AM Splatter Killer

SplatAway the AM Splatter Killer >> How to get it

Wideband receivers like Winradio etc. that also receive the adjacent channels allow new methods to improve AM reception you could only dream of before. One of those is the reduction or near elimination of splatter from adjacent channels that fall within the bandwith of the wanted signal. Especially useful in split channels conditions or during trans-atlantic MW-DX when 9 and 10kHz spacings overlap. This method uses the redundant information of an AM signal, namely the two indentical sidebands of the unwanted adjacent signal to elminate each other, so that the wanted signal becomes clearer. All is done in software, no extra hardware is required.

Typical Example

In North-Germany my wanted 590 kHz signal (VOCM) is splattered by 594 Hessischer Rundfunk, Germany and 585 BBC Scotland.
I also found that one sideband of each interfering signal is clean. So I can use that clean sideband to kill its mirror that causes the splatter on the wanted 590.
Method:
We are tuned to 590, the lower sideband (LSB) of 594 reaches into the upper side band (USB) of 590.

The USB of 594 is clean and we subtract it from 594-LSB.

590-USB has now been cleared.
The upper sideband (USB) of 585 reaches into the lower side band (LSB) of 590.

The LSB of 585 is clean and we subtract it from 585-USB.

590-LSB has now been cleared.
The 590-LSB/USB is now clean and can be AM demodulated.

An easy theory, but needs a lot of sweat to implement in software, impossible to do in hardware. Eventually I found the solution in Spectrum Lab. It contains a Lab Components window where you can wire all sorts of analog function blocks. I quickly got it set up there and only then discovered that the conventional analog approach has a lot of problems. A new approach had to be found.

Why not just subtract the spectral lines that are equidistance from the carrier? Easy on paper but how to put that into practice? Spectrum-Lab came to help again. Its author Wolfgang Büscher implemented an interface for a FFT-Filter-plugin. I created such a pluging called SplatAway and now have access in real time to all the spectral components, manipulate them and hand them back to SL, which converts them back into an audio signal (time domain). So all is being catered for, I only needed to concentrate on my ideas.

As you can see an IF-bandwidth of >20 kHz is required, and the whole range has to be linear. Winradio has about 22 kHz but its roofing filter rounds off the edges, so max. 19 kHz can be used. For this reason and to simplify operation I only clean up one sideband and use USB or LSB demodulation to recover the audio, all done in the plugin.

The result

For testing I fortunately have a stable split channel available on LW: 177 kHz (DLR) and 183 kHz (Europe-1)
In the left spectrogram you see both, the right part shows the resulting audio spectrum ( 3 kHz wide ).
The setup uses:
177 kHz USB 3kHz wide,
183 kHz LSB 6kHz wide, subtracted by 183 kHz USB

In the waterfall at marker 12:53 the splatter-killer was briefly turned off to show what you normally would hear. The right audio waterfall shows a lot of high frequency (1..3kHz) splatter is audible. Listen to this audio clip taken at the same time. It starts with SplatAway turned on. From :12 to :25 it is turned off

In audio clip #2 DLR carries music. SplatAway is turned on & off. In between you hear sections sounding muffled. In those cases only the static standard small bandwith (as in any radio) was used to reduce splatter. SplatAway adds to it the higher spectrum parts cleaned from splattering. The result sounds much better.

It works best when the unwanted is strong and stable with equal strength sidebands, thus optimal for locals that cause the most headache.

Ever wanted to "switch off" your local MW transmitter ?

SplatAway can work as Local-Away or Pest-Away

When carefully eliminating the sidebands of a local AM tx, then you can hear the tx underneath it on the same frequency
This audio clip demonstrates the elimination of AFN on 873kHz (150 kW) just 30km away from me and then reveals from Spain SER R. Zaragoza

It starts with suppression = ON. AFN becomes loud each time the suppression is briefly turned off. Towards the end SER becomes stronger and you hear it also in the background of AFN when suppression is off. At the end AFN carries music and suppression works quite remarkable Believe me, the antenna was not touched at all

This audio clip is even more impressive, because it was recorded with SDR-IQ that has no AGC and does not attenuate on strong signals and thus the levels of weak signals are retained. The local 900 WGHM Nashua (recorded by Bruce Conti, Nashua NH) is completely suppressed and reveals clearly the signals underneath it. Bruce writes: The English under WGHM is CHML Hamilton, Ontario. The Spanish vocal is probably R.Progreso, Cuba.

This audio clip uses Perseus as a receiver with its higher dynamic range. It is tuned to 531 Faeroers that hits here like a strong local despite its 1000km distance. In the background you hear Spain. The clip starts with suppresion=Off and you hear Faerors with a man singing. During suppression=On there is still remains of Faeroers audible, but a lot weaker

Note:
The released SplatAway V1.0 does not support this new mode, so don't try it.
The newer version has a few more parameters to be adjusted. It now uses an auto adaption scheme that makes it easier to operate in everydays usage

A big Thank-You goes to Wolfgang Büscher the author of Spectrum-Lab !
Without his help and him expanding Spectrum Lab to handle FFT Plugin this project would not have been possible.

How do I get SplatAway?

