Any experience with doppler rdf (radio direction finders)?
Philip Kahn
has an article that says they can only do well to about 5 degrees.
Have you heard of systems or ways to do it that gives better results?
thanks,
phil...
Gary Coffman
Doppler systems theoretically can give better results than that, but
the practical limitations of precision antenna arrays, the switching
speed, precision reconstruction filters, and the radio's response all
conspire to reduce practical accuracy to around 5 degrees.
For best accuracy, you need to:
1) Space the antennas as precisely as possible in a square. For 5 degree
accuracy, spacing errors must be less than 1.3%, or about 0.24 inch at 2
meters. One degree accuracy would require array errors be held below 0.053
inch at 2 meters.
2) The higher the switching speed, the greater the doppler, and the easier
small angle changes are to measure. So you want to use as high a switching
rate as you can manage. This also relates to 3 and 4 below.
3) The reconstruction filter must accurately smooth the discrete antenna
switch samples to a sinusnoid so that the precise point of negative going
zero crossing can be determined. A 4 pole antenna array only gives 4
sample points to define the waveform. Theoretically that's enough, but
it's easier with an 8 or 16 antenna array.
4) The wider and flatter the phase bandwidth of the radio, the more precisely
it will pass the doppler shift information on to the resolver.
This latter is a problem when using typical ham receivers, so switch speeds
are typically held down to 2 kHz or less for a 4 pole array. This corresponds
to a rotational rate of 500 rps. A cumulative error of 0.27% in the array,
in the receiver passband phase response, in the reconstruction filter phase
accuracy, and in the resolver will allow a 1 degree readout accuracy. That's
tough to accomplish in practice.
Gary
--
Gary Coffman KE4ZV | You make it, | gatech!wa4mei!ke4zv!gary
Destructive Testing Systems | we break it. | uunet!rsiatl!ke4zv!gary
534 Shannon Way | Guaranteed! | emory!kd4nc!ke4zv!gary
Lawrenceville, GA 30244 | |
Tom Bruhns
: I have been reading up on doppler RDF's. The Amateur Radio Handbook
: has an article that says they can only do well to about 5 degrees.
: Have you heard of systems or ways to do it that gives better results?
I've been using the "Dopplescant" described in May 78 QST for about
that long, off and on, and have a couple comments kind of tangent
to what Gary C. noted in another followup:
First, though you can build a doppler system that accurately reads
apparent direction with better than 5 degree accuracy, I see little
point, at least for the typical bunny hunt where the RDF receiver system
is mobile (vehicle or on foot). Multipath makes it quite unlikely
that better than 5 degree accuracy at the receiver is worthwhile, and
the mobility means that in practice, you move toward the bunny and
simply don't worry about small errors.
Second, though you can undoubtedly achieve high accuracy with
careful construction, another way that should be as applicable to
doppler antennas as to many other areas is calibration. For a given
system, if you understand how the errors will affect the readings
and calibrate often enough and at enough points (which could be one
thorough calibration and an occasional single-point check), you can
get the errors quite low. You can't resolve ambiguities that way,
so the raw (uncorrected) output has to have an appropriate d(out)/d(theta)
at all angles.
Finally, for best results, use equipment with adequate basic
performance, and spend some time learning to use it under a variety of
conditions. You will find that the best results are obtained by
using multiple techniques and using your head to understand what the
several sensors are telling you.
73, K7ITM
Jeff Liebermann
In 1976, I helped design the AN/SRD-22 doppler direction
finder for Intech. The USCG specification was 1 degree
resolution. 3 degree total variation over a wider temperature
range. This is radically better than the typical Roanoak
or Doppler systems design. This accuracy was achievable,
but in my estimation worthless. Reasons to follow.
In order to maintain a stable phase shift from the antenna,
through the receiver, and finally to the inevitable phase
comparator, some compensation is required. The following
are the basic error sources and the solutions required.
The antenna is 4 vertical tubes with pin diode attenuators
(not switches) in series. These attenuators are driven
in quadrature with a modified sine wave to give a low
distortion diode resistance vs time waveform. Any distortion
in the drive waveform cause errors at points BETWEEN quadrature
points.
