Using the
ZL2AFP PSK Sounder

Installation and Help Information
by
Murray Greenman ZL1BPU
March 2008
Software, Requirements & Installation     Using the Transmitter     Using the Receiver     Tuning Technique    

Introduction

Ionospheric sounding (detecting the presence, properties and location of different refractive layers in the ionosphere) can be achieved using several techniques. The most common are: Chirped sounders are widely used, and they measure propagation over a wide frequency range. Using high power phase modulated pulses, Over the Horizon Radar is also able to measure and account for ionospheric effects. PN-sequence phase modulated data transmission sounders operate only on a single frequency. The ZL2AFP PSK Sounder uses a similar (single tone fixed frequency) approach, which is more suited to low power Amateur operation.

The single tone approach has three principal advantages:

Because the transmissions are continuous, operation of transmitter and receiver at the same site (as is common with pulsed sounders) is not practical.

Cross-Correlation

PSK Sounder uses a modulation technique which is widely used by high speed HF radio modems. Using BPSK modulation, a pseudo-random binary (PN) sequence is used to identify an exact point in the transmission from which the data can be synchronized. A cross-correlator is used in the receiver to locate the one point in the whole message where the sequence matches up with the local copy of the sequence. As opposed to an auto-correlator, which tries to find repeated patterns in an unknown message, the cross-correlator works with a known pattern to look for, and is a very much more powerful tool.

These radio modems use the PN sequence technique to enable complex high speed data to be decoded accurately - the ranging information is in this case not generally used as a tool for propagation study. However, in PSK Sounder the ranging information is primary, and data transmission of secondary importance.

Let's take a simple cross-correlator example. Imagine we send a message 'ABCDthe quick brown fox' (where the 'ABCD' represents the pseudo-random number sequence, and the rest is user data. Lets say the next message is 'ABCD jumps over the lazy dog'. Then, at the receiver, we see: 

ABCDthe quick brown foxABCD jumps over the lazy dog
The receiver can work out where each message starts by comparing the known sequence 'ABCD' with every step in the message, moving along one character at a time. Let's portray a match of all four characters in the example known sequence as a score of 4, a partial match as 3, 2, 1 etc, and no match as '.'. So, we get this result: 
4......................4............................
and so it is very clear where each message starts.

Further, when the message is noisy (there are many errors), you can still detect the message start. For example, if we received: 

AbCDthe swickAbrown foxABcDdeumps oler tge lABy dog
the scores would be: 
3............1.........3....................2......
and provided we know the messages are always the same length, we still won't be confused by partial scores caused by errors in the message. Partial scores such as are illustrated here (caused by errors in the message looking a little like the special sequence) occur much more rarely in real life, because a larger special sequence is used, with a much lower probability of this occuring.

PSK Sounder uses this idea in a binary fashion, using a 31-bit PN sequence with one chance in two billion of a perfect score being caused by noise. It uses 80 symbols (modulation time slots) of this sequence in a 256 symbol frame. We can represent the cross-correlation of the signal over 256 symbols (512 samples) by drawing a graph where horizontal position represents sample time starting at an arbitrary point, and vertical position represents the cross-correlator score measured at that each sampling time:


Graphical representation of a real cross-correlation

Further, the 31-bit PN sequence is sent about two and a half times in each frame, giving a unique pattern with a main peak and two side peaks (see picture above). We have borrowed this feature from the STANAG 4285 modem waveform, which means that we can practice with the software on the many military transmissions we hear on HF.

In comparison with a pulsed sounder (which might send a 500µs pulse at 1kW, the PN-sequence achieves 500µs resolution using a 31-bit sequence spread over 15.5ms. This gives us an immediate power gain of 31 times, implying that the same sensitivity would be achieved with a transmitter power of only 32W. In fact there is further coding gain achieved by this technique (for example PN sequence repetition, rejection of non-correlating signals and noise and frame averaging), which makes it a very sensitive technique.

PSK Sounder (and STANAG 4285) always use messages of the same length, 256 symbols. The cross-correlator allows us to locate each message block (which we call a frame) exactly in time, making decoding so much simpler. By comparing the time of the message block with a clock in the computer, we can also measure changes in the propagation delay of the received signal over time.

