Tuning range and linearity improvements for the SW20+

A quick inspection of the capacitance/voltage plot of one of the MV1662 devices shows that there is a definite bow in the plot. I imagine, although I'm not an RF engineer, that these device characteristics, combined with the other LC elements of the VFO produce the tuning nonlinearity. This has been documented before. In the book QRP Power, authors KB6FPW and AE6C discuss several improvements in their builds of the 40-40. This rig has a VFO design similar to the SWxx+ series rigs. Also, N7XJ wrote an article in the ARS Sojourner about modifying his SW40+ tuning range. I began my experiments based on these articles.

One thing I should point out here: I have a Small Wonders Labs FreqMite installed in my SW20+, which made theses experiments very easy to do. It also makes my ultimate solution possible without having to get involved with dial scales like my 30 meter rig has (hint: think multiple scales; to be made evident later). My first modification was to expand the range from the 33 Hz I currently had so that I had coverage below 14.035. I also wanted to be able to predict how to make similar changes for my upcoming build of the SW80+, since I want to be able to cover the 80 meter QRP calling frequency at 3560, the local NoGa net frequencies, and the 3720 target frequency of the Straight Key Century Club (SKCC). In the SWxx+ rigs, C7 sets the operating point of the VFO, for lack of a better term. It determines the high end range limit. When building the rig, the last step is to try several values of C7 to acheive the desired segment coverage. Changing C8 from its standard value (27 pF in the SW20+) will increase or decrease the span of coverage.

So, step 1 was to try a different value for C8. I changed it from its initial value of 27 pF to 51 pF. This increased the tuning span from 33 Hz to 83 kHz, but the top end had shifted down from 14.068 to 14.058, and the bottom end was now at 13.975. Doing some quick scaling of the 27 pF and 51 pF results, I next I tried 39 pF, and the observe span after changing the cap was 14.008 to 14.065. Getting there.

Step 2 was to try to address the linearity issue. After reading the QRP Power article, I ordered a temperature-compensated constant-current source IC, the LM334CZ from Mouser. This chip, and 3 other parts...two resistors and a 1N4148...are connected between +12 V and the wiper of the tuning pot. In the article, KB6FPW was able to improve the linearity significantly. Here are my before and after results of this mod at the 10 positions on the dial scale (dots on the front panel):

Uncompensated	Compensated
14.008	14.009
14.010	14.010
14.013	14.013
14.016	14.018
14.022	14.025
14.030	14.035
14.040	14.045
14.052	14.055
14.063	14.064
14.065	14.065

There is a noticeable improvement of linearity from 14.025 to 14.064 such that each dial division is 10 kHz. The extremes left much to be desired, however. This is especially important, because I already knew that I wanted my high end to be ultimately around 14.070, and I wanted good no compressed-frequency effect around the QRP calling frequency of 14.060. By the way, I did a voltage vs. dial position check of the pot itself, and I found that the pot that Dave Benson selected for these rigs is very linear.

Step 3 was to use information from N7XJ in his ARS Sojourner article. Bob was able to expand the scale, if you will, by adding resistance to each end of the pot; i.e., between the "hot" end and the 8 V regulated supply, and between the "cold" end and ground. I disconnected both of these points and used four jumper wires and clips for subsequent experiments. I quickly found that adding resistance to the hot end had the undesired effect of lowering my top end frequency.

Step 4 was to get closer to the actual range I needed prior to further resistance/compensation chip experimentation. I changed C7 and C8 to reset the overall frequency range higher and the span a bit less, and got 14.034 to 14.082. C7 at this point was 30pF and C8 was 33pF. I then lowered the operating range by changing C7 to 33 pF, ending up at 14.020 to 14.067, close to my final objective.

Step 5 was a series of compensated (with the chip in the circuit) vs. uncompensated (chip out of the circuit) data sets, with various values of resistances connected to the pot ends. Note that the addition of this resistance "magnifies" the range. Here is the set of data that I took using 33k at the hot end and 330k at the cold end:

Compensated	Uncompensated
14.057	14.045
14.058	14.045
14.059	14.047
14.060	14.049
14.061	14.051
14.062	14.053
14.063	14.056
14.064	14.058
14.065	14.060
14.065	14.061

That was a real eye-opener! The near-perfect linearity with the LM334 in the circuit came at the price of frequency span: only 8 kHz. Uncompensated, the span was 16 kHz. The thing that I notice from this test was that even uncompensated, the linearity wasn't too bad. I immediately began to envision switching resistance in and out of the circuit of the pot to have a "normal" and "magnified" range capability.

Knowing from beforehand that adding resistance to the hot end lowered my upper frequency, I jumpered that end of the pot to 8V (normal connection), and I changed the cold end resistance to 470k, and subsequently to 680k. This resulted in the following data (both sets uncompensated):

470k	680k
14.053	14.055
14.053	14.056
14.055	14.057
14.057	14.059
14.059	14.060
14.061	14.062
14.063	14.063
14.065	14.065
14.066	14.067
14.067	14.067

Both data sets looked pretty good; delta-F on 470k is 14 kHz, and 12 kHz fo 680k.

I continued to experiment with different resistors, and compensated/uncompensated for several hours. However, with some values, compensated or uncompensated, the range was not linear enough, or the span was too compressed, compared to the data set for 470k uncompensated. The best compensated test was with the normal 8 volt direct connection to the hot end, and 220k on the cold end, as follows:

220k Compensated
14.054
14.055
14.057
14.059
14.060
14.062
14.064
14.065
14.067
14.067

The results were very similar to uncompensated using 680k; my conclusion was that while compensation using the LM339 was desirable, I wanted the ability to switch between the normal range and magnified range, and the two were to an extent mutually exclusive without some more complicated switching. For example, 220k uncompensated did not result in a good span or range, and 470k compensated did not either. Compensated at 680 k resulted in only 8kHz of frequency span. Therefore, I decided that I would simply use a 470k resistor connected to the cold end, and simply switch it in and out of the circuit with an SPST panel-mounted switch.

Step 6 was another investigation that I wanted to do before I soldered things in for good. One technique used in the article in QRP power, and one I've seen described elsewhere, is to put a fixed resistance in parallel between the hot end and the wiper of the pot. For the 40-40, KB6FPW tried this with a 100k resistor. Here is the data that I took for a 10k, 20k resistor and a 100k resistor paralleled to the wiper (cold end connected to ground, normal):

10k	20k	100k	No Resistor
14.020	14.020	14.020	14.020
14.024	14.023	14.021	14.021
14.044	14.034	14.025	14.023
14.052	14.043	14.030	14.028
14.057	14.049	14.036	14.033
14.060	14.053	14.042	14.041
14.062	14.057	14.049	14.049
14.063	14.060	14.058	14.058
14.066	14.066	14.066	14.066
14.067	14.067	14.067	14.067

While the 10k results show some value in that 14.060 is near the center, with good resolution and linearity above it, the resolution and linearity degrade below that frequency. It is preferable to not having any resistor attached, but the results are not as good as the 470k resistor in series with the cold leg.

Final comments: Once I had soldered the resistor in place, I found that I needed a minor tweak. I must have switched resistors, because my range had shifted slightly. I compensated by soldering in a 2.7M resistor in parallel with the 470k. My final dual-resolution SW20+ VFO results as as shown:

Normal	Magnified Scale
14.020	14.055
14.021	14.056
14.023	14.058
14.026	14.059
14.031	14.061
14.038	14.062
14.047	14.064
14.057	14.065
14.066	14.067
14.067	14.067