849 Microphone three, yeah, hello. 1387 00:52:09,850 --> 00:52:11,969 Yeah, it's a simulation of 1388 00:52:11,970 --> 00:52:13,829 the attendant right in the ceiling of a 1389 00:52:13,830 --> 00:52:16,079 building, but you said it would be 1390 00:52:16,080 --> 00:52:18,569 too, you know, expensive to actually 1391 00:52:18,570 --> 00:52:20,909 refurbish a building that way. 1392 00:52:20,910 --> 00:52:22,800 What happens if you, you know, 1393 00:52:24,210 --> 00:52:26,669 construct a new building and plan 1394 00:52:26,670 --> 00:52:29,429 to build a scene from from the get go? 1395 00:52:29,430 --> 00:52:30,929 Would that be more reasonable? 1396 00:52:30,930 --> 00:52:32,609 Yeah. 1397 00:52:32,610 --> 00:52:35,009 So it's probably not so much refurbishing 1398 00:52:35,010 --> 00:52:36,059 the building, which would be so 1399 00:52:36,060 --> 00:52:38,129 expensive. It's just the physical array 1400 00:52:38,130 --> 00:52:39,359 of antennas itself. 1401 00:52:39,360 --> 00:52:40,379 Unfortunately, 1402 00:52:42,660 --> 00:52:45,059 if things grow with a square of pixels, 1403 00:52:45,060 --> 00:52:47,399 it quickly gets very expensive. 1404 00:52:47,400 --> 00:52:49,529 So we did some estimates of 1405 00:52:49,530 --> 00:52:51,599 it. Even if these if each of these 1406 00:52:51,600 --> 00:52:54,479 antennas would cost you only ten cents 1407 00:52:54,480 --> 00:52:55,929 and you wanted to have a thousand by a 1408 00:52:55,930 --> 00:52:58,079 thousand pixel area, then you would end 1409 00:52:58,080 --> 00:53:00,329 up with 100000 euros of investment 1410 00:53:00,330 --> 00:53:02,409 only for the electronics 1411 00:53:02,410 --> 00:53:04,359 to acquire the signals. 1412 00:53:04,360 --> 00:53:06,479 So that would be a very high, 1413 00:53:06,480 --> 00:53:08,699 um. Well, Barria, I 1414 00:53:08,700 --> 00:53:10,889 think to a practical implementation. 1415 00:53:13,950 --> 00:53:15,629 Microphone for yeah. 1416 00:53:15,630 --> 00:53:18,319 Given that it's relatively easy to 1417 00:53:18,320 --> 00:53:20,729 a wireless fire into 1418 00:53:20,730 --> 00:53:23,939 continuous wave mode, 1419 00:53:23,940 --> 00:53:25,919 do you think this would improve the 1420 00:53:25,920 --> 00:53:28,109 signal quality and this is something 1421 00:53:28,110 --> 00:53:30,689 you've tried so setting 1422 00:53:30,690 --> 00:53:32,999 the router to some continuous mode rather 1423 00:53:33,000 --> 00:53:34,320 than downloading a video? 1424 00:53:35,520 --> 00:53:36,809 We tried both. 1425 00:53:36,810 --> 00:53:38,699 And for our scheme it doesn't matter 1426 00:53:38,700 --> 00:53:40,739 because we throw away all this bit 1427 00:53:40,740 --> 00:53:42,959 pattern information anyway. 1428 00:53:42,960 --> 00:53:44,799 So we started with that. 1429 00:53:44,800 --> 00:53:46,319 Then we bought a model where we couldn't 1430 00:53:46,320 --> 00:53:47,699 do it for some reason and then we 1431 00:53:47,700 --> 00:53:50,039 switched to downloading a big file. 1432 00:53:50,040 --> 00:53:52,139 But it actually does 1433 00:53:52,140 --> 00:53:54,209 for us. Doesn't matter for future 1434 00:53:54,210 --> 00:53:56,489 experiments. Probably one very, very 1435 00:53:56,490 --> 00:53:58,169 powerful thing would be to move to 1436 00:53:58,170 --> 00:53:59,879 dedicated emitters, obviously. 