PA8W Amateur Radio

Wil, PA8W,  E-mail:           

The PA8W Doppler RDF

NEW: Building the reworked Version 2.3 doppler RDF

Building the old Version 2 doppler RDF

The old Version 2 Main diagrams

Doppler Antenna Driver

Homebase Array

Mobile Array

Operating a doppler RDF

8-antenna high performance Doppler

November 2013: improving UHF performance

Januari 2014: adding elevation measurement

May 2015: The new RDF-40 project: a microcontroller based RDF

Why a circle is supposed to be round...

By the end of the year 2011 I started the development of the PA8W Doppler Radio Direction Finder.
Wherever possible I used available knowledge found on the internet.
I studied, filtered, adapted and enhanced existing designs until my dopplers finally performed the way I wanted them to. 
As a result, the PA8W Dopplers have a few features that I missed in existing designs.
For example, is uses soft switching of the antennas, giving a massive improvement compared to the generally used hard switching.
It has an automatic gain control to make it highly tolerant to input level variations. 
It also features an antenna test mode, in which all antennas are selected in a slow sequence.
And last but not least, I optimized the necessary antenna array using modelling software.
So, despite its simplicity, the "PA8W dopplers" really have a few advantages over existing amateur designs, 
resulting in surprisingly good accuracy even on weak signals.
As soon as a FM signal is strong enough just to be readable, the RDF is providing a pretty stable and accurate bearing.
The RDF also produces steady bearings on signals with strong modulation.
This soon led to the development of Version 2, which, due to its simplicity and yet very good performance, 
has been reproduced by a large number of Radio Amateurs since its birth in 2012.
And in december 2012 a 8-antenna high performance Doppler was added to the range.
Since then further additions and refinements were made in all three designs, resulting in a makeover of Version 2 in the first months of 2015. 

But first let's get back to the beginning:
Initial Design Goals:

Simple design, easy to reproduce, using very common components, mainly based on the buffered C-Mos 4000B logic series.
Single sided PCB's only.
No software necessary, no PIC's. Lots of people still don't like handling bits and bytes...
It should however have an output available to feed data into a computer, because there's very nice software freely available for presentation in computerized maps.
Very good performance, avoiding the pitfalls of earlier designs.
Autonomous operation from 12VDC, to support use in cars or all kinds of simple temporary setups.
It should work with practically all FM receivers from say 120MHz up to 450MHz.
It should come with optimized antenna designs for base-operation and car-operation.

Features of Version 2,3:

32 LED pelorus display (compass rose) "wobbled" for better than 5,6 degrees dynamic resolution.
Pelorus freeze after a signal has disappeared. 
Drives an array of 4 antenna elements.
Wide band capability: 25MHz to 500MHz using appropriate antenna arrays.
120MHz to 170MHz coverage using a 145MHz array.
Very good signal conditioning using a high grade digital filter with extremely high, variable Q.
Automatic gain control.
On board loudspeaker, since a lot of small receivers ony offer a headphones output, switching off the internal speaker when used.
Full calibration range, to compensate for all possible antenna configurations and receiver properties.
Antenna check mode, selects the four elements in a slow sequence, and the display shows the selected antenna element. 
Soft control with extensive overlap instead of hard switching of the antenna elements, 
since hard switching produces a lot of noise and unwanted mixing products.
Hardware accuracy: better than 5 degrees.

How It Works:

First, let's get to understand what Doppler shift is.
Everybody knows of the pitch change of a fast moving object, like a passing car.
The motor or horn sound of a car at constant speed seems to be pitched up as long as it approaches, and it pitches down as soon as it has passed.
This is called Doppler shift.

The same doppler effect would occur if the car would be static and you would be running by at very high speed.
Or imagine you would take a microphone and swing it around at its cord.
As long as it swings towards the car, the microphone will pick up a higher pitched tone of that car.
In the opposite part of the swing, it will move away from the car, pitching down the sound of the car.
There's only two points in the circle, where the pitch is correct, one point going from pitched too low to too high (the closest point), and the other point will be found going from pitched too high to too low.

A Doppler Radio Direction Finder does exactly the same thing; it uses a circular arrangement of antenna elements which are activated one after the other in a very fast sequence, simulating one antenna element being swung around in a circle at very high speed.
This means that the received carrier of the tracked radio signal will be moved up in frequency as long as the antenna element swings towards the transmitter, and shortly after that the frequency will apparently be lowered below the true frequency due to the antenna element moving away again.
The antenna array is rotated electronically at around 500 revolutions per second, so a FM receiver connected to this electronically rotated antenna will produce a 500Hz audio tone.
The zero crossings of that tone (plus or minus some phase shift etc. etc. ) will mark the points where the doppler shift is zero.
These points are easy to detect electronically, and after the necessary compensation for phase shift mentioned above, one of these points is used as a trigger to "freeze" the LED display at the correct point in the circle.
You may imagine that the cirular LED pelorus display copies the circular movement of the antenna array and only the LED that is exactly at the trigger point is allowed to light.


