Did you see the weaker sideband copies of this mystery signal which are -425, -434, -2167, -2620, and -5717 kHz to the left of it? There several other even weaker ones that could be copies too.
I've also noticed that the noise floor drops 2 dB across the band when this signal hits and then it rises to normal afterwards. I didn't think there was any sort of AGC in the ATA signal chain. Is there?
There is no intentional AGC in the system (although the gain is adjusted before observations begin). However, strong signals can cause nonlinearity in analog amplifiers and/or overflow in the digital processing. Such "clipping" causes weaker signals at other frequencies to be depressed. Imagine you hit the upper or lower rail of our processing system at some point in the chain. At the voltage rail (or max value of integer in digital processing). The signal stays at the rail, unchanging for a period of time. During this time, weak signals that should appear are "clipped" from the signal. So the weak signals are proportionally depressed.
2 dB suppression is consistent with spending about 1/3 of the time on the rail. I wonder if this is visible in the time series?
If there is a sine wave also present in the raw data (usually there are multiple sine waves present), then hitting the rail generates new signals (intermodulation products) that look like products of the strong signal with the sine wave, creating false images with spacing related to the sine wave frequency. This is again consistent with clipping.
We adjust our system prior to observations so that clipping does not occur for most signals, but a very strong signal can cause clipping. Given the number of amplifiers and DSP blocks in our system, it is likely that this particular signal has clipping prior to sending to disk (unrecoverable). It could even be happening in baudline!
(You may know much about this subject already, but I'm going back to basics here so everyone is on the same page.)
No, it is not visible in the time-series. Here is a plot of power in the setiQuest data covering a period a bit before, during, and after the ~12.7 second event. Note the little blip where the power is temporarily restored during the event which corresponds with a gap in the signal.
Yes, I did see those sidebands.
No, there is no AGC. The total power integrated over the relatively wide bandwidth of this signal may have driven some part of the signal chain outside of the linear region.
Watch and listen to baudline process this usual modulated signal collected at 1424.79 MHz for the setiQuest project at the Allen Telescope Array (ATA). It has a strange sine frequency modulated carrier with bursts of data. An unusual 605 baud pulse modulation is being used. The odd startup frequency chirp (drift) looks RC and is likely caused by something in the transmission line charging up. This signal is a mystery and we do not know what the source is.
In this movie the mystery modulated signal is played twice. First at a faster playback speed and then at half speed. Note how the audio clarity is increased when the low and high pass filters are enabled.
Three instances of the baudline signal analyzer are used in this listen demonstration. One baudline to play and adjust the playback speed and filters. A second baudline is recording and displaying a Fourier transform. The third baudline is recording and displaying a correlation transform with different color aperture settings and sample rate. The FM sine modulation, the data bursts, and the pulse coding make this an extremely interesting signal to watch and listen to. The "baud'ing" is clearly visible in both Fourier and correlation domains at the slower playback speed later in the movie. Next step, demodulation.
If you have the bandwidth make sure you watch this movie in 720p HD otherwise 480p isn't bad. I apologize for the dimness of the spectrograms, YouTube did this when it compressed it, the original file is much brighter, next time I'll anticipate this and open up the aperture a couple more notches than I think it needs.
What do you think it sounds like?
Thanks for posting this. This really shows off the utility of baudline. Would it be OK if I used a copy of this movie in some presentation? Just some seconds at the beginning would be enough.
Thanks for the analysis. Video and sound is a great way to impedence match my brain and the data.
Yes, it would great if you used the movie in a presentation. Let me get you a copy of the original higher quality file, the resolution and colors are better. The file is about 2 GB but the codec is Photo-JPEG so if I transcode to MPEG4 it should be a lot smaller.
The best way to view it though would be to run baudline. That would give the best resolution and a smooth 75 FPS frame rate. For the movie I had to crank the frame rate down to 28 FPS because the screen capture and compression program (QT player) consumed the vast majority of my CPU resources.
This video is amazing. It captivated me. I was hoping you would find interest in this signal and attack it with baudline. The sound made me reminisce a bit over tuning past RTTY transmissions on my shortware, but this signal is more complicated that that. Looking forward to hopefully further baudline analysis. Wish I had time to look at it further...
Glad you liked it. It will be interesting if we can figure out what the source is. The pre-data carrier being a sine wave FM modulated by a sine wave that also has AM modulated pulses is just bizarre. It seems way more complicated than a training carrier needs to be. I don't understand why anyone would do that. This just makes it even more mysterious. I have an exploration movie of this signal in the planning stages.
It is a captivating sound. I was playing it in my office and my girlfriend walks in and asks if we're being invaded? This is setiQuest so it's not such a crazy question and it does sound like classic B-movie flying saucer.! (:
I remember that a few years ago, during an "all sky" transient survey with the ATA, one satellite was observed to be transmitting right in the middle of the protected band. With a 12.7 second crossing time and a primary beam width of 2.5 degrees at 1422 MHz, it suggests a satellite with an angular velocity of 30x siderial (LEO). But then, the cutoffs at the top and bottom of the signal's appearance, are too sharp for a constant amplitude signal passing through the beam.
SIgblips' reported suppression of gain when the signal is reminiscent of effects we've seen when someone turns on a cell phone (cases > 10 dB of gain suppression). Perhaps someone could have turned on a cell phone at the dorm?