The largest cause of phase errors in the receiver is the IF
crystal filter. The AN/SRD-22 uses a 4KHz drive signal where
the 2nd harmonic and above (i.e. any distortion) ends up outside
the IF crystal filter bandwidth. Slight changes in frequency
yielded monsterous changes in phase shift through the filter.
The receiver needed an automatic frequency control to insure
that the ADF tones would always land on the same place in the
IF xtal filter.
With weak signals, it was found that signal strengh affected
the phase shift through the receiver. The final limiter and
quadrature demodulator was the largest contributor. An
automatic gain control was required to fix this.
The demodulated 4Khz tone was filtered after demodulation by
a commutating filter. The initial bandwith was about 1Hz.
Clock leakage and switching transients became a large source
of errors. This required very carful board layout.
To give a rough approximation of the distortion requirements,
1 degree total accuracy is 1 part in 360 or 0.28% total error
or distortion. At 1Hz bandwidth, it may take 20 seconds to
obtain a stable reading.
The reason the Roanoak and Doppler Systems switched antenna
systems function is that the environmental sources of error
far exeed the instrument errors. A 130ft vessel or moving vehicle
is NOT a stable platform and is subject to errors induced by
ground or sea reflections, multipath, antenna tilt, polarization,
faraday rotation, mechanical damage, and multiple transmitter
sources. Any ONE of these can induce enormous errors. Multiple
transmitters (including intermod, garbage, grunge, rfi, emi,
computer noise, skip, etc) are the worst as they generate totally
false readings. Another major botch is the 0-360degree digital
display. Since the readings tend NOT to be stable, getting
an accurate bearing is like reading a digital watch that can't
decide what time it is. Error induced by drive waveform distortion,
signal levels, and such are reasonably small. There are other
problems introduced by switching type antennas, but they do not
affect the basic accuracy. Only the crystal filter errors are
large enough to justify an AFC. My guess is that +/-5 degree
overall accuracy for a fixed base is the best that can be expected.
+/-15 degrees is about the best for a mobile.
This technology is now 20 year old. Were I to do it today, I
would use DSP to filter and demodulate the tone, statistical
algorithms to reduce erronious readings, beam steering to deal
with reflections, adjustable filter bandwidth, and intelligent
display control.
--
# Jeff Liebermann Box 272 1540 Jackson Ave Ben Lomond CA 95005
# 408.336.2558 voice wb6ssy@ki6eh.#nocal.ca.usa wb6ssy.ampr.org [44.4.18.10]
# 408.699.0483 digital_pager 73557,2074 cis [don't]
# je...@comix.santa-cruz.ca.us scruz.ucsc.edu!comix!jeffl
Jeff Liebermann
>2) The higher the switching speed, the greater the doppler, and the easier
>small angle changes are to measure. So you want to use as high a switching
>rate as you can manage. This also relates to 3 and 4 below.
One problem with switching as opposed to sine wave driven attenuation
is that switching creates a "comb line" of modulation spectra that
extends well beyond the bandwidth of the crystal filter. Put
differently, 33.3% of the energy in a square wave drive signal is
in the odd harmonics of the fundamental. This results in signifigant
spurious responses in the adjacent channel area. Doppler Systems
used DG-MOSFETS with a "soft" switching characteristic to reduce
these spurs.
One of the major reasons that amplitude based, rotating yagi or
quad antenna direction finders are becoming popular is that they
can seperate multiple carriers on the same frequency. Another
advantage is that they can visibly show a false reflection or
multipath signal. The doppler direction finders go insane when
faced with multiple signals. Rotating quad do not have the aquisition
speed of a doppler system, but potentially are a more accurate and
better system.
If you are interested in maximum accuracy, I suggest a "Lorenz"
style antenna system. This was used by the Germans during the
Battle of Britain for guiding the bombers to their target. Two
directional antennas were aimed in the general direction of the
target. The two antennas had a -3db beamwidth of about 15 degrees
and were aimed 15 degrees apart. A single transmitter was alternately
tone modulated and switched between the two antennas. When switched
to the left antenna, it would send morse "A". The right would be "N".
The dots and dashes were inteleaved. If the bomber was exactly between
the two beams, the signal strength from the two antennas would be
identical and a continuous modulation tone would be heard. Any
deviation from centre would cause the letter "A" or "N" to be heard
depending upon direction. The beam was said to be 500ft wide at
a distance of 100 miles on about 60Mhz. See "The Wizard War" by
R.V. Jones.
