Getting real performance out of an SDR: dynamic range, noise floor and where the dongle ends and the antenna begins
- Replies 2
- Views 13
- Created
- Last Reply
Top Posters In This Topic
-
Mod_Chrissy 1 post
-
Mod_Nancy 1 post
A lot of SDR questions come down to the same thing. People compare ADC bit depth and sample rate, then find the real limits are overload, a noisy shack, or an antenna that doesn't suit the band. This thread is meant to separate those factors. Add corrections if I get something wrong.
Start with dynamic range, because the marketing numbers can mislead. An RTL-SDR dongle has an 8-bit ADC, and the usable range is further limited by the tuner's gain stages. A strong FM broadcast signal can overload the front end and produce phantom signals across the band. Those images and intermod products are not weak signals you could dig out with better software. The overload happens before the data reaches the PC. Lowering gain, adding an FM broadcast notch filter, or using a bandpass filter ahead of the dongle all work in the analog domain, where the problem occurs. Receivers like the SDRplay RSP series, the Airspy HF+ and the Airspy R2 have more ADC resolution and better front-end design. Even so, 14 bits does not make a receiver immune to a strong local transmitter. Filters, attenuation and AGC settings still matter.
Gain is where many setups go wrong. Maximum gain makes the waterfall look busy and lively, but it often just raises the noise floor and pushes the ADC toward clipping. A better method is to tune to a quiet part of the band, then raise gain until the noise floor in your software rises noticeably. Back off a little from there. Check again at the busiest time of day, because overload often shows up when strong broadcast stations are at their best. If the noise floor rises sharply when you connect the antenna, the antenna or its environment is the limit. If it doesn't, the receiver's own noise is the limit, and a low-noise amplifier close to the antenna might help on higher bands.
HF needs separate treatment. The RTL-SDR v3 has a direct sampling mode that lets it receive HF without an upconverter. It is simple and cheap, but it is a compromise: the ADC samples the signal with little front-end filtering, so strong shortwave broadcasters can swamp weak amateur signals. The traditional alternative is an upconverter, which shifts HF up into the tuner's range. The software must then be told the converter's LO offset or the frequency display will be wrong. A converter adds its own mixer, oscillator and spurs, so you need to know whether a signal is real or a product of the converter. Dedicated HF receivers like the Airspy HF+ family or SDRplay models that cover HF directly avoid some of this, at higher cost.
On the software side, the main choices are these. SDR# is popular on Windows and has a large plugin ecosystem. SDRuno is the native software for SDRplay hardware. HDSDR is an older Windows program that uses ExtIO drivers and handles CAT control and panadapter-style setups well. GQRX is the common choice on Linux and macOS, built on GNU Radio and gr-osmosdr. CubicSDR is cross-platform and uses SoapySDR, which makes it handy for LimeSDR and HackRF. The HackRF One and LimeSDR are more flexible platforms, with wide tuning ranges, transmit capability and very wide bandwidth. They are not automatically better receivers for a narrow weak signal. HackRF's 8-bit ADC makes it prone to overload, and the LimeSDR needs careful gain staging.
For digital modes, the usual chain is SDR software demodulating audio, a virtual audio cable passing that audio to a decoder, and sometimes a CAT or rig-control link for frequency tracking. WSJT-X, fldigi, multimon-ng, dsd-based tools and ADS-B decoders all work this way or take IQ data directly. Decoding problems usually come from the signal, not the decoder. Check for wrong sideband, wrong filter bandwidth, audio levels that are too hot, and a sample clock or frequency offset that has drifted. Cheap dongles can have a frequency error of a few parts per million, which matters for narrow modes. Check the ppm setting against a known reference signal.
Antennas matter more than most upgrades. A discone or wideband whip is convenient for scanning but compromises sensitivity everywhere. A resonant dipole, a properly sited loop or a purpose-built antenna for the band will usually beat a wideband one. On HF, local noise from switch-mode supplies, LED lighting and USB 3 devices often swamps the receiver's own noise. A ferrite choke on the feedline, a good distance from the PC, and a clean power source can help more than a better dongle. An active loop or a small magnetic loop can also reduce received noise compared with a short whip near the house.
Panadapters and bandscopes are the last category. A panadapter feeds the radio's IF to an SDR so you can see the spectrum around the frequency you're tuned to. That is useful, but the SDR sees what the radio's roofing filter and IF stage let through, so the picture is not the whole band. A bandscope built into a transceiver is more tightly integrated but usually has less flexibility than PC software.
Questions for the thread. How do you decide when gain is set correctly? What filters make the biggest difference to a low-cost receiver? And is there a case for direct sampling over an upconverter on a budget?
Link to comment
https://www.hamradiobase.com/forums/topic/11802-getting-real-performance-out-of-an-sdr-dynamic-range-noise-floor-and-where-the-dongle-ends-and-the-antenna-begins/Share on other sites