ASYMMETRICAL PEAK LIMITING
Bob Bruhns, WA3WDR
This article expands upon my earlier articles in the June and July 1988 AM P/Ex on
asymmetrical peak limiting and bass-phase adjustment. An improved version of the
asymmetrical peak limiter (APL) is presented, which offers bass phase adjustment and
high quality variable clipping as well.
PEAK LIMITING
This version of the APL has an asymmetry adjustment. The peak modulation level will
depend on the waveform and the degree of asymmetry selected. If you set up for 2:1
asymmetry, with 200% positive and 100% negative modulation, then sine waves,
sibilents and whistles will modulate to 100% negative, and will typically not exceed
100% positive, because they are symmetrical. With the same settings, an asymmetrical
sound waveform such as your voice could, if properly polarized, modulate up to 200%
positive.
BASS PHASE AND VOICE ASYMMETRY
A variable phase equalizer in this APL advances the phase of the lowest frequency
components of a voice signal in order to optimize the polar asymmetry of the composite
waveform. I call this circuit a bass phase aligner (BPA). Fig. 2 illustrates the effect.
The BPA is unique in that it can adjust polar asymmetry without introducing harmonic
distortion.
CLIPPING
The clipper in this APL is unusual in that it is part of the gain control servo-loop. The
clipping threshold is the same as the peak limiting threshold in each polarity, so it can
take advantage of the full modulation capability of the transmitter. Waveforms with high
peak to average ratios are amplified and clipped harder, while waveforms which already
have low peak to average ratios are not clipped as heavily. High frequency distortion
components resulting from the clipping are attenuated before being introduced into the
audio path, so low pass filtering of the resulting audio is not required.
THE IMPROVED ASYMMETRICAL PEAK-LIMITER
Refer to Fig. 1. Input gain is set by R1. U4a is a buffer. Bass phase is adjusted by R29
(if S3 is closed); the low frequency phase advance produced by C13, R28 and R29 is
doubled by U5a, yet frequency response stays flat. S4 selects input polarity.
U1a can provide over 30 dB of gain. Q1 reduces the effective gain by shunting the
input of U1a. R22, R23, R24 and C11 compensate for any leakage in C1 so that there
is no DC shift in the output of U1a when Q1 conducts.
The output of U1a feeds U3, which drives T1. R9 and R10 set the gain of this stage.
C15 compensates high frequencies, and C16 keeps RF leakage out of U3.
T1 is driven so that the audio voltage at either end of its secondary winding is about two
times the audio output voltage from U1a. The output from T1 is full wave rectified by
diodes CR1 and CR2. With S1 off (open), the rectifier will be equally sensitive to peaks
of either polarity (no asymmetry), while if S1 is switched on (closed), and R34
(asymmetry) is turned all the way up (zero ohms), the sensitivity will be cut in half for
waveform peaks of one polarity (2:1 asymmetry). The resulting full wave rectified
waveform is peak detected by Q2, filtered by C17, R35, C5, R17, C8 and R16, and fed
to Q1.
Q1 (a MOSFET) serves as a variable shunt that works with R3 to divide the audio from
C1 before feeding U1a. Think of Q1 as a voltage variable resistor. A MOSFET will
shunt low AC voltages with relatively little distortion; Q1 will see about 30 mV p-p audio.
For simplicity, imagine S1 is open. If the audio level at CR1 or CR2 exceeds about 3V
peak to peak, the peak detector output causes Q1 to begin to conduct, reducing the
audio fed to U1a. Any further increase in input level will only result in a slight increase
in output level, producing correspondingly greater gain reduction. Thus the output
signal is peak-limited to about 1.5V peak-to-peak.
If S1 is closed, the peak detector will be desensitized to peaks of one polarity, by an
amount determined by the setting of R34. S4 (input polarity) is used to select the input
polarity so that the higher input peaks are fed to the desensitized peak detector. S2
(output polarity) sets the polarity of the output so that these peaks produce positive
modulation.
If R35 (clipping) is turned up, gain control voltage due to instantaneous peaks will
immediately charge C17 and pass through to Q1, but this will not immediately charge
C5. The result is a much faster release time for peaks, so fast that the peak limiting
action begins and ends with the peak. This makes it peak clipping! C17 extends the
release time enough that peaks corresponding to frequencies from about 3 KHz and up
are not clipped, but are peak limited with an extremely short release time. That
attenuates high frequency clipping distortion components and reduces intermodulation
distortion of high frequency sounds such as sibilents. It also reduces blocking of high
frequencies by louder low frequencies that are clipping. See Fig. 3.
R8 sets the output level to the modulator.
SETUP
Line levels to and from the APL are expected to be about 0.2V rms. Higher input levels
can be used by reducing the setting of R1. Higher output levels can be produced by
turning up R8. The modulator is expected to have flat frequency response and good
phase response following the APL. Setup is best monitored with an oscilloscope, and it
can be done over the air if necessary.
