That’s 150 mA vs the airspy advertised 50 mA of biastee power available.
Would recommend getting an external biastee anyhow to supply the rtl-sdr LNA.
(not that you have a choice probably)
2 Likes
@wiedehopf @janoskonya Thanks both for your suggestions.
The Airspy has been providing power to the RTL-SDR for the last 5 years! Maybe it decided to give up during the really hot weather? I did stick a few heat sinks on it when I first set it up. I do have an external bias t somewhere. I’ll dig it out and let you know.
A cheap one is here, without housing:
https://www.amazon.nl/-/en/Module-Stable-Filter-Board-500mA/dp/B0B5YPWR21
You can build the necessary low-noise DC adapter by picking a stabilizer IC e.g. one of the 7805 series.
See: https://www.tme.eu/en/details/ua7805ckcs/unregulated-voltage-regulators/texas-instruments/
The minimalist style circuit is really simple:
Capacitors are ceramic type. Less noise can be measured when capacitors are soldered close to the IC.
Voltage drop of this IC is 2V, so the “optimal DC in” (before regulation) is about 7-9 V DC.
Higher value may cause some more - unnecessary - heat dissipation.
1 Like
@janoskonya I’ve got this one now rigged up in my kitchen with a bit of coax core stuffed in the end of the LNA SMA 
Power comes from a 5v Pi power supply. It seems OK so I’ll get it in the loft tomorrow. Thanks for your suggestions.
2 Likes
That subject was discussed here a while ago
That would explain why it worked and then didn’t, when it got a bit too hot, maybe…
1 Like
Nice. 
…Just to have the optimal result:
Keep in mind that RTL’s filtered 1090 LNA has a mountain of amplification (~26dB if my memories serve well), thus designed to be used with some lenght of coax cable between antenna and the receiver.
Check the high and low values on RSSI and Airspy Noise graphs on “Graphs1090” to make sure that your receiver is not overdriven.
My coax is 16m long, and the receiver (R2) works optimally with Gain=13
Without coax cable, I am not sure that I could go low enough with gain setting…
1 Like
All up and running again after a few false starts. I have an external bias-t as the one in the Airspy Mini seems to be broken. How do my Noise and RSSI look?
OPTIONS= -v -t 90 -f 1 -w 5 -P 10 -C 60 Gain=auto (currently 15) Sample Rate = 12
1 Like
The values read from the graphs seem correct, although I haven’t used my station (R2) with 12 Mbps sampling for years - so I don’t have any recent memories of it.
The optimal minimum signal level for me is around -42. Then, by manually increasing the Gain, I no longer achieve an additional RSSI increase. Of course, by reducing the Gain, the upper value of the RSSI decreases, so it is proven that the upper signal range is not compressed.
As a result of the above settings, I can reach a state where the noise is already low enough (weakest is18), but I lose airplanes not because of the insufficient signal level, but only because of the distance.
By the way, your -C 60 looks a bit strict.
In the absence of a specific counterargument, you can safely set -C to around 90.
FWIW, here’s some Airspy graphs with some flip flopping between 12 and 20 MSPS, among other things, for comparison.
@janoskonya Thanks for the tips about gain, Sample Rate and -C. I’m tweaking my numbers as we speak! My weakest noise is now 18.2 I’ll give it 24 hours and have another look.
What antenna do you have?
@triggers
I have a FlightAware antenna stick on the roof, 9m AGL.
I have built some antenna versions before, but I had to realize that weather resistance, mechanical strength and the simultaneous (!) fulfillment of high-frequency parameters rarely occur in the same type. Also, I don’t want to climb the roof for 0.5% extra aircraft position. 
I’ve had many good antennas (diy ones) but they were somehow not completely satisfactory.
1 Like
@janoskonya Sample Rate is now 20, -C = 90. I messed around with Gain but came back to Auto. I think @prog and @wiedehopf have optimised auto gain.
Thanks for your help.
1 Like
Auto-gain is pretty much adequate when the antenna and the filtering are good. I would be more interested in cases where it deviates by a large margin from the optimal.
2 Likes
Yeah, I agree. Auto works pretty well. Devs are ace
Still, I have a bit better result when I use the manual gain trying to use wisely the given dynamic range, to get a noise minimum for the receiver - instead the RSSI (max) tuning. Having cavity filter and relative low altitude overhead flights near an airport, here it works slightly better. The rate of favorable change is really small (less than 1%), but it is noticeable.
If anyone wants to do some testing on gain, a while ago I wrote a couple of scripts which will cycle through various gain settings and produce some graphs detailing the results. It changes gain every 10 minutes and runs for 48 hours.
does the testing, and
produces the graphs at the end by pulling data from the graphs1090 data. The data collection is over 48 hours because that’s the period graphs1090 maintains the higher resolution data for.
I haven’t run it in a while so not sure if there’s any issue with current python or library versions.
Here is some discussion of it previously: Signal Strength Heatmap - #526 by caius
Here is some results from when I ran it originally: HOWTO: Airspy mini and Airspy R2: Piaware / dump1090-fa configuration - #3621 by caius
2 Likes
Thanks for sharing, got a Pi4 64bits with an AirSpy for ADSB and 2 RTL-SDR’s with pager decoding, all runs fine.
Which values do you have configured in /etc/default/airspy_adsb? Mine are: OPTIONS= -v -t 90 -f 1 -w 5 -P 8 -C 60 -E 20
It confuses me a bit that while the powergouvenor is set to power save, the CPU increases? Need to say the power consumption drops 1 watt.
There is some impact at the preamble.
This was the main trigger to write you, my Pi 4 is running at 55% cpu (ondemand) / 140% cpu (power save) and yours at 250% are you able to explain this to me? Thanks in advance.
The CPU runs at lower frequency and CPU usage is relative to the available computing power at that frequency.
You can increase -C to 90 and it will probably keep the preamble up even with powersave.
If you want any last bit of performance, you can run -E 60 which disables the limit to 20 preamble.
Unlikely you really see much difference though, just means usually 0.5% maybe 1% more messages / positions decoded while using lots more CPU
1 Like
Thanks for the help that clarifies a lot! Thanks for your great support for this community.
I used the automatic installation script but I noticed it installed stretch instead of the bookworm version.
I’m running 64-bit bookworm on Pi 5:
Is this an error in the install? Do I need to use the manual install instructions?
Is there a significant performance improvement with the bookworm airspy driver vs the stretch airspy driver?
Thanks!
show the output for this please.
ldd --version
ldd --version | grep -i glibc | grep -o -e '[0-9.]*$'