 Hi everyone, my name is Paul Boven, call sign Papa Echo 1 November uniform tango like many good people. I'm a radio emitter and I have a slightly bigger dish to play with than most other ones. It's not just mine and I'll get into details about that. This is the Dwingelow Radio Telescope. It was built in 1954, opened in 56 by our Queen Juliana and at that time at 25 meters diameter it was the largest radio telescope in the world for about a year and a half. I'm not going to go into all the numbers because we've lost three minutes. I'm a member of Kamras which is foundation named after C.A. Muller, Lex Muller, a former head engineer and professor worked on the telescope and was also occasionally using it for hemorrhage activities and doing very well of course. We are now in 2007 taking over from the dish after a distinguished scientific career. We started using it basically for amateur radio and for radio amateur astronomy or amateur radio astronomy as you say. We have at least 300 sponsors working on the telescope and teaching programs on outreach, on observing with it, on doing hem radio with it. So this is just a picture of it's already a little bit dated of some of it at the reopening of the dish in 2014 after it had undergone a complete restoration where even the dish was lifted off decades of protective coating and painting and looks much better now almost like brand new apologies to anyone watching this on the stream. C.A. Muller Radio Astronomy Station we have three goals making the Dwingler radio to the communities of amateur radio astronomers and radio emitters. So yes that means you can actually in a certain circumstances call and use it. Stimulating the interest in science and technology in particular for the youth by providing access to the telescope and preserving and maintaining it as an industrial and scientific monument. But let's get into the technical details. This is our dish, low noise actually right here in the focus. We mix it down with the signal generator because the receiver that we have only goes so high it doesn't actually go high enough for what we want to do. And then we go into a back end which is the part that digitizes the signal and makes it ready for processing. And then contact the internet and we store it on a server with lots of hard disk rate configuration etc. All of that is locked to a rabbidian clock so we have good timing and frequency. We can capture at once about 25 megahertz of spectrum. And this is what it looks like in the early reality. I've drawn some colored arrows. These are the antenna connectors. From there we go to the mixer, then we go to the receiver and then we go into the back end. And the back end is basically a software defined radio. And then here are the LO signals and over there you have the rabbidium and that is our astronomy PC. I already had the HPCB on my doorstep. These are for scale SMA connectors. Pin pitch is here a half millimeter. You need something to drive it. I didn't feel like hand soldering an FPGA. So I opted to use a silence development kit. This is the FPGA itself. It's a Spartan 3A because it's a Spartan 3A DSP. We can actually use this chip. This chip is the gigabit internet driver and the internet connector and the rest of the thing we don't really use. To build it all together on my kitchen table, signal generator for the mixing. This is actually simulating the hydrogen line. This is the receiver. And here it goes into the AD converter through a little bit of rainbow flat cable into the FPGA and then the internet. And you actually see the spectrum. Oh great. I thought I had killed all software on this thing. I apologize for that. Give away. It took about two years to get it into a proper box. It took a lot of complaining from the other volunteers. You can't just have it laying around on a piece of cardboard. So again, AD converter, FPGA kit, JTAC for programming it's power supply. A little display so you can see what it's doing. And that's when it's an operation and you have a little spinner here that shows that it's actually getting the clock signals it needs. We have a lookup table with the FFT window function. So every incoming sample gets multiplied against the window function. FFT block, which I didn't make myself. The output of that are real and imaginary, which gets squared and added. So then you actually have the power per frequency. I keep integrating the received power. The reason for that is that otherwise the input here is already 700 megabits. The output there would be a lot much a lot higher. And that would actually exceed the throughput of the internet connector. So we do 64 integrations here and then we dump the spectrum. So develop three personalities for the back end. Pulsar mode, where it takes 512 samples, applies the window, does the FFT. That gives you a spectral resolution of 137 kilohertz. And 2000 and something spectra is paid out per second. Looking at the hydrogen line and other spectral lines you want more resolution. So we use the biggest FFT I could fit into the FPGA, which gives a 17 kilohertz resolution and a 267 spectra per second, a bit more sedate. And then if you really want the fire hose approach to radio astronomy, you simply try and get all of the samples out over a gigabit. That came a few years later when I had learned enough FPGA to actually be able to pipeline the CRC calculations, etc. And that gives you a gigabit of data, which you don't have to offline process. A little bit of radio astronomy. A Pulsar is the remnant of the core of a star that