5

Is it possible to have a network of separate (but co-located) 'receivers' (antenna + adc) act as a phased array receiver?

Supposing one can get very precise timestamping on each receiver, couldn't all the receivers be assembled as a phased array?

I'm thinking about putting a bunch (4 for starters, at least 16 after) of uwb antennas and space them on a panel pointing all in the same direction, then for each antenna having a rtl-sdr or something better if necessary to acquire the signals with a very precise clock.

I want to receive specific signais, timestamp them precisely and then record them. Later in post processing, for some specific timestamp, I need to check the presence of the signals (using the whole array) and perform some analysis on them.

Touisteur EmporteUneVache
  • 61
  • 4
  • 2
    Hello, and welcome to this site! :-) What kind of antennas did you have in mind? You can [edit your question](https://ham.stackexchange.com/posts/18870/edit) and include that and any other details. – Mike Waters Aug 30 '21 at 21:48
  • 1
    Oh thanks for the kind words. I've tried to expand a bit. – Touisteur EmporteUneVache Aug 30 '21 at 23:22
  • 2
    Yes, the radio telescopes all over the world that get 'hooked up' together to do interferometry have their recordings timestamped by hydrogen maser clocks. It can be done, just do it carefully and to the required precision. – Neil_UK Aug 31 '21 at 16:34

2 Answers2

9

In short: yes, it's possible, but it's a bit harder than that. You want to drive all of the receivers from a shared clock, and sample them in phase with one another, to form a "coherent receiver". This requires some hardware work; you can't just plug multiple regular RTL-SDR sticks into a computer and make it work.

There are some coherent RTL-SDR projects that you can find by searching, though. Some of them seem to be available for purchase, and others are research projects. Some of them have code to illustrate how you can take the streams from those multiple receivers and use them to do things like direction finding or passive radar; studying them would probably be helpful if you want to figure out how to do similar things on your own.

I've seen something related from people who are doing VLF/ELF experimentation. Since they're working with frequencies under 100kHz, they can use commercial audio interfaces (sound cards) as direct-sampling receivers. If they set up two antennas at 90° angles from one another, and connect them to two channels of the same audio interface, they get a phase-coherent recording of both of them. Then they can phase and combine the two recorded channels to "electronically steer" the antenna after the fact of receiving a signal, to maximize the signal-to-noise ratio and the decode probability. Pretty crazy, isn't it? We're probably not far from being able to do similar things with HF using off-the-shelf hardware.

hobbs - KC2G
  • 11,469
  • 15
  • 28
  • Thanks. I understand better now. The part I'm missing here is the 'coherent' part. If my receivers are not synchronized but every sample is very precisely timestamped, do I need to sample each receiver exactly in phase? Couldn't I resync/re-phase them later on? I'm missing something but I'm not sure what. If I have a sample at time t on receiver 1 and another sample at the same time t on receiver 2, it's not enough? – Touisteur EmporteUneVache Aug 31 '21 at 00:12
  • 1
    @TouisteurEmporteUneVache no, it's not enough to timestamp the samples, you need the analog mixers to be frequency- and phase-locked to one another too. – hobbs - KC2G Aug 31 '21 at 00:20
  • Is the problem that, to work as a phased array, the samples for every channel needs to be performed at the same time? It wouldn't work if rx1 acquired sample X at time t, and rx2 had acquired the 'same' sample at t+(1/(Fsampling*2)) ? One would have to timestamp precisely enough and eventually perform some transform, not based on content, but based on non-exact alignment of the timestamps? Am I making any sense? *edit* sorry my comment crossed yours* – Touisteur EmporteUneVache Aug 31 '21 at 00:25
  • If I do direct sampling (which most cheap sdrs won't do, OK) http://www.panoradio-sdr.de/direct-sampling/ those problems still exist? Am I back to 'just' the timestamping? – Touisteur EmporteUneVache Aug 31 '21 at 00:36
  • 1
    @TouisteurEmporteUneVache if you were direct-sampling *and* you had a precise timestamp for every sample (which is harder than it sounds when the clocks are wandering independently and samples are transferred in blocks) then you could resample everything to a common timebase and work from there. – hobbs - KC2G Aug 31 '21 at 00:52
  • Thanks a lot @hobbs - KC2G – Touisteur EmporteUneVache Aug 31 '21 at 01:13
  • 2
    Keeping your time-stamp mechanism accurately sincronized enough across multiple receivers would be more difficult than just doing the sampling in phase with a single distributed clock. – hotpaw2 Aug 31 '21 at 03:49
  • 2
    Great answer. A (single) noise source fed (in phase) to all receivers provides a way to remove phase uncertainties. So as long as your receivers use the same base clock, you can correlate and find the noise, find the N sample / 180 ambiguities /thermal drift in the RF. Coherent multi-SDRs are cheap now. All Direction Finding antennas, and radio telescopes, use noise to synchronise several antennas, down to the last degree of phase where even thermal drift of components matter. – tomnexus Aug 31 '21 at 05:12
4

In Radio Astronomy, arrays are phased either by a distributed local oscillator, or by reconstructing the timebase (local atomic clocks synched to GPS) during analysis by also listening to known strong signals.

enter image description here

Source

Data sampled at a baseband of 2 GHz, recorded on hard drives, air shipped to a central location where interferometry is reconstructed from independent local oscillators using an algorithm exploiting known stronger signals in the same field of view.

This also allows for off-line geometrical corrections and light-path corrections (ionosphere, water vapor)

See also

For a distributed local oscillator via fiber optics, see answers to

https://www.youtube.com/watch?v=lbjCq8-ksBQ

uhoh
  • 543
  • 2
  • 16
  • 1
    Oh this is very interesting thanks for all the links! I'm now wondering what datation accuracy was necessary. I'll check the papers out. – Touisteur EmporteUneVache Aug 31 '21 at 06:44
  • 2
    If I recall correctly, with 3 VLB signals (with high-quality local timing) you can determine the relative phase offsets of the signals. With more signals the problem becomes "easier" but more compute intensive. – Jon Custer Aug 31 '21 at 14:10