Linrad

Linux radio is an ongoing software development by Leif Asbrink, SM5BSZ, that makes clever use of the processing power of a PC,
combined with a soundcard, to visualize a portion of the RF-spectrum and to manipulate its signal content to the profit of the user.

Here, Linrad is running on a secondhand 1.8GHz desktop with the free Linux Mandriva 2006 operating system. (EOY2006)
Midwhile, Linrad is running on an assembled Dual Core AMD processor board (EOY2008)on free Opensource Ubuntu releases.

A high performance, M-audio Delta 44, multi I/O-soundcard (4-in/4-out) forms the heart of the DSP-machine.
The hardware, to convert the RF-channels to the audio spectrum, consists of the WSE-modules.
These modules consist of multiple dual channel receivers to move down from 2m or several hf bands to 70/10.7/2.5 MHz modules to feed the audio board.

Linrad can also be used together with other existing SDR hardware such as the Time Machine, the Softrocks or SDR-14 transceiver. (EOY2006)

A typical Linrad-screenshot showing the abundant information available on a 800*600 screen resolution is shown, hereafter.

It was taken on 20/10/2006, around 20h45 local time on the low end (CW-portion) of the 40m band.
We can see multiple stations calling ST2BF on 7004.89 kHz, in a pile-up higher up in frequency, around 7005.52kHz.
At least nine stations are calling the rare station from Sudan.

From the top left to the bottom, the following large subscreens are present:
-the main waterfall
-the main spectrum (with blue horizontal lines)
-the high resolution screen (with red horizontal lines)
-the baseband waterfall (minimized so hardly visible)
-the baseband screen (with green horizontal lines)

A great amount of horizontal information around these screens are frequency information.
A majority of the vertical figures are level scaling and timing information.
The multiple arrows allow to adapt the look of the different screens to the wish of the user.

Linrad allows the user to tailor not only the screens but also the many settable parameters.
Beyond the on-screen parameters, the Linrad-toolbox excels by an extra set of parameters (on accessible consecutive windows)
that allow for more specific usage of the software.
They go along with the mode of operation, the hardware and the goals of the user..

Back to the 40m screenshot:
Two antennas were used (one feeding each channel); an attic mounted non-resonant 30m dipole and a short piece of wire out of
the Velux window.

The Sudan station is very weak. Its signals are identified as a weak peak under the grey vertical line in this screen in the high resolution screen.
It now and then pops out of the atmospheric noise, allowing the copy of its call in CW, with report and his tx-ing of "qsx up".

After taking the screenshot below, the S-meter plot (see the subscreen in the top right corner) was calibrated to a HP8640A generator.
The noise floor with connected antenna lies around the S4-level. The ST2 peaks just a couple of dB's out of this background noise.

The S-meter also shows that receiving the signal as left-hand circular gives about 6 dB more compared to right hand circular.
Interestingly enough, the caller on 7005.65 has the opposite rotation.
By clicking on "S" next to the downpointing arrow, the scale can be turned into a dBm or other logaritmic scale.

There is much more to be read from these traces but not all is well understood, today.

Looking to the baseband spectrum we can see yellow dots under the frequency indication.
These dotted lines show the passband plot of the IF-filter which is preferrably sited around the decoded signal. (here, it is just 20Hz wide!)
Pulling these yellow lines apart with your mouse, increases the IF passband bandwith, moving them closer to one another, narrows the IF passband.
And this in software without the annoyance of filter ringing, within limits.
In a classic hf-transceiver this would mean the aquisition of very costly, cascaded CW-filters.
With the arrival of software defined radio's, this is definitely past time!

Today, 7/11/2006, we hooked up, for the first time, the 144MHz WSE setup to the 12 el antenna (2.87 wl) on the roof and this coincided, nicely, with the 2m SAC event
(Scandinavian activity contest).
The cable end that is usually going into the 2m PA was now connected to the shiny, high gain, low noise figure, 144MHz LNA, described elesewhere on
these web pages.

The LNA output went into about 15m of Aircom+ cable and then straight into channel 1 of the 2m software defined receiver system.
By the way, the shack is on the attic and the computer room resides at the other end of the house, one floor lower...

The first picture shows a screenshot that I have long wanted to see and to be shown...
It represents what I hear on the lower portion of the 144MHz band, centered around 144.050MHz +/- 45kHz.

The signal is centered on SM7UYS (see the high resolution screen and baseband screen) located in JO65MN and he's very easy copy here. (on 144.056MHz)
The IF-filter is about 800Hz wide.

But more important in the screenshot (see the waterfall) are all the successive carriers between the bottom end of the band and about 144.050MHz.
A pretty disappointing sight at this QTH of too many, potential QRM-carriers.

Nowadays and luckily, the 2m band activity is very rare on this (lower) portion of the VHF-band but it certainly shows the mess, I've been struggling with...and what
has made me to abandon all EME activities at my parent's QTH.
There, I've once achieved 47 QSO's in one ARRL EME contest but if I tell you that I usually missed KB8RQ (at that time always to be found on 144.018) due to
a strong TV-carrier product on that frequency, working more than 50 random stations in CW would have been possible!
Of course, that offending carrier was never alone on the band...There were always plenty of such birdies around.

That scandinavian activity night, we also heard plenty of DL, PA0, OZ and SM and G's with the beam in northeastern heading both in CW but many more in SSB.
The SSB activity is visible in the picture below:

The SSB activity is visible in the picture above:

This time, the screen is centered higher up in the band up and around 144.350MHz +/-45kHz. (90kHz bandwidth)
This portion of the band is certainly less populated with unwanted carriers...until the inverter of the photovoltaic cells wakes up...
The waterfall shows the local time of recording and the typical waveform trace of SSB signals on a SDR-receiver.
It is amazing what can be heard with as little that can be seen on the waterfall screen...!

