Spotty Accuracy When Tracking Transmitter

Discuss KerberosSDR - 4x Coherent RTL-SDR
RJsaurus
Posts: 8
Joined: Sun Feb 14, 2021 4:16 pm

Spotty Accuracy When Tracking Transmitter

Post by RJsaurus » Wed Mar 31, 2021 3:06 am

Hey forum!

Just posting here about a project I'm working on for school and some issues I've been having. We have our Kerberos and Pi and everything correctly calibrated and set up(I'm pretty sure). But when we're stationary(and probably also moving) we can't get a stable DOA. I'm wondering if there's an explanation for the different confidence values and such to see if that's the problem?

Our spacing element for this testing session was just .1 and it's a 121.775 MHz signal so we are going to test Thursday with an array with larger spacing and see if that works. If not, then our issue with getting DOA is probably settings related.

Any input is appreciated!! I didn't want to infodump but I can provide screen recordings/anything else about our setup if you believe it's not just a DOA value page issue.

Thanks so much,
RJ

ckoval7
Posts: 93
Joined: Fri Sep 27, 2019 8:54 pm

Re: Spotty Accuracy When Tracking Transmitter

Post by ckoval7 » Wed Mar 31, 2021 10:38 pm

The easiest way to explain the confidence is to take a look at the DOA graph. If you pay attention to the scale on the side, it ranges from 0 to some auto-scaling negative number. The tallest (and hopefully only) peak is the DOA value. Confidence is derived from the difference between the peak and the trough, the bigger this difference, the better.

The main things that will mess with your accuracy are:
  • Reflections -
    If there are objects near your antenna array, especially metallic ones, the signal will reflect which will show up in the DOA graph as multiple peaks or a small difference between the peak and trough (low confidence). I've also seen it where the reflected peak is taller than the real signal. You really have to be mindful of what's around you before trusting the DOA value.
  • Poor calibration -
    Leaving the antennas attached or just forgetting to do the calibration will yield poor results. You need to recalibrate any time you change frequency or gain.
  • Differences in cable length -
    Just because you purchased four of the same antennas doesn't mean the cables are all the same. There was over a 6 inch deviation between the cables on the antennas I purchased. I lined them all up, cut the cables to matching lengths and put new SMA connectors on them.
If you haven't done so already, practice on a strong signal. Pick an FM station or similar to hunt down and make sure you're comfortable with your gear.

If you set everything up correctly you'll get great results: https://twitter.com/corey_rf/status/1345526183375613954

RJsaurus
Posts: 8
Joined: Sun Feb 14, 2021 4:16 pm

Re: Spotty Accuracy When Tracking Transmitter

Post by RJsaurus » Thu Apr 01, 2021 2:53 am

Thank you so much @ckoval7!

We're heading out in 12 hours with a larger array and what you told me in mind.

I didn't look very closely at the DOA graph so I think that will help me a lot. Your description actually makes sense.

I think I did calibration right, and our cables seem to be identical.

So we will try and practice on a big FM station and then move to fine tuning to track the transmitter we have. Will update you!

rtlsdrblog
Site Admin
Posts: 2895
Joined: Mon Nov 19, 2012 11:54 pm

Re: Spotty Accuracy When Tracking Transmitter

Post by rtlsdrblog » Fri Apr 02, 2021 6:04 am

A spacing factor of 0.1 is quite low. When it's this low multipath will heavily skew the results as any multipath tends to "pull" the bearing away from the true result. I know the array will be huge at 121 MHz, but try to keep the spacing factor around 0.3.

Make sure you calibrate with the antennas disconnected. And as ckoval noted, the cables should be identical in length.

RJsaurus
Posts: 8
Joined: Sun Feb 14, 2021 4:16 pm

Re: Spotty Accuracy When Tracking Transmitter

Post by RJsaurus » Fri Apr 02, 2021 1:32 pm

Thanks we did calibrate with the antennas disconnected!

As for the spacing factor... yeah, we can't really afford for it to be much higher than 1.5 due to space constraints. Are there any other ways to account for multipath rather than just a larger array?

RJsaurus
Posts: 8
Joined: Sun Feb 14, 2021 4:16 pm

Re: Spotty Accuracy When Tracking Transmitter

Post by RJsaurus » Sun Apr 04, 2021 4:59 pm

Crazy idea: non uniform circular array??? Like an oval?? That could give us some more space but not sure if it's feasible

RJsaurus
Posts: 8
Joined: Sun Feb 14, 2021 4:16 pm

Re: Spotty Accuracy When Tracking Transmitter

Post by RJsaurus » Thu Apr 08, 2021 3:35 am

Here is a screen recording we took of what the DOA graph looks like when we drive the transmitter around it in a circle:

https://youtu.be/mY7HhzyAsIQ

I think I put in everything right and did the calibration right(with the antennas disconnected) and I even raised the spacing element to .25 instead of 1! Should I be doing something different with algorithm stuff ? Or change to the compass mode and change confidence numbers? Just a lil confused as to what this might look like to some people more familiar with it.

rtlsdrblog
Site Admin
Posts: 2895
Joined: Mon Nov 19, 2012 11:54 pm

Re: Spotty Accuracy When Tracking Transmitter

Post by rtlsdrblog » Wed Apr 14, 2021 11:11 am

Don't use MEM, I would recommend using the MUSIC algorithm first.

Can you provide some pictures of your setup? What are you using to test?

Are your calibration graphs looking nice and straight with the antennas disconnected?

RJsaurus
Posts: 8
Joined: Sun Feb 14, 2021 4:16 pm

Re: Spotty Accuracy When Tracking Transmitter

Post by RJsaurus » Wed Apr 14, 2021 1:44 pm

Ok, we switched over to MUSIC. In that video we used a spacing factor of .25

This isn't the exact setup from the video but it's pretty close: Image

We have a strong 121.775 mhz transmitter from Pointer. That's what we're tracking.

And yes, the calibration graph is completely straight with antennas disconnected and terminators on.

rtlsdrblog
Site Admin
Posts: 2895
Joined: Mon Nov 19, 2012 11:54 pm

Re: Spotty Accuracy When Tracking Transmitter

Post by rtlsdrblog » Fri Apr 16, 2021 3:17 am

First point is that the antennas you are using are way too small for 121 MHz. Coherent DFing can withstand quite a bit of tolerance in antenna element sizing, but you're currently using an 800 MHz antenna for 121 MHz, which is quite major.

If you have them, use the long whips that came with the set I think you purchased. Calculate the required length for 1/4 wavelength. and get as close to that as possible. For 121 MHz you're looking at about a 0.61m tall antenna.

The system might not work well if the antenna sizing is significantly wrong for the frequency you're working with as the incorrect sizing is going to cause all sorts of multipath issues.

Can you test first with a higher frequency first, say over 400 MHz? At least to get some results and practice with the system.

Secondly. if the array is sitting on the ground, you might have issues if the transmitter is higher. You'll want to raise the platform off the ground.

Thirdly, try to keep the cables neat and all heading in the same direction.

Fourthly, experiment with the transmitter polarization. You'll get best results when it's vertically polarized (antenna vertical as in pointing up) because you're using vertically polarized antennas. If your transmitter is horizontally polarized you could be seeing multipath effects becoming dominant.

Fifth, after calibration, did you reduce the bandwidth via decimation and FIR filtering down to only the bandwidth of the ELT signal?

Sixth, you mentioned that the signal is very strong. Please check the gains and spectrum graphs to ensure that the signal is not too strong, and is actually overloading the receiver. If it's really strong you'll want to run with lower gain settings.

Hope this helps.

Post Reply