As an example, imagine we are receiving a signal from an Amateur Radio station some distance away, which is providing both ground wave and F-layer ionospheric signals - the classic NVIS situation. The ground wave will be delayed by just a millisecond or two (300km per millisecond at the speed of light), while the F-layer signal, travelling a further 300km up to the ionosphere and another 300km back, will be delayed an extra 2ms. Both these signals arrive at the receiver, and the cross-correlator has to work with the combined signal. So, it will in fact find two peaks, about 2ms apart, and we can display or plot this information, and measure the delay. Very often the propagation is even more complex, especially at greater distances, and of course very often the ground wave signal is not received at all.


Multiple cross-correlation responses with multi-path reception

It is important to realize that the timing in the computer is completely arbitrary, and so because it has no knowledge of the time of transmission, the actual path delay is unknown - all we know is the variation in path delay and the difference between various paths. Only if the user can recognise the ground-wave signal as being present, can the actual total path delay of each path be determined.

Message Waveform

The PSK Sounder message frame consists of 256 symbols, of which the first 80 are the preamble. The preamble consists of ~2½ BPSK repeats of the 31-bit PN sequence, one bit per symbol. The remaining 176 symbols contain the 'payload', our ID message. The PN sequence header, modulation and frame length are identical to STANAG 4285, but the payload is different. (The military payload has extra propagation correction bits, error correction and no doubt very strong encryption).


The PSK Sounder / STANAG 4285 message frame

Like the STANAG 4285 system, PSK Sounder can operate at 2400 baud, using a sub-carrier frequency of 1800Hz. However, to fit the signal into a normal amateur transceiver IF, it is usually operated at 2000 baud using a 1500Hz sub-carrier. Think about the waveform a moment - in a 3kHz bandwidth we have 2400 baud data modulated on a 1800Hz sub-carrier. None of the data symbols contain even one complete cycle of the sub-carrier! However, once the signal is on the air, the sub-carrier is translated to a radio frequency and all is well.

Unlike the PN sequence header, the payload uses 8-PSK modulation. This way each symbol can carry one, two or three bits of data, as required. You would expect 8-PSK performance on HF to be inferior to BPSK (2-PSK), but through the use of the correlation technique and an equalizing system, in fact the performance can be very good.

PSK Sounder sends one data bit per symbol, but it still uses the STANAG 4285 scrambler, which makes the data seem more random. This turns the data into three bits, and generates 8-PSK. The scrambler is not used to hide the data (it is NOT an encryption, since the receiver knows the method), but in fact because it has several advantages:

We know from experience that using 2-PSK, or going without the scrambler, gives much worse results.

Payload Message

The PSK Sounder message is very simple. It is an ID message in standard 8-bit ASCII. There are 176 data bits, so the message will accomodate 18 ASCII characters and a CCIR16 CRC (error check). The user can set any suitable message, for example 'ZL1BPU RF72ku 1W', and the CRC is coded automatically. The same frame of data is repeated over and over, at about eight or nine frames per second, depending on the baud rate.

To receive the ID text message data, the receiver uses clever (and processor intensive) software called an Equalizer, which takes all the time and phase information from the header, and uses it to predict the timing and phase of the rest of the message. This technique uses a least mean squares extrapolation technique which learns as it goes, correcting ionospheric, transmitter and receiver errors on the fly. It is a type of Kalmann filter, and the user can adjust the power of the equalizer to suit the available processing performance. Since we don't use the STANAG 4285 probe bits in the payload, with a shortish Equalizer the timing and phase correction is sometimes best closest to the header, at the start of the payload.

Because, as just stated, if the Equalizer has a weak setting, the accuracy of decoding is best close to the PN sequence header (timing can deteriorate later as the extrapolation fails to correct later changes), the first few characters of the message tend to decode more reliably at the receiver, and so the first few characters should contain the most important information - the user callsign.

You can't use the system to send free-form text or make a QSO. It just sends the one message over and over again. The data rate is quite high (equivalent to 1375 bits/sec, 171 characters/sec or about 1700 WPM!) Of course as a message ID system this is a crazy speed, so the receiver makes use of the multiple repeats of the message to be sure it has the message correct.