On the other extreme, some signals are so loud that they get out of the actual scale settings. (see the main screen)

Being so far with the setup, I couldn't resist to hook up the WSE-setup again to check, what could be copied with my small 2.87wl antenna,
pointing to the horizon, during the 2nd part of the ARRL EME Contest on 11/11/2006. (always on my moonrise)
Unfortunately, I could not transmit nor could I stay very long in the shack.
The most important screens to look at are: the main waterfall, the main and high resolution spectrum.

The screenshot, hereabove, pictures the following 2m EME CW stations:
144.045.6: IK3MAC (that evening, Grazziano was the loudest signal heard)
144.046.8: SV1BTR
144.049.3: OK1MS (Stan was calling CQ in the same fashion as when I first worked him 15 years ago- I was listening to Stan while capturing the screen)
The baseband (IF) filter width amounts to about 15Hz.

Above and on 144.058, a clear frequency, Marko, LZ2US is solely calling CQ with a good signal level of 429.

On 144048.7, RA3AQ is calling CQ. It is the first time, I've heard this station from Russia.
Some other signals are visible but were not clicked on.

The 144MHz beacon band with the antenna beaming East (19-11-2006)
The picture was taken around sunset meaning that the photo voltaic cells (producing electricity in our sun house) were inactive.

The above picture has an improved screen resolution of 1024 x 768 because today, I finally mastered the 800x600 monitor problem on Linux with my screen!
The 90kHz occupy nearly the full screen width.

The main waterfall shows the strong, local beacons in ON, F and DL between 144.415 and 144.428MHz. The widest signal on 144.018 is ON0VHF
(in white, see the high resolution graph for colour comparison). It is actually 1kHz higher on the Linrad screen than in the beacon band frequency list.
The following beacons were copied from further away: (the locator here is JO20FV)
There is a small frequency offset between the WSE indicated frequency and true frequency.

144.437 DA5DUB in JO40AQ (for some reason this beacon transmits call +locator and 10u ERP instead of 10W ERP...)
144.445 DB0KI in JO50WC
144.476 DB0SGA in JN69KA

The main waterfall shows some interruptions during which some tests were performed in the antenna line.
Remember that a SDR like this can double as in-band spectrum analyzer avoiding very costly measuring equipment.

As can be seen, the birdies are still present but can originate from different sources...

The following picture shows the high CW activity on the 40m band during the above event. The CW activity expands way above the usual first 40kHz.
The periodic CQ calls from the memory keyers can easily be seen. (the short and repetitive vertical lines in the main waterfall graph)
Sometimes, more than 4 stations can be found packed on one single kHz of bandwidth!!
Most FFT1-settings are far from optimized, not to speak about the odd settings of the waterfall (average and gain) resulting in a nice art picture... ;-)

The following day, the parameters were corrected but unfortunately also activity decreased...
The noise blanker settings of the high resolution screen are not optimum and should be lowered some 25dB.

The Dubus CW EME activity day, 02-12-2006

Again very little hobby "playing time"... "Sinterklaas" passed by at my parent's in law QTH and some 10 young children received gifts from that holy man with his black helpers...
A yearly event not to be missed...

I took the following screenshot on that windy, rainy Saturday afternoon, some 20 minutes into my Northeastern moonrise. Remember, this is a relatively quiet QTF at this QTH.
(some 8km NE of Brussels)

The picture shows the center of activity to be between 144.045 and 144.050. The antenna is still the "terrestrial" 12el M².

The following stations were rather easily identified and subsequently heard: SV1BTR, SP7DCS, OK1MS and LZ2US.
There were some pile-ups audible on the first two stations but unfortunately, I did not record the raw data for later investigation. (Linrad has a built-in recorder option)
I did, however, record LZ2US's CQ with long "tune", here:

. Marko's signals grow stronger and stronger with time...
I hope next perigee to collect some SWL-pictures of the WSJT-segment...

LINRAD calibration part 1, 06-02-2008 (in blue some helpful statements of typical settings)

Looking at the date of my last Linrad report, it seems like an eternity since I've last played with the marvellous operator assistance tool. (Linrad)

The Linrad system here, early 2008, is composed of the following: Ubuntu 7.10 + OSS v4 1012 with Leif's latest 2.34 on the same secondhand PC. (P4 @ 1.8GHz)
The soundcard remains the high-end M-audio Delta 44 at an input sampling rate of 96ksps, the output is provided by the onboard soundcard which
only works at a sampling rate of 48ksps. (which is different with ALSA)

Linrad calibration was the next logical step, after finding out that the Belgian beacon on 144.418, ON0VHF in JO20HP, had just 32dB of mirror rejection.
I do not recall at what distance from the center frequency, the measurement was made...Anyhow, such a low value is not acceptable and would lead to false signals appearing at the mirror frequency.
Eg. SSB signals in the CW portion of the 40m band and vice versa- (a situation to be avoided)

So I hooked up my trustful HP8640A-signal generator with 144 MHz output [@ abt. -50dBm for 55% AD(0) and AD(1)] at exactly the same spot where the antenna cable is entering the shack,
i.e. the input of the main Tx/Rx relay.
Following this coaxial relay, I still have the high performance PA3BIY-LNA feeding the Kathrein bandpass filter A (see here -chapter 8) and subsequently going into the WSE 144MHz input.
As such, the true transfert function could be drawn and compensated by Linrad on the real live antenna line.
As determined elsewhere most of my antenna line components are rather broadband except for the Kathrein bandpass filter, for which the levels fluctuate up and down through the lower VHF band.
(non-flat insertion loss)

I decided to run the calibration routines [in 64 segments] from 144.080^- to 144.160⁺ MHz as shown hereafter... I increased the FFT1 size from my optimum settings to 16384. (because 2048 was too small)

It can be seen that the initial I/Q-rejection to be also about 32dB, the same value as noted some 300kHz⁺ away from the VHF beacon band...

