SDRPlay2RSS

This program acquires time-domain data from the SDRPlay receiver, computes spectra, and pipes the spectra to Radio Sky Spectrograph (RSS). The former is a software-defined radio (receive only) covering frequencies from near zero to about 1.7 GHz and having up to 8 MHz of useful bandwidth. The latter is a spectra acquisition and display tool (Windows based) used by the Spectrograph Users Group (SUG), a group affiliated with the Radio Jove. Besides piping integrated spectra to RSS for live display, the program can also save integrated and full-rate spectra to RSS's native SPS files, and also save time-domain data to binary files.

Please refer to the screen shot of the program window. In that figure the antenna, amplifier, and first bandpass blocks depict the installation's antenna and rf signal conditioning. 'Mixer 1' in the SDRPlay receiver sets the rf center frequency, which is followed by a band-pass filter, which sets the analog bandwidth to be processed. IF1 is a roughly 200-MHz IF frequency. Mixer 2 translates IF1 to I and Q base-band signals, which are then quantized by the dual ADC and transported to the host computer via USB2.

FFT and further processing are done by the computer. The frame size under the FFT block sets the sample count that are Fourier transformed as a block. The running sum depicts the channel-by-channel absolute squaring of the Fourier transforms and their accumulation. There is a great deal of data reduction in this accumulation. The offset and gain are applied to each integrated frame (each spectral channel), and the result is passed on to RSS. This is the baseline functionality of SDRPlay2RSS.

Basic usage

This baseline functionality is initialized by the 'Integrate' button. When armed in this way, the button's color is yellow and RSS acquisition may be started and stopped using RSS's start button repeatedly without rearming. Hit the 'Integrate' button again to disarm (and stop RSS acquisition if acquiring).

Most receiver settings must be set prior to arming. To change these settings, one must disarm SDRPlay2RSS if armed, change the setting(s), and then rearm. Exceptions to this rule are the gain reduction, offset, and gain settings, which may be changed on the fly.

There are four SDRPlay settings (see the next figure).

• The center frequency is typed into its box in MHz.
• The analog bandwidth to be processed is selected in the drop-down list box from among the values supported by the receiver. The values are in kHz.
• The sample rate f_s is entered in MHz. Its value must be greater than the analog bandwidth, or else an error message is printed on the status line and nothing else happens. The SDRPlay API does not allow undersampling.
• Gain reduction is a schedule of attenuations inserted at three places in the SDRPlay's analog signal path. It is expressed in decibels and is incremented and decremented using the two buttons. The current value is displayed.

Computer processing starts at the FFT block, where the time-domain data are Fourier transformed. The first setting is the Fourier frame size N, in the figure shown at 2⁹. The drop-down box allows selection of a power of two. A fundamental relation is that the spectral resolution Δf is

The absolute squares of those transforms are then computed (not shown), and integrated by the 'running sum' block for the interval entered in the 'integration time' box.

Although the SUG typically uses 0.15-s integration times for Jupiter and solar observations, much longer and shorter times can be used. The low end is limited by the computer and RSS being able to keep up with the data rate. Five-millisecond (0.005 s) integrations result in useful spectrograms on my computer. The high end is limited by the buffering required. See the notes below.

The offset setting may be adjusted on the fly using the 'U' and 'D' buttons. They increment and decrement, respectively the offset by the adjustable increment in the box just above the buttons. The current value of the offset is displayed above the offset block. The gain may also be adjusted on the fly by entering a value in its box. You must hit carriage return to register a new gain value.

The 'Span' setting sets the frequency span of the spectral channels that are actually sent to RSS or otherwise recorded. When the RF bandwidth is changed the span is set to its default, which is the rf bandwidth. But it can subsequently be changed to other values that are not more than the sample rate. The actual number of channels n_RSS sent to RSS is

nRSS = Span/fs × N

Streaming spectra to disk

RSS is able to accept integrated spectra at a rate in excess of 100 Hz (10-ms temporal resolution) on my computer. This rate can be exceeded by streaming integrated spectra directly to .sps files (and later opening the files using RSS). This is done by selecting the 'FD->SPS' option of the selector associated with the 'Integrate' button. Integration times down to the unintegrated rate (f_s/N) can be handled this way.

The duration of acquisition saved to a file is the 'Integrate record time' figure of the 'Integrate' menu. The following spectrogram is of Jovian S bursts at 1-ms intervals.

Streaming without integration is used for the highest temporal resolution (N/f_s). It is started using the 'Stream' button using its default 'FD->SPS' selection. Like with integrated streaming, the data file is of the SPS format, and SDRPlay2RSS streams to each file for the duration specified by the 'Streaming record time' menu item under the 'Stream' menu. Once activated, spectra are written to new files until the 'Stream' button is pressed again. File volume piles up quickly in this mode, particularly at high sample rates.

Note that in unintegrated acquisition, the temporal resolution and time resolution are reciprocals of each other.

where Δf is given above.

