These three articles are about using off-the-shelf software or an EXCEL spreadsheet for pulling out signals 30 dB below the noise. You will not have to have advanced degrees in Math or Electrical Engineering to employ this DSP technology (some money, yes). It is intriguing to me that there are many astronomers dedicated to getting the signals, while others are interested in what you do with them once you get them; this is not a criticism, only that there is a segment of scientists who relish having the weak signal data for processing. You may have to understand simple signal concepts and algebra, but all the real work being done in this article will be done by software.
Here is a overview of each of the three articles which will appear on this web site:
In this article: DSP Part 1 - Click Only DSP Processing Using Spectra Plus
Let us say that you have your Earth Moon Earth (EME) signal or the radio star Cass A full-power audio signal on stereo tape, Digital Audio Tape/Digital Compact Cassette (DAT/DCC) tape, or directly captured in a wave file AND you know what you sent on EME transmission or what the expected wave form should look like on the radio star Cass A. This article shows you how to see if your expected signal is in the noise with some wave file manipulation using pointing and clicking in Spectra Plus. If your signal is well above the noise then most audio spectral software packages readily produce spectrums.
In a later article: DSP Part 2 - Click Only DSP Processing Using EXCEL
First we will show you a more complex "dig out" by "time phasing" the source audio wave signal against a reference signal on the other channel; this is a follow-up to the DSP I article. Secondly let us say that you have your EME moonbounce audio signal on stereo tape, DAT/DCC tape, or directly captured in a wave file and you do NOT know what was sent on EME transmission; but you do know that it was repeated over and over. The remainder of DSP Part 2 of shows you how to see if a repeated transmission, for which you do not know the pattern, is in the noise. With some pointing and clicking, first using Spectra Plus followed by an EXCEL spreadsheet, you can pop out the weak signal pattern.
This third paper is also on this site and deals with a statistical process to identify signal in the noise using the concept of isolating monophonic signals in the noise, without having knowledge of any signal patterns. The software algorithms employ signal runs test analysis for noise reduction and signal identification.
DSP Part 1 - Click Only DSP Processing Using Spectra Plus
Now let us move on to introduce you to the SASER© Project more fully, and show you our communication strategies for sending microwave signals to the Sun and analyzing the return signal products. We introduced SASER© at Greenbank in 1996 and had our proposal reviewed by six experts in radio astronomy. The central theme of the reviewers was the weak nature of the return signals, if they exist at all. We felt that we needed to address the rear end first... thus the impetus for these DSP experiments.
Finally, you at least really need to know why we are interested in this DSP so intently. We are putting together a Solar Project which can be summarized as:
Our project explores the radio characteristics of the solar spectrum during solar maximum. The plan for SASER©, Solar Amplification by Stimulated Emission Radiation, is to investigate the effect, if any, that weak, transmitted microwave signals have on the solar radio continuum with the possibility of receiving these signals back on Earth. SASER© will attempt to bounce or mix or amplify (or some combination) a weak 400 watt microwave signal with the solar radio continuum and receive signal products back on Earth. Simply go to www.jupiterspacestation.org to see the details.
The following is a table of contents for this DSP Part I article so you might go directly to the part that fancies you the most:
A spectrum analyzer is an instrument used to convert a signal from the time domain (amplitude vs time) to the frequency domain (amplitude vs frequency). If you are familiar with an ordinary oscilloscope you know what a time domain display looks like. A frequency domain display is known as a spectrum. Unless you are measuring a single tone, an oscilloscope provides little in the way of frequency information; however, a spectrum analyzer clearly reveals this information. An Audio Spectrum Analyzer, by definition, is limited to processing signals in the audio band. The specific frequency limit is determined by the capabilities of your computer's sound card.
An Audio Spectrum Analyzer is very useful for measuring the fundamental frequency components which are contained in an audio signal. It can accurately measure the frequency of single or multiple tones and the frequency difference between them.
In radio astronomy when we capture the signal of a radio star the traditional data collection methods are to either put the signal energy on a chart recorder or convert the signal energy from an analog voltage to a digital value (often called A/D conversion). These digital values can be then plotted. While capturing the data in the more traditional methods we have to establish an integration time which amounts to collecting signal in a bucket until it rises above the noise, then writing the resultant value onto the recorder or into the database. During this integration time the integral is additive, but not the noise component. More significantly, you only get one shot at it; you can not later change the integration time after the data collection has concluded.