After [a 4-pass] calibration and at roughly the same frequency seperation from the center frequency the following plot was taken:
The amplitude of the belgian beacon is some 45dB louder than previous screenhot by turning the antenna to this beacon.

The mirror rejection, after calibration, now amounts to 100dB. A very respectable value and in line with Leif's findings on his web page. This is a power difference of 10 to the power 10.
(Order of magnitudes that are encountered only, when measuring passive intermodulation on antenna line components in cellular networks..)
Even at the band edge, the rejection was still found to be higher than 70dB! (this is not shown in the above picture)

With the given calibration parameters, a plot was taken higher up in the beacon frequency band, on ON0VHF at 144.418. See the picture below:

The amplitude level of ON0VHF reaches 80dB; the noise floor sits under a level of 20dB.
Many carriers were clicked at or near the mirror position but no audio trace of ON0VHF could be found, meaning that the mirror level of this beacon to be below the noise floor,
corresponding to more than 60dB rejection.
If the beacon signal would have been bigger, a better mirror indication of mirror rejection would have been possible with an external signal.
Also note the increased noise floor in the southern heading (towards this beacon), as compared to the bare noise floor.

As a side note, Linrad recognized the calibration effort by ommitting the statement of "I/Q phase not calibrated". (in the main spectrum = the spectrum with the blue horizontal lines)

The above picture shows that there is still about 4dB (+/-2dB) variation in the 100 kHz+ passband.
The purpose of the second calibration part is to further flatten out these amplitude variations (minimize ripple) in order to have the same (lowest) sensitivity throughout the selected passband.
Moreover, tradeoff's are provided (by the Linrad calibration routines) for the passband edge filter skirts and passband width.

This calibration is also needed to have an effective smart noise blanker. This will be needed since the radio environment is rapidly changing, here:
In the eastern headings, a train corridor is being dug (to quickly travel from Brussels airport to Antwerp). In the western headings, more and more houses continue to grow and there is very high
noise coming up from these headings. The risk for TVI is high in an environment that is every day day looking more urban than suburban...!

Leif, SM5BSZ has adapted the calibration procedures in order to be perfomed with low frequency square waves. I used an old homebrew function generator from the magazine Elektuur (Elektor).
(based on the same Exar IC (XR2206) as I used to have as the 1.5kHz tone gererator to make HSCW with the SSB Tx input on 2m )
The low frequency square waves were injected at the 10.7MHz level of the WSE-converters.

The following pictures are the results of the second calibration attempt and these are still subject for improvement.
You'll note that the text "FILTERS NOT CALIBRATED" is not visible any more in the main spectrum, indicating that Linrad succesfully calibrated the RX-line.

The following is visible (in the weak signal CW mode) in the passband of the main spectrum:
-bandwidth of about 95 kHz
-relative flatness with less than 1dB ripple except at the band edges..

AUTOMATIC SPUR CANCELLATION
Another interesting, recently introduced feature in the Linrad program by SM5BSZ, is visible in the plot below: the automatic spurious cancellation and 26 spurii were treated at the moment of the screenshot...
(see the top right corner of the main waterfall, above the yelllow letter c)

A good view on what the automatic spur cancellation is doing here on my "corrupted" radio spectrum in the 2m beacon band is visible from the following plot: (taken on Saturday afternoon, 23-02-2008)

At the time of the screenshot, 26 spurs are treated (read rejected) by the automatic spur removal feature.
What the feature really is doing is best noticed between 15:12:37 and 15:14:25, where the automatic spur cancellation was temporarily switched off. The spurs appear abruptly...!
Quite amazing sight, isn't it... Well done, Leif!

This is one of the main differentiators with similar SDR-programs at this QTH.

BTW: most of the visible spurs are generated by the DC/AC-inverter which is part of the house's sun electricity production plant based on photo voltaic cells.
The inverter produces a relatively clean 230V ac sine wave from the 100V (and more) dc voltage produced by the photo voltaic cells and this in a very efficient way. (>95%)
Unfortunately, the dc/ac chopper produces a lot of discrete carriers, as well as wide band noise throughout the hf and lower VHF spectrum!

Linrad in uncalibrated SSB mode during the 2m Nordic Activity Contest, 04-03-2008

Again a snapshot in the 2m SSB band portion during yesterday's NAC. The antenna heading was 55°.
Tropo conditions were not so good. During the 20 minutes of listening, I counted a handful of OZ's, just two SM's and good activity from northern and other parts of Germany.

This is a brief breakdown of the identified stations with their frequency:
144.362: DL-station; .344: OZ1BEF JO46, .334: ???, .326: SK7MW JO65, .317: DF8XC JO41, .296: OZ9KY JO45, .282: DL6BCT JO43
No traces of overload were found, not even with nearby belgian high power stations.

Important findings of last week Rx-activities indicate that the SDR setup, based upon the WSE converters (with additional gain/filter/attenutaor blocks) with Linrad to be:
- as sensitive and
- at least as good in strong signal handling as the TS850S+transverter setup!

Next on the agenda will be the calibration of the CW mode...

Linrad in calibrated narrow CW mode during the Dubus EME activity event, 07-03-2008

The Dubus CW EME events are 2m moonbounce activity events that allow the newcomer to test the performance of his Rx-system.
Usually, a good number of CW EME stations are QRV with often some big stations easily copyable by us, single yagi stations. (helped by the selected near perigee activity times)
This event was the first time, the Linrad system was used as Rx and the TS850S+transverter used as Tx to make CW EME QSO's.