At the receiver, the ID message is treated as 176 separate bits, and they can be averaged individually over 10 or 100 frames. Only then is the result assessed as ASCII data - any bit averaging less than 0.5 over 10 or 100 frames must be a '0', and any 0.5 or more must be a 1. While not as good as true forward error correction, this technique is good at eliminating noise, and also (over 100 frames or about 10 seconds) rides through fades quite well. When the CRC is enabled, the reconstructed message is not displayed until the received CRC matches a version calculated from the decoded message. The likelihood of spurious reception is very low. One nice feature of this option is that if you are monitoring for a very weak signal, just set CRC on, and even if the ID decodes only once correctly in an hour or more, it will remain displayed. You only need one good reconstructed message to be sure that the station is identified.

We know from experiments that you can decode the ID message reliably with 10 second averaging at -10dB S/N, and the correlation peaks can just be detected at -20dB S/N. The reception limit is not in any case the lack of error correction, or weakness in averaging, it's simply the ability to reliably locate the correlation peak. The bigger problem is locating the weak signal in the first place!

Doppler Measurement

Once the correlation peak is located, the receiving software knows where the message frame starts, and so can measure the phase of the sample at the start of the header. It can also then move along the header and measure the phase difference from this reference at each sample. The software knows what the phase should be because it knows the PN sequence. The difference in average phase from the start of the header to some point later in the header is a direct measurement of sub-carrier frequency. If there is no phase shift between samples, the receiving software oscillator (NCO) must tuned exactly to the audio frequency (1500 or 1800Hz) of the received signal. The measurement range is about ±32Hz (38Hz for 2400 baud), and beyond that the phase becomes ambiguous, as one or more complete phase rotations occur between measurements.

The receiving software can measure Doppler over four, eight, 16 or 32 symbols. The fewer symbols used, the more responsive the Doppler measurement, but the more noise influences the measurement.

The Doppler measurement could be made independently on every received header bit, so in theory it is possible to portray the Doppler shift for every step in the cross-correlation, and even display propagation delay versus Doppler shift, as was done in Peter G3PLX's very clever SBSTANAG software. However, we make only average measurements, which gives the most sensitive results.

The Doppler measurement display makes a handy tuning indicator, and also shows how the strongest propagation path is changing (it may not be the only path). Layers in the ionosphere change density and charge over the course of the day, changing the refractive index of the air, and so the apparent point of reflection moves up and down several hundred km from day to night, and while we can measure the height, the actual movement causes the Doppler shift we see.

The Doppler measurement also includes receiver and transmitter drift and frequency offset, so unless both are extraordinarily stable, you may not be sure which effect you are seeing. Good results can be achieved with a modern amateur transceiver with 10Hz stepped synthesis, and even better if the transceiver has the high stability option. Most of the military transmitters have very high stability and are carefully adjusted to frequency. In our experience, almost all these transmissions sit on an exact 100Hz frequency multiple, to well within 1Hz. The STANAG 4285 specification calls for only 10ppm accuracy.

PSK Sounder Software

PSK Sounder consists of two separate programs, one for transmit (TX_data_CRC.exe), the other for receive (PSK_sounder.exe). By using both of the programs at the same time, you can test the performance and become familiar with the receiver controls, all inside the same computer. Normally, however, a user would run only one program or the other, one station acting as transmitting station the other (or others) receiving.

Another small program (PTT toggle.EXE) is provided in order to control the transmitter where VOX is not available. It uses the same PTT control setup as in most Amateur sound card programs. All you need to operate the programs is the usual digital mode setup - interface and cables - and the same sound card settings.

The sounder programs share a common setup file (STANAG.SET). This is a text file which contains the operating sampling rate (baud rate x 4), the sub-carrier frequency in Hz, the user's ID message, the equalizer length, the display speed, and finally the sampling rate correction value (which is only used by the receiver).

Editing Setup
At present there is no way to change the operating baud rate or display speed except by manually editing the STANAG.SET file, and then restarting the program. To edit the setup, you must use the Windows™ Notepad applet. Right-click on the STANAG.SET file in Explorer, and select 'Open With...'. This brings up a dialog (see example on right), where you can scroll down and select Notepad. You can safely check the 'Always use this program to open this kind of file' tick box, so that next time you want to edit the STANAG.SET file, you only need double-click on it.

