Comments about the "Averaging"
and its comparison to the switched-capacitor filter:
Note: Version 7A of the alternate firmware has the ability to make this filter an adaptive filter: See below for more information.
The existing filter:
The Montreal Doppler 2 unit uses the MAX492 dual rail-to-rail op amp as a 2-section high-Q "peaking" filter. The main purpose of this filter is to pass energy in the general area of 500 Hz and amplify it while rejecting energy elsewhere in the audio spectrum. Because the amount of energy at that frequency is relatively low compared to the cumulative amount of energy "everywhere else" in the audio spectrum, this filter has quite a bit of gain (over 50 db) at the center frequency to make it a bit easier to bring the audio up to several volts peak-to-peak and extract the phase information from the received signal.
In the presence of a lot of modulation, mobile flutter, and other noise the 500 Hz information can be somewhat "masked." This often has the effect "noising up" the bearing information, causing it to vary in a seemingly random way.
Under these conditions the quality of the bearing may be ameliorated by increasing the number of samples taken for each reading. This, of course, slows down the updates of the bearing (averaging 32 bearings results in a new reading every 1.5 seconds or so.)
The effect of this integration is substantially the same as narrowing the peaking filter's bandwidth. If an extremely narrow filter is used, however, the update rate can remain the same (although the speed of response to change in bearings is slowed in a manner commensurate with the filter's bandpass) but the effect is more that of a "sliding average" rather than an update occurring in "chunks." In actual use, the effect is similar to averaging, but the response is somewhat quicker.
This particular design of switched capacitor filter has a long legacy: It has been an integral part of the "Roanoake" design. It is quite simple- yet very effective, and its "Q" (or "Damping") may be easily adjusted without affecting the center frequency or the filter's delay. Perhaps most important of all, it's center frequency is precisely dictated by the clock frequency driving it: If the driving frequency for the filter is derived from the same source as the switching clock for the antennas, then the two will track exactly.
Why add this filter? It provides an added degree of flexibility: It has the effect of "smoothing" erratic bearings (cause by modulation or noise) but still maintaining a reasonably high update rate - faster than comparable "smoothing" using the averaging. It does have the disadvantage, however, of reducing the ability of the unit to quickly detect when a signal drops, owing to the extremely high Q of the filter: The "ringing" of the filter maintains an output signal for some time after the original input signal goes away. (Jacques has added such a filter to his Doppler-III unit.)
How it works:
Refer to the hand-drawn schematic.
Major circuit revision: This circuit was revised on 1/13/05 to accommodate two major changes:
Because the center frequency of the switched-capacitor filter is
by its clocking frequency - and because this filter must by
to the antenna's rotation speed - a suitable clock must be
The 'EMM Doppler-II wasn't originally designed for this, so a suitable
clock must be generated:
Frequency synthesizer (required when using the original firmware):
The synthesizer circuit is required only if using the original Doppler II firmware: The alternate firmware can output a clock that eliminates the need for the frequency synthesizer.
This circuit operates from the main supply voltage - filtered by R8 and C9. A "mid-supply" voltage is supplied by the +5 volt regulated supply on the 'EMM board - and this is filtered using R3 and C5, and a "virtual ground" at this voltage is thus created.
The 500 Hz antenna switch signal (from any of antennas 1-4) is applied to pin 14 of U3 (a 4046) via C1. A VCO on-board U3 operates at 8 times the original frequency, or 4 KHz. The VCO output is applied to U4, a 4040 binary ripple counter and the divided-by-8 output from U4 is fed back into the other input of U3 via pin 3 thus forming a phase-locked loop, with C2 and R9 as the frequency-determining components and R1, R2 and C4 form a loop filter. The ultimate result is a 4 KHz signal firmly locked to the original 500 Hz antenna switching signal.
Using this filter with the alternate firmware:
If the alternate firmware is used and the appropriate modification is made, a pulse train from Pin 9 of IC3 (the main processor) that clocks the filter, completely eliminating the need for U3, the frequency synthesizer. This signal is applied to U4 (through a level shifter using an NPN transistor) for filter clocking.
A minor modification is required to obtain the clocking signal from the alternate firmware. Details of this of this modification may be found on the Modifications for new firmware page.
The filter itself:
U4 also provides a 3-bit binary counting output that is applied to U2, a 4051 8-channel multiplexer - the heart of the switched capacitor filter. How this filter works may be a bit difficult to understand - but here goes: Assume that the input signal is exactly 500 Hz. For every cycle, each of the 8 capacitors (C10-C17) is connected to the "virtual ground" for 1/8 of the cycle (500 times-per second, for 4000 Hz / 8 = 500 Hz.) Because - in this example - the input signal is exactly the same frequency as that that the capacitors connected to U2 are "scanned" each capacitor sees the same portion of the input 500 Hz signal each time - and thus, it charges to the voltage of the waveform each time.