Software spectral analysis offers a powerful alternative at a low cost. Let us pose a radio astronomy question: "Can we say in this article that there exists a direct, if not equivalent, relationship between the signal energy in the original RF signal (before the detection phase) and the signal energy represented in the audio spectrum?" Obviously there are various radio receiver IF and AF gain stages which change the absolute energy level by significant amounts; however, the spectral characteristics of the signal of interest remains the same. Receiver IF and AF filters, as well as AGC, may only shape the signal. You would prefer a high quality receiver with AGC turned off.
Another question: "Can we say that if a radio has a bandwidth of 12.5 KHz then the signal spectrum (and energy) in that audio spectrum bandwidth is a map of the spectrum about the center frequency ±12.5 KHz?" The detected signal is essentially a tunable window of the total RF spectrum, however, the pre-amplifier, filtering and detection stages are never exactly linear so you would be more accurate to say that the audio window is an approximation of the RF window. The powerful feature here is that all the spectral information is recorded and one can later manipulate the wave with software to extract the signal of the original source, even below the noise.
The spectral analysis software works in conjunction with the sound card on your computer. Plug the audio signal to be measured into the Line-In or Microphone jack on the back of the sound card. Spectral analysis software such as Spectra Plus then uses the card to perform an "Analog-to-Digital" conversion on the audio signal. This digitized audio is then passed through a math algorithm known as a Fast Fourier Transform (FFT) which converts the signal from the time domain to the frequency domain. The CPU on your computer and the spectral analysis software is used to perform this transformation.
FFT is an abbreviation for a Fast Fourier Transform. The Fourier Transform is a mathematical tool developed and named after Jean Baptiste Fourier (1768 - 1830) and is commonly used to convert a signal from the time domain (amplitude vs time) to the frequency domain (amplitude vs frequency). A frequency domain plot is also known as a spectra plot.
The Fast Fourier Transform (FFT) is a computationally efficient implementation of the Fourier Transform developed by J. W. Cooley and J. W. Tukey in 1965. The Fast Fourier Transform is limited to block sizes which are even powers of 2.
For example, if a FFT is performed on a pure sinusoidal signal, the resulting spectra would be a single peak (line). Real world signals are a composite of many sinusoidal signals; examining the signal spectra clearly shows the individual frequency "tones" that could be used to produce the original signal, which is what we are seeking in radio astronomy.
Many of you will think we are partial to Pioneer Hill Software (Spectra Plus) or Microsoft (EXCEL). We just happen to have been users of these products for several years. You need to be aware that other fine products like SETIFOX (by Daniel B. Fox) or FFTDSP42 (by Mike Cook, AF9Y) and other well written articles in the SARA Journal by M. E. Valdez are additional tools for understanding and employing DSP for radio astronomy. Also in the February, 1997 issue of QST is an excellent comparison article about different DSP software packages so hams can "see" their signal.
Here at the Jupiter Space Station our Pentium computer uses a Sound Blaster card, however we are seriously considering the Fiji sound card by Turtle Beach which permits the output on the card to also be duplexed with the input to the card for signal generation applications, and this card features a high performance 20 bit A/D chipset on the front end. The card also provides an optional Digital I/O feature for use with DAT/DCC recorders or other devices supporting the SP-DIF (Sony/Philips Digital InterFace) output. The SP-DIF provides a digital audio datastream that can be fed directly to digital audio devices such as DAT (Digital Audio Tape) or DCC (Digital Compact Cassette) recorders, or even into a digital signal ready amplifier. The sound quality provided by this output is of the highest level and is limited only by the device connected to it. The audio performance of this card is a level above the sound blaster/clone cards on the market. With a dynamic range of well over 90 dB and Total Harmonic Distortion (THD) of <0.0025%, you are getting professional quality measurements for a very affordable price ($275).
