|
Home
Up
Updates
Current Products
Prior Products - no longer available
Documents
Book
Software Updates
Softrock Lite 6.2
Adventures in Electronics and Radio
| |
|
|
Peak and Average AGC
and AM Mode
Revision History
04 March 2009: Prepared
05 March 2009: Added modulation percentage / peak formula
05 March 2009: Corrected typographical errors found by Dave, G3TJP
In working on measuring my Elecraft K3 receiver in AM
mode, I ran across something I didn't understand. After reflecting upon it and
making measurements of other receivers, I now understand what I observed.
Whether the observations on this page make a real
difference in AM recovered audio quality I don't know. I do know that when
measured with single tones via a signal generator, there's a major
difference.
Let's start with fundamentals. Suppose we have an AM
receiver connected to a signal generator with variable modulation depth (percent
modulation). We connect the receiver's speaker output to a dummy load (in this
case, a GR 1840A audio wattmeter) and also connect an AC voltmeter (I used an HP
3468 digital multimeter) across the speaker load, as illustrated in the block
diagram below.
|
 |
In a theoretically perfect system, with no implementation errors in the receiver
and test equipment, the audio output level is precisely linearly related to the
modulation percentage. If, for example, 100% modulation results in the voltmeter
reading 1.00 volts, reducing the modulation percentage to 50% causes the
voltmeter to read 0.500 volts. Setting the signal generator modulation depth to
10% corresponds to 0.100 volts etc. This relationship is illustrated below.
|
|
Notional relationship between receiver audio output and
AM modulation depth.
|
 |
This sketch shows what I expected to see when I ran the test on a receiver. What
I actually observed is plotted below.
|
 |
Interesting, isn't it?
- At 0 modulation, there's essentially no voltage
output, as it obviously should be.
- At 100% modulation, I've set the two curves to
match--by "normalizing" the voltage outputs to be 1.000 regardless of what
the actual voltmeter reading is, so that different receivers, different
volume control settings and the like can be put into a consistent form for
comparison. (The HP8657A signal generator I use has a maximum modulation
depth of 99%. This is sufficiently close to 100% that we will ignore the
difference.)
- At all other modulation depths, however, the K3's
audio output is significantly greater than it should be, based on how AM
works. With the signal generator set at 40% modulation, the audio output
level corresponds to a perfect receiver with an input signal modulated at
60%. This means the recovered audio output is non-linear with respect to
this particular test setup. (We'll consider later on this page whether this
test setup corresponds to real AM signals.)
To remove the possibility of test equipment malfunction, I
checked the HP 8657A signal generator's modulation depth calibration using a
spectrum analyzer to compare the carrier level and the modulation sideband level
and compute the modulation depth from the ratio of carrier to sideband level. As
the plot below reflects, the 8657A shows no significant error. |
 |
I decided to look at several other receivers and see if they
show the same characteristics:
- Icom R-7000 (25 MHz - 2000 MHz receiver)
- Kenwood TS-940 HF transceiver
- Elecraft K3 with AGC on and AGC off
- Heath AJ-29 high fidelity AM/FM tuner
The plot below shows the K3, Icom and TS-940 share the
"bowed" response. The AJ-29, in contrast, is quite close to the
theoretical perfect response. When the AGC is turned off, the K3's response is
also quite close to the theoretical response, with the divergence at low
modulation percentages being due to noise. (I used the 13 KHz bandwidth filter
and in order to keep the K3 in the linear range, a relatively weak signal
level.) |
 |
|
The fact that the K3 looks much closer to the theoretical
response when the AGC is disabled is, as they say, a "clue" or in the
alternative spelling of our friends across the Atlantic, a "clew."
(Dave, G3TJP, writes to say he has never run across this use of "clew" so the
time or two I've run across it in reading British mysteries likely represent an
intentional use by the author of the word.)
Traditionally, AM receivers use an "average signal" based AGC
system, whereby the AGC level—and hence the receiver's gain—is adjusted based on
the average value of the signals within the AGC passband (which may differ from
the main passband in some designs.) SSB receivers, however, use a peak-based AGC
system, with the gain being set by the peak signal level within the passband.