Now, suppose that the inputted audio frequency were "slightly"
In this case, the charge of the capacitors would change on each
cycle. If the frequency was "close" then the voltage wouldn't
particularly fast and the capacitors could respond. If it were
off frequency by a significant
the capacitors' charges wouldn't be able to be changed quickly enough
and as the frequency moved farther away from center, each capacitor
would get less and less charge and hover near "zero."
In this circuit, R5 adjusts the time constant associated with capacitors C10-C17: The lower the value of R5, the the faster the capacitors may be charged and thus, the wider the frequency response (and the lower the "Q".) In this circuit, R5 is used to set the actual bandwidth - with the lower limit of Q being set by R6. This lower limit is necessary to maintain circuit stability.
The voltage on the charged capacitors (the filter) is then buffered by U1B - another unity-gain buffer amplifier. The circuit consisting of C19 and R10 blocks the DC voltage from U1B and biases the voltage to 2.5 volts, for the input of IC3.
A few comments:
Setting R5 to its minimum value does, for all practical purposes, remove this filter from the circuit (that is, it will have little or no effect) and because of this, a "bypass" switch is not provided. This is likely to be the "normal" mode of operation, as the "quality" indication operates very quickly and functions as originally designed. When the "Q" control is set to maximum, the 3db bandwidth of this filter is approximately 0.5 Hz and when properly constructed, changing the "Q" does not change the amplitude or phase of a signal centered in the passband (such as the switching signal.)
When trying to resolve very weak or noisy signals, a high Q is helpful to further stabilize the bearing indications - without slowing the display update as much as simple averaging might. It does have the potential problem in that with high a high Q, the voltage response is relatively slow and the filter output persists for a short time (a second or so) after the signal has disappeared - which has the effect that several (bogus) readings may be produced under no-signal conditions before the filter's output drops below the "quality" threshold. Note: The alternate firmware provides a means to quickly detect the disappearance of a signal and instantly stop display updates, "freezing" the last bearing obtained when the remote transmitter unkeyed (for example.)
Will this filter make "good" bearings out of "bad" ones? It
depends on what made them "bad" in the first place: If the signal
is very "multipathy" the answer is probably no - the signal
already degraded (although this isn't always true if either the source
and/or you are in motion.) If the difficulty in obtaining
a bearing is actually a result of the signal just being weak (noisy) or
very heavily modulated then this extra filtering can make all of the
in the world - with the obvious sacrifice of response time.
An "Adaptive" filter
(Version 7A and newer of the alternate firmware only):
One advantage of this filter is that it may be adjusted to provide a
very fast response or a slow response. A fast response (low
damping or "Q") is often desirable when the signals being monitored are
of fairly good quality while a slow response (high damping or "Q")
provides better filtering of noise and modulation and can be helpful
weak or noisy signals.
Occasionally, one may wish to use the "slow" filter but encounters
difficulty when trying to determine the bearing of signals that last
only a short time, such as a "kerchunk." If a slow response is
used, the "q" of the filter may be such that it does not register the
signal before it is gone. Another problem is when a
short-duration signal such as this immediately follows another
signal: In this case, the filter will still "contain" the bearing
from the previous signal and may not be able to read a correct bearing
until enough time has passed for the "old" bearing to be purged from
This "adaptive" filter option causes a "fast" mode to always be
automatically selected when the signal goes away, reverting back to the
selected response time shortly after a signal is detected. In
this way, the "old" signal is purged from the filter immediately and
when the new signal appears, the filter is "charged" up for a brief
moment (between 50 and 100 milliseconds) at the "fast" rate before the
"slow" rate is re-enabled once again.
With this feature, it is more likely that even very brief
transmissions (especially if they closely follow other transmssion) can
be analyzed - even if the damping/Q of the switched-capacitor filter is
set to a fairly high value (e.g. a "slow" filter.) For more
information, see the section on Adaptive Filtering
in the New Firmware manual as well as this brief description of
This modification involves the isolating of pin 5 of the main
processor on the Doppler II to disconnect it from the +5 volt
supply. Once this is done, the firmware detects that this
modication is present and when the "Average Clear" indicator is active (see
Filtering" section in the New Firmware manual) this pin will be
set at +5 volts. This signal may be applied to a 4066 analog gate
(after necessary logic level conversion if the switched-capacitor
filter described above is not being operated at 5 volts) that, when the
5 volt signal is present, "shorts out" R5 (the "damping" control) in
the schematic above. Note that once this is done, it is likely
that C18 (the 150-180 pF capacitor) will have to be readjusted to
accomodate the added capacitance of the 4066.
More specific circuit details of this modification will soon
follow. If you would like more information now, use the email
Note: Neither the author or UARC officially endorse any vendors mentioned above. The level and satisfaction of performance of any of the above circuits is largely based on the skill and experience of the operator. Your mileage may vary.
Note: This page (and other pages on this site) are not "official" pages of VE2EMM. These pages are simply set up to aid those who have built or might build the described equipment.
Do you have any questions on this or other DF-related topics? Go here.Return to the KA7OEI ARDF Page.
This page updated on 20060510