4x dump1090 on KerberosSDR

Discuss KerberosSDR - 4x Coherent RTL-SDR
Post Reply
Stromeko
Posts: 43
Joined: Wed May 01, 2019 6:32 pm

4x dump1090 on KerberosSDR

Post by Stromeko » Wed May 01, 2019 6:57 pm

Finally received my kerberosSDR yesterday. :D

I don't have the USB-C PSU yet, but running four dump1090 instances just works when connected to a USB3.1 port directly on the PC motherboard (on the frontpanel connector I get it to run, but the devices frequently drop and then reconnect when running the fourth instance). The configuration currently is:
  • * PCB antenna
    * ADSB filtered preamp 38dB
    * passive power splitter (1:2 into two 1:2)
This is already great for determining the best gain value to use, even if I'm not using the coherency. Later on I hope to figure out how to determine the sample offsets (or sync the receivers) so I can merge the four data streams without breaking MLAT.

Stromeko
Posts: 43
Joined: Wed May 01, 2019 6:32 pm

Re: 4x dump1090 on KerberosSDR

Post by Stromeko » Sun Mar 01, 2020 9:04 am

So this took way longer than I've had hoped for various reasons, but yesterday I've recorded ADS-B in parallel on all four channels into an IQ dump file and then used GNU Octave to align the samples. That actually worked as expected, so I can now play around with the IQ files a bit before coding it up as an RT program.

hjj9777
Posts: 1
Joined: Sat Mar 07, 2020 4:20 am

Re: 4x dump1090 on KerberosSDR

Post by hjj9777 » Sat Mar 07, 2020 4:35 am

Curious to know your thoughts on combining dump1090 data with passive radar capability of the KerberosSDR and then mapping all of it in realtime, such that any objects that can't be accounted for by ADS-B positioning data could be marked/colored as anomalies on the map. Obviously, not all objects in the sky are required to emit ADS-B, so the lack of corresponding positioning data would not be definitive of anything in particular. However, if an audible alert were produced (or perhaps an SMS message sent via smtp) it could prompt an 'enthusiast' to run outside for a visual inspection just for fun! If I weren't already working 12 hrs/day, this is the project I would find most interesting. Any thoughts?

Stromeko
Posts: 43
Joined: Wed May 01, 2019 6:32 pm

Re: 4x dump1090 on KerberosSDR

Post by Stromeko » Sat Mar 14, 2020 11:59 am

hjj9777 wrote:
Sat Mar 07, 2020 4:35 am
Curious to know your thoughts on combining dump1090 data with passive radar capability of the KerberosSDR […] Any thoughts?
In a sense that is what MLAT already does, so I'm not sure what I'd gain (except more data to increase the confidence level or precision).

Stromeko
Posts: 43
Joined: Wed May 01, 2019 6:32 pm

Re: 4x dump1090 on KerberosSDR

Post by Stromeko » Sat Mar 14, 2020 12:11 pm

Stromeko wrote:
Sun Mar 01, 2020 9:04 am
I can now play around with the IQ files a bit before coding it up as an RT program.
So I ran out of memory a few times while doing that… :lol:

I can correlate the four data records to a stable alignment that looks correct when checking various signals. Remember that I work off a 1:4 passive splitter, so I had hoped that I could combine the four streams (in IQ space) into a single one and have the noise drop 6dB, but that's not happening and any combination of data actually decodes fewer ADS-B messages than the original ones. Looking at certain points in the data it seems that the magnitude signal on the four channels shifts around one or two samples depending on which message I look at. I have not yet tried to combine the data in magnitude space and the synthesize it back to IQ for dump1090, that might actually work better. It's still odd that averaging the four channels doesn't seem to work as it should when the samples are correlated.

Stromeko
Posts: 43
Joined: Wed May 01, 2019 6:32 pm

Re: 4x dump1090 on KerberosSDR

Post by Stromeko » Sun Mar 15, 2020 8:34 pm

Stromeko wrote:
Sat Mar 14, 2020 12:11 pm
Looking at certain points in the data it seems that the magnitude signal on the four channels shifts around one or two samples depending on which message I look at. I have not yet tried to combine the data in magnitude space and the synthesize it back to IQ for dump1090, that might actually work better. It's still odd that averaging the four channels doesn't seem to work as it should when the samples are correlated.
Note to self: I've likely fallen prey to the default dithering here… I need to find an IQ recorder that allows me to switch off dithering or compile my own rtl_sdr with that option patched in.

Stromeko
Posts: 43
Joined: Wed May 01, 2019 6:32 pm

Re: 4x dump1090 on KerberosSDR

Post by Stromeko » Tue Mar 17, 2020 5:57 am

Stromeko wrote:
Sun Mar 15, 2020 8:34 pm
Note to self: I've likely fallen prey to the default dithering here… I need to find an IQ recorder that allows me to switch off dithering or compile my own rtl_sdr with that option patched in.
Done that and it looks much better as long as I only look at the magnitude of the four signals (or alternatively force the angle on IQ to zero before averaging). But the angles of the raw IQ data are all over the place, so averaging directly kills a lot of signal and not just the noise.

Post Reply