This is a screenshot that was taken from the 2m CW bandsegment after the peak of the second ground gain lobe. (>5° moon elevation)

On the main waterfall (on top), you'll easily see the traces of :

Jimmy, SV1BTR at 144.050.7
Chris, SP7DCS at 048.6
Tom, DK3EE at 053.3 (tropo and propably EME with 250Hz doppler shift, hf)

Remarks:
-there is a frequency offset of about 1 kHz between the Linrad -setup and the actual frequency
-SV1BTR topped 16dB over the noise when I qso-ed him prior to the above screenshot was taken.
-SP7DCS's frequency is above the green vertical line in the main spectrum and was easily copied some 2-4dB over the noise
-DK3EE's massive tropo signal also shows some 280Hz higher in frequency, his EME echoes (from his 4x 2M8WL array) are some dB's below Jimmy's moonbounce signals.

Leif, SM5BSZ (the author of Linrad) suggested, based upon the above screenshot, to change the parameter called: FFT2 attenuation N, with unplugged Delta 44 soundcard.
The screenshot hereunder shows the actual situation:

FFT2 attenuation N = 10, corresponding to: an fft2 amplitude margin of 56 (visible after typing 'A')

The Floor value always stayed at 1.09 (for a fixed FFT1 amplitude=1000=default) throughout the next screenshots

The main waterfall shows that the sensitivity varies (colour variation) within the passband and is far from symmetrical. (this is no good)

New value for FFT2 attenuation N = 5, resulting in fft2 amplitude margin = 28

The main waterfall has now become very regular over the 95kHz passband and symmetrical around the center frequency.
Not visible due to the size reduction, the altered waterfall parameters (from 0.5/3 to 3/1) to better visualize the waterfall pixels.
Further on, the parameters were kept constant to 3/1.

New value for FFT2 attenuation N = 3, resulting in fft2 amplitude margin = 15

I do not see much improvement compared to previous value.

Lowest possible value for FFT2 attenuation N = 2, resulting in fft2 amplitude margin = 10

Yesterday, I've had the Linrad system running with the above FFT2 att N-setting (=2) where at a certain moment 15 or so replica of the strongest signal appeared (at 144.418) in the f-domain. (Ough!)
Sorry, I forgot to make a screenshot...

I suspected a hardware problem (oscillation, IMD, etc.) but after reducing the visible passband to about 25 Khz the odd spectrum disappeared. (don't ask me why...)
After re-reading Leif's page: http://www.sm5bsz.com/linuxdsp/install/dlevel.htm , I found out that as low FFT2 att N-settings of 2 and 3 on the full Linrad system to be producing FFT2 amplitude margins
of 0dB, which may have caused the anomaly. So I came up with the following setting:

FFT2 attenuation N = 7, resulting in fft2 amplitude margin ~ 13dB
I overlooked Leif's recommendation for the fft2 ampl margin to be in the range of 5 to 25dB for the complete converter system.

As you can see, the RF-spectrum as visualized by Linrad in the 2m beacon band and at a QTF of 45° is far from clean... (even with 37 birdies removed by the Linrad auto spur removal feature)
Leif, SM5BSZ has implemented on the Linrad cockpit where exactly these birdies have been withdrawn in the passband. (see the, many, tiny vertical lines on the frequency scale)

Early findings indicate the multiple carriers to be picked up by the antenna... more experimenting is needed to see whether any (radiating) source is located indoors...
In some QTF, the dipole is just 2m from the shack PC and Linrad setup!!

The green vertical line is positioned on DB0SI in JO53, a beacon in Northern Germany that is often at the edge of copy at this low elevation QTH, so ideal for monitoring while experimenting with Linrad...

Linrad 02.47 with the default parameter settings in uncalibrated state on 40m CW, 16-06-2008

Linrad together with the WSE converters has often surprised me by the easiness CW or SSB callsigns could be copied and QSO's be heard.
Common examples when listening with the above configuration are exotic callsigns calling CQ with hardly any european feedback.
We assume local QRM/QRN in Europe to be the main reason resulting in the dx-signals sitting in the mud...

Last night, a piece of wire was connected to one end of the indoor 15m wire dipole and feeding the usual Linrad setup through the WSE HF converters.
The following screenshot resulted. The S-meter graph was set in such a way to easily identify the dit's and dah's of the CW-stations on the bottom end of the 40m band.
Given the non-resonant antenna, I cranked up the WSE HF converter gain by 10dB.
On 7027.8, YI9WV was found calling CQ for several minutes, without callers... ;-(
Compared to other screenshots on the page a number of innovations were added, such as indications of the mode, Rx-frequency, frequency of the associated Tx (CAT-controlled), IF-filter bandwidth,
network status, etc.
The calibration files were not yet passed on to the new Linrad 02.47-directory, hence the lack of the smart noise blanker... (FFT2)

WSE and 70MHz, 25/06/2008 (1) A second part follows further below.

On the 25th of June 2008, I was lucky enough to catch my first 4m Es (sporadic E) opening with the SDR-setup.
The setup was very basic and consisting of the WSE modules connected to an attic mounted folded dipole (under the roof) for 70MHz.
The signals were strong enough for the 70MHz WSE converter input to be well visible and easily decodable. (all in SSB)
There was no LNA in the line so the NF is probably around 10dB.

The 4m band is not allowed for ham transmission in Belgium by the local regulator so here follows my SWL-report in a very visual way.

One of the reasons Tx-ing is not allowed on this nice band is no doubt, the still active data signals (from taxi's) on the band, on discrete frequencies such as 70.150 and 70.175MHz.

I read on the web that these data signals are often heard on Es openings throughout Europe... so now, one knows, where they originate from.
They are quite loud here, north of Brussels in JO20FV.
You can easily spot the 12kHz wide data signals on the left hand side of the following screenshot: (taken with an uncalibrated, non-optimised Linrad 02.47, eg. the smart blanker was not operational)

On 70.175MHz, the wide data signals cannot be missed.
A similar data carrier is located 25kHz lower on 70.150MHz. Resulting key clicks from these carriers can be seen up and around 70.2MHz.

