Thursday, 26 September 2013

Passive radar with $16 dual coherent channel rtlsdr dongle receiver


My previous post describes the $16 dual channel rtl_sdr dongle hack. In the last few days I've done some more testing and it turns out I can use the system for passive radar! I didn't expect this, because the receiver only has 8 bits and passive radar requires a lot of dynamic range.

Airplanes and occasional specular meteor echoes. 
I hooked up the two channels into yagi antennas that we have used with Echotek and USRP receivers for passive radar. One of the antennas was measuring the transmit waveform, and the other was measuring the echoes. I ran a measurement, and to my great surprise, it worked just fine.

I did tweak the signal levels a bit in order to ensure that I optimally use the dynamic range. I also had the bandwidth set to 2.4 MHz, giving me about 4.5 bits extra dynamic range after filtering the signal to 100 kHz in single precision floating point.
Two log periodic antennas used to passive radar with the dual coherent RTLSDR R820T dongle.
This really does give us a glimpse of the future where high end digital receivers will cost $10 per channel. The low end ones are already in that price range. Think of all the potential science that can be done!

21 comments:

  1. Awesome! Keep up the good science....

    ReplyDelete
    Replies
    1. Thanks. This was still a technical demonstration. I'm hoping to do some science with this too at some point. I wish I was at a high latitude, closer to the aurora where I'd see field aligned irregularities on an almost daily basis. I might also try moving to lower frequencies at some point.

      Delete
  2. This is really cool. Is there anywhere I can get more information? for example what software are you using?

    ReplyDelete
    Replies
    1. Hi Helio,

      I'm using software that I've written myself. This requires removing the strong ground clutter and direct path signal. After this, there is an incoherent scatter radar processing step to see the weaker echoes, like airplanes.

      I'm in the process of writing a book about radar signal processing for CRC Press. The working title is "Practical Software Defined Radar and Radio Remote Sensing". It will have a lot of code examples written in Python, and radar data sets that can be downloaded from the web. Due to a lot of interest on this passive radar case, I'm probably going to write a chapter on how passive radar works, with a code example and this RTLSDR data. The planned release date is still 1.5 years away, but I'll keep you guys updated.

      Delete
  3. Very cool indeed. I'd love to understand it better. The software and also the antenna gain / link budget considerations.

    And might a network of these setups be fused to derive target position and track?

    ReplyDelete
    Replies
    1. I hope to write up an explanation about the signal processing, and associated example code.

      Sure, multistatic observations would do the trick. Passive radar also opens up other exotic multistatic configurations: you can listen to multiple radio stations at different locations. Then you would only need one receive location to obtain target trajectory. Interferometry is another possibility.

      The antennas don't have a lot of gain (<10 dB at best). The signal processing achieves approximately 140-150 dB noise floor to direct path ratio. I want to try double precision floating processing at some point, just to see if I can get more dynamic range.

      Delete
  4. Wow! What software do yuu use to plot those dopplers from FM? I have not yet experience about FM aircraft scatter (AS) radar setups, but AM signals work easily with standard audio spectrum software.

    Some experiments here with AS dopplers on OIRT 1 TV carriers: http://www.oh7ab.fi/foorumi/viewtopic.php?f=21&t=295&sid=a65f5873eb64887213bfe66af0786a51&start=80#p1174

    'Meteor scatters' are regular on mid-VHF FM, and still more frequent on low VHF like the 6m TV band. However, I have studied their spectrums, and found no doppler evidence about fast object movement that a meteor or its emerging trail ionization should create.

    Instead, the supposed 'MS' do have spread spectrums, which are usually distinctive to ionizing electric discharges in the air. It also appears, that the higher the discharge is, the longer it lasts, and the wider its spread spectrum is. Instead of 'MS' I am now calling them 'EDS' for electric discharge scatters.

    Pics of EDS here: http://forum.flightradar24.com/threads/3402-Aircraft-Scatter-Experiments?p=40325&viewfull=1#post40325

    EDS and Aurora scatter screenshots: http://peditio.net/fmtvdx/index.php/topic,17.msg56.html#msg56

    Regards, - OH7HJ

    ReplyDelete
  5. Auroraheijastuksen erikoisen kähisevän soundin selitys keskimmäisen kuvan kuvatekstissä:

    http://www.oh7ab.fi/foorumi/viewtopic.php?f=15&t=225&p=1271#p1271

    ReplyDelete
    Replies
    1. These are most probably scatter from Farley-Buneman irregularities caused by auroral currents. One of the geophysical uses for passive radar is to study this so called "auroral scatter" using inexpensive means.