Where the full temporal resolution might be needed is, for example, for resolving emission by lightning, such as in this 12-ms spectrogram.

Being able to resolve more distant lightning emission can bring out what in this spectrogram is likely strong scattering by the ionosphere.

Background subtraction

There is a simple facility for setting a background spectrum and subtracting it from the spectrogram as it is being acquired. When the 'background subtraction' menu item under the integrate menu is clicked, it averages the next several spectra to determine the background to be subtracted. The program subsequently subtracts that background from spectra before they are sent to RSS or to a file. The number that are averaged is set by the 'background count' menu item. Once enabled, the menu item 'bgr' appears, which when clicked refreshes the background.

This capability is necessarily a bit clumsy and an otherwise imperfect technique. One must start the averaging not knowing that the roughly empty background at the start of the averaging will persist. And it abandons an antenna-temperature calibration, if implicitly present. But it is nevertheless useful at times for bringing out faint features.

Logarithmic scaling of spectra

Spectra may be delivered with either linear or logarithmic scaling of intensities. The log setting is useful for displaying a larger range of intensity scales, including down to the noise level of the spectra. Select one or the other using the 'Log scale' menu item under either the 'Integrate' or 'Stream' menu. Changes to this setting may be made on the fly (no restart needed). When in the logarithmic mode, the offset (offset block) is measured in decibels. To display many decibels in RSS, a small value of the gain is generally required to reduce the scale. The log-scale setting is saved and restored in configuration files.

Applying a windowing function

A block-by-block Fourier transform has a sinc-function response with its characteristically long 1/f tails. Around strong lines, these tails spread into nearby fourier channels, often obscuring them or at best contaminating them. The problem is often especially evident when the spectra are logarithmically scaled. A window function has the effect of suppressing those long tails. The price paid for this benefit is widening slightly the central peak. These effects are illustrated in the spectrogram to the right, where to the left of the gap no windowing is applied, and on the right it is applied. The lower line shows the suppression of the sidebands, and the upper line shows the widening of the line.

In SDRPlay2RSS, a Blackman window function is applied to the Fourier frames prior to the Fourier transform when the 'Window function' menu item under the integrate menu is checked. The setting is saved and restored in configuration files. Windowing can be applied in both streaming (non-integrating) and integrating modes.

Poly-phase filter bank

As was mentioned elsewhere, the baseline Fourier processing divides the time-domain data stream into blocks that are individually Fourier transformed, absolute squared, and accumulated (integrated). A windowing function may optionally be applied to the blocks as a means to suppress artificial sidebands of intense lines, sidebands that arise from the blocking of time-domain data. This method is effective, but it also degrades slightly the spectral resolution.

SDRPlay2RSS can alternatively insert a filter ahead of the Fourier transform that has the effect of simultaneously suppressing those side bands and preserving the spectral resolution. This configuration is called a poly-phase filter bank (PFB). Engage the PFB through the 'Integrate' menu.

A few other facts about this function.

• It cannot be run on the fly; the currently running acquisition must be stopped and restarted.
• The PFB and windowing function cannot be applied simultaneously.
• The frame size can be at most 1024 due to the difficulty computing taps for frames larger than this number.
• While a PFB is a win-win in terms of performance, it is computationally demanding. It may not work well at higher sample rates on some (or all) computers.
• The PFB setting under the 'Integrate' menu is saved and restored in configuration files.
• Finally, PFB use in the streaming mode is not supported.

The graphs to the right show a 20-MHz test signal without and with the PFB engaged. The vertical scale is logarithmic and the span is something like 50 dB.

Capture of raw samples

There is the ability to stream time-domain data (ADC samples) to disk at the full data rate. To use it set the streaming selector to TD->file and click the stream button. The files written have a 72-byte header followed by the samples. The header has the following structure:

1. The 16-character string "SDRPlay receiver".
2. The SDRPlay center frequency in MHz (double).
3. The SDRPlay rf bandwidth setting in kHz (int).
4. The SDRPlay sample rate in MHz (double).
5. The SDRPlay IF bandwidth: the value 0 (int).
6. 32 unused bytes (not necessarily zero).

The samples consist of a sequence of I/Q pairs. Each sample of each pair consists of a signed 12-bit word sign extended into its 16-bit word.

Samples are written to a single file and the file continues to grow until acquisition is manually stopped. This means that this mode of acquisition is not suitable for unattended operation. For example, at a sample rate of 12 MS/s, the file grows at a rate of 50 MB/s

Other capabilities under development

In the integrating mode there is a circular buffer of time-domain I and Q samples that is continually being filled. Its size is fixed at 10⁵ × the raw packet size set by the receiver. When the 'T' button is clicked, this buffer is written to an SPS file with some samples before the trigger recorded. The idea is to allow manually triggered capture of transient events for off-line processing. It has not yet tested. Let me know if you need it.