SplatAway is available for free provided your insert the links (shown on the right) of my other web sites on your web pages. I need them for search engines to better rate those sites. You can place the links at the bottom of your page in order not distract your visitors. Please make sure that Google etc. finds and lists your web pages. Verify it by entering in google: site:www.yoursite.com/yourwebpage.htm
It must be real links, not just plain text, and no nofollow attribute please.
If you don't have a web site yourself then ask a friend who has one to do that for you or even your employer.

Send me an email containing the url. I will check if it is ok and then send you SplatAway by email. SplatAway will run for 3 months and has to be renewed after that period, provided that my links are still there and google lists them by that time.

The unrestricted old version 1.0 only works with real IF (not I/Q) and cannot do Local-Away mode is available here

Try out how SplatAway works by loading this short IF file in Spectrum Lab

File / Audio Files / Analyse and Play (with DSP)

It is ZIPed wav file recorded around noon.

12kHz is the "noise" 183-Europe1
6kHz is the wanted DLR

Links

http://fewo.3sdesign.de

http://zugang.3sdesign.de

http://zeiterfassung.3sdesign.de

Descriptive Text

Ferienwohnung Burhave, Nordsee
* Vacation home at Northsea coast

RFID Zutrittskontrolle
* RFID Access Control

Zeiterfassung mit RFID
* RFID Time and Attendance

* Use German or English words, or please translate them into the language of your web page

Installation

copy

Splat-Away.usr into C:\Spectrum\configurations\

Splat-Away.dll into C:\Spectrum\

Operation Instruction

With "Quick Settings" load splat-away.usr This will set up most of the needed configuration

Open circuit window

click on the light-blue FFT filter

This rotating pointer in the small window shows the noise carrier vector, it should be adjusted to a near standstill by entering a -2.99 .. 2.99 Hz value in the green Osc field.

The symbol = indicates that your are excact on the carrier, otherwise it changes to < or >

The above parameters are used to clean up the USB of 177 (IF= 6kHz) from splatter caused by LSB-183 (IF= 12 kHz). When the TRT-180kHz carrier shows up you will hear
a 3kHz het. Set the notch filter to 3 kHz to get rid of that.

Due to some quirks in the b20 Version of Spectrum Lab, it may be nessecary to restart SL when you don't see the rotating pointer window. This will hopefully be corrected in the next SL version.

Click on tab FFT Filter plugin

The grid holds all needed parameters, enter them in the Value column.
Note:
Values only become effective after pressing the return key

Enter carrier and sideband width of the wanted signal. Positive sideband values select USB, negative values LSB.

Enter carrier and sideband width of the noise signal (the one that causes the splatter). The carrier frequency value has to be accurate, monitor the rotating vector in the small scope window. It should rotate slowly, set it to a near standstill by adjusting the value of the green Osc value in the above circuit window. Normally that value will be in the 0..4 Hz range, enter 1 or 2 digits after the decimal point when needed.

The noise sideband is the one that reaches into the wanted and is to be removed. Make sure that its mirror sideband is present with the same width and is clean.
Set reduction level to 0..1 (decimal value). 0 = no reduction, 1 full reduction. It depends on how unequally strong the noise sidebands are. If the clean one is stronger than use only 0.5 or 0.7, if it is weaker than enter 1.5
At the bottom Zero-Clipped should not go above 40..50% If it does then reduce reduction level
Grid line 5 and 6 hold two notch filters which come for free, because they were easy to implement. Enter the tone frequency that you hear and want to suppress. Set to 0 if not needed.

Brief instructions

To make operation easier, this small user interface was created. Start it after SL was started. It is loosely coupled. To fetch the current SL settings click on the update button.
It was written in a hurry, so it has some quirks. It is now a tool and not fool proof.
The number fields are used as divider for the spectrum display values, the smaller the value the larger the spectrum becomes. Make sure NOT to set it to zero, it will crash. Tip: Leave always a 1 at the end of the string
The slider increase when moved upward. Will be fixed when I have time.

The white field shows the carrier spectrum, the little marker indicates the position of the entered noise carrier. It should sit on top of carrier peak as shown. However if there is no real peak or it is rapidly moving, then more than one carrier are present. You need to increase the FFT-size of the FFT-filter or reduce sample rate. Otherwise it will not work, SplatAway needs to lock on the carrier. Finally move the left slider until the rotating vector gets to a standstill.
The left slider controls the reduction level, adjust it around 0.9 - 1.0 for maximal suppresion.
The central slider controls the angle, normally it is 0°, but sometimes carriers may have a phase difference, especially when receiving a signal from the back of a directional antenna. If you can't get reduction, then try this slider. However you need to constantly readjust, as the phase difference changes with propagation. See this as a last resort. You listen how the frequency distribution of the audio changes

All sliders can be changed by up/down keys or page-up/down keys with a larger step size.

Best way to proceed when to remove splatter

enter values of noise carrier and wanted carrier to the carrier to be removed

enter bandwith of both to the one that causes splatter on the wanted

adjust to minimze the unwanted signal

enter value wanted carrier to the wanted signal and set its sideband accordingly

If no splatter reduction is obtained, then switch on Amplitude-Only mode. It will remove all spectrum lines destructively, i.e. an underlying signal is also killed. This is a last resort to reduce splatter at adjacent channels, and often gives good results.

After a change of setup in SL, it may appear that it does not work well. In this case it helps to restart SL

Two .USR files are included, one for I/Q-radios and one for non-IQ as used by Winradio or soundcard input. Load the appropiate before you start. You need to enter the sample rate in SL yourself