      Perhaps the best example of the use of field aligned irregularities for geophysical remote sensing is SuperDARN, which is a network of radars that are used to study currents in the high latitude regions:

      http://en.wikipedia.org/wiki/Super_Dual_Auroral_Radar_Network

      Delete
  6. Hi Juha,

    Thank you about the remarks about Farley-Buneman irregularities! However, I guess I am rather ignorant about them, because I did not manage to find how they would explain the aurora scatter spread spectrum. Perhaps I did not quite understand what you did refer by 'These'..?

    Neither did I find any hint that they would explain the radio scatter spread spectrum effect of the short scatters this far thought to be caused by meteors, but which in reality bear no doppler proof of fast linear movement related to the usual explanation of meteors trails.

    The doppler spread spectrum of these short scatters is somewhat similar to aurora scatter spread spectrum, except that their spread spectrum is narrower, and the discharge time is shorter. This would suggest their origin as a high voltage ionizing discharge similar to that of aurora, except that these short scatters appear to happen in higher pressure than aurora, meaning of course that they would occur in lower altitude.

    Further, those short spread spectrum electric discharge scatters, which I would rather call as 'EDS' instead of real meteor scatter 'MS', appear be born quite independently of aurora scatter, because they keep happening constantly, no less when there is no aurora activity present at all. Also, the EDS appear be born in all directions, not only in the direction of aurora purge.


    Regards,

    - Juha OH7HJ

    ReplyDelete
    Replies
    1. There is classic paper on meteor trails [1] which describes that the scatter from meteors at long enough wavelengths is from the column of ionized plasma or the trail of the meteor, which is seen as a specular reflection. To first order, this has the Doppler shift of the neutral wind.

      With larger aperture and more power, it is also in some cases possible to see echoes from the meteor head, ie., the dense cloud of plasma around the moving meteor. This has a much smaller radar cross-section than the specular trail. The meteor head has the Doppler characteristics of the moving meteor (between 0..72 km/s).

      [1] G. R. Sugar, Radio Propagation by Reflection from Meteor Trails, Proc. IEEE, 1964

      Delete
  7. Hi Juha,

    That paper appears to correctly introduce the appearance of meteor scatters. To my experience, it tells reasonable basics about real meteor scatters with their ionized trails, causing a narrow spectrum doppler to the scatter by ionized trail drifting in the air by wind.

    Real meteor scatters behave like those described in the paper. However, here is the this far unexplained observation: If you check spectrum images of the so called 'meteor scatters' usual and frequent on 6 m band, their radio spectrum and dopplers are quite different to those meteor scatters described in the classic paper.

    The spectrums of these short scatters fail to show the narrow band scatter of a meteor trail. Neither is there any evidence of a fast linear momentary movement in the very beginning of the scatter, which should be there, if the trail were produced by a fast meteor generating a trail behind.

    During years of 6 m band aircraft scatter doppler observations and measurements my Spectrum Lab strips regularly record from tens up to to hundreds of these spread spectrum short scatters in an hour. They bear spectrum different to those of real meteor scatters, which appear very rare, compared to these rife short spread spectrum scatters.

    So these short spread spectrum scatters appear far too frequent to be caused by rather rare meteors with ionized trails. They also appear to occur all the time, without relation to meteor showers. Their frequency and spread spectrum has made me to suppose, that their birth may rather be related to atmospheric electricity, than to meteors.


    Regards,

    - Juha -

    ReplyDelete
  8. Hello Juha,

    Mr. Ashcraft is an observer in New Mexico specialized in high altitude lightnings called sprites. He has not only managed to get clear photos of them, but he also has spotted simultaneous radio scatters with sprites on the 50 MHz TV band - 'Radio Sprites'.

    Here are some of his very interestin images of narrow-waistes sprites: http://www.heliotown.com/Radio_Sprites_Ashcraft.html

    So our observations appear to support the discovery, that high voltage atmospheric or ionospheric electricity discharges may be a cause for radio scatters. So EDS's may be born in similar electrostatic fashion to aurora scatters, except that the EDS's last for a short period only in their lower altitude.

    Here is an extract from my earlier e-mail about sprites and radio scatters to Thomas Ashcraft, Heliotown, New Mexico:

    "... Your photos reveal the sprites as vertical discharges which have narrow electric arc 'waists', spreading upwards and downwards as streamers, which end up to luminous 'clouds' above and sometimes visible also below them. The pics make sense: I can now see from them how a high voltage electric discharge between two semiconducting ionized layers of atmosphere actually happens!


    Discharges In Ionosphere

    Ionosphere consists of alternate insulating and semiconducting atmospheric layers. By my understanding, the semiconducting layers have considerable high voltage potential between them, with the positive terminal up and negative down on earth.

    The easiest way for a high voltage discharge to jump through the insulating air layer between semiconducting layers is as an arc. I suppose that the 'waist' of the sprite is this discharge arc. Next your photos show how the arc divides to multiple streamers into the semiconducting ionosphere layers above and below, to collect high voltage charges from a wider area of the semiconducting layer.

