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

E-mail: Jack.Smith@cliftonlaboratories.com
 

Mike, KS7D, has developed a program K3 Filter Tools to measure and plot the crystal and DSP-based filter response of an Elecraft K3 transceiver, with a minimum of external test equipment. The program is available for download at http://mysite.verizon.net/ks7d/ .

I've helped Mike with some algorithm discussions and also tested some pre-release versions of his program. The concept is similar to the one I developed to measure the K3's filter response, described at my K3 filter page, but Mike's program is much more polished and user friendly. And, it requires a minimum of test equipment--nothing more than a strong, stable signal.

Before measuring the K3's filter response, it's necessary to determine the RF gain setting that produces maximum signal, just short of gain compression within the K3. Mike's program includes an automated way of testing the K3's gain linearity versus RF gain setting and when working through the algorithm with Mike and testing the K3 Filter Tools software, it became apparent that the K3's manual gain control has some quirks or, perhaps a better term is "oddities." In the ordinary course of events, none of these oddities are apparent to the K3 user, but they surfaced when working on  the filter software.

In a purely analog receiver, the RF gain control can be implemented in a variety of ways. In the crudest method, it could be a simple potentiometer on the antenna input port, capable of reducing the signal level applied to the receiver. This approach has many infirmities, but it's occasionally used in very simple home brew receivers.

Most modern analog receivers implement RF gain in conjunction with the receiver's automatic gain control (AGC) function. A DC voltage is used to vary the gain of one or more stages within the receiver. When in manual gain mode, the RF gain control potentiometer sets the control voltage. When in AGC mode, the control voltage is set automatically, based on the strength of the received signal.

In fact, in many receivers, the gain of the true RF stage is not altered by the gain control, whether manual or automatic, and the gain reduction is performed purely at the IF stages. Some receivers, such as Racal's RA6790/GM, label the control "IF Gain" in recognition of this fact. And, even if the RF stage has its  gain vary with the control voltage, the IF stages also have their gain controlled, as it's impossible to obtain the 100 dB+ gain control range required by varying a single receiver stage. It's also desirable, from optimum signal-to-noise and intermodulation performance considerations to allow a receiver's RF amplifier stage to run at maximum gain, or, if it is to be controlled, to not apply control voltage to the RF amplifier until after considerable gain reduction in the IF stages has already taken place, a concept called "delayed AGC."

It's reasonable to expect, therefore, that when in manual gain control some analog stages will have gain control applied and, perhaps, some gain control adjustments will be made in the DSP as well.

The extract below from the K3's Owner's Manual shows analog AGC (and manual gain control when in AGC off mode) is applied to one IF amplifier stage operating at 8215 KHz.

The image below shows the associated schematic.

Q1 is the 8215 KHz IF amplifier subject to gain control. It's a J309 FET operated in grounded gate. Q1's gate bias is obtained by the DC drop across R26 and R30, plus whatever voltage is added or subtracted by U3A's output, which can range between 0V (maximum gain) and 3V (minimum gain).

Control voltage at U3A's output can be generated from two sources.

• First, a DAC controlled by the K3's DSP and onto the VIFGAIN1 line. U3B is a unity gain buffer. (VIFGAIN1 is generated from the DSP based on the RF gain control value in manual gain control mode, or derived from the signal level by the DSP when in AGC mode.)
   
• Second, via the "HAGC" or hardware AGC circuit. This consists of diode rectifiers fed from U15, which is an active filter and buffer amplifier operating at the second IF frequency of 15 KHz. The rectified output voltage has a threshold voltage hold-off. Once the 2nd IF signal level crosses this threshold point, the HAGC circuit provides an alternative control voltage input to U3A and reduces Q1's gain. I've written about the HAGC circuit at Elecraft K3 AGC and S-Meter page. Note that the HAGC is not under software control and cannot be disabled absent breaking out the soldering iron. The HAGC's purpose is to prevent overloading the ADC, which would cause major distortion problems.

D22 and D33 form a soft voltage regulator, with a bit less than 3V to be expected at the junction of D21 and D22 (0.4V drop across D22, a Schottky diode, and 1.8 to 2.0 V or so across D33, plus a small voltage developed across R25.

