Development and experimentation of a new receiving system based on the correlation method.
By Andrea Dell’Immagine IW5BHY

1- Abstract
The recent release in Europe of the new VLF band (137 Khz) to the radio amateurs, offers some interesting and difficult challenges. The very low power and the minimal efficiency of the areals to be used, introduce the radio amateurs to a very low signal to noise ratio world. I have been studying the DSP analysis of the very weak EME signals for some years, thus I have been struck by the possibility to apply some new techniques to the VLF band that offers more semplicity in the RF section construction. This effort produced a whole year long work that gave a receiving system capable to dig out signals buried something of -40 dB under the noise floor. One prototype is ready and fully operative, in this paper will be introduced the results obtained in the reception of the VLF band. The application of this system to the higher weak signal bands for the moment brings to some tecnological problems that may be can be solved in future if enought interest will develop in these techiques.

These are the system features:

· Digital modulation ("zero" and "one")

· Can be used with a signal to noise ratio down to – 40dB

· Very low data speed from 3 bit/minute to 12 bit/minute

· Bandwidth less than 1 (one) Hz

· Fully automatic

· It needs an auxiliary signal from a frequency standard transmission

In my prototype I used the reception of the nearest medium wave broadcasting carrier wich I knew is controlled by a cesium reference.Here is the national RAI 1 channel on 900 Khz which can be received in my QTH (far 200Km) all day/night long with good strenght. This transmitter covers great part of Europe during the night. We have not to be discouraged by the very low data speed: in many weak signal activities like EME and Meteor Scatter the normal the normal sked lenght is 1 hour.

2- Principles
We know that the possibilty to receive a signal buried in the noise is only function to the speed of data transmission. In other words the lower will be our data speed (bit per second), the lower will be the signal to noise ratio of possible reception: the latter is only limited by practical issues, in my case I can receive at a data speed of 3 bit/minute a signal 40 dB below the noise. The kind of modulation can be CW like (0= carrier OFF, 1= carrier ON), or two tone FSK.
For Its simplicity I chose CW, with a bit lenght from 5 seconds (12 bits per Minute) to 20 sec. (3 bits per minute).
The receiver task is the following:

"given the signal to receive during the time of a bitlenght, decide if a carrier is present  (1) or not  (0) in the noise".

 In order to decide there is need of:

  1. A local oscillator (L.O.) that must be on the same frequency of the L.O. of the distant Transmitter, with a very high precision around +/- 10 mHz (1 mHz = 1/1000 of Hz), and this is the real technical challenge.
  2. A circuitry  that makes the comparison between the signal received and the local oscillator , (correlation).

  3. The circuit that performs the correlation is shown below
Picture 1 Correlator


The L.O. and the received signal are on the same frequency, thus the mixer converts the signal in a direct voltage (frequency=0). The Integrator gives a voltage that is proportional to the mixer output area calculated in the time gate.

The final output of this system is the sum of two different items:

  1. A component due to the direct voltage, so to the input signal. It is proportional to the input signal level and to the time gate.
  2. A component due to the noise . It is proportional to the amplitude (RMS value) of the input noise and to the square root of the time gate.
In other words, if the time gate doubles at a parity of input signal to noise ratio, the relative output bound to the signal doubles while the noise output is multiplied only by 1.41. From here we realize that we can inprove the S/N ratio increasing the bit lenght. This beaviour is valid if the L.O. has the same frequency and phase of the received signal.
This conditions applied to the phase is practically impossible to achieve becouse of the inevitable phase shifting intruduced by the radio waves propagation. For this reason the block schematic shown in the above picture must be doubled, giving so two different channels called I (in phase) and Q (in quadrature). That structure is shown in picture 5. The I and Q channels have the L.O. with a phase shift of 90°. This bring to a situation were when the I channel gives no output, the Q channel gives the maximum output and viceversa. The interpretation of the I and Q is easy. The vector wich has coordinates in I,Q represents the signal: the modulus measures the amplitude, the phase measures the phase shift relative to the L.O..
An alternate way to see this system is to consider it like a very narrow bandpass filter: the signal will enter the filter only if perfectly centered, while noise will be enourmously attenuated. This gives the possibility to receive signals at a very low signal to noise ratio.

