HOWTO: Airspy mini and Airspy R2: Piaware / dump1090-fa configuration

Nice :slight_smile: much like mine

Have there been any updates to airspy_adsb v2.2-RC30? Is this still the best of the code changes?

Hi,
As an RC, airspy_adsb v2.2-RC30 is the latest - but there is a 2 month younger test version of it. This latest has slight improvements, at least on my station and circumstances.
See: GitHub - wiedehopf/airspy-conf: Configure airspy_adsb for use with readsb or piaware.

sudo bash -c "$(wget -O - https://raw.githubusercontent.com/wiedehopf/airspy-conf/master/update-binary.sh)"
sudo bash -c "$(wget -O - https://raw.githubusercontent.com/wiedehopf/airspy-conf/master/test-binary.sh)"

If youā€™ve already installed the airspy-conf stuff ā€¦ you can choose your poison.

2 Likes

Hello @wiedehopf ,

  • Is there a natural cap of msgs/sec rate around 2400-2600 at 20MSPS ? (higher is rarely spotted here)

Not inherently.

My suspicion is also that ATC radars will start reducing interrogation rates as the message rate ramps up.
Then you have more overlaps as the number of messages increases.
Do some math :wink: (iā€™m not sure exactly how to even approach that haha).
But you have some message rate and a message is a certain duration: Radartutorial
Note that there are short and long messages depending on DF type. (64 and 120 us i believe)

For 120 us long messages the absolute maximum theoretical rate is 8333 but the probability to get some overlap is obviously present at lower rates.
I have no stats on how many ModeAC messages are present which are on the same frequency.

2 Likes

Thank you for taking the time to respond.
I wanted to avoid studying the math background like any typically lazy guy, hoping youā€™re past the calculations and have a ready answer. :slight_smile:
ATC can actually reduce the query rate over a given amount of traffic, as this is for security, but beyond this phenomenon, the tendency is that the number of decoded DF17 packets may increase further, but the number of positions per second peaks at a certain value. Typically, a value of around 150-160 (RHS) can be read during the day.

Well you can do some very simplified models in your mind.

Say we have a rate of 2400 messages per second and theyā€™re all long and there is one message, then some free air. (for simplicity)
That will cover 29 % of available airtime.

So if you send an extra message at a random time, you have a 29 % chance of overlap.
Again thatā€™s much simplified but it gives a decent ballpark figure i think?

airspy_adsb tries decoding multiple phase offsets and due to the higher sample rate has a better chance to decode overlapping messages with phase offset.

If someone feels like doing some better math or wirte a simple program to just simulate the whole thing with random numbers, go ahead :slight_smile:

The simplified model you have outlined is perfectly sufficient for me to associate an illustration with it in an easy imagination.
lol Itā€™s better to save on expensive paper :slight_smile:

Indeed, the chance of finding the preamble is proportional to the sampling rate - and the high number of overlaps and the imperfect signal-to-noise ratio have the opposite effect. ā€¦ in addition, as traffic increases, the number of ACAS (DF0 and DF16) increases, further increasing the number of overlaps.

I think, we can simply enjoy the fact of having a working decoder - thanks for the great dev work behind.
For some reason, I am always tempted to complicate my otherwise quiet work day. :slight_smile:

Thank you

1 Like

The current math is heavily optimized. The small tweaks in the internal parameters (filters) can have different effects depending on the receive conditions. We have a good average, I would say. To get better results, we will need a completely different approach, from the hardware to the DSP. As of today, I donā€™t think itā€™s economical.

1 Like

If we skip the economic factor, what kind of improvement do you think is possible? And would there be different solutions for a busy site close to Heathrow compared to a site in a less aircraft-dense area?

Biggest improvement you could get is this in my opinion: 530, 6 Sector Squitter, Ground Station Antenna Antenna | dB Systems Inc.

Using 6 antennas that cover a different sector of the sky each.
Then running the same setup for all of them.

1 Like

I guess that antenna costs as much as a fairly decent used car :grin:

Then we can add the cost of 6 airspys etc. Probably yet another decent used car.

But you would be the antenna king in the ADS-B community :smile:

As you mentioned: ā€œif we skip the economic factorā€ :slight_smile:

  • Without cost limit - Iā€™d build a space station with 10-12 beam antennas aimed to Earthā€¦ and also an other one on the opposite side of the big ball. (full coverage)

Hereā€™s what I see at my pretty busy station near London:

The highest message rate Iā€™ve seen is 3115/s, but a more typically see a maximum of around 3000. It really depends on the conditions, as you can see that after a certain point more aircraft does not necessarily mean more messages.

I keep looking for a reasonable cost directional ads-b antenna, but itā€™s quite a niche so thereā€™s not much around. I donā€™t think you need to go as extreme as 6 sectors, even 3 or 4 would work quite nicely I think.

1 Like

But even if you sink the cost of a couple of used cars, you still wonā€™t see aircraft that arenā€™t there.

You can make the best of what you have, but you canā€™t make something from nothing.

1 Like

Probably using direct (under-)sampling hardware with heavy filtering to suppress the alias responses. Then use something more elaborate to detect the bits and do the correction based on ML. It can be a fun project, but I doubt the extra few frames can justify the effort/investment.

3 Likes

There is a more linear section on my station until the message ceiling is reached, where the msg/plane value decreases from about 20 to 10 (inversely proportional to the number of planes). Near the ceiling, the number of decoded messages barely increases.
This ceiling may vary due to changes in the physical conditions of wave propagation. eg humidity, electrical noise (natural and industrial), etc.
I think, this cap mostly depends on overlapping and changes in SNR.
Noise reduction at least on our site may help a bit but there is no miracle.
(Really low noise LNA, perfect rf lines, battery feed for LNAā€¦ and even better analogue and digital filtering on receiver side.)

I donā€™t think using a sector antenna (group) would improve the quality or quantity of fed datastream.
An already existing antenna with a gain of 5-6 dB + filtered LNA can supply the input of our receiver with signals in the appropriate dynamic range (even beyond the theoretical maximum distance).

Further increasing the gain of the antenna can also cause disadvantages. (eg reduced reception capability at higher elevation angles, phase errors in the overlaps of the common areas of sector antennas, when outputs are added together. // ā€¦ not to mention the not justifiable redundancy and possibility of malfunction resulting from the use of sector antennas as stand-alone stations.)

Stacked-phased antennas have the same gain as sector antennas -->> without the above mentioned possibility of phase problems with sector antenna-group as source.

Youā€™re avoiding all the overlap stuff we discussed.
Thatā€™s why you sectorize.
Also the noise floor can be lower.

Due to the reflector design of the antenna i linked the high elevation reception will be just fine.
If youā€™re concerned about that you just add some circular polarized antenna for the area directly above.

1 Like