If U3A's output is 3V, D21 is reverse biased and the RF signal applied to Q1's source via  D1 will be significantly attenuated as D21 looks like a high impedance. This range, as the schematic note indicates, corresponds to minimum gain through Q1. In addition, as U3B's output increases, Q1's drain current Q1 drops, also reducing Q1's gain.

If U3A's output is 0 V, then D21 is forward biased and hence conducting and the IF signal has little loss through D21 and Q1 can develop it's full 10 dB gain. As U3A's voltage  ranges between 0 and 3V, therefore, Q1 operates in the range between +10 dB gain and a significant negative gain.

I tried a simple SPICE simulation of Q1 and its gain control mechanism. It shows a maximum gain a bit over 10 dB, and a minimum gain around -50 dB for a total control range of 60 dB. That's likely more than might be realized in the circuit, or at least that might be acceptable without excessive distortion, but a gain control range of 30 dB or more should be achievable.

The image below shows Q1's output (at T5, pin 1) for 0.1V peak (one-sided) 8215 KHz input, with the control voltage ranging from 0V (maximum signal) to 3V (minimum signal) in 0.25V steps. I've made a number of educated guesses about impedance levels in the simulation, so it could well be off but the overall concept should be valid.

A couple of points are notable. First, there's very little change in gain whilst the control voltage is between 0 and 0.75V. Then, a major change is seen for the next few steps. Once the control voltage drops below 1.75V or so, it's not possible to easily resolve further steps.

I've extracted the numerical signal levels from the simulation and plotted them versus the control voltage below. The plot also shows Q1's drain current. I've made some simplifications in the simulation, but the data below should provide at least a rough sense of what's going on with Q1 as the control voltage changes.

The next voltage step, to 2.0V, drops the stage gain to absurd levels, -220 dB or more.

The plot does a better job of presenting gain change non-linearity versus control voltage. For all practical purposes, the simulation shows the useful gain control is concentrated over a control voltage range of a volt or so, from 1.75V to 0.75V.

The test starts with a relatively weak signal, -115 dBm, and the K3's RF gain at maximum, RG=250. The K3's internal audio measurement is set to 0.0 dBV.

The program then steps the RF gain down one unit at a time, 250, 249, 248, etc. At each changed gain step, the signal generator's output is increased to restore the K3's audio measured report to 0.0 dBV. The signal generator's output level is saved to a disk file, along with the associated RG setting.

As a safety measure, this process was discontinued when the signal generator reached +10 dBm.

The plot below shows the result. A couple of things stand out in this graph:

• Note the discrete jumps for RG values in the range 250 down to 120
• The gain versus RG setting is highly non-linear.
• The break point around RG=70 is due to HAGC action.

With only the plot data and a general knowledge of how DSP works, the best explaination I can devise is that the high nibble of RG is used in the DSP to adjust gain in 2 dB steps and that all 8 bits of RG go to a DAC to develop VIFGAIN1, the analog control voltage that ultimately adjusts Q1's gain.

There are 16 major jumps in incremental gain, which matches using RG's high nibble value to control DSP gain. These gain jumps range in value from 3 dB down to 1.8 dB.

The total gain reduction is then the sum of the DSP gain and Q1's gain. The analog gain control voltage RG parameter has 255 steps maximum (250 used) and gives somewhere around 0.4 dB / step average, but with individual steps ranging from 0 dB/step to around 0.9 or 1.0 dB/step, depending on where you are at on Q1's gain versus voltage curve as illustrated in the earlier plot. The DSP kicks has 16 steps of 2 dB each, based on the high nibble of the control. Since the actual reduction is the sum of both the DSP and the analog, the high 4 bit control gives between 2 and 3 dB gain variation, representing 2 dB in the DSP and something between 0 and 1 dB in the corresponding analog gain control section.

As I said at the outset, none of this matters to the K3 user. Rather it's one of those curiosities that surfaces when you try to do something out-of-the-ordinary, such as look at RF gain control linearity.