You must not use a word processor or other type of editor to change the setup, as they will most likely add unwanted information to the file or alter the way it is saved, which would confuse the programs.

The default contents of the setup file is as follows: 

9600                  Soundcard sample rate          (8000 for Amateur use)
1800                  TX & RX sub-carrier frequency  (1500 for Amateur use)
NOCALL Nowhere        Beacon ID message              (no commas and max 18 characters)
10                    Kalman filter length           (multipath spread in symbols)
10                    Correlogram time scale factor  (slows display down if >1)
.0000052              Soundcard samplerate offset    (do not change this manually)
Before using the transmit program, you should set the user beacon message. Otherwise, you need not edit anything to get started. Later in this help information reference will be made to changing various values here.

Sampling rate coarse adjustment is made by setting the first value in the STANAG.SET file up or down slightly from the nominal 9600 or 8000. Since you can't tell by receiving your own signal, set this by checking out a range of military transmissions, which will all be very accurate. If you find you consistently need 100ppm or more Clock correction in the receiver (see later), try tweaking the sampling rate. When using 2000 baud, there's no simple way to calibrate, so just make sure your transmissions are consistent with those of others. It may or may not be appropriate to scale the 8000 value by the same amount used for 9600!

If you wish to use both military and amateur speeds frequently, install a version of the program in two different folders, with different settings in each STANAG.SET file.

By the way, you can run two instances of the receiver, or the receiver and transmitter at the same time, but your computer will likely be confused if you have the programs set for different sampling rates. It can be useful to monitor signals using two instances of the receiver running at different display speeds.

Installation
Installation is simple - just unzip all the files into a new folder, unzip this help information into the same folder, edit the STANAG.SET file if necessary, and make shortcuts to the programs. Then carefully read this help information! The help can be called up from the receiver program.

Computer and Hardware Requirements

  • A Pentium 750MHz computer or better, with Windows 98, 2000 or XP software. For Equalizer settings greater than 10, use a 2GHz computer. One serial port (for transmit control), a good quality 16 bit or better sound card or system.
  • An HF receiver or transceiver with 10Hz step synthesis or better, and high stability. Fine tune to 1Hz (e.g. using RIT) is required. For 2000 baud, use a conventional 2.4kHz SSB filter. For 2400 baud, use a 3.3kHz filter. Use USB, or (amateur only) LSB below 9MHz. Set AGC to fast.
  • Radio sound interface with isolation transformers and isolated PTT control.

Operating the PSK Sounder Transmitter

When you first start the program, it brings up a small box showing the ID message and its CRC. Clock OK and the program itself will start.

There is only one control on the transmitter (see picture). This turns on and off the sound from the sound card. If you use VOX, when the sound starts it will also turn the transmitter on and off. If no VOX is available, you should also run the PTT toggle.exe program, set the correct COM port for control, turn the sound on and then use the TX and RX buttons on the PTT toggle program to control the rig.

Use the sound card Volume Control applet in order to set the transmitter power level, using a combination of the WAVE level and the main volume level.

While you can probably operate at up to about 50% of the rated transmitter power, not only is high power unnecessary, but transmitting for long periods will be very hard on the equipment. It has been found that effective sounding can be achieved with 1W or less, while 20W works well on 80m under noisy conditions. Since the transmission is single tone, and phase change keying is limited by the transmitter bandwidth, it will not matter if the transmitter is driven hard (harder than would be usual for PSK31, for example). Listen to the sound of the signal on another receiver, to make sure that it sounds clean and steady.

While you can use the 2400 baud option with a ham rig, some of the signal is lost in the filters, and performance will be superior in most respects (sensitivity, correct ID decoding, constellation diagram) if 2000 baud is used. Remember to change the sub-carrier frequency as well as the sample rate. The preferred values are:

Mode Baud Rate Sample Rate Sub-carrier
Amateur 2000 8000 1500
Military 2400 9600 1800

Unfortunately the program structure does not permit changing these values on-the-fly. You must close and restart the program after editing the setup file. Amateurs use 2000 baud, 1500Hz sub-carrier, and you should operate your correctly calibrated SSB transceiver with the (suppressed carrier) frequency display on an exact 1kHz or at least 100Hz step (which makes tuning easier for other stations).