Loading and saving program configurations

Almost all of the settings of the program's user interface can be saved and restored. These settings are also saved when the program is closed, and restored when the program starts again. Use the file menu's 'Load', 'Save', and 'Save as' to load a configuration, save the configuration to the currently loaded file, and save to a new file, respectively. These files are saved and loaded from the program's current directory. They are human-readable ascii files made up of XML-tagged parameters. The program's current directory at run time should be explicitly set in the desktop icons and/or links that are used to start the program.

The idea of these configuration files is to customize settings for different kinds of measurements, including the directories to which data are written. For example, Jupiter measurements settings, incuding data directories, are saved to one configuration file, while meteor forward scatter measurements have settings and data directories saved to another configuration file, streaming settings for S bursts can go to a third, very high resolution settings for resolving lightning to a another, etc.

File paths

Paths to which spectrograms are written may be configured. There are paths for both integrating and streaming files, the former set through the 'Integrate' menu, the later through the 'Stream' menu. There is also a path for the observatory data accessed through the file menu, but its utility is questionable (just the path and not the data itself). These paths are saved and restored in configuration files.

Installation

• Install Radio Sky Spectrograph and the update, the current version of the latter as of this writing is here. When RSS is first run, from the 'Receiver menu', set the receiver type to 'RTL dongle receiver', and on the color menu, load the AJ4CO-Rainbow.txt color file.
• Go to the SDRPlay Start here page for the driver and tools installer, and for the SDRPlay receiver documentation. Download and run the SDRPlay installer.
• Establish a desktop and/or start menu icon that explicitly sets the current directory at run time.
• Install the filter taps from the zip file in a directory of your choice and when first running SDRPlay2RSS, set the taps-directory setting under the 'Integrate' menu to that directory.
• On my WinXP laptop, RSS initially failed to run with an error pointing to MSCOMM32.OCX. If this happens you may try replacing the existing version with version 6.1.98.16, as I did, which was present on my Win7 machine. It is probably necessary to register it at a command prompt with the command:

  regsvr32 mscomm32.ocx

The rf input does not require much signal intensity. It is easy to over power the input to the degree that it looks like there is no signal at all. I suggest that you insert sufficient attenuation at the input so that the working range of the gain-reduction setting is ≤ ~20 dB.

Other info

• The stream and integrate selectors next to their buttons have the cryptic A->B symbolism. In this symbolism, A is either 'FD' for frequency-domain source data or 'TD' for time-domain source data. B refers to the destination of the data, 'RSS' for sending to Radio Sky Spectrograph, 'SPS' for writing to RSS files, and simply 'file' for writing a binary-file format. So, for example, the 'Integrate' selection FD->RSS is the baseline functionality of SDRPlay2RSS that sends integrated spectra to RSS. To route the integrated spectra to disk (to an SPS file), select FD->SPS. Use this setting to acquire integrated spectra at higher rates than RSS can handle. There are analogous selections associated with the 'Stream' (non-integrating) button.
• SDRPlay's minimum sample rate seems to be 1 MHz.
• There is a status line at the bottom of the window. It displays certain actions taken, errors that may or may not cancel operations, and dropped sample counts when they occur.
• Controls that can be changed without stopping acquisition are windowing, log scale, gain reduction, offset, and gain. All other controls require that acquisition be stopped, the parameter changed, and acquisition restarted.
• SDR is computationally intensive. This program repetitively calculates spectra from ADC samples using multiple threads running on the available processor cores. As a result the CPU can run hot. On one occasion my laptop clicked off without warning because of over heating. So it is important to keep the CPU's internal heat sink free of dust that normally accumulates with use. Disassembling the computer and cleaning the heat sink would be best, but switching off the computer and blowing into the exhaust vent can help. When I did this a cloud of dust emerged from intake holes and the overheating has not recurred.
• Memory usage. When writing .sps files, buffers are allocated to hold the spectra to be recorded per file. In integrate mode, the number of spectra buffered is the integrate record time divided by integration time, and in streaming mode the number is the streaming record times the sample rate divided by the Fourier frame size. This memory usage limits the duration between file writes. Check memory usage through the task manager.
• Observatory infomation must be available when writing .sps files. It is loaded from the file ObservaotoryLast.obs in the program's current directory if it was previously There can be errors when it is not defined, and the diagnostics are not always clear.
• There are two exception logs that record the exception type and a stack trace. They are SDRPlay.log and SDRPlay2RSS.log. They are cryptic, but on occasion they may help identify problems.

If you are new to SDRPlay, some patience may be needed while getting to know the receiver. There is a lot of analog processing happening in the device that can give rise to harmonics, mixing products, and images in crowded frequency spans. It takes time to acclimate to the signs and learn to best manage these issues while preserving sensitivity.

Acknowledgements

Thanks to the people who have provided bug reports.

FFTW is a capable Fourier-transform package used by SDRPlay2RSS.

Nathan Towne
towne56 at ownmail dot net
6/2016