Clifton Laboratories 7236 Clifton Road  Clifton VA 20124 tel: (703) 830 0368 fax: (703) 830 0711

E-mail: Jack.Smith@cliftonlaboratories.com

Table of Contents
Introduction
AGC_SLP_and_AGC_THR
Test_Setup
Fast/Slow_AGC;_Preamp_On/Off
Adjusting_AGC_THD
Adjusting_AGC_SLP
S-Meter_Calibration
AGC-F,_AGC-S_and_AGC_HLD
AGC-F_Changes
AGC-S_Changes
AGC_HOLD_Changes

Original page 17 August 2008
Revised 19 August 2008 to include AGC-F, AGC-S and AGC HOLD parameters
Revised 24 August 2008 to correct legend in plot showing fast/slow AGC
Revised 25 August to correct typo on AGC F/S differences; added several paragraphs with practical differences

This page explores  the effect of adjusting five AGC-related parameters of the K3's dozens of user-settable configuration parameters.

AGC Threshold and Slope are discussed first, followed by AGC decay time (for fast and slow settings) and AGC hold time.

 AGC SLP and AGC THR

The K3 Owner's Manual provides a terse description of these two parameters:

The data is from my K3, operating with firmware revision 2.22. It's possible, of course, that other firmware releases will alter the behavior of these parameters. My K3 was assembled and calibrated by Elecraft and came with the AGC SLP and AGC THR settings illustrated above, i.e., 12 and 5.

Before looking at the data, it's first useful to see a "notional" view of what we expect to see. You may also wish to read my earlier page on AGC measurements, Receiver AGC Curves in conjunction with this page.

First, AGC means automatic gain control and its function is to keep the receiver's audio output more or less constant as the RF input signal changes level. Designing a useful, stable AGC system is a non-trivial task.

It's convenient to divide the K3's AGC characteristics into four regions, as illustrated in the conceptual diagram below.

1. For extremely weak signals, the audio output is mostly determined by noise and hence does not not change much as the signal increases.
2. When the signal level is somewhat greater than the noise, the audio output increases on a 1:1 basis, i.e., a 1 dB increase in input signal level results in a 1 dB increase in audio output. This relationship is more apparent as the signal to signal plus noise ratio increases above 10 dB or so.
3. At a point defined by the AGC THD setting, the K3's software AGC begins to operate and increases the audio output at a slower rate than represented by the increase in RF signal level. Increasing the RF signal level 10 dB might, for example, increase  the audio output by 2 dB. This input versus output relationship for signals above the threshold point is governed by the AGC SLP setting.
4. Once the input signal level increases above approximately -43 to -48 dBm (with the K3's preamp off; 10 dB weaker levels if the preamp is on) an independent hardware AGC circuit takes effect. The hardware AGC threshold and slope are not user settable, other than by making component changes within the K3. (Earlier K3 production runs had a different hardware AGC threshold and an update kit is available.)

The description of AGC defines how one measures it. Since its function is to adjust the audio output as the RF input varies, we simply apply an RF test signal of known level to the K3 and measure the audio output after a brief pause for levels to stabilize. Increase the signal level and repeat the audio output measurement.

To do this for a couple dozen combinations of AGC SLP and AGC THD, each involving 100 or more individual measurements would be tedious, to say the least.

Fortunately, computer controlled test equipment permits us to automate the process. The figure below shows the arrangement I used to collect the data on this page. The M6300 laptop computer runs software I wrote to control the HP 8657A signal  generator and to read  the Agilent 34410A digital multimeter. In addition, the software reads the K3 report on the number of signal strength graph segments displayed. The data is saved  to disk for post-collection analysis and plotting with Origin software data plotting software.

RF signal levels went from -140 dBm to -20 dBm, in 1 dB increments. The K3 is operated in CW mode, AGC at slow, preamplifier off unless otherwise noted. Bandwidth is set at 600 Hz. The 34410A digital multimeter is a true RMS reading meter and was set for AC response.
 

K3 Signal Level / S-Meter Check / AGC Audio Out
Run Date/Time: Aug 16, 2008 10:55:40
Transceiver ID ID017;
Transceiver Info IF00007000010 -000000 0003000001 ;
Filter BW Info FW0000;
AGC Speed / Status GT004;
Mode MD3;
Preamp Mode
Attn Mode RA00;
RF Gain RG239;
Prologix Card Ver Prologix GPIB-USB Controller version 5.4

Agilent 34410A ID Agilent Technologies,34410A,MY45000265,2.21-1.10-0.09-46-6
Start Level: -140 dBm
Stop Level: -20 dBm
Step Level: 1 dBm
Frequency 7 MHz

AGC=SLOW PRE=OFF THRESHOLD=5 SLOPE=09

dBm Bar Graph Audio
--- --------- -----
-140 0 +2.05011334E-03
-139 0 +1.87790216E-03
-138 0 +1.85180895E-03
-137 0 +2.06282250E-03
-136 0 +2.07184504E-03
...
-22 18 +7.74713052E-02
-21 18 +7.75875510E-02
-20 19 +7.76909983E-02

Before adjusting the AGC SLP and AGC  THD, I first measured and plotted the AGC characteristic curves with the default values of 12 and 5, respectively, with AGC fast and AGC slow and the preamp on and off.