 3–System Architecture
The system architecture is shown in picture 2. The L.O. is drived by the carrier of the reference transmitter, and thus matches our frequency tollerance requirements. The VLF signal comes throught a preselector/amplifier, and then goes to its input to the correlator. Time gate represents a bit lenght of 5,10 or 20 seconds. The result of the correlation is represented by two direct voltages (I and Q) which are acquired and processed by a microcontroller who makes the decision about the received bit (0 or 1). Meanwhile it is possible to calculate some reception parameters as the signal strenght and phase.
 Picture 2 System Architecture

4–System performances
In order to characterize the system performances we must define some parameters, the most important are:

· -3dB bandwidth . It is the system total bandwith

· Noise equivalent bandwith . It is the badwidth of an ideal filter (rectangular window filter) that gives in output the same noise power of the system.

· Signal to noise ratio for a fixed error probability .Given an error probability of 1/100 (a wrong bit every 100), is the signal to noise ratio that yield this result. We take as convention a calculated noise over the typical SSB bandwidth (2.5Khz).
In the following chart the above paramenters are shown in three different system conditions:

· Bit lenght = 5 seconds (12 bit per minute)

· Bit lenght = 10 seconds (6 bit per minute)

· Bit lenght = 20 seconds (3 bit per minute)

A bit lenght longer than 20 seconds in not practical , in this design at least, due to the instability of the L.O.
  bandw -3dB (mHz) Noise bandw eq. (mHz) S/N Ratio
12 bit/minute
6 bit/minute
3 bit/minute

NB. The frequency unit is milliHertz (1/1000 Hz)

Instead of numbers , we can use a practical system to evaluate the performances of such system: we can compare it with the most efficient tradional modulation method: CW.
We suppose that we are trasmitting a telegraphic signal wich is being received almost at the minimum copy limit, even using narrow filters (e.g. DSP).
A "correlation receiver" in this conditions would offer an excellent  reception, and to obtain again the previous condition we should decrease our trasmitted power.

Following the estimared  power reduction (in times) for the three different bit/rates:
 12 bit/minute
6 bit/minute
3 bit/minute

These are quite surprising results, such reduction ratios, if applied to E.M.E. (Earth-Moon-Earth) communications, would bring to the possibility to make a two way contact between two stations equipped with a single yagi and 70 W in the 144 Mhz band.
However the application of this method is very difficult to VHF and higher for some reasons.
The necessary frequency precision (few mHz) is far in those bands from the possibilities of the radio amateurs (n.d.r. for now), the doppler effect and the signal spreading due to the lunar reflection are additional problems.

 5-System description
The receiver is divided into three different parts:

· 900 Khz reference signal receiver

· Local Oscillator (L.O.)

· Correlator

Reference Receiver
The block diagram of the MF reference signal receiver is the following:
It is a classical superetherodyne receiver . Applying this design we have some problems becouse the L.O. should have the same frequency precision of the wanted reference signal itself. The solution is to add a second mixer : the first one works by difference, the second one by addition. We obtain so a 900 Khz signal again but the L.O. error is cancelled. The last stage is a comparator wich converts the 900 Khz signal at TTL logic level
Picture 3 Standard Reference Receiver


Local Oscillator (L.O.)

This is the most critical part of this system. The L.O. must :

· Generate a carrier with frequency from 0 to 500Khz drived by the 900Khz reference.

· Garantee the best possible frequency stability between the stated limits (few mHz).

· Whitstand the signal strenght variations and the noise on the 900Khz reference.
Picture 4 L.O. block diagram

The circuit is build around a D.D.S. This device can deliver a variable frequency carrier, digitally programmable in very little steps, starting from a reference oscillator.
In our circuit this reference is generated by a normal 20 Mhz xtal.
The precision of the output is the same of the reference crystal at 20 Mhz, that is quite below the one required.

My trick is now explained: the 900 Khz stable reference signal is used as the timebase for a frequency meter wich measures the real frequency value of the 20 Mhz quartz. With this measured value, a microprocessor re-programmes the DDS making the necessary correction every 10 seconds to balance the quartz tolerance and thermal drift.

The employed algorithm can detect, if due to signal fading or noise on the 900 Khz reference, the current frequency update is valid or not, and computes a current precision estimation. This value in mHz is shown on a display.

The correlator is the "real" receiver, and is composed of the following parts:

· Front End. It is a variable gain selective amplifier.It is connected to the antenna and it is tuned on the chosen VLF frequency to receive.

· Dephasing Filter. It is a circuits that splits the signal from the L.O. into two identical signals , with a phase shift of 90°. (quadrature).