The following stations were heard during the above Es-opening and their signals can be seen from the screenshot::
-70.190: I0JX (Tony in JN61GW)
-70.195: IK0SMG
-70.204: ???
-70.208: SV2DCD (he was heard for long periods around 18:20 and back on 19:00 with good signals)

Here is another screenshot from the Greek station showing the very signals in Belgium... (remember my modest, indoor setup)

If you have a nice QTH for EME, I'm happy for you. Here, the situation remains challenging, to say the least...
Luckily, there is Linrad that remains my biggest asset allowing me to copy weak signals with my modest 6m long 12M2 antenna, sometimes very close to the ever-present, annoying birdies....

Hereunder, you'll find a Linrad-screenshot (taken on Saturday, 20-09-2008) showing what stations that were heard/worked on my moonrise...
Traditionally, Graziano IK3MAC went first (loud as usual) and easy in the log, closely followed by Alex, RU1AA.
Remarkable in the QSO with RU1AA was that I received a 549 report, where on the Rx-side, Alex was much weaker than the QSO's in the past.

IK3MAC: 144.045.8 MHz (the strong, white vertical line)
RU1AA: 144.053 MHz (a bit weaker with some red colour) (in an alternating sequence)
G3ZIG: 144.047.3 MHz ( tropo sigs only)
JH0WJF: 144.043 MHz (it is the first time, I heard this station so well)

Since these screenshots were taken after the QSO's with Graziano and Alex, the levels were some dB weaker than during the QSO's...(higher elevation usually means lower signal strengths, remember I have no
elevation)

At the top right hand side of the main waterfall and above the yellow c, you see a number indicating the amount of spurii that were cleaned up from the spectrum, and this automatically without any operator action. (32)
They appear as tiny tick marks between the frequencies on the top waterfall frequency axis.
I do not know of any other weak-signal program providing this very helpful feature.

On Sunday night, before midnight, it was fun seeing G3ZIG calling CQ just above 144.043.3.
His tropo and EME signals were visible for many consecutive periods, seperated by some 100Hz of doppler shift...
On my moonrise (0-20° of moon elevation), the lunar echoes are usually higher in frequency than the tropo signals.
G3ZIG's EME signals are actually stronger and more visible than his tropo signals!
On the screen, you can see 2 one minute sequences of a long row of CQ's. The green vertical line in the wide graph shows the spot where the echoes are located...

This morning, 11-10-2008, the Linrad built-in FM-mode (frequency modulation) was used, for the first time, to receive an active FM-repeater on 145.775MHz. (ON0LB-JO20SR)
This repeater is located in Tongeren, in the Limburg province, in the northeastern part of Belgium, about 80km from this place.
In order to hear some signals with my horizontal yagi antenna, I had to bypass the large bandpass filter after the first LNA.

In Linrad, the FM-mode is selected from the main menu. The standard settings from Leif, SM5BSZ were used. ON0LB's signals were about 30dB above the noise floor.

The signal was audibly affected by fast-fading and this can also be seen on the S-meter graph. (top right side of the Linrad "cockpit")
The baseband filter was set to nearly 6kHz which is a bit on the low side. Nevertheless, the audio was very good.
At least, as judged by someone, who rarely listens to this portion of the 2m band...

Linrad surprises, 25-10-2008 (1)

While playing around with the experimental 02-53 Linrad version, last saturday evening, 25-10-2008, I discovered YL3AG calling CQ on 144.305. The time, I had rewired the Linrad setup to the transmit chain, the station from Latvia had disappeared into the noise.
Some time later, a PA0, an ON-station and several DL-stations were calling him much lower in frequency. (144.286)
The DL station that worked him was located in JO32 and was S9+ here with my antenna towarrds KO26AW. At that time, there was an opening to SM7 and SP, so not so surprising...
You can hear Vitauts 51 signal (using a 14el + LNA on top of a 9 floor high building with 500W output power), here:

(some 4-5dB over the noise).

The best part is at the end of the 13 seconds sequence where he signs "CQ CQ Yankee Lima Three Alpha Golf, over" with no station replying...

The best, I heard YL3AG must have been some 3dB stronger. Not bad for a signal that has travelled some 1439km! The QTH is just 23m asl!
The wav/mp3 file was recovered from a recording made from standard features available in Linrad.
The following screenshot shows YL3AG's signals some 4-5dB over the noise at the exact moment of the wav file recording.
How the weak SSB signal looks like can best be seen in the high resolution graph (with red horizontal lines) and the baseband waterfall and baseband spectrum graph. (with green horizontal lines)

Linrad & the Marconi Memorial november contest 2008

This November contest is the yearly highlight for the enthusistic CW and VHF-addict and also a commemoration of this great Italian experimenter.

Hereafter, follows my SWL-report.
The screenshot was taken around noon time on Sunday 2/11/2008 with exactly the same set-up, as described higher.
The Linrad waterfall starts at 144.055MHz and ends up some 90kHz higher.
Tuning 25kHz lower and higher in frequency learns us that stations were still working below 144.039 and above 144.155MHz!!

So much frequency span on the 2m CW band has become rare. Such a band occupation only happens with large-scale aurora openings, which are uncommon at this moment of the sunspot cycle.
As can be seen, the activity was very high. The 12el antenna is pointed to an azimuth of 95°. (towards OK1)
There were many, loud signals on the band and these were not necessarily originating from local stations!
Some OE and HB9-stations were heard on this heading, some DL's were heard working G's and the OK's from Bohemen were loud as usual.

I'm always glad to hear and work OK1KCR in JN79VS. I operated this clubstation back in 1997 during the September contest... I only speak a handful of Czech words so I operated in CW and enjoyed it very much!
Thanks to Ivo, OK1FGM for being our guide and host in this part of the wooden, hilly Czech republic where at night, one cannot see far...

One station worth mentioning is OK2KYZ in JO80NB, he was peaking nearly 20dB above the noise, while working DL5FDP. Listen for yourself:

This station is at exactly 90° Az and nearly 900km away. Not bad for the small and low antenna at 33m asl!
No, I did not work him. I was called away for dinner and I did not come back afterwards...