1437 00:53:59,880 --> 00:54:01,739 So if we could have an ultra wide bent 1438 00:54:01,740 --> 00:54:03,899 emitter, these pictures would look 1439 00:54:03,900 --> 00:54:06,179 way better right from the start. 1440 00:54:06,180 --> 00:54:08,309 And this is the follow 1441 00:54:08,310 --> 00:54:10,049 up project would definitely be a thing to 1442 00:54:10,050 --> 00:54:11,050 do. 1443 00:54:13,200 --> 00:54:15,209 Sorry, one quick follow up question. 1444 00:54:15,210 --> 00:54:17,579 Couldn't you just maybe print antennas 1445 00:54:17,580 --> 00:54:19,829 and a cable and then just lay 1446 00:54:19,830 --> 00:54:22,019 cables on the top off like a big 1447 00:54:22,020 --> 00:54:23,349 warehouse or something like this? 1448 00:54:23,350 --> 00:54:25,709 Yeah, you might be able to find some 1449 00:54:25,710 --> 00:54:28,079 way to make that work, 1450 00:54:28,080 --> 00:54:30,119 but as soon as you need to at least 1451 00:54:30,120 --> 00:54:31,979 switch the antennas to select a 1452 00:54:31,980 --> 00:54:34,139 particular one, you end 1453 00:54:34,140 --> 00:54:35,579 up with a switch and that will be a 1454 00:54:35,580 --> 00:54:37,109 semiconductor device and that will be 1455 00:54:37,110 --> 00:54:38,279 expensive. 1456 00:54:38,280 --> 00:54:40,139 So I'm not an electrical engineer. 1457 00:54:40,140 --> 00:54:42,299 I don't see a straightforward way 1458 00:54:42,300 --> 00:54:44,429 to do it in a completely 1459 00:54:44,430 --> 00:54:45,839 passive scheme where you only have 1460 00:54:45,840 --> 00:54:47,729 antennas and nothing else, it might be 1461 00:54:47,730 --> 00:54:49,829 doable. And then that array could 1462 00:54:49,830 --> 00:54:51,959 be a viable thing to do. 1463 00:54:51,960 --> 00:54:53,639 But as soon as you need only a little 1464 00:54:53,640 --> 00:54:56,069 switch, at every point, it's probably 1465 00:54:56,070 --> 00:54:57,070 too expensive. 1466 00:55:00,120 --> 00:55:01,120 Microphone three, 1467 00:55:02,520 --> 00:55:03,520 yes. 1468 00:55:03,990 --> 00:55:06,149 What do you think it is feasible to turn 1469 00:55:06,150 --> 00:55:08,339 this around and use a transmitter 1470 00:55:08,340 --> 00:55:10,529 array like I want 1471 00:55:10,530 --> 00:55:12,929 with a hundred ISPs, uh, 1472 00:55:12,930 --> 00:55:15,389 which sent rapidly in succession 1473 00:55:15,390 --> 00:55:17,519 of frame and receivers through one 1474 00:55:17,520 --> 00:55:20,129 receiver to get 1475 00:55:20,130 --> 00:55:21,830 the hologram from Jack? 1476 00:55:23,270 --> 00:55:25,669 Sought to invert that, to have something 1477 00:55:25,670 --> 00:55:28,189 like, uh, a holographic 1478 00:55:28,190 --> 00:55:30,439 array that could create any wireless 1479 00:55:30,440 --> 00:55:31,670 waveform that you want 1480 00:55:33,770 --> 00:55:36,919 to transmit, array like transmit. 1481 00:55:36,920 --> 00:55:37,920 OK. 1482 00:55:38,840 --> 00:55:41,149 Mm hmm. And then 1483 00:55:41,150 --> 00:55:43,069 have one receiver and. 1484 00:55:43,070 --> 00:55:46,009 Yes. And look at it. 1485 00:55:46,010 --> 00:55:48,379 Well, yes, I think that in principle 1486 00:55:48,380 --> 00:55:50,479 should be doable since, as 1487 00:55:50,480 --> 00:55:52,609 you say, it's just the inverse of our 1488 00:55:52,610 --> 00:55:53,610 approach. 