· Mixer for the I and Q channels.

· Integrator for the I and Q channels.

Being at relatively low frequencies, all the circuits above are built with operational amplifiers, this brings to an easy control over the total gain of the whole chain.The I and Q outputs are then acquired by a microprocessor (see picture 2), which calculates the reception parameters and controls the whole system.
The results are shown on the LCD display and sent to a RS232 interface, so they are available on a P.C. to be processed for other uses.
 Picture 5 I-Q Correlator
  Click on the images to enlarge

6-Results and Measures
All the measures were made using the signal of the time/frequency standard reference DCF77 on the frequency of 77.5 Khz.

Other information about the DCF77 are:

· Trasmitting site : Mainflingen, 25 Km from Frankfurt Germany.

· Frequency precision: < 1E –12 over a day period

This trasmitter allowed very thorough measures verifying the L.O. precision, and thus was experimentally confirmed my theoretical base.
It is very interesting to study the stability and precision of the synchronization between the L.O. and the 900 Khz standard reference . This is obtained measuring and processing the correlator output and then estimating the phase of the defined vector having coordinates I and Q. The phase variation speed represents the L.O frequency error compared to DCF77.
This is a worst case situation , in fact the phase shift due to the radio wave propagation is not considered. The following picure shows the L.O. instability measured over a 15 minutes interval .
The maximun frequency drift is about +/- 5 mHz (5/1000 of Hz !). As stated above this value is to be considered worse than the real.
Picture 6 L.O. frequency drift

Another important parameter was the signal to noise ratio during the DCF77 reception. I made this measure in the following way:

· The L.O. frequency has been moved of about 10 Hz away from the DCF77, therefore only noise could be received.

· The RMS value of the I-Q vector has been computed. This needed some tenth of bits, so brought to a some minutes long reception time. This operation allowed to compute the noise power or rather what is called the  reception "noise floor".

· The L.O. frequency  has been set in order to receive the DCF77 transmitter.

· The RMS value of the I-Q vector has been computed again. This operation allowed to obtain the DCF77 signal power.

· The ratio between the two computed values (S/N) has been computed  and converted in dB.

Picture 7 shows the measured S/N ratio in dB over a 15 minutes time base. The values oscillating between +33dB and +40dB due to propagation fading, confirmed the teorethical estimate.
The areal used was a simple long wire 10 meters long. In this conditions, this signal heard from a normal communications receiver was clearly there but affected by strong noise (only few dB above the noise). A great amount of data about the DCF77 reception were recorded over few months .
A special reception case is shown in picture 8. It is a 80 seconds long periond when due to very negative fading the DCF77 signal was not audible in a standard SSB communications receiver.
The graphical representation on the I, Q planes shows clearly the phase shift of the received signal every bitlenght (5 seconds). The vector with I, Q coordinates (signal strenght) appears to overcome several times the noise floor, giving so a flawless reception.
Picture 7 Signal to Noise ratio of DCF77
Picture 8 Reception of DCF77


7- The VLF Beacon Project
Together with the Correlation Receiver I developed a suitable beacon which will be used for propagation tests. Its data are the following:

· Frequency controlled by the 900 Khz carrier standard reference.

· Programmable frequency between 0 and 500 Khz in 4 mHz steps.

· Transmitted data: 1 minute carrier on, 1 minute carrier OFF, callsign in slow CW.

· Output Power : abt. 30W on 50 Ohm.

It is interesting to note that a "correlation"beacon system can offer the following advantages:

· Trasmitter can be identified by his frequency.

· Channel bandwidth < 1 Hz.

· Completely automatic reception and data recording by mean of a P.C.

· Automatic reception measurements of reception parameters  (signal strenght, phase shift etc.).

 8- Future developments
A system capable of receiving signal at a very low S/N has been discussed.
This method , nowadays seems to be the only that can be used for efficient VLF communications over average/long distances.

Some ideas could be developed in future:

· Usage of a  TV carrier from a DBS satellite as the reference standard signal.

· Long distance reception experiments (1000-2000Km).

· System integration

- totally DSP technology correlator

- Receiver/trasmitter set easily reproducible.

· Definition of a protocol for two way data exchange (QSO).

· Correlation method applied to other bands (for E.M.E. use ?).

I would like to thank the following people:

Maurizio Gragnani IK5ZPQ for the the prototypes and his valuable help
Andrea Ghilardi IK5QLO for this english translation and WEB publishing.

I can be reached at .