The figure below shows the same four regions in our conceptual discussion. It's easier to observe the match between measured and conceptual looking at the data taken with the preamp off, i.e., the red and black curves. (It's also immediately obvious that there's no real difference between fast and slow AGC speeds for our purposes.)

We then start to see the signal rise from the noise and maintain a 1:1 relationship to the point of AGC  threshold. For example, at -120 dBm input,  the audio output -45 dBV (dB with respect to 1 volt RMS). At -110 dBm input, the audio output is -37 dBV, an 8 dB increase, not quite 1:1. (At -120 dBm input, the noise component is contaminating the measured audio to some degree and likely accounts for some of the difference.)

The software AGC threshold is visible at about -104 dBm. For a 20 dB signal increase (from -90 dBm to -70 dBm), the audio output increases 1.5 dB (from -30 to -28.5 dBV) so the slope is 1.5:20, output change : input change, in dB.

At -43 dBm, we see the effect of hardware AGC.  The hardware AGC manifests itself not so much in a clear intercept / slope change as it does in removing the ripples in the AGC response. I don't know the source of the ripples, but one possibility is that the K3's DSP uses an approximation to a logarithm function optimized for speed rather than accuracy. In any event, the ripples exhibit relatively small amplitude changes, on the order of 0.5 dB and are likely undetectable when using the K3 for normal purposes.

The effect of the preamp is mostly to shift the curves about 10 dB, the preamp gain. It also increases the noise by 10 db and this is easily seen in the difference in shape of the curves for very weak signals.

I ran signal sweep plots for all possible values of AGC THD, while keeping AGC SLP at 12. In addition, I ran a sweep of my K2 as a comparison point.

I've based these values on the point where I see a clear divergence from the 1:1 slope,  but reasonable people could read these numbers a couple dB either way.

The K2's  AGC threshold point is between -80 and -90 dBm, but the AGC application is not abrupt, making it difficult to define a single "threshold."

I next looked at varying the AGC slope whist keeping  the threshold AGC  THD constant at the factory default setting of 5.
 

Based upon these values, I've prepared a table showing how much the audio changes for a 10 dB change in RF signal level, versus slope setting. I've shown these values to two decimal places, but that does not imply the data is accurate to that degree of precision.

The K2 data shows a 5 dB change in audio for a 50 dB change in RF level (-20 to -70 dBm), for a slope of 0.50 dB for 10 dB change, corresponding most closely to the K3's AGC SLP 13. However, the K3's hardware and software AGC systems make it impossible to duplicate the K2's characteristics.

The two plots below show how my K3 displays S-meter bars as the RF signal level changes. I've presented the data separately in terms of dBm and µV. The S-meter offset and slope values are the ones my K3 was shipped with.

To the extent there is a standard, S-9 corresponds to 50 µV and one S-unit corresponds to 6 dB change in signal level. With the preamp engaged, at 7 MHz, my K3 ticks over to the S-9 bar exactly at 50 uV and tracks the 6 dB per S-unit objective quite closely.

I did not collect data in the "independent" mode where the S-meter reading is independent of preamp or attenuator setting.
 

In addition to the AGC threshold and slope parameters, the K3 has three user-adjustable parameters governing how the AGC releases.  As with the earlier AGC parameters, the K3's Owner's Manual provides only a brief discussion of the three parameters:

Why have different AGC decay speeds? AGC attempts to preserve the dynamic range of the input signal without overloading the receiver. Consider an SSB signal--the maximum gain possible before receiver overload is determined by the strength of the voice peaks, so the AGC must adjust the overall gain based on the peak signal. In order to react to these peaks, the AGC needs a fast attack time, on the order of a millisecond or two.

Now consider the release or decay time constant. If the AGC has a fast decay time, it will adjust the receiver's audio output upwards on low voice levels and downward on high voice levels. This may be acceptable under certain conditions, but overall it removes dynamic range from the voice audio. The result is similar to the effect of the transmitting station using an aggressive speech compressor and can prove annoying to listen to. In addition, during brief pauses, fast AGC  will bring the background noise level up to nearly the peak speech level, not a good thing at all.