collapsed after it went through all of its fuel. A very violent collapse called a supernova. It's still heavier than the sun, but the gravity is so much that it completely gets squished together. The atoms can no longer exist. You only have neutrons left. And as it gets smaller, it actually spins up due to conservation of angular momentum. And you get a sort of a lighthouse beam at two ends that rotates around. And if the Earth happens to be in line with one of the lighthouse beams, we will see a very small increase in the amount of noise that we detect. And this is a recording that I did. You can see how regular they are. Bright enough for us that we can see single pulses. You can also see they're slanted because due to the interstellar medium between the Pulsar and us, the lower frequency actually gets delayed. And that's actually one over F squared. So it goes really like so. So the lower you get, the more delay you get. But out of these, you can measure lots of interesting facts about Pulsars. And I can actually show you, you can see the slanted there and there. You can actually see the slanted arrival of the Pulsar signal in time. And of course, if you don't have quite that long an antenna, what you take is you observe for longer and then knowing the Pulsar ratio, you can actually fold them together. But having a 25 meter dish is sort of having cheap mode for detecting Pulsars, of course. And in the telescope, we can actually listen to them live, which is a great demo for outreach when we have people visiting. A hydrogen, I'm not going much in the theory, but in the universe, in space, it exists in two states. And Dutch physicist discovered that it would actually radiate a 21 centimeter. That's the famous 21 centimeter lie. And the frequency is extremely accurate and stable. And that means if it's not on that frequency, that is because it's moving towards you or away from you due to the Doppler shift. This is a scan of our galactic plane. And if you download the slides later, this will look a little bit better. Projector is not quite doing it justice. What you see here is basically the rest frequency of hydrogen. Anything that's higher is coming towards us. Anything that's lower is going away from us. And this actually, these structures you see here, these arcs actually correspond to the spiral arms of our own galaxy. And that is the thing that the discovery of the hydrogen signal allowed us to show that our own galaxy is actually a spiral galaxy. We can also, just to an also see two signals and we see them significantly above the noise. This is without using a preamplifier because we don't have a preamplifier for that frequency. This is actually the civilian signal. And this is the wider spread military signal. This is the other band where you only have the military signal. We can actually, from this signal, detect the individual ones and zeros, so to say the individual BPSK values before the spreading. And then write your own software and decode it, etc. But in the interest of time, by then we're getting like three years ago. And I had started to learn to learn about GNU radio and actually got my own SDR and wanted to play with it and wanted to introduce it also to the Ringelow telescope. And for the existing back end that I showed you, I made a new GNU radio compatible mode. I already found out that if I put in 70 mega samples per second, RPC didn't keep up. So I made a mode where using a two-stage fur filter, we take the center five megahertz of our pass band, convert that into I and Q interleaved samples, spit it out over internet so that it gives us about 160 megabits which the PC can keep up with. And you can just straight plug that into GNU radio into the GDP receiver block. And one of the things you would like is more sensitivity. If you do the naive FFT window, FFT window, then at the edges of the window, you're basically multiplying your samples with values that are near to zero. So you're not actually using the samples. So you want a much larger window and not throw away your data. And there's a method in a radio astronomy called weight overlap at to do that where you have your input samples you multiply them into much longer sample than your FFT size. The next thing, mathematical trick is that pieces and then add them piecewise together and then do the FFT. So instead of doing a large FFT and then folding it together, you do the folding together first and then the FFT. And if you look at the difference in shape, how much more rectangular the PFB is compared to the regular FFT, you simply get much nicer shape in your frequency domain and that actually relates to the fact that you're using more of your samples. Now the radio has a polyphase waterbank block used. Unfortunately, that outputs the PFB bins as single outputs instead of as a vector. So if you want to do 4,000 of them, you need to draw a hell of a lot of lines. So instead, I implemented, sorry, the radio and this is actually from this parallab and the link is at the bottom and these are all the values that you need to fill in to make this work. But if you are going to do radio astronomy with an article as the air for instance, this is a great way to get just a little bit more sensitive to looking at a hydrogen line. So this is the UDP where the samples come in. I convert them to complex, so interleave shorts to complex and then it's just a normal flowchart where you got the four different chains, each with a different delay and then they're all added together. We do the FFT with no additional windowing