Before operating, make sure you set the ID message in the STANAG.SET file! The maximum length is 18 characters, and any further beyond 18 will be ignored. The message must not include commas (,). Make sure you put the callsign first, then the locator, then anything else which might fit. It makes no difference to timing, transmitter power or reception whether you use all the available space for characters or not. The message will be padded out with blanks anyway. A checksum will be added to the end of the message.

Since not many amateurs (especially those not digitally aware!) will recognise the signal or be able to identify it, it will be useful to switch off every 15 minutes or so in order to make a voice identification and listen on the frequency. The transmissions are reasonably immune to most forms of intereference and also to man-made noise and static, although these will of course affect the sensitivity of the receiver, acting on the AGC and thereby attenuating the sounder signal.

An interesting point is that if you have the transmitter signal going into your transmitter at a low level, as well as your voice at normal level, the other stations can measure propagation, determine your ID, and listen your voice at the same time!

You must check your local regulations and any operating
conventions before running this transmitter software.

The ITU mode definition is 3K00G2B (military, 2400 baud) or 2K40G2B (2000 baud Amateur).

Some countries have Amateur restrictions on baud rate, either on certain bands, or within certain segments. This transmitter operates at 2000 baud. Other countries may not permit in-mode automatic ID (as in this program) without other manual voice or Morse means of identification. The program has no built-in ID method other than the ID in the message. The mode transmitted does not include encryption for the purpose of hiding meaning or content, and text is transmitted uncompressed in plain language.

Operating the PSK Sounder Receiver

The receiver is the heart of the system, and by far the most complex part of it. You need to completely understand the characteristics of the STANAG 4285 waveform, and the PN-sequence timing technique before you start. Read this help file carefully.

There are several different areas of the receiving screen that need to be considered and understood, in order to make the most of the sounder. There are five graphical displays:

  1. The Correlation Trace
  2. The Dopplergram
  3. The Correlogram
  4. The Constellation Diagram
  5. The Tuning Trace
Each of these traces (except the Constellation Diagram, top right) is labelled.

Correlation Trace
At the bottom of the receiver screen is a blue flickering trace, rather like an oscilloscope trace (see examples below). This is a line representing all 512 measurement points in the cross-correlation between the message frames and the known PN sequence. The left hand side of the trace represents an arbitrary point in time at which the PC software starts sampling the frames, and does not represent any particular signal delay. In fact, if you stop and restart the program, or the transmitting station stops and restarts, the arbitrary point will be different.

When the receiver is correctly tuned, three spikes, one taller, with two smaller ones on either side, will be prominent on this display, although the location of these spikes will be random. You need to tune VERY slowly, in the smallest available steps. Once tuned correctly (spike height maximized, Doppler centred, see later), adjust the gain controls again so the biggest spike fits in about 75% of the available height for this trace. When correctly adjusted (on a good signal) you should achieve a trace something like this:


Correctly tuned signal (horizontal position of peaks is arbitrary)

Notice how the noise between the spikes is less than it is elsewhere. When the signal is more noisy, the peaks and quiet patches will be less easy to discern. When propagation is affected by multi-path, you see a trace more like this:


Multiple cross-correlation responses with multi-path reception

In order to practice tuning, try some of these frequencies (heard in South Pacific):

4271.0         8625.0         12666.5 kHz
4279.9         8632.6         12708.1
4280.6         8098.2         12727.6
4401.6         8646.0         12728.6
4576.2         8650.0         13001.2
9912.7 (freq approximate - this one drifts!)
You are sure to find others in your area on nearby frequencies. They are most common around 4MHz, 8MHz and 12MHz.

The frequencies quoted are 'dial frequency' for USB mode on a conventional SSB transceiver (where the display frequency represents the zero beat suppressed carrier), and are accurate to better than 10Hz. The true centre frequency for STANAG 4285 transmissions will be 1800Hz higher than the dial reading.

Dopplergram
During each frame, the phase difference between two symbols in the header is measured. Since the time difference is known (number of symbols) the measurement represents frequency difference. A single value is measured at every frame (the result is seen in the Tuning Trace, see below). The Dopplergram (labelled 'Doppler', circled in red on the picture) is a continuing record of these Doppler frequency measurements, in the manner of a grey-scale distribution graph against time. Because each of the propagation paths received can have different movement, the Dopplergram represents all of these within the limitation that the frequency measurements represent that of the strongest path at the time.

