It’s in the writeup on Github: GitHub - wiedehopf/airspy-conf: Configure airspy_adsb for use with readsb or piaware.
Max value is 21 and each step down is -3dB.
-e value sets the utilization and in my experience is the most important, needs to keep it about 90% (one core that is).
Check utilization with “top”
Check for lost samples: “journalctl -en100 -u airspy_adsb”
Doesn’t matter too much … tweaking for the lowest signal around anywhere from -25 dBFS to -38 dBFS works fine.
But even way too high gain usually works fine.
Really all you need to know should be in the readme:
GitHub - wiedehopf/airspy-conf: Configure airspy_adsb for use with readsb or piaware.
Yes, meanwhile i took a step back and read before asking stupid questions over and over again 
I simply left the adsb config as it came from your script.
So far it works, but without a significant change in the message rate compared to the blue FA stick or the AirSquitter
Reducing the gain lower is also reducing the messages/s
Currently 45 aircraft with a message rate of 770
Waiting for higher traffic now to see how it performs
Full graphs
Airspy installed where the downtime is documented. The CPU load increased after i applied the -e switch suggested by @SoNic67. Lost samples are not documented.
The drop at 14:15 was a temporary switch of the AirSpy gain down to 15
1 Like
FYI Heres my airspy_adsb settings
these are well used and refined like a squeezed lemon yours will at first look different
sampling will be 12 or 20 ( not -m 24 )
Maybe start off with options= -v -f1 -e9.5 -w3 -t120 -g21 -m20
He has the Mini, the 20 SR is kind of experimental on that one… like 24 on the R2.
Also the options have spaces after the letter.
Not that much change (added the spaces between characters and numbers:
But @wiedehopf stated already at an earlier time: Maybe there is no further room for improvement even if an AirSpy is used, and i should not be surprised seeing no significant change
EDIT: At least the size compared to the FA gives me the option to power the Uputronics via USB from Raspberry without additional cabling to the Airspy
1 Like
Check with “top” the CPU usage.
And for the lost samples, like I shown above.
I did already.
the -e option increases the CPU utilization, from your journalctl command no lost samples were reported.
What does the -e switch improve?
Sorry, i am not that deep so far in this…
-e looks “deeper” into the captured data stream to select what can be ADS-B messages. Increases the message count with higher numbers.
Of course, at some point, its possible that all the available messages are decoded, so increasing the -e doesn’t get any more messages.
2 Likes
That would explain that not more can be squeezed out of my setup.
After 1 1/2 hours of comparison, the FA and Airspy are side by side from the performance while the AirSquitter has a very liiiiiiitle higher sampling rate per second clearly with the disadvantage of the price and the limited networking features (full access requires additional money).
Another improvement is that the number of tracks with a single message went down to almost zero.
But as @SoNic67 already said, this is not a general statement. Every setup and location is different.
I leave this running for a while and then decide which to put where, next updates to follow here
I will have more problems once my wife detects i took another wooden box from her stuff :
2 Likes
I used a flexible USB extension between my Airspy and the port, I didn’t want to mechanically stress the USB connectors.
Also i used a short, flexible, cable extension at the RF input for the same reason. But then, I have a bias-T feeding the LNA.
I’ve been using 20MHz for over 18 months now without issue. If you have the cpu to support it, eg a pi 4, there is little reason not to use it as it does provide a performance advantage.
Yeah, that’s what i used before with the FA stick. Problem is that i only have two meters which fills the box immediately.
Once i decide how it will look at it’s final position, i will change it anyhow
Temperature can be a reason, or?
In my setup i am still on m12 and it’s already much warmer than the FA stick. I do not have a heatpipe here which fits.
EDIT:
Quick check now shows message rate of almost 800 with 42 aircraft only. That value i never reached with the FA stick.
What i also see on the map is a little better coverage to the east, where my own roof is in the way.
2 Likes
I have it in the loft which can be over 40C or in the summer and the airspy has never missed a beat. It only gets warm to the touch.
The Airspy can be pushed harder, is limited only by your decoder processing power. I realized that my Pi3B was limiting my Airspy… for my number of airplanes (pre-COVID).
My experience was that I needed to keep the SR at 12 with an Pi3B.
Thanks both for your input. Very helpful.
The Airspy gets pretty warm, but no problem to touch it also for longer. Maybe i will need some heatsink if i keep it where it is.
Yes, currently the traffic is still low. Since Covid, none of my devices registered more than 80-90 aircraft while pre-Covid it was up to 150.
Maybe I will run another comparison once the traffic is back.
EDIT:
Switched to 20, but no directly visible change on the map. This is the log, so far no lost samples:
May 31 16:03:16 Raspy airspy_adsb[17553]: airspy_adsb v1.85
May 31 16:03:16 Raspy airspy_adsb[17553]: Listening for beast clients on port 47787
May 31 16:03:16 Raspy airspy_adsb[17553]: Acquired Airspy device with serial B06C68Dxxxxxx
May 31 16:03:16 Raspy airspy_adsb[17553]: Decoding started at 20 MSPS
May 31 16:03:20 Raspy airspy_adsb[17553]: Push client connected to localhost:30004 (beast)
Another chart, not really a comparison, but some observations:
it shows the peaks of aircraft and messages, looks like the Airspy overall performs better. Difference in Message of peaks is smaller than on the aircraft chart.
But that’s really a snapshot without claiming completeness
1 Like
I run my adsb cpu utilization level @ 90 %
and cpu pi also 90 %
That’s with differing antennas and stuff or do you have 3 antennas in the same spot now?
(and if it’s tested not at the same time you can’t tell anything from the numbers)
2 Likes