    This high voltage discharge area inside ionosphere layer sometimes appears in your photos as a luminous 'cloud'. These discharge clouds appear brighter above the sprite, possibly because air pressure is lower up there. The discharge current is same in the above and in the below parts of the sprite. Below the higher atmosperic pressure collects more gas mass for the sprite discharge current to heat it up than above the sprite in the lower pressure. So the thin air above glows brighter than the dense air below, if I have reasoned it correct. ..."


    High Atmosphere Discharges

    It is a pity that we can seldom clearly see and photograph many varieties the high discharges from ground, like sprite halos, C-sprites, blue jets, gigantic jets, blue starters, and ELVES.

    However, because there are many varieties of EDS's by their duration and spectrum, so it would be tempting to think that each of them represents one of those TLE's or high atmosphere electric discharges or lightnings of different altitude and shape.

    If thinking further to understanding their relation to ground striking lightnings, as well as to other atmospheric electricity, we are close to understanding source of thunderstorm electricity and high altitude lightnings and aurora borealis, to mention just a few..! ;)


    Regards,

    - Juha -

    OH7HJ@hotmail.com

    ReplyDelete
    Replies
    1. Thank you for the interesting link.

      I have seen a few papers that describe scatter of VLF from sprites, but I hadn't come across VHF and UHF radio scatter off of ionized columns of sprites before. The videos are very beautiful and convincing. It would be interesting to do a passive radar or radar experiment with a co located camera.

      Delete
    2. Your cheap RTL sticks are very suitable for spotting both VHF sprite scatters and other still less known EDS's. On the OIRT 1 49750 kHz TV channel I very usually see simultaneous EDS's of separate distant TV carriers.

      By their relative strength on carriers of each TV Tx, I can tell their approximate direction. They off course are likely to appear in the direction of the Tx that they scatter most effectively!

      This of course only if the TV carriers are identified, which is not easy at all, because exact freqs of eastern TV stations are only partially listed.

      So in case of identified Tx's one can tell the approximate direction and sometimes even distance of the sprite or other atmospheric electricity discharge, by the EDS strength on each carrier from Tx's in different known locations.

      Now as you have a more advanced coherent setup of the stick RTL SDR's, there is a lot better choice for you with the numerous 4 m band eastern FM stations. That is because they are a lot easier to identify by their frequencies which are wide apart from each other, and because they are listed a lot better.

      Also our own 3 m band FM transmitters becomes available for EDS spotting with your coherent Rx setup, with Tx locations very easy to identify by good lists avaiable from SWL's.

      The low-VHF and also mid-VHF bands are very good for EDS spotting, because it appears that not only there are lots of powerfull FM BC transmitters, but also because low VHF frequencies appear to be scattered easier by these sporadic ionizing discharges than higher VHF freqs.

      By adding some mathematics, the approximate sprite etc. discharge locations should be possible to be calculated by software. A new kind of lightning radar, for the very high discharges, first of its kind..? ;)


      Regards, - Juha

      Delete
  9. Wow! That is crazy I have never hear of someone getting a working BPR system using less than 16 bits (well except for the guys who use crazy analogue direct path cancellation).

    Can I just ask if the listening antenna you used was shielded from the direct path? or at least if there was a large attenuation in the direction of the fixed path

    ReplyDelete
  10. Hi Joshua,

    We were also a bit surprised that we could do this with 8 bits. In fact, the results aren't that much different from our 16-bit receiver (USRP N200).

    The transmitter is quite far away (we are in Westford MA, and the TX is in Providence, CT), so yes, the direct path is somewhat attenuated.

    I did carefully adjust signal levels going into the rtlsdr dongle. In addition, to increase dynamic range, I used 2.4 MHz bandwidth and worked with single precision floating point when filtering the bandwidth to 100 kHz.

    The direct path is cancelled by estimating a 1 Hz wide the direct path complex amplitude at several different multipath configurations (I believe I had about 70 range gates), so there is in a way a 2.4 MHz -> 1 Hz bandwidth reduction, which provides the increased dynamic range. There is typically around 140 dB of difference between the noise floor and the direct path signal.

    The antennas are shown in the picture (the two log periodic antennas on the roof below the discone). I'd say there was perhaps <10 dB of gain difference between the two antennas on the direct path.

    ReplyDelete
    Replies
    1. Nice, that is seriously impressive! Hope to to see more great stuff from you. I love implementations where people are able to bring the hardware to a really simple level.

      Delete
    2. I would not be surprised that the 8-bit data is nearly as good as the 16-bit data. In the early 1960's a student at MIT named Sander Weinreb discussed how much information is lost if you go as low as one bit digitization (i.e., just clip the data). It is a long time since I read his thesis but I recall being very surprised at how little it was. His thesis was published as Tedchnical Report 412 of the MIT Research laboratory of Electronics in 1963.

      Delete