Although it shows the relatively long term changes in the ionosphere, the Dopplergram will also show any drift in the transmitter or receiver frequency. The Dopplergram and the Correlogram (see next) move to the left at the same speed, so corresponding changes can be seen. The speed is set in the setup file. The number represents how many frames are averaged for each update of the screen displays.

The Correlogram
Imagine the spikes of the Correlation Trace rotated vertically, so elapsed time is upwards. Then imagine the height of each point plotted as brightness on a waterfall display. This is in effect what the Correlogram shows (circled in red on the picture). The display is slowed down by a time scale factor (anywhere between 1 and 500 or so), so that only one in (time scale factor) frames is a new line of points plotted on the waterfall. However, all the data between these events is averaged, and so nothing is lost and the waterfall display (if water falls sideways!) shows a detailed grey-scale version of the time-varying effects of the ionosphere.

The Correlogram is a magnified view of the signal, and can only show about 10% of the total data frame period. It is arranged to show the important area around the main Correlation Trace spike, through the simple expedient of clicking on this spike with the mouse. (In the picture, the part of the Correlation Trace depicted in the Correlogram is marked in red, and the arrow shows where the mouse was clicked).

The vertical range of the display is 20ms (delay increases from bottom to top), and a few minutes to many hours elapsed time horizontally. Spike strength is portrayed as blackness, and so multiple paths, with different delays and multiple different main spikes, cause multiple responses on the display. At the fastest speed, changes that occur in real time (second by second) can easily be observed, while with very slow speeds (say time scale factor > 50) allow longer term effects (such as onset of F-layer at sunset) to be viewed.


Correlogram from 4.8MHz around sunset at ~3000km range
(Time scale factor set to 250)

The vertical position of the trace (i.e. the apparent delay) is completely arbitrary, as it depends only on where you clicked in the Correlation Trace.
Constellation Diagram
This diagram is two constellation plots in one (red circle, on right). As each frame is received, the phase of all the symbols is measured and stored. Once the equalizer has run, the phase of the header and payload symbols will have been corrected. The header phase data is then plotted in yellow using a relative vector length of half the payload phase data. The payload data is then plotted in white.

When reception is very good (ionosphere very stable) the Constellation Diagram can be very sharp, and the two header and eight payload phase positions are clearly seen. As conditions deteriorate, the ability of the equalizer to correct the phase and timing is reduced, and the plotted points become more fuzzy. The yellow header dots remain grouped together better since the equalizer already knows what to expect and can correct well. As extrapolation through the payload proceeds, the ability to correct for ionospheric errors is reduced and the white dots become more fuzzy.

The Constellation Diagram is for interest only. It is of no practical use, although it helps identify STANAG 4285 compatible signals and gives some general idea of propagation conditions.

Tuning Trace
Labelled 'Tune', the small red trace to the right of the Dopplergram (circled in red) shows the instantaneous frequency changes of the signal. It is used to provide exact tuning. Simply adjust the receiver to centre the trace.

When off tune, it will be quite difficult to locate the tuning point, and the Tuning Trace will be very noisy. It becomes much cleaner when near the correct tuning point. Centering the trace will also maximize the spikes on the Correlation Trace.

Minor Functions
The MIXER button (top left) activates the Windows™ Recording (sound input or Mixer) control so you can set the correct sound input and level.

The ? (HELP) button (top left) activates this help information if the help archive has been installed in the same folder as the program.

The Station ID window (top centre) shows the ID (if any) of the received station, if DFE (Decision Feedback Equalizer) is on. When DFE is off, you can type your own message in this box, for example to identify station, frequency and date before taking a screen shot.

The DFE button to the right of the Station ID window enables or disables the Decision Feedback Equalizer. This corrects timing and phase based on the timing and phase trends measured in each frame header. Since it is processor intensive, it is best turned off when not used. When it is turned off, recognition of Station ID is also off, and you can type in the Station ID window - otherwise the Station ID is printed once it has been recognised.

The next button, labelled CRC wii cause the ID message cyclic redundancy check to be checked on incoming messages. With CRC on, only valid messages (CRC of data matches transmitted CRC) will be displayed. As the button is pressed, any previous message will be cleared. Further, nothing will be displayed until a valid message is received, and this valid message will stay displayed until another valid ID is decoded.

