Sensitivity issues with bladeRF 2.0 micro A4 compared to RTLSDR

[matprec]

Hi all,

we recently acquired a bladeRF 2.0 micro A4 for evaluating its performance on Mode S secondary radar signals for PDOA.
Our initial focus are the ADS-B extended squitter messages because their format and signal are well known and widely supported.

Using various antennas, we have significant sensitivity issues with the bladeRF, especially compared to a standard RTL-SDR v3.

To reduce potentially problematic variables like cabling etc, the test setup is reduced to two Nuand Tri-Band antennas directly connected to the SMA Port of the respective SDR. The SDRs are taped in the same height to a sky-facing window, with spacing >lambda/2. Distance to other metallic elements like window frame is also sufficient. LNAs (2x BT-200) are removed from the setup.

Using the latest source from dump1090 from flightaware for a freshly built and modern dump1090, latest version of the ads-b FPGA build (v0.11.1) from the homepage, freshly built ads-b publisher (https://github.com/Nuand/bladeRF-adsb/t ... adeRF_adsb), libbladeRF version is system package manager supplied 2.4.

Both devices run simultaneously to rule out runtime differences.

Using RTL-SDR, i get plausible and frequent ads-b messages, randomly verified against ads-b exchange.

Using bladeRF, i get not even a usable fraction of the messages. Many messages are scrambled and wrongly "recovered" by dump1090. Correct messages, if they come, have RSSI of roughly -50 dBm.

My thoughts so far:
That the RTL-SDR performs well with the same antenna makes antenna or environmental factors unlikely to me.
There is a telco tower in direct line of sight of the office window though, roughly 200-250m according to maps.

I'm aware that there are better suited antennas for ads-b, i choose the tri-band because it should be available to you as well for testing.

I did not conduct comprehensive tests from homeoffice yet, but at least on a quick but comparable setup, it showed the same symptoms.

Tuning to FM stations works with SDR++, although there are signficant broad artifacts, even with analog bandwidth reduced to 2MHz. Offcenter tuning the SDR and virtually tuning to the station produces way better signal, moving the SDR tune in 1kHz steps moves artifacts in way larger steps. This suggest aliasing to me. Same with a telescopic whip antenna "correctly" tuned for FM (tbh, FM needs no antenna tuning, but still). Larger sampling rates help a little bit, but not much.

At 1090 MHz, there are not many artifacts in the waterfall with SDR++, at least not directly on 1090 or vincinity. Using off-center tuning to 1085 for ruling out a dc spike, i used fft hold 0 to show maximum signal energy per bin, a bin size 8k at 10MS/s as a sweet spot for having enough frequency resolution but also enough samples to average the signal energy over (randomly timed 120us pulses). This shows that there is some energy at 1090 MHz with SNR at least a few db over the noise floor (~-90/-100dB @ gain 38).

Do you have recommendations for further troubleshooting steps?

Greetings from Germany,