1489 00:55:54,690 --> 00:55:56,689 Yes. If you if you if you could switch 1490 00:55:56,690 --> 00:55:58,609 them one by one, probably, I think that 1491 00:55:58,610 --> 00:56:00,289 would work. But of course, it would be 1492 00:56:00,290 --> 00:56:02,179 way more expensive because emitters are 1493 00:56:02,180 --> 00:56:03,649 more expensive than receivers. 1494 00:56:05,840 --> 00:56:07,969 I know, at least in principle, I think 1495 00:56:07,970 --> 00:56:10,139 it should be feasible and 1496 00:56:10,140 --> 00:56:12,349 you get for three years, 1497 00:56:12,350 --> 00:56:14,659 put 100 percent on the want and 1498 00:56:14,660 --> 00:56:15,709 you have one. 1499 00:56:15,710 --> 00:56:17,449 I think so. 1500 00:56:17,450 --> 00:56:18,979 Right. Yeah, that's true. 1501 00:56:18,980 --> 00:56:20,869 I mean, this is maybe maybe in a way, 1502 00:56:20,870 --> 00:56:22,939 this is what our colleagues did in 1503 00:56:22,940 --> 00:56:24,739 the other schemes where they actually had 1504 00:56:24,740 --> 00:56:27,679 a raise of Amita's 1505 00:56:27,680 --> 00:56:29,569 and they could face them in a way that 1506 00:56:29,570 --> 00:56:31,669 you could scan the beam 1507 00:56:31,670 --> 00:56:33,739 they looked at reflexion for rather than 1508 00:56:33,740 --> 00:56:35,479 receiving the signal at one point. 1509 00:56:35,480 --> 00:56:38,029 But this is something that you probably 1510 00:56:38,030 --> 00:56:39,619 could do the other way to. 1511 00:56:40,800 --> 00:56:42,869 OK, you're 1512 00:56:42,870 --> 00:56:45,029 not to modify your 1513 00:56:45,030 --> 00:56:47,099 setup for a different lifestyle and it's 1514 00:56:47,100 --> 00:56:49,499 like a triple eight or two, 11 AC 1515 00:56:49,500 --> 00:56:51,719 or and or any of the others. 1516 00:56:51,720 --> 00:56:52,649 No, we wouldn't. 1517 00:56:52,650 --> 00:56:54,779 And we actually did these experiments 1518 00:56:54,780 --> 00:56:56,939 both with two point four gigahertz wi fi 1519 00:56:56,940 --> 00:56:58,380 and five gigahertz wi fi. 1520 00:57:00,300 --> 00:57:02,159 The five gigahertz wi fi looks a bit 1521 00:57:02,160 --> 00:57:04,109 nicer and the resulting images because 1522 00:57:04,110 --> 00:57:05,429 the wavelength is smaller. 1523 00:57:05,430 --> 00:57:07,559 So you can see more details. 1524 00:57:07,560 --> 00:57:08,909 But the scheme doesn't make any 1525 00:57:08,910 --> 00:57:11,009 assumption on the standard employed and 1526 00:57:11,010 --> 00:57:13,259 we can transfer to any standard we like. 1527 00:57:13,260 --> 00:57:15,059 And this is actually an interesting 1528 00:57:15,060 --> 00:57:16,559 prospect for the future. 1529 00:57:16,560 --> 00:57:18,749 So both the bandwidth and 1530 00:57:18,750 --> 00:57:20,939 the frequency of these wireless 1531 00:57:20,940 --> 00:57:23,159 communication systems is 1532 00:57:23,160 --> 00:57:25,559 probably going to be increased 1533 00:57:25,560 --> 00:57:27,569 in future implementations. 