When CRC is off, the messages are always displayed (filtered as selected), and the CRC is also displayed at the end of the message.

The Data bit average box (top right) provides three radio buttons, to turn off averaging, average over 10 frames, or 100 frames. The decoding of the Station ID is slower with more averaging, but more secure as it is filtered over more frames. These buttons have no effect if DFE is off.

Under the Constellation Diagram (top right) is a pair of up/down buttons, the Clock offset control. Beside them the current clock offset is indicated. The sound card sampling rate is defined in the setup file, but since there is usually a moderate frequency error in the computer hardware clock, this software adjustment is provided. It does not change the sampling rate - it performs a mathematical process on the sound card data, interpolating or decimating to add or remove samples to give the correct rate. If the Correlogram trace has an upward trend, press the lower button, and vice versa.

The cyan coloured box under the Clock offset indicates the Delay between the currently point on the Correlogram (as indicated by the mouse pointer) and where the mouse was last clicked. To measure a delay, click on one Correlogram feature, move to another feature, and read the time difference in milliseconds.

The Doppler bits box (bottom right) gives a choice of four symbol spans over which the Dopplergram data can be calculated. The lowest number (4) gives the fastest, most detailed response, since the Doppler is measured over only 1.6ms, but the result can be noisy. The highest number (32) gives a longer average slower response with less noise. The default setting is 4. Choose the setting for each station which gives the best looking Dopplergram.

The Gain control (bottom right) can add further gain to the receiver signal input. This is in addition to the normal sound card controls, and is a handy adjustment for weak and fading signals. Make sure the main spike of the Correlation Trace (just to its left) stays within the available display range.

Tuning the Receiver

To tune in a STANAG 4285 or Amateur PSK Sounder signal, set the receiver bandwidth to maximum (or 3kHz if you have a choice). Select upper sideband. Use lower sideband below 9MHz for Amateur transmissions.
  1. Tune the receiver so that the signal is centred around 1500Hz (Amateur) or 1800Hz (military). This is rather a tall order, but you eventually get used to the sound. You can also use a spectrogram program (such as Spectran in a wide setting) to approximately centre the signal.
  2. Start the BPSK Sounder receiving program. Use the sound card Record applet (may be called Wave In) to select the correct input and set the sound card level. The applet can be called up using the 'Mixer' button at the top of the receiving software. You should see 'grass' on the blue Correlation Trace at the bottom of the display.
  3. Next, very slowly and carefully, tune the receiver back and forth until spikes jump up out of the grass on the Correlation Trace. You will need to tune in 10Hz steps or less. Adjust the tuning (fine tune, clarifier, or RIT if necessary) to maximize the height of the main spike. Ideally, your receiver should have a tuning resolution of 1Hz. Unless the signal is local, or is received via a single path, the spikes may not be very distinct.
  4. Watching the Tuning Trace, adjust the tuning to centre the trace at zero. The Tuning Trace has a range of ±15Hz, and it is quite easy to tune in the wrong place, 30-odd Hz to one side or another, where the spikes will be less distinct. Using the Tuning Trace rather than the Correlation Trace may be easier, as there can be considerable processing delay in the Correlation Trace, which will cause confusion unless you tune extremely slowly. When the tuning is adjusted correctly, the Tuning Trace will show a steady (if noisy) line, rather than a broad band of noise.
  5. Adjust the height of the spikes on the Correlation Trace to about 75% of the available height, using the receiver or sound card Record applet controls.
  6. Now click the mouse on the tallest spike in the blue Correlation Trace display. This will choose an arbitrary time delay which becomes the centre of the Correlogram. Now the Correlogram will start to display the propagation responses around this time, with the main spike details centred on the waterfall. You can measure the time difference between individual responses, by clicking the mouse on one response, then reading the cursor value shown in the cyan 'Delay =' box (centre right of the screen) while hovering the mouse over a later response.
Hints:
  • If you are within ground wave range of the station, the first trace (earliest in time) will show no time variation (although it may not be the strongest, by any means), and you should use that as the basis for measurements. Very often you will see a trace that appears to be ground wave, because it is steady - don't be fooled - this could well be E-layer response, and checking over a long time period will confirm this (it may fade or have slight changes in delay). See the example below.