of course and then complex to magnitude and then we start to integrate and then we get in trouble because in radio astronomy, we think in integration times of seconds for instance, you integrate up to say a second, then by the time the vector sync actually wants to show it to you, that's like two and a half minutes or so because it takes a long time to fill this buffer. So what we're actually doing is we integrate it and whenever it is full, we actually repeat the output 40 times just to flush just to flush it so that our display is a little bit real time. So that's a useful trick to low band with new radio stuff. And what you can actually do with this is for instance, we looked at M74, which is a spiral galaxy that we see at face on, which means that all the rotational Doppler is basically normal to a line of sight, so it's zero so we don't see it. So all the 21 centimeter radiation is just almost on the same frequency, which allows us to actually detect it at a distance of 30 and a bit million light years away. And that took to get this signal to noise, even with our telescope took an hour and a half of looking at the source, looking next to it, looking at the source and repeating that every 100 seconds. Another thing we do with the same backend and then the new radio interface that we made, we now have a flow graph to do SETI. And SETI, you might all know the SETI at home screensaver. We're working together with SETI Institute and we are now going to do SETI at cameras, where we record whether on telescope at the locations of interesting new, newly discovered planets. And then you'll be able to actually process the data on your own PC. And because it takes lots of time to do it, we do what happens often in radio astronomy is we use very few bits in our sampling. And actually the SETI at home data is one bit sampled, zero or one, at two and a half mega samples. So the backend gives us five megahertz. So I throw away half of it, decimated by two, complex to float, I interleave them, binary slice them. So I have only one bits, pack the bits into bytes and that goes into a file. Basically, we have now written all the software where we can actually feed this to the SETI at home and we can actually run this flow graph is actually run on our own data. What you see here is O and zero EME. This is actually a beacon in Belgium that is aimed at the moon and then we aim our dish at the moon and then we can easily see it. And we're hoping to start off to kick off our public site of our SETI at home this year. It's a Chinese satellite. It is in orbit around the moon. And before it got launched, they asked us if we could be a ground station, help them out because we have quite a few square meters of antenna for this. What we often do is we track this satellite as it is around the moon. We add the moon because our beam is not much bigger than the moon. So we actually know how far the satellite is from the moon. And one of our volunteers wrote code that actually calculates the offset. So we point a little bit next to the moon so they have the best signal. And then we actually stream this on our website and everyone can just listen into it. You can try and decode the package yourself. You can actually also do it with a reasonably big antenna. The software for decoding DSLWP is actually public. It's made by the Chinese and it is on GitHub. So you see it's just a new radio out of three module that you could use. And if you want to look at the images that we download and the telemetry that lots of people are downloading from the satellite and all uploading to this website. Once again, these slides will be online I think on Monday because I left the password to the Pental Barf account at home. And this is a picture that we downloaded on last Wednesday. So this is the student CMOS camera which is the amateur radio payload. And it sends packets out which we decode and occasionally packet goes on. And then somebody in Germany repeat that block. And then that block is repeated and we receive it. And they from all the packets from all the receivers around the world, they on their website that built this nice little image. And this is actually in the Chinese mythology. The moon is a little bunny and the Chinese students were very happy to have seen this bunny from very close by. And you might have actually seen a very famous image that they took where the earth and then where you see the moon and then you see the earth behind it that was really big in the news a few months ago. So finally, radio astronomy and resolution. So the resolution that you have in a in a telescope is depending on two things, your wavelength and the diameter of your of your mirror. So the Hubble Space telescope is 2.4 meter diameter, but works at the wavelength of 600 nanometers. So this point is 0.06 arc second resolution. So full circle is 360 degrees. Then one one sixtieth of a degree is an arc minute. One sixtieth of that is an arc second. Compare that to a 25 meter telescope, which is 10 times as big, but you're working at a wavelength of say 6 centimeters. You certainly you have a resolution that is 10 times 10,000 times worse, which makes radius for me much less interesting in sense. So if you want to have comparable resolution to what the optical world is doing, you actually would need a dish with a size of 240 kilometers. And then we get to what to do as a day job, which is I work for the for Jive, the joint institute for VLBI in Dwingelow, the Netherlands. And we actually coordinate the European VLBI network. We