Email the author

First test signal

At the same time as I uploaded WU2WAV to the page, the SETI@home client program downloaded a new workunit, 16my00aa.16831.834.11088.102. That workunit turned out to be one of the test-signal workunits (identified by SETIMonitor). What made it so special was that I with the help of WU2WAV knew before I started processing the unit that it was special.

In the wav produced by wu2wav, there was a faint tone that lasted for almost 2 seconds. When I started the SETI@home client (Windows GUI 2.70) the signal appeared on the screen very strong.

SETIMonitor confirmed my suspicions, it was indeed a test-signal I was seeing and hearing:

Now the frequency reported by both SETISpy, an analysis of the wav and the graphics produced by SETI@home were identical, except for 2 smaller signals at the same time, but at a different frequency. Here's a screenshot of the frequqncy analysis of the wav file produced by WU2WAV:

As you can see, the frequency is the same in SETI@home's graphics, the analysis and SETISpy's info below.

What immediately caught my interest was the 2 weaker signals next to the main strong one. I tried waiting for further graphs from SETI@home, to see if the two weaker ones would turn up.

As the SETI@home client moved on to smaller FFT's, it became harder and harder to read the graphs. It could very well be that WU2WAV doesn't convert the WU's correctly, thus the two weaker signals.  Finally, I analyzed the WU with a small program I made in Linux. It analyzes the WU's in a more 'native' WorkUnit format and at an FFT length of 131072 points, here's the screenshot from it. It's hard to make out the graphics since it is scaled linearily. It does seem WU2WAV has a slight error in the converting routines, still, it's the best program released so far for converting WU's. 

Well, I guess you all came here to hear the test signal. I've got two versions, one that is the raw output from WU2WAV, the other is a filtered version. I used Noise reduction technologies available in both CoolEdit and GoldWave. I did this only after I had found the test signal, and could hear it. When I knew the position of the test signal, I marked all the noise to the right of it and had CoolEdit analyze it at the highest FFT setting, 24000 points. Then, I marked an area twice the time as the signal itself, and engaged the noise reduction. I did this twice. At this point, you can hear the FFT 'noise' in the background, that noise can not be reduced any further without loosing the signal too.

 Unfiltered WAV  (82,7KB) Listen to this first, several times. If you still can't hear the signal which is almost two seconds long, then try the Filtered WAV.

 Filtered WAV (91,5KB)

It is my personal opinion that the noise reduction in GoldWave and CoolEdit can not be used sucessfully until you've found the signal in the unmodified wav. So, just because you've filtered a 'normal' WU 12 times and now can hear something that reminds you of a monkey playing piano, doesn't mean that it's anything special.

Additional resources:

• outfile.sah (30KB)