Use a voice signal, or another asymmetrical signal, for tests. The APL should be
adjusted to match the modulation capability of your transmitter, then the BPA should be
adjusted to match the asymmetry of the voice waveform to the asymmetry of the APL.
Insert the APL between your audio preamp and your modulator. Set S1 (polar
asymmetry) on (closed). Set S3 (BPA) off (open). Set R1 (input gain) about mid-scale.
Set R8 (output level) for about 50% modulation. Set R34 (asymmetry) all the way up to
2:l (zero ohms). Set R35 (clipping) all the way down (zero ohms).
Now try both positions of S4 (input polarity). One of them should permit larger
modulation than the other; this is the setting you want. Set S2 (output polarity) so that
the higher peaks are in the positive direction. Set R8 (output level) for maximum
positive modulation or 100% negative modulation, etc, depending on your particular
station setup.
R8 (output level) affects both positive and negative peak levels, while R34 (asymmetry)
affects only the positive peak level. If your modulator is not capable of 200% positive
modulation (it probably isn't), then you may want to reduce peak limiter asymmetry by
turning R34 (asymmetry) down. It may then be possible to increase R8 (output level).
If the asymmetry of the voice waveform is not adequate, try switching S3 (BPA) on, and
adjust R29 (bass phase) for the best waveform. Start with R29 turned all the way down
(maximum resistance), and gradually turn it up. See Fig. 3. You may find that you need
to reverse S4 (input polarity)!
Now try the servo-clipper by turning up R36 (clipping). Voice signals will sound
increasingly louder, with little audible distortion until the loudness improvement is quite
substantial. I set R36 (clipping) around the l0:30 position (about 1/3 of the way up).
You may want to adjust R8 (output level) slightly, because you will be better able to see
the peak modulation level after introducing the clipping.
Note that clipping sounds best on a crisp audio signal. If your audio is bassy or muffled,
you should try a bright mike such as a D-104, or add high frequency boost, or both.
Once the APL is set up with a transmitter, you should use only S4 (input polarity) to
reverse phase as you change mikes or play with equalization, bass phase and clipping.
Close-miking is advisable, for several reasons: room acoustics are usually bad, room
reverberation reduces articulation when clipping is used, directional mikes sound
"warm" up close, and RF, hum and noise problems are minimized when you reduce
preamplifier gain.
A VERY SUBTLE SETUP ADJUSTMENT
With heavy compression, this APL produces a soft rounding down of peaks in one
polarity, increasing or decreasing asymmetry by as much as 10%. The polarity that is
rounded down can be selected by moving the connection of S1 between R13/CR1 and
R14/CR2. If you change this, you must also reverse input and output polarity.
My own station can only provide about 150% modulation, so I set up my APL so that
this effect reduces positive modulation. The effect is subtle; unless you have an
oscilloscope, don't bother with it.
IN CASE OF TROUBLE
The audio chain following this audio processor must be of good quality. Significant
phase distortion or frequency response variations following the compressor will defeat
its purpose. Equalization should be done in the preamp feeding the APL.
If square waves from the clipper are canted due to restricted modulator low frequency
response, try decreasing C4 and C10 to minimize the canting. See Fig. 3.
In the crowded HF bands, amateur AM sounds best when it has bright, crisp audio. This
is especially true if clipping is used. If your audio is bassy or muffled, try a bright mike
such as a D-104, or boost highs, or do both.
If you expect to achieve voice-polarity supermodulation, your transmitter and modulator
must be able to handle the large peak powers involved. If they can't, it won't work.
This processor has a lot of audio gain. If you get RF leakage into your audio, if your
modulator "talks back" to you, if you have loud background noises, etc, you will have
problems. Shielding, filtering, grounding and room acoustics become more important
when advanced techniques such as compression and equalization are used.
IF YOUR MODULATOR IS NOWHERE NEAR FLAT...
... you can still have APL by using the circuit shown in Fig. 4. It may also be necessary
to increase C17. R8 will no longer control the negative peak levels it will act as a gain
control, and distortion will result if it is set too low or too high. You may have to move S1
from CR2/R14 to CR1/R13 for proper APL action.
CONVERTING A TUBE TYPE PEAK LIMITER
Use push-pull peak limiters only. With push-pull audio and common-mode AGC, control
voltage changes are balanced out of the audio, permitting the extreme control speed
required for good peak limiting and proper operation of the servo-clipper.
Simply add a resistive voltage divider (see Fig. 5) as necessary to reduce peak detector
sensitivity to one side of the waveform. You can add servo-clipping by inserting the
equivalent of R35 and C17 in series with the main AGC loop filter capacitor.