1534 00:57:27,570 --> 00:57:29,669 So people talk about even moving up 1535 00:57:29,670 --> 00:57:31,769 to 60 gigahertz also. 1536 00:57:31,770 --> 00:57:33,539 And that would with that, you would be 1537 00:57:33,540 --> 00:57:35,579 able to see pictures of nearly optical 1538 00:57:35,580 --> 00:57:37,409 quality. So with millimeter scale 1539 00:57:37,410 --> 00:57:38,410 resolution. 1540 00:57:39,210 --> 00:57:41,399 And if the bandwidth grows along with 1541 00:57:41,400 --> 00:57:43,619 that, with much less speckles than we 1542 00:57:43,620 --> 00:57:45,719 had, so then 1543 00:57:45,720 --> 00:57:47,849 the whole security thing might 1544 00:57:47,850 --> 00:57:49,710 be worth another thought to. 1545 00:57:52,100 --> 00:57:54,499 Could you, for example, play a small 1546 00:57:54,500 --> 00:57:56,749 array of antennas at every 1547 00:57:56,750 --> 00:57:58,879 corner of a building and then 1548 00:57:58,880 --> 00:58:01,219 record different frequencies 1549 00:58:01,220 --> 00:58:03,619 or waves which enter this 1550 00:58:03,620 --> 00:58:06,919 race and calculate the position 1551 00:58:06,920 --> 00:58:08,989 of a source in the building 1552 00:58:08,990 --> 00:58:11,809 from the different signal strengths 1553 00:58:11,810 --> 00:58:14,389 of different arrays, 1554 00:58:14,390 --> 00:58:15,799 receivers? 1555 00:58:15,800 --> 00:58:18,199 So without a race, this is actually 1556 00:58:18,200 --> 00:58:19,099 being done. 1557 00:58:19,100 --> 00:58:21,259 This is how these many of these indoor 1558 00:58:21,260 --> 00:58:23,479 positioning systems work that 1559 00:58:23,480 --> 00:58:26,329 you measure signal strength to specific 1560 00:58:26,330 --> 00:58:28,279 routers in the building. 1561 00:58:28,280 --> 00:58:30,799 Um, so 1562 00:58:30,800 --> 00:58:32,539 we could do that. And but I think that 1563 00:58:32,540 --> 00:58:35,389 actually blowing up these 1564 00:58:35,390 --> 00:58:37,609 antennas, antenna arrays is 1565 00:58:37,610 --> 00:58:39,529 a very interesting thing to do. 1566 00:58:39,530 --> 00:58:41,689 So, as I said, a kind of a hybrid 1567 00:58:41,690 --> 00:58:43,579 scheme where you have some strategically 1568 00:58:43,580 --> 00:58:45,319 placed points, but on each of them you 1569 00:58:45,320 --> 00:58:47,659 have a small antenna array 1570 00:58:47,660 --> 00:58:49,849 that could be the way to go to to make 1571 00:58:49,850 --> 00:58:52,129 it really viable scheme 1572 00:58:52,130 --> 00:58:53,769 for commercial applications. 1573 00:58:55,640 --> 00:58:57,709 Still, microphone one, this is 1574 00:58:57,710 --> 00:58:59,129 going to be the last one. 1575 00:58:59,130 --> 00:59:00,949 Sorry, it just one quick question. 1576 00:59:00,950 --> 00:59:02,729 What you do sounds very similar to a 1577 00:59:02,730 --> 00:59:03,919 radio telescopes, too. 1578 00:59:03,920 --> 00:59:05,989 So did you actually talk 1579 00:59:05,990 --> 00:59:08,149 to people from Lofa or České 1580 00:59:08,150 --> 00:59:09,619 or something like that? 1581 00:59:09,620 --> 00:59:11,719 Uh, not with respect to that 1582 00:59:11,720 --> 00:59:12,949 project. 1583 00:59:12,950 --> 00:59:14,749 It's true. So it's very similar. 1584 00:59:14,750 --> 00:59:16,399 They essentially do the same thing. 1585 00:59:16,400 --> 00:59:18,229 They actually do it on a global scale. 