A similar antenna derrangement could be constructed and rotated.
Instead of generating the signals, the direction finder would be
a sychronous antenna switch and AM demodulator. When the signal
levels are equal, the carrier is half way between the antennas.
The Intech AN-SRD/21 homer direction finder worked on similar
principles, but without a rotating or directional antenna. Instead
the entire vessel was rotated until the signals strengths were equal.
A 1 degree resolution was easy. However, mechanical considerations
prevented achieving this on a 43 ft vessel. +/-10 degree was typical
and adequate.
>A 4 pole antenna array only gives 4
>sample points to define the waveform. Theoretically that's enough, but
>it's easier with an 8 or 16 antenna array.
That's what I thought until I tried it. The problem is that the
"unused" (i.e. the elements that are turned off) get in the way.
A 16 element array worked just fine when both the transmitting and
receiving antennas were exactly vertically polarized. Tilt one
antenna slightly, and other elements acted as a polarization filter
and drastically reduced the received signal. In the "attenuator"
flavour of antenna (as opposed to the switched flavour), the
resultant distortion and noise was intolerable.
>4) The wider and flatter the phase bandwidth of the radio, the more precisely
>it will pass the doppler shift information on to the resolver.
True. However, the wider the IF, the better chance a signal on an
adjacent channel will mangle the readings.
>This latter is a problem when using typical ham receivers, so switch speeds
>are typically held down to 2 kHz or less for a 4 pole array. This corresponds
Nope. The reason is that the closer the tone is to the center
carrier frequency (i.e. lower modulation frequency), the less effect
the effects of the carrier being off frequency. Group delay
(phase errors) increases toward the IF filter band edges and
are flattest in the middle.
Another interesting method of direction finding is to use the
same technique as the satellite (forgot name) which monitors the
ELT (121.5Mhz) frequency. It uses doppler shift for locating.
The satellite follows an exactly known path. A transmitter on the
ground creates a doppler shift that changes from high to low as
the satellite passes. The rate of change during this transition
can be used to locate a line of position.
The same method can be done on the ground. I visualize a GPS
receiver into a laptop. The received signal and doppler shift
are measured exactly while roaring down the freeway. Increased
shift means you're approaching the transmitter. Decreasing means
you're going away. The carrier frequency, doppler shift, direction,
and speed are all known. The rest is number crunching.
Ugh. Back to taxes...
Gary Coffman
>In article <1994Apr11.1...@ke4zv.atl.ga.us> ga...@ke4zv.atl.ga.us (Gary Coffman) writes:
>
>>4) The wider and flatter the phase bandwidth of the radio, the more precisely
>>it will pass the doppler shift information on to the resolver.
>
>True. However, the wider the IF, the better chance a signal on an
>adjacent channel will mangle the readings.
>
>>This latter is a problem when using typical ham receivers, so switch speeds
>>are typically held down to 2 kHz or less for a 4 pole array. This corresponds
>
>Nope. The reason is that the closer the tone is to the center
>carrier frequency (i.e. lower modulation frequency), the less effect
>the effects of the carrier being off frequency. Group delay
>(phase errors) increases toward the IF filter band edges and
>are flattest in the middle.
What I said.
Re Wizard War, those arrays were *huge* fixed arrays to get the
necessary narrow beam. I don't think they'd be much use in a car.
>Another interesting method of direction finding is to use the
>same technique as the satellite (forgot name) which monitors the
>ELT (121.5Mhz) frequency. It uses doppler shift for locating.
>The satellite follows an exactly known path. A transmitter on the
>ground creates a doppler shift that changes from high to low as
>the satellite passes. The rate of change during this transition
>can be used to locate a line of position.
>
>The same method can be done on the ground. I visualize a GPS
>receiver into a laptop. The received signal and doppler shift
>are measured exactly while roaring down the freeway. Increased
>shift means you're approaching the transmitter. Decreasing means
>you're going away. The carrier frequency, doppler shift, direction,
>and speed are all known. The rest is number crunching.
Your car must be faster than mine. The SARSAT is moving at 5 km/s
to get that doppler. Poking along at 0.027 km/s, I'm not going to
generate much doppler shift at 2 meters, fractional Hertz. That's
not going to pass through a typical amateur receiver, nor are typical
PLLs that stable.