As you can see, the recording was made in a baseband bandwidth of 80Hz.
Luckily, the AFC (automatic frequency control) permanently corrects the signal's frequency to keep it aligned inside this narrow bandwith.
Check out the high resolution graph (at the bottom left with the red horizontal lines) to see the offset, between the green vertical line and the long grey vertical line.
A handy feature for the weak-signal operator. (and not only to compensate the doppler shift on EME...)

The uncalibrated WSE SDR-setup passed the test with flavour and the receiver was never overloaded while still very sensitive!

The above screenshot shows a 2 minute recording of the above event which started at 15h UT, whereas my moonrise happened 1 hour earlier...
The level of the signals are somewhat lower than what can be expected from within my first ground gain lobe...
The top waterfall shows the CW traces of the following active stations:

One of the many CQ's from JØrgen, OZ1HNE can be heard here:

. Some QSB is noticeable. He peaked some 12dB above the noise floor. OK1MS is some dB louder and rather consistant.
Also SP7DCS came in fine:

...you can hear him sending many OOO's to IT9CJC, his call and two K's.
Other stations heard were IT9CJC, DL5MAE and G3ZIG.

There is very little relevant info on this screenshot since it just shows the noise floor.

The above screenshot shows a 10 minute recording of the above event. This corresponds with my first ground gain lobe, i.e. close to the maximum possible ground gain and signal levels in contrast
with previous event. Two stations were visible here:

On 144.049,4: OK1MS and 050,2: OZ1HNE

I first answered OK1MS CQ call (at the bottom of the waterfall), but he selected another station so I looked for other CQ's on the 2m band...
A bit higher in frequency, JØrgen, OZ1HNE was heard with nearly the same signal strength as Stan. So I gave JØrgen a call...

The white 1cm wide horizontal bars in the top waterfall is what is left of my Tx signal after the isolation of the main Tx/Rx-relay in the shack.
By playing with the waterfall zero and gain (at the bottom right of the main waterfall) one can set the background such that copyable signals are visible and not the birdies at non useful signal levels,
close to the noise floor...
Compare this waterfall with the waterfall of the 07/02/09-event. Quite a difference, isn't it. The effort to eliminate some shack interference has paid off, too...
At the top right of the same waterfall, the autospur feature has removed 14 stable birdies.

The AFC (automatic frequency correction) at the bottom right was again put to good use, in order to keep the Linrad signal tracker, centered inside the 20Hz IF-filter passband.(horizontal green lines)
This periodical Rx-tuning activity, immediately after a 1 minute Tx-period, is a thing of the past since automated by Leif, SM5BSZ so the CW-operator can concentrate solely at decoding the morse code.

Back to the Dubus event...after a QRZ, OZ1HNE first sent some ?? in front of my call so I sent him both calls and O-report.
This was his reply:

. You should quite easily hear: many ON7EH's, de, followed by some OZ1HNE's, several RO's and a K.
The following period final RRR's were exchanged with the usual 73 and GL (good luck) to complete another CW EME QSO.

I was glad to have worked JØrgen so easily. This time with just one single 2.87wl antenna, after having worked him, more than a decade ago with the large 4x17el at my parent's home.

The above screenshot shows the massive direct and reflected moon signals from Tom, DK3EE while he worked IT9CJC during my first an best groundgain lobe. Tom is using 4x 8WL M² antennas and EME power.
IT9CJC was seen calling Tom and he was easily heard giving RO-reports... (see below)
A QSO with this smaller 4 yagi station would be possible from this QTH, especially when in a sked.

The uncalibrated Linrad S-meter hereabove (useable for relative values, only) indicates DK3EE's direct signals to peak some 40dB (=83-43) over the noise floor on the actual azimuth...(left part of the screenshot)
It also shows (after clicking on them) the lunar echoes peaking some 23dB (=83-60) over my noise floor! (right hand side of the screenshot)
A very loud signal !
The middle part represents my noise floor.

The high resolution screen shows DK3EE's direct signals and his moon reflected echoes to be about 270Hz apart, corresponding to the doppler shift at that specific moment.
The doppler lies between both peaks (left and right) of the above screen. (In Linrad, the green line indicates the selected and it is left of the moon echo)

WSE on 70MHz and Linrad 03.03, 06-06-2009 (2)
The 4m band is not allowed for ham transmission in Belgium by the local regulator. (In 2010 it is...)

This morning, I copied for the first time a 70MHz beacon (by means of sporadic E propagation) some 15dB above the noise with Linrad 03.03.
The setup consisted of a folded 4m dipole and WSE modules on a well-screened attic.
A LNA with some 20dB in-band gain was added and its addition now allows me to copy also the mobile response to the very strong base station data carriers!
This was not possible without this low noise gain block.
The beacon's frequency is 70.166MHz and it is sitting nicely between the two very strong data carriers on 70.150and 70.175 MHz. The activity of these carriers was rather low at the moment
of the screenshot. (see the waterfall below)
As can be seen from the EME-window (bottom left), the beacon is at 222° Az and 1652km from my locator JO20FV.
The CS5BFM in IM59rd A1 transmission can be easily identified:

It transmits with 10W ERP, it is directed 45°/135° and its doublet antenna is just 12m agl.
(Linrad 03.03 was used in a non-optimised SSB mode to tune the CW with a reduced bandwidth baseband filter.)

The calibrated WSE stereo receiver and Linrad 03.07 (10-09-2009)
For some time now, we've been running the WSE modules on both HF, 70MHz and 144MHz with 2 channels, in uncalibrated state.
The picture below shows the Linrad "cockpit".