Spectra Plus Version 4 ($497) is now out, however the materials in these articles were developed with Version 3. The $497 figure includes the base price and two options which were necessary for this work. We have been using the product for 3 years and the new version has some new features which we are going to try out right away: merging wave files, time delay processing of one channel against another, and dynamic external linkage from application programs like Visual Basic. A complete list of specs and a thirty day free evaluation copy (without features disabled) is available from:
The signal processing capabilities are astounding and some of the specs are listed here to give you a feel for the technology employed in signal processing:
In our first encounter with Charles Osborne he gave us an excellent tool for approximating antenna system performance. It is a "slide rule" type of calculator that provides many antenna specs with a few inputs. Upon aligning the frequency of 1300 MHz with the 10' dish mark (assume a 60% antenna efficiency), the calculator says the antenna gain is a little over 30 dB. If you needed to acquire a signal that required 30 dB more gain, and wanted to achieve that goal by building a bigger dish, the dish would be beyond the 200' mark. You can just imagine what your spouse and neighbors are going to say when you start building that dish! We have successfully done 30 dB down and we are now moving to 40 dB down in the noise.
You have to have some signal 30 dB in the noise to play with so we waited for the Sun to pass at the zenith over our loop antenna mounted inside our 10' dish, and we recorded the solar noise at 24.5 MHz onto digital tape. Everyone agrees that natural solar noise is basically random and quite strong.
Next a small C program was written to dub the theme from Close Encounters of the Third Kind (CE3K, 5 notes, a mistake we might add) onto the other channel repeatedly at fixed intervals. The C program played the 5 notes through the PC speaker, into a medium quality mic, through a mixer, and onto the digital tape. The mixer was connected to the line in (stereo) of the sound card running under Spectra Plus. One could see (in real time) the solar signal waveform on one channel and the CE3K music waveform on the other channel. Finally using the signal strength y-axis of the Spectra Plus Spectrum Plot, the CE3K music waveform was dropped 30 dB using the volume control on the mixer. The CE3K music was mixed with the solar noise on one channel and recorded by itself on the other channel.
One can see in Figure 1 that the left channel is on the top and it has the solar noise and the music from CE3K mixed in, 30 dB down. The 30 dB down music from CE3K is simultaneously recorded on the right channel. Figure 1 shows a single snapshot in time of a simple Spectrum Plot. Again we use this real time Spectrum Plot in Spectra Plus to adjust the mixer volume levels.
Another significant piece of information is that the CE3K music is dirty. It has a lot of noise introduced because of the chain of recording equipment and the mic was even picking up all the room background noise and echoes. It is not a clean reference signal.
Now in Figure 2 we have a continuous map of the music being played in what is called a 3D Spectrum Plot. The time goes from 0 to 8.5 seconds in this 3D plot and both left and right channel are displayed. The humps on the right channel spectrum are the musical notes from CE3K. Also if you look at the long black band on the right side of the left channel (solar noise mixed with CE3K) you can clearly see the signal strength is above that of the right channel. This is our favorite view for examining signals.
Immediately some are saying that makes this too easy because the reference signal (CE3K) is in time phase with the solar/CE3K mixed signal. Well it does make the solution of pulling the CE3K signal out easier, however we will address the more complex issue of "time phasing" in the next DSP Part II. Besides any time we can get easy in DSP, we will take easy. And in case you are also wondering, you cannot hear CE3K music in the solar noise of the left channel.
So, how do we pop the "CE3K hidden signal" out? Well, you ask for (click on) a right channel/left channel display with Spectra Plus software and this pulls the CE3K signal out like lightening bolts against a black sky. This feature in Spectra Plus is called a real transfer function which is a logarithmic ratio of the two channels. The results of this can be seen in Figure 3 and the concept is portrayed in Figure 4.
What we are dying to do is make a wave file of the transfer function results and listen to it through the soundcard; well it is on my wish list anyway.
Here are the Spectra Plus settings we use to do a transfer function:
Next let us move on to examine the transfer function concept in Figure 4.
Wish We Had Not Done That Section
Using the five notes from CE3K was not a good idea because the 5 notes created too much of a delay between the repetition of each note, and some of the notes created awful harmonics which clouded the transfer function process. It is far easier to create a signal pattern with dit dah as in a cw transmission. The transfer function did pop the CE3K out nicely, however, in signal averaging (DSP Part II) we had a hornets nest on our hands, which we will share with you in the DSP Part II article.
You can try out Spectra Plus free for 30 days by going to:
Again, you can download demonstration wave file used in this article from our web site:
and clicking on the DSP Signal Processing link.
With this trial software and our wave files, you too can feel like Flash Gordon at the control panel of Dr. Zarkov's rocket ship.