1586 00:59:18,230 --> 00:59:20,629 They link together telescopes on 1587 00:59:20,630 --> 00:59:23,029 different continents to get a 1588 00:59:23,030 --> 00:59:25,039 very sharp picture and they think about 1589 00:59:25,040 --> 00:59:27,199 building kilometer scale arrays of little 1590 00:59:27,200 --> 00:59:28,759 antennas to do that. 1591 00:59:28,760 --> 00:59:31,189 Um, we have been talking 1592 00:59:31,190 --> 00:59:32,899 with these people yet 1593 00:59:33,950 --> 00:59:35,719 because they probably will be doing 1594 00:59:35,720 --> 00:59:37,159 precisely that. 1595 00:59:37,160 --> 00:59:39,479 So they, 1596 00:59:39,480 --> 00:59:41,539 in a way, will 1597 00:59:41,540 --> 00:59:43,669 use it in very much the same way, maybe 1598 00:59:43,670 --> 00:59:45,019 in a slightly more restricted way, 1599 00:59:45,020 --> 00:59:47,389 because they look at very focused sources 1600 00:59:47,390 --> 00:59:49,489 and they, in a way, end up with beam 1601 00:59:49,490 --> 00:59:50,629 scanning. 1602 00:59:50,630 --> 00:59:52,249 But it's a it's indeed a similar 1603 00:59:52,250 --> 00:59:53,250 approach. 1604 00:59:55,930 --> 00:59:58,419 We're now moving into tomography 1605 00:59:58,420 --> 01:00:00,219 and like twenty one centimeter signal and 1606 01:00:00,220 --> 01:00:02,139 so on, and that sounds very similar to 1607 01:00:02,140 --> 01:00:03,399 what you were trying and I guess they 1608 01:00:03,400 --> 01:00:04,989 have very advanced algorithms because 1609 01:00:04,990 --> 01:00:07,929 their signal to noise ratio is abysmal. 1610 01:00:07,930 --> 01:00:10,089 Right. And so you might be actually 1611 01:00:10,090 --> 01:00:11,799 that. Yeah, this is definitely a good 1612 01:00:11,800 --> 01:00:13,359 idea. There is one different one 1613 01:00:13,360 --> 01:00:15,249 important difference, though, that they 1614 01:00:15,250 --> 01:00:17,289 only look at signals that are infinitely 1615 01:00:17,290 --> 01:00:20,229 far from the receiver. 1616 01:00:20,230 --> 01:00:22,329 So it's a good 1617 01:00:22,330 --> 01:00:24,039 idea and probably we should look at that 1618 01:00:24,040 --> 01:00:25,090 a second time. 1619 01:00:26,380 --> 01:00:28,569 So as when we did it, it looked 1620 01:00:28,570 --> 01:00:30,909 like looking at the coherent 1621 01:00:30,910 --> 01:00:33,069 optics papers, um, 1622 01:00:33,070 --> 01:00:35,229 would be more useful once it gets 1623 01:00:35,230 --> 01:00:37,299 to reconstructing three 1624 01:00:37,300 --> 01:00:39,339 dimensional things that are very close to 1625 01:00:39,340 --> 01:00:40,840 the receiver. 1626 01:00:41,860 --> 01:00:43,839 But they probably have very good they are 1627 01:00:43,840 --> 01:00:46,179 probably very good in terms of 1628 01:00:46,180 --> 01:00:49,329 spa sampling and, um, 1629 01:00:49,330 --> 01:00:51,759 and well, estimation 1630 01:00:51,760 --> 01:00:54,249 schemes if you have very noisy data. 1631 01:00:54,250 --> 01:00:56,459 So, yes, definitely. 1632 01:00:56,460 --> 01:00:58,659 Right. So the next week will be here in 1633 01:00:58,660 --> 01:00:59,679 15 minutes. 1634 01:00:59,680 --> 01:01:01,359 It's about machine check the proofs in 1635 01:01:01,360 --> 01:01:02,409 everyday software and hardware 1636 01:01:02,410 --> 01:01:03,850 development. Let's give him a hand.