The main spectrum below the waterfall plot shows the amplitude response in the frequency domain from a fixed level originating from a wideband noise generator.
If we do not consider the "bumps" at passband edges, we observe some 2dB signal variation. The waterfall also reflects this response by a variable green colour flavor.
The non-flat curve is far from optimum for locating weak signals buried in the noise. The acceptable (more-or-less flat) bandwidth for operation is just 75kHz.

So the Elektor pulse generator was dug up and connected to the 10.7 MHz WSE module and calibrated following the procedure outlined on Leif's webpages and
assisted by the "F1" help files that can be found in every Linrad program and pointing to the window/area/setting for which more information is required...

The screenshot below shows the noise floor after calibration.: The wideband noise generator was again connected to the WSE 144MHz setup.

Again, only the waterfall and the main spectrum plots are relevant...

The flat passband widened and smoothened up, remarkably.
From 75kHz, it went up to more than 92kHz!! (the soundcard=the M-Audio Delta 44 at a 96kHz sampling rate)
The amplitude variation reduced from some 2dB to less than 0.5dB!
The waterfall colour (green) essentially remains constant throughout the passband.

We can say that the I/Q amplitude and phase calibration to be successful. (the calibration steps also includes the removal of the unavoidable center discontinuity)

Another major benefit of having the hardware calibrated is the availability of the Linrad smart noise blanker.
This feature has not yet been found in other traditional and commercial equipment.

The blankers are necessary at this QTH and no doubt at many other QTH's worldwide, with the many households ( with their electrical apparatus) nearby, causing sometimes high levels of
interference as high as 144MHz and beyond.

Without the Linrad automatic spurious removal and the Linrad noise blankers, this OM would have long moved to stamp collecting...

Linrad and the WSE modules integrated in the 2m VHF system
See here for the bloc diagram.

Linrad 03.12 with different hardware (10/05/2010)
The popularity of "software defined radio" within the ham community is rapidly increasing and more and more hams have adopted them for one or another shack-application...
(improved receiver, spectrum analyzer, bandscope, ...)

Just a handful of commercial ham SDR transceivers are known to exist and they are not cheap.
So most hams have turned cheaper receiver SDR's, such as the KB9YIG Softrock receivers, into transceivers. (KB9YIG also sells kits for transceivers)

The following describes one way to combine an existing commercial transceiver (in my case a secondhand, Kenwood TS850D) into a SDR-machine with the Linrad software.
Given the great time, we had, playing with a Softrock Lite II on 40m, we decided to order a Softrock Lite II IF 8.812 MHz for use with the intermediate frequency of the hf tranceiver.
The TS850D has a 8.83MHz IF ouput connector, which was used as the antenna input for the Softrock.
A 20dB gain block, with settable attenuator, followed the Softrock to the Delta 44 soundcard in the development dual-Core Ubuntu Linux PC.

Running the hybrid transceiver, as described above with the Softrock II Lite IF, is special.

First, the exact offset between the Softrock LO and the TS850D IF frequency should be determined.
This proved to be easier than anticipated, thanks to a small C-program for Linrad, written by Pierre, ON5GN, to control the transceiver frequency for the A-VFO, in my case.
(different new and older Trx from different brands can also be controlled by this small program)
The CAT frequency control window can be seen on the 2 Linrad screenshots above and the one below.

A bloc diagram of the setup is here.

The first thing to do is to put the transceiver in transmit mode (SEND).
This produces a small signal in the Linrad GUI (graphical user interface) that corresponds to the exact transceiver Tx frequency. (the small signal is leaking through the Tx/Rx relay)

The mouse is put on this carrier and clicked.
This should be done very accurately in the high resolution screen, checked in the baseband window center and finetuned with the mouse-roller.
See the screenshot above with signal peak under the grey vertical line in the high resolution screen. (with red horizontal lines)

Clicking on the blue "CALIBRATE" in the CAT frequency control window (visable at the bottom of the screenshot above) determines this frequency offset and shows its value in kHz and Hz. (in white)
Finally, push the letter "Q" on the keyboard, to transfer the corrected Linrad value to the hf transceiver via the PC-serial port, the CAT-interface and ACC1 of the Kenwood transceiver.

The right hand side of the screenshot shows the correct transfer of the TS850D-frequency (yellow frequencies) and the offset between the TS850D and Linrad. (some 15.384kHz)

Normal CW operation with the HF-transceiver is easy:

-for fast Tx/Rx transition, the Linrad parameters were adapted for low delay times. In the above example, the total delay from antenna to audio amounts to 140ms.
-turn down the transceiver audio and let the Linrad-PC generate the audio..
-tune a signal using the TS850D VFO dial to fall within the center of the baseband filter! (not the usual "click and listen" of normal Linrad operation!)

The above procedure and small delay time worked fine to contact SE6Y, Rolf in Göteborg (100W+2x21 doublet and 100W) and
DM2BRF/P (Didy with 5W and a 30m long wire, 4m up) on the 40m band, with:
-the indoor 15m dipole under the roof,
-a 60 year old handkey and
-some 50Wout of the TS850D.

WSJT 9.0 under Ubuntu Linux 10.10 (26-04-2011)

The easiest way to download and install WSJT, the popular digital communication SW package from K1JT, Joe Taylor under Ubuntu 10.10 is through the synaptic package manager.
You'll then get an older but fully-packaged and functional, WSJT version 5.9.3.

For more recent WSJT-versions, e.g. you've to start with locating and installing missing prerequisites for a succesful compilation, as indicated on K1JT's Linux information page.
Download the latest WSJT Linux version (in my case, into my home directory /home/michel) from here.
Launch the terminal command from /home/michel/WSJT: "sudo dpkg --instdir=. -i wsjt_yy_rxxxx_i386.deb" (in my case: yy=90 and xxxx=2136)
To start WSJT from the installed directory, just type: "./WSJT"
Make sure to select the right sound cards in the WSJT setup menu.
With thanks to Peter, OZ1PIF for the advice!