Any system that won't work for that bane of the repeater owner, very
brief bursts of interference that kerchunk the machine, isn't going to
cut it as a mobile direction finding technique. The switched doppler
displays do work for that case, giving at least a general bearing on
the rosette. Mechanical arrays aren't generally quick enough to catch
the burst, and any system that depends on a continous pure carrier is
out the window. (Typically the offender is going to be a late model
Motorola repeater that keys while the PLL is still sweeping into lock.
Hunting them down seems to be a regular passtime around here.)
Tom Bruhns
: Any system that won't work for that bane of the repeater owner, very
: brief bursts of interference that kerchunk the machine, isn't going to
: cut it as a mobile direction finding technique. The switched doppler
: displays do work for that case, giving at least a general bearing on
: the rosette. Mechanical arrays aren't generally quick enough to catch
: the burst, and any system that depends on a continous pure carrier is
: out the window. (Typically the offender is going to be a late model
: Motorola repeater that keys while the PLL is still sweeping into lock.
: Hunting them down seems to be a regular passtime around here.)
Direction finding is certainly part of the passtime. A method that
also has promise is "fingerprinting," characterizing the brief
burst in ways that differentiate it from other sources. A local
repeater here does this; I believe the system is available for
purchase (see small ads in the back of recent QSTs). If you want
to get really serious about this aspect of the solution, my
employer sells an instrument that can time-capture a tremendous
amount of raw data and allows analysis of the carrier and modulation
to your heart's content. The carriers and tone modulations have
distinct amplitude and frequency characteristics as a function of
time that differ from one transmitter to another, even for the same
model. (I believe the operator of the repeater mentioned above
can even lock out responses to particular transmitters based on
this information. You "kerchunk" a few times, he locks you out,
and you will need a different transmitter to access his repeater--
or get your apology accepted.)
Sorry for the drift from the original topic.
steve Butler
: Any system that won't work for that bane of the repeater owner, very
: brief bursts of interference that kerchunk the machine, isn't going to
: Gary Coffman KE4ZV | You make it, | gatech!wa4mei!ke4zv!gary
: Destructive Testing Systems | we break it. | uunet!rsiatl!ke4zv!gary
I've wondered why the GPS technology can't be used in reverse. Put 3-4
stable receivers up on hills about 30-50 km apart (maybe even 100 km if
the hills are high enough). Clock the anomolies of the incoming signal
with something akin to an atomic clock and forward from all receivers to
a centeral processing point. The anomolies would include 1) start of a
signal, 2) stop of a signal, 3) non-repeating characteristics (DTMF
start/stop), 4) etc.
Time stamp the anomolies at each repeater. The centeral processor could
command the receivers to another frequency to do a quick check of the
propogation characteristics (the central site transmits a known signal,
each receiver time stamps it, then each receiver transmits in sequence
with the other receivers timestamping. All timestamps get sent to the
central processor. The central processor knows everybodies exact
location and can thereby deduce the clock timing differences for each
site. Ergo, we now have adjusted timings for the original signal.
It falls down to simply solving a set of simultaneous equations involving
the locations of all receivers and the timestamps of the anomolies.
Given that the unknowns are X0, Y0, and T0 (coordinates of LID and times
signal was transmitted) with knowns X1, Y1, T1 (coordinates of receiver 1
and time signal arrived there), X2, Y2, T2 (ditto #2), and X3, Y3, T3 (#3)
you get these three equations (assuming only three receivers):
(X1 - X0)^2 + (Y1 - Y0)^2 = (T1 - T0)^2 * c^2
(X2 - X0)^2 + (Y2 - Y0)^2 = (T2 - T0)^2 * c^2
(X3 - X0)^2 + (Y3 - Y0)^2 = (T3 - T0)^2 * c^2
Three equations, three unknowns ==> should be solvable. After three
pages of hand written stuff I've dropped a term or two. Anybody care to
simplify?
--Steve Butler, KG7JE
Dick Bingham
If you want to know about the 'fingerprinting' system contact Phil
Ferrell - K7PF. He holds the patent and runs the Seattle Repeater
(146.28/88). He also mfgs and sells these systems.
de w7wkr