Why does this make a difference? Consider a signal AM
modulated at 100% as illustrated in the oscilloscope image below. This signal is
close to 100% modulated with a 1 KHz sine wave. The peak value of the signal
(peak-to-peak) is 1.284 volts. |
 |
|
Now consider the same level carrier, but with the modulation
depth at 0, i.e., no modulation. It has a peak level of 652 mV. In
theory, the 100% modulated signal should have peak of twice the un-modulated
signal. Our measured data has a modulated peak of 1.97 times the un-modulated
value, which is close enough to 2.00 for our purposes. |
 |
|
Now, what is the average of both these envelopes? In fact,
they have an identical average. In the symmetrical about zero waveforms shown,
both have an average value of 0. (If your aesthetic sensibility is offended
generating AGC voltage from an average signal level of 0, imagine a constant
offset of some voltage to the waveform such as might result from rectifying the
carrier and low pass filtering the result.) The
difference between average and peak in these waveforms and their corresponding
AGCs have a profound difference in receiver gain when an AM signal is received:
- An average-based AGC system will keep the
receiver gain constant for any modulation level between 0 and 100%, because
the average value of the AM envelope is constant regardless of the
modulation percentage. (This assumes the modulating waveform is symmetrical,
of course.)
- A peak-based AGC system, such as is required
for SSB, will reduce the receiver gain by 6 dB as the modulating percentage
increases from 0% to 100%. This is because the peak power in the AM signal
increases proportionally as the modulation depth increases. Hence, a
peak-based AGC system reads the increased modulation depth as a stronger
signal and hence correspondingly reduces the receiver's gain.
To verify that the three receivers with peak-based AGC
reduce the gain 6 dB as the AM signal modulation depth increases from 0 to 100%,
I developed a test setup using a second "probe" signal using the setup
illustrated below. The probe signal is an unmodulated carrier offset 1.5 KHz
from the AM modulated generator, at a level 30 dB down from the modulated
signal. The audio output is then viewed with an HP 3562A Dynamic Signal Analyzer
(low frequency spectrum analyzer). The strength of the 1.5 KHz beat note between
the two signal generators will be directly proportional to the receiver's gain.
|
 |
The plot below shows the HP 3562A's output for the K3, with the AM source
modulated with a depth of 50%. The probe signal is at 1.5 KHz with a level of
-64.46 dBVrms (decibels below 1 volt RMS.) |
 |
|
A receiver with a peak-responding AGC system will, in theory,
cause a gain reduction as modulation percentage increases and a gain
increase as the modulation level decreases. The relationship between the peak
value and modulation percentage can be easily determined as:
Peak = 1 + Modulation Percentage/100
The 100 divisor is necessary to change the modulation
percentage to the range 0..1. The leading 1 is the un-modulated carrier
level.
Expressing this in dB and flipping the range to reflect
gain increase with reduced modulation depth. the table below shows the expected
gain increase as modulation depth decreases from 100% to 0. (These are voltage
values, so the decibel conversion uses 20log10, not 10 log10.)
|
AM Modulation % |
Gain Increase (dB) |
|
99 |
0.04 |
|
50 |
2.50 |
|
25 |
4.08 |
|
12 |
5.04 |
|
10 |
5.19 |
|
0 |
6.02 |
|
The figure below shows the measured gain increase for three receivers as
modulation depth reduces below 100%.The K3 and R7000 are relatively close to the
theoretical reduction values, with the K3 being within 1 dB or so and the R7000
within 1.5 dB. The high fidelity AJ-29 tuner, as well as the K3 with AGC off
show much less gain variation with signal level. (In theory, with the AGC off,
the K3 should show no gain changes.) |
 |
|
Now, the question to be resolved is whether the peak-based
AGC causes distortion or audio artifacts when compared with either the AGC off
or, preferably, an average-based AGC system. One
argument says that a voice modulated AM transmitter will have audio peaks at
100% occurring frequently enough such that a fast attack, slow decay AGC system
will more or less act like an average-based AGC system. Where receiver testing
is performed with a signal generator as described in the first setup, an
artificial condition exists as a single tone level persists for some tens of
seconds during which time the receiver's AGC reacts.
On the opposite side of the equation is the
unquestioned fact that an average-based AGC system will provide close to perfect
AM demodulation, without regard to how often voice peaks occur or what
modulation level the peaks are. In essence, this view says that 70 years of AGC
design in AM receivers use an average based system, so there must be a reason
for it.
In the case of a DSP-based receiver such as the K3, a
logical AM AGC system would be average based, with the gain set so as to provide
a reasonable amount of "headroom" to allow for modulation peaks. At a minimum, 6
dB of headroom is necessary, but some AM broadcast stations operate with greater
than 100% positive modulation. with 125% being the current FCC maximum. (100%
remains the maximum negative modulation limit, as the carrier cannot be reduced
below zero watts on negative modulation peaks.)
One of the most highly regarded receivers of recent years,
the AOR-7030, implements
both peak and average AGC systems:
AGC is carrier derived with peak detector in SSB modes
and mean detector in AM modes.
|
|
|
|
|
|
|
|
|
|