process the data that all these telescopes that you see over here generate and this whole array of telescopes together can work as one virtual telescope that has like a diameter almost the size of the earth. And therefore you get unsurpassed resolution. I think it's still the highest resolution that you can get in in astronomy. You can see many famous telescopes like Effelsburg 100 meter Archibo 300 meter diameter. The Americans have their own VLBI network. They have lots of telescopes as well, of course, there are many more, but these are the core telescopes of the European VLBI network. And one of them is missing, I think, and that's the Dwingelow telescope. So, and this is where my hobby and my work start to completely intersect and become indistinguishable. A few months ago I wrote a VLBI front end or VLBI flow graph for the Dwingelow telescope. So we take the samples. This is 310. Now we have an actual USRP. Now we don't only have our homemade FPGA thingy. So with the TwinRx, we sample the data. We bring it down to a standard bandwidth that we use in radio astronomy. Using a polyphase channelizer, we bring it down to four channels. These are complex. So over here I simply turn them into upper side bands. And then over here we sample them into two bits. Because, again, due to storage limitations, we don't want to use too many bits. So we use two-bit sampling. Run into a bug in GNU radio where if you try to do a low number of bits, then it gets a little bit confused about how it actually wants to do the rounding. Most of the time it uses one way of rounding. Occasionally it uses different, but that is being addressed. At the moment I just use the full rate of a float and then I use a lookup table to bring it down to two bits. Then I put on the VLBI headers that tell you what the telescope is and, very importantly, what the time is. And then it goes to file. Because I can't do this in real time. But still it works. And on August, which is a professional radio telescope, Jolliubank in the UK, the Mark II telescope, and the Dwingelok telescope. And what you see here is the cross-correlation product on each of these baselines. And you can see you get a nice signal to noise out of it. That in itself is just an intermediate product. What we actually ended up doing is taking the Fourier transform of how it of that in all the different constellations as the Earth rotates. And out of that you can work back to the image. That itself would be a two-hour talk. Not going to do that. Things I still have to do. At the moment we're using a rubidium. We have a wide-wrap link. It goes all the way from the maser over here to, well, almost here. We're like 300 meters short. We've already dug the fiber. We've put the fiber in. This Wednesday we're going to do the fiber welding and then have it complete. And the other thing, which is a big challenge, because I need to have this done by early March, is it doesn't work in real time. So I'm currently taking this flow graph and turning it into a RFMoc flow graph, which is a bit of a challenge. So if you're interested in cameras, what we do, keep in touch. I'll put the slides online, including this one. We have a website. We have a mail address, GitLab for all the stuff that we make, which is open. Including the VHDL design of the backend. But I would not encourage anyone to look at it, because I'm not good at VHDL. And also it's a 10-year-old design. We have a Twitter and we have an Instagram. We also have public observation data, if you want to look at Pulsar data process yourself, if you want to look at diesel WP data, for instance. And especially if you're from around Holland, we welcome new members. And if you're interested, get in touch. Thank you. I was in the GPS spectrum picture. I saw a few spears, and I think you know where they come from. Well, first of all, this is just a process in the backend. This is just an FFT picture in the backend. Well, also it has to do with statistics. As in, over a long time this should be the nice sync shape that you would expect. But this is just done over 512 samples and then averaged over 40 seconds or so. And the other thing is there are actual imperfections and nonlinearities. And you see that, for instance, here in the place where it actually should be zero and you get these little, and that actually has to do with the fact that if you square the signal, you collapse it in VLBI. So in VLBI we actually need to have the phase at each telescope the same, or at least a known offset. And what currently happens is that every of those participating telescope has its own hydrogen maser, because that's really the only frequency standard that has sufficient stability. One of the things I'm working on at the moment is upgrading wide-rabbit, improving wide-rabbit so that it actually, and not just me, but together with colleagues and collaborators, that it can actually be transported over fiber. And we currently have 169, I think, 170 kilometer link wide-rabbit running. And in the next few months I'll be showing that it is sufficient to do VLBI by actually doing VLBI between the Dwinglow and the Westbrook radio telescopes. Time and phase. Yeah, well, the one in the UK is still on its own maser. Actually in VLBI what we do is we don't really care about the absolute phase. So at the beginning of an optimal project, because your hydrogen maser, they're very stable, but they're not very accurate. So you measure the frequency of each of the different hydrogen masers and then you keep using that for your observation.