    Sunset correlogram on 4588kHz. Note the lowest trace is E-layer,
    and there are two close F1 traces (O and E refractions).

    At this distance from the station (probably >3000km), the angle of entry and exit into the ionosphere is very low, and F-layer penetration (and delay) is small.
  • The header portion of the signal is sideband independent (like PSK31), and so you can receive a transmission on either sideband. The Correlation will happen, and the program will operate correctly, except that on an Amateur transmission the ID message will not decode to a sensible message.
  • It's best to write down the receiver tuning to the maximum available number of digits, so you can easily return to the station later.
  • Once the Correlogram trace is running, you will likely see that the trace has a slope on it. If the trace slopes upwards, click DOWN (the lower button) on the Clock offset control (right, below Constellation diagram) until it runs straight. If it slopes downwards, click UP (the upper button). You may need to re-click on the Correlation Trace main spike to recenter the Correlogram.
  • As you slow down the speed (increasing time scale factor) you will probably have to fine tune the Clock offset again. At present you need to stop the program and edit the time scale factor in the setup file to change speed. The fine tune Clock offset value is stored and reused next time you start the program. For best stability, leave the PC running for some time before starting a recording. The resolution of the Clock offset adjustment is sufficient to give straight traces for several hours if used with care.

ID Message

If the signal you are receiving is an Amateur transmission, you should be able to identify the station by the 'Station ID' message which appears at the top of the screen. If the signal is strong and the propagation good (clean Constellation) the message should decode with no averaging.

To switch on ID decoding, press the 'DFE' button (top right) to turn on the Decision Feedback Equalizer and ID decoder. If there is an ID message, you should see it. If it is stable but reads rubbish, you may be on the wrong sideband. If the message is erratic or has errors, use a longer averaging setting.

The Decision Feedback Equalizer is quite processor intensive. If the program freezes or misbehaves when DFE is turned on, the computer may be too slow. Try increasing the time scale factor and try again. If it still misbehaves, reduce the Equalizer length (a value of 10 works on a 750MHz Pentium). Windows 98, 2000 and XP should be OK.

Even quite weak or noisy signals can be decoded correctly, because you in fact never need to receive a complete message - the averaging builds up each bit of information independently, and the displayed message may eventually show a valid ID from the averaged information. If you turn on the CRC, the message will only display when it is known to be correct, and will stay on screen until a new signal is received or the CRC button is pressed again.

If the signal is extremely weak, it will be difficult for the software to locate the correlation reliably (even if you can see it on the waterfall, and the Doppler is stable), and so no amount of averaging will recover the message. (If you can't even get a correlation trace on a moderate to strong signal, you probably have the wrong sampling rate or sub-carrier setting).

Hints:

  • When receiving military signals, leave DFE off. The program no longer reads and displays the Station ID message, and the box in which it is displayed can be used to type your own ID message.
  • The receiving screen, with Station ID or hand-typed ID message, can be copied to the clip-board using ALT + Print Scrn, and then pasted into a graphics program to be saved as a file.
  • If you can achieve a spike on the Correlation trace, or see a trace on the Correlogram waterfall, you can be sure you have a real signal. No other signals have this characteristic. Conversely, some signals which sound similar, do not show a correlation spike or meaningful tuning.
When you tune around the HF bands, you'll find many similar sounding signals. Most of them are STANAG 4285 (or we hope soon, Amateur sounders), but there are a few with 2400 baud BPSK, which do not have the unique combination of the same PN sequence in the header, and the same frame size, and the same scrambler. We have no ideas what there are!

Military Transmissions

The STANAG 4285 waveform on which this program is based is widely used by military, government and non-government organizations. You must check your local regulations before running the receiver software on non-Amateur signals. In your country it may not be permitted to receive signals that are non-Amateur or not public broadcast signals.

The software is not equipped to demodulate or decode any data from non-Amateur transmissions. All it does is recognise the cross-correlation sequence, frame size, modulation technique and scrambler, all public domain information from the STANAG 4285 documentation. It is not possible to read any traffic from these non-Amateur signals.

Copyright © C. Wassilief and M. Greenman 2008. All rights reserved.