The PA8W Version 2.3 Doppler has the following controls:

Threshold, to adjust to the "freezing" level of the pelorus, given a specific output level of your receiver.
Speaker on/off.
Calibration, to make the RDF point into the right direction.
Phase switch, to enhance calibration with 180 degrees shift.
Filter-Q, to set the signal conditioner for fast or slow response.
Antenna check switch, selects the four elements in a slow sequence, and the display shows the selected antenna element.
(In this mode the 500Hz tone is absent so you can listen to the modulation of the tracked transmitter, and check for equal performance of all 4 antenna elements. 
I added this test mode mainly for developing purposes of the antenna arrays, but I kept it in since it presents an easy way to make sure all antenna elements are ok.

The simplified Block Diagram:

Top-left we have the 4-element vertical dipole array, meant for non-mobile applications.
(for mobile applications there's a similar arrangement of 4 magnet mount whip antennas)
The antenna array is "rotated" by the antenna multiplexer, which activates only one element at the time.
From the antenna array, a coax runs to our FM-receiver. The audio output of the receiver is fed into a series of audio filters.
Next, a zero crossing detector marks the point of no phase-shift, and, after a calibration pulse shift section, this triggers a data latch.
This data latch then copies the current state of the address lines, and presents it to the LED driver, which lights the corresponding LED in the pelorus display.
Of course, the entire process is timed by the central clock + address counter.


One could think of a LED pelorus display with 64 or even more LED's to get better display resolution. 
Knowing that a 4 antenna array is capable of offering a maximum accuracy of around 5 degrees, it would make sense to use 64 LED's giving the pelorus a resolution of 5,6 degrees. 
However, this would make the RDF much more complex.
So I designed a 32 LED pelorus with wobbled response, filling the gaps between 2 subsequent LEDs, giving a better than 5 degrees resolution.
 And if you really want more, a hookup to a computer will give all the features you could want, even a direct plot of the transmitters direction on an electronic road map.
Two or more RDF stations can exchange this information over internet to get a cross bearing of the tracked signal!

Keep in mind that the simple 4 element array does not exactly simulate a smooth rotating antenna.
It will produce 4 phase jumps every cycle, or even only two jumps per cycle, worst case, when two elements are at equal distance from the transmitter.
Obviously this is pretty hard for the RDF to analyse with some degree of accuracy.
A 8 element array already does a better job, and professional systems often use 16 or even more antenna elements.

As for practical reasons we will have to stick to 4 or 8 elements max (in my Version3 doppler), 
our RDF will need to have an incredibly good signal conditioner, and that's where the digital filter and additional low-pass filters step in: 
They integrate the incoming audio, they don't allow for fast signal jumps to pass, but they average everything into a sinus-like waveform.
To achieve this the applied digital filter has an extremely high Q. It has a bandwidth of less than 1 Hertz.
The extremely narrow bandwidth will force about any signal shape into a sinusoidal shape, and it
also effectively cuts off modulation on the tracked signal, which would otherwise blur the display severely.
As far as I know this is the only type of filter this narrow, that is not depending on very high component accuracy.
Its center frequency is determined by the clock frequency only.
In this design, the same clock is used to rotate the antenna, and if the clock would deviate, the filter would track that deviation automatically.
No crystal controlled precision necessary. This is why this type of filter is the eureka part of the design!

This signal smoothening process actually fills-in the gaps between the jumps from one antenna element to the next, giving the RDF a highly improved accuracy.
And in fact the softened activation of the antenna elements plus the considerable overlap on every transition from one element to the next also add to that process.
The soft switching has an even bigger advantage: It reduces the level of spurious in the radio band so your reception of the weaker signals will improve by a large amount.
Developing the PA8W Doppler RDF, I experimented quite a bit with a variable timing and overlap for the antenna elements to find the best spot, so I assume we have achieved pretty much the maximum performance one can squeeze out of this simple antenna configuration.

However, we should realize that particularly a fixed (base) RDF accuracy will suffer from objects in its vicinity, 
which may introduce bearing deviations much larger that the intrinsic accuracy of this RDF.
A mobile RDF has the same problem, but since it changes location constantly, the changing deviations will be averaged into a much smaller bearing error.


The following screenshots will illustrate the excellent signal conditioning in the heart of the RDF:

First of all the red wave shows an audio signal right out of the receiver.
It is a bit noisy because it was a rather weak carrier. 
But since it is a clean carrier without modulation, this signal is easy to process by the RDF.
A modulated signal would be total chaos to the eye, and still the RDF is able to pick out what's important.
This is achieved mainly by the digital filter, of which the output is shown in yellow.
It resists to all rapid changes in the signal, so what remains is information that is constant over some period of time, like the doppler information.

After the digital filter two low pass filter stages are responsible for filtering out everything above the doppler frequency.
So the output is only the base frequency, an nice 500Hz sinusoid shown in blue, very well suited for a precise phase measurement.
Note there's a considerable phase shift between the yellow and the blue signal, due to the low pass filters. 
This is of no meaning as long as it is a constant factor, since the calibration will compensate for this automatically.

Accuracy measurement:

I performed accuracy measurements with the base array on a rotatable glassfiber mast.

The, hardware mean accuracy turned out to be around 5 degrees, 
which is in accordance with what has been stated by professionals about 4-antenna doppler RDF's.

Please don't hesitate to email me if you have any comments or questions.

73, Wil.