QIRX is able to work in the following environments:
This environment might also be useful in case the installation of the USB driver (e.g. the Zadig driver) is not desired on the machine running QIRX. No particular speed requirements exist for a PC with rtl-tcp as I/Q data server.
The sampling rate error is always corrected.
As mentioned, the operating principle of QIRX is based on very simple Fast Fourier Transform (FFT)-based down conversion (Decimation) of the incoming I/Q data.
For down conversion, a suitable part of the available spectrum of about 2MHz is cut out such that this spectrum piece satisfies the necessary bandwidth. To proceed with the following inverse Fourier Transform it needs to have a length of a power of two.
An Inverse Fast Fourier Transform (IFFT) switches from the frequency domain back to the time domain, providing samples directly suitable to be fed into the demodulator. The demodulator provides the audio samples to be presented at the headphones output of the PC.
The disadvantage of the method of course is its lack of flexibility. In order to preserve its simplicity, one has to stick with very few available bandwidths. Bandwidth and Demodulators cannot independently be selected.
For a more detailed description, please consult the Technical Report available in the download section.
The picture shows a nearly ideal spectrum of a DAB+ ensemble, 20-fold averaged (the degree of average is GUI-controllable).
In previous versions of QIRX (until 0.9.1) the center frequency has been found by inspection of the spectrum shape
and searching for the central "dip" (suppressed central carrier). While this usually worked well, the method has been
omitted in order to speed up synchronization also in more adverse receiving conditions like strong distorted central parts
of the spectrum due to multipath.
Starting with version 0.9.2, the coarse frequency synchronization is performed by scanning the receiving frequency around
a reasonable (e.g. 50kHz) range of
the center frequency, and searching for a correlation peak with the Phase Reference Symbol. With steps of 1kHz and some interpolation
the accuracy achieved is about 250Hz.
Coarse timing and frequency checks are only performed after a complete synchronization loss.
Usually the constellation is displayed showing the bits as dot heaps in a polar diagram, like in the
following picture, from a report by Andreas Müller, ETH Zurich.
The dot heaps show the single bits, the heaps being separated by 90 Degrees.
QIRX uses a different, linear display for an improved visual control of the synchronization accuracy.
The picture shows an example: The spectrum (upper part) shows regions of strong multipath reception reducing the signal strength around 178 MHz. The constellation (lower part) shows – for each subcarrier – how well the bits are arranged on their correct position. Each dot corresponds to a single bit value. The reduced signal strengths around 178MHz clearly shows a much larger scattering of the bits off from their correct values. In a polar display it would not be possible to assign regions of large scattering to the frequency regions (subcarriers) with reduced signal strength.
This kind of constellation display can be used to obtain additional information.
In QIRX, the sampling rate error is permanently corrected.
The four red-framed boxes in the spectrum all contain the same coded information about the transmitter. This information consists of a "Main Id" and a "Sub Id". The Ids are displayed in the table of the dialog.
Only a single transmitter contributes to the received signal. The number in the "Strength" column is an approximate value how much an average signal strength exceeds the selected "Threshold" value.
It indicates that two or more copies of a signal of identical origin arrive at the receiver, having taken different paths from origin to the receiver. The signal could have arrived on the direct way ("Line of Sight", LOS), and also could have been reflected by e.g. mountains or buildings and arrived later due to the reflection. At the receiver, both signals arrive as one signal consisting of the LOS part and the delayed part, added together.
A DAB network consists of stations at different locations transmitting exactly the same content, all on exactly the same frequency, synchronized down to the last microsecond. This is called a SFN (Single Frequency Network). As a consequence, a receiver located at different distances from say two transmitters encounters the same situation like in a multipath environment: It receives the identical information from two sources, entering its antenna as a signal added together from two sources. Without additional information like the TII decoding it cannot be decided whether a multipath situation is due to multiple paths of one transmitter or identical signals originating from two or more transmitters.
The software tries to find the overlapping region in the guard range of all three signals (In the picture the region between the red dashed lines). The "FFT Start" is then taken as the middle of that region (red dash-dotted line). As this calculation is only made after each frame and not after each symbol (what would be preferred but needs too much performance), it is expected that the relative delays do not much change from frame to frame. In case such a range can be found, there will be no degradation due to inter symbol interference (ISI), thanks to the cyclic nature of the FFT.
In case no overlapping region can be found, there will ISI, reducing the probability of a correct symbols decoding. ISI manifests itself in a larger scattering of the decoded bits in the constellation (see above).
It may be noted that in the commercial world the solutions with respect to finding the optimal "FFT Start" position seem to be kept confidential. This has been pointed out in the paper by Roland Brugger (IRT) and David Hemingway (BBC) covering this and related problems, titled "OFDM receivers - impact on coverage of inter-symbol interference and FFT window positioning" . This paper found its way into a technical report titled "SFN Frequency Planning and Network Implementation with Regard to T-DAB and DVB-T" latest edition of 2013.
The second spectrum was already shown above and is here repeated for convenience. The two notches in the spectrum are due to a true multipath reception and correspond to a delay of about three samples, corresponding to a "deviation" of the delayed signal of about 400m. A more quantitative analysis will be presented separately.