A fast attack, slow decay system for SSB will adjust the gain based on peaks, but hold the gain constant for several seconds after the peak. Thus weaker voice sounds produce weaker audio output and stronger voice (but not quite equal to the peak) produce stronger audio, thus preserving the SSB signal's dynamic range. During speech pauses, the background noise is likewise reduced.

AGC operation may be easier to understand if we think of CW mode. During key-down either fast or slow AGC will produce the same result - there's no amplitude variation to speak of in the CW signal during key down, except under rare conditions such as fast polar flutter. However, when the key is up, fast AGC will bring up the background noise but slow AGC will suppress the background noise as it maintains the gain at a level suitable for the signal. One problem with slow AGC is that a noise spike, such as from lightning crashes, will charge up the AGC system and reduce the gain for the decay period. A fast AGC may be desirable for these conditions. (The K3 has a slow AGC mode with spike suppression to avoid these problems.) Slow AGC will make it difficult to copy both a strong and weak signal in fast sequence as the gain will be governed by the stronger signal until the decay period passes.

To observe what happens as we vary the three AGC speed setting parameters, we'll use an AM signal, modulated with a square wave, with a modulation depth of approximately 13 dB. We use a reduced modulation depth because it allows us to easily measure the key parameters with a single test waveform.

I apologize for the sketch below—I'm not an artist and none of the drawing software I have is suitable for smooth exponential curves. The illustration is for CW mode but the same concepts apply to all modes in which AGC is used.

1. When the test signal amplitude drops 13 dB, the K3's audio output drops 13 dB (assuming the RF and audio levels are set to be within the K3's software AGC range and out of saturation.)
2. If the AGC is set for slow, with AGC HOLD > 0, the audio output will stay -13 dB from the full amplitude level for the holding  time defined by the AGC HOLD parameter.
3. When the AGC HOLD  time passes (or immediately if AGC HOLD is 0)  the AGC begins to increase the gain so as to increase the -13 dB signal's level. In a vacuum tube analog  receiver, the AGC would go less negative, or decay from the negative voltage level established  by the full amplitude signal. As the AGC voltage became less negative, the receiver's gain increased. For this  reason, the period from the hold interval until the new AGC level is fully asserted is called the "decay period." (Generally the decay period was based upon an RC time constant, so the shape of the decay was determined by the negative exponential of an RC circuit. Designers are free to use other approaches in the world of digital signal processing, but it seems that Elecraft's K3 emulates an RC network
4. When the new AGC level is fully asserted, the audio output will be a bit less than observed for the 100% signal level because the AGC's sloped  response means that a weaker signal will have a slightly weaker audio output even when within the AGC range. (See earlier discussion on this page for more details.)
5. When the signal increases back to 100% amplitude, there is a brief delay due to the digitization and mathematical computations in the K3's DSP. I may measure this at some point but for now we'll take it as a  given that a few milliseconds of DSP delay is present.
6. The AGC then will begin to bring  the 100% level signal down. The time it  takes for the AGC to assert itself when the signal increases is called the "attack time" and in the current K3 firmware the attack time is not a user adjustable parameter.
7. AGC Hold functions only when AGC-Slow mode is engaged.

The K3 is set for 7.000 MHz, data mode, with 600 Hz bandwidth. AGC SLP and AGC THD are set at the default values, 12 and 5 respectively.

100% modulation corresponds to -70 dbm. The modulating square wave frequency is around 0.1 Hz for the AGC-S and 1 Hz for AGC-F. The SG-100 has a DC-coupled modulation input so it accepts low frequency modulation signals.
 

AGC-S with AGC HOLD = 0. The decay time is measured between the two cursors as 772 ms, measuring from change in signal level to AGC fully stabilized. This corresponds to AGC-S = 20. There's an argument, of course, that the measuring period should be based upon the 10% - 90% envelope. I'll stick with the 0 - 100% points as they are easier to measure.
 

You may also note trace 1, which shows the modulation voltage applied to the SG-100 function generator. The time between trace 1's high-to-low transition represents the K3's DSP delay. It's approximately 18 ms for this particular mode and filter selection.
 

Rather  than present a series of oscilloscope captures, I've plotted the measured AGC decay speed versus AGC-F parameter settings.

Likewise, the plot below shows the AGC decay time associated with the range of permitted AGC-S parameters.

The final plot in this series shows how the AGC Hold time varies with the AGC HOLD parameter. As the Owner's Manual states, the hold time in milliseconds equals the AGC HOLD parameter times 10. (The manual actually says the hold time equals the parameter. It's really 10x the parameter.)