Linrad 03.22 and WSJT 9.0 under Ubuntu Linux 10.10 (26-04-2011)

For a long time, I've wanted to connect sound card output and inputs together by virtually connecting them. (avoiding the need for an extra soundcard and connecting cable)

The goal would be to combine:
- Linrad (fed by a wideband SDR-source) with its noise & interference combatting features as the input (frontend) program and then run
- other communication programs such as WSJT (JT65 or FSK441A) or WSPR as backend program.
In Windows, this function exists by means of payable SW, called Virtual Audio Cable. (VAC)
Again Linux shines compared to Windows since this function is "open sourced"...

The key to this internal Linux audio routing is the ALSA snd-aloop function. We started by following Leif, SM5BSZ's guidelines, here.

Important is to set and check the correct sound board order for a logical audio flow between the different audio devices.
Snd-aloop has created new sound devices called Loopback, which will function as our virtual sound connection.

These are the settings on the development Linux PC:
-Linrad:
INPUT     = Maudio Delta 44 (device 2)
OUTPUT = Loopback 1 (device 3)
-WSJT:
INPUT      = Loopback 2 (device 4)
OUTPUT = pulse (device 12)

Since WSJT requires a sampling rate of 11025, Linrad's parameters should be set for an output sampling rate of 11025Hz.

Make sure you start up WSJT first and only then Linrad!

Some relevant screenshots:

Linux WSJT9.0 debug screen (in the terminal) showing (from the 8th line onwards) the WSJT-version, the identified and finally the selected WSJT audio devices. (see higher)
The WSJT sound board selection is done under the WSJT Setup menu and checked in the WSJT debug window.
This WSJT debug log above shows some errors like: "bt_audio_service_open: connect() failed: Connection refused (111)".
These bluetooth related errors are not harmful to the combined Linrad-WSJT operation and can be disgarded.
More important is the message: "Audio streams running normally".

Linux Linrad 03.22 "U"-menu showing the "frontend" sound card selection, the M audio Delta 44 as first (low noise @ 96kHz sampling rate) sound board.
And Loopback 1 as the Linrad output "sound card".

The WSJT9.0 screenshot above shows that audio is getting through from the broadband Softrock Lite on 40m, through Linrad to the WSJT9.0 SW..
Due to the absence of WSJT-signals on the 40m band, there were no meaningful decodes...!
The correct WSJT audio level can be set by the Linrad audio cursor or via the sound mixer. (In the example it is some 15dB too high, hence the red color)

With big thanks to Pierre, ON5GN for his expert guidance!

Linrad 03.22 and WSPR 2.11 under Ubuntu Linux 10.10 (28-04-2011)

The same setup as for WSJT (Softrock 40m kit, fed by an untuned 15m dipole on the attic) was used to try out another interesting digital program, called WSPR.
(WSPR 2.11r2263 was used )
WSPR stands for "Weak Signal Propagation reporter" or whisper to reflect the weak signals. A lot more can be read on the web: http://www.physics.princeton.edu/pulsar/K1JT/wspr.html

The installaton of WSPR on the Linux Ubuntu 10.10 development PC (without additional prerequisite packages) was very straightforward and followed the same guidelines, as described above.
Important is to change the Linrad output sampling rate to 48000Hz and to have Linrad in SSB mode with USB decoding settings. (the red vertical lines in the baseband window)

Linrad 03.22 + WSPR 2.11screenshot.

The Linrad 03.22-screen shows the required USB-mode settings in the baseband window.
The green vertical line under the waterfall sits at 7038.6 MHz, the center of WSPR-activity on the 40m band. The DDS-VFO was hung close to coax of the antenna line, to calibrate the frequency.

The WSPR-screen indicates 4 stations to be decoded in the last 2 minute sequence: HB9CZF, M5DND, G4ZKV and OZ1PIF, Peter, who has helped me with the installation part! (see before)
Some more pruning is necessary to get the font right, to have the columns aligned in the bottom right WSPR-window. (minor issue)

Surprisingly enough, several decoded stations, today were the same callsigns as visible in the screenshots in the WSPR 2.0 User's Guide on I-net!
As predicted by K1JT, levels as low as -28dB were decoded, the strongest level reached 6dB.
The last columns indicate the stations' callsign, Maidenhead locator and the transmit output power, expressed in dBm. (30dBm equals to 1Watt, 17dBm to 50mW)
Good time synchronisation is of paramount importance to allow decodes. Internet time servers were used for this.
The latency (signal processing) through Linrad should be not too high to report acceptable delays through WSPRnet. Timing delays of 3sec through Linrad still produces good decodes in WSPR.
The above screenshot had less than 0.5sec of Linrad-latency.
Timing delays of 3sec through Linrad still produce good decodes in WSPR.

Linrad 03.22 with WSPR 2.11 under Ubuntu Linux 11.04 (30-04-2011)
After solving the usual NVDIA non-compatibility (through a BIOS action and a startup in safe mode), we successfully installed the latest Ubuntu 11.04 release on the development Linux PC.
We tried Linrad 03.22 + WSPR 2.11 with it and this combination still decodes fine.
Starting up two terminals through Unity did not seem possible. We found that with "Ctrl-Alt-T" this was possible
We can now enjoy the new GUI Unity, Firefox 4, LibreOffice 3.3 and the Banshee music-player.

Linrad 03.22 with WSPR 2.11 under Ubuntu Linux 11.04 on the Softrock Lite II IF on the Kenwood TS850D (02-05-2011)
With our Softrock 40m Rx, we were limited to a 200Hz span around 7.0386MHz.
We replaced the 40m Softrock with our 8.8MHz counterpart that get its signals from the TS850D IF-output.
This allowed us already to decode signals on the 40m and 20m band.
We're sure higher frequencies will be possible with our compromise indoor antenna later this year...
Usually, the transceiver dial is put some 9 kHz lower than the WSPR Rx frequency which works fine to avoid the Softrock LO-carrier interference.