Low noise VFO oscillator with internal amplitude stabilization

[neazoi]

Hello,
Here is an interesting vfo circuit in figure 26. I would like to ask a few things.

1. The author states "The base is tapped only one turn up from the cold end." What does it mean, where should the one-turn exist, at the variable capacitor side, or at the other side?

2. I am thinking of using an FT50-43 (ferrite) core for the 25uH, is that ok, or do I need to use an iron powder?(T50-2 for example)

3. The 3:1 transformer at the output, what number of turns should it have?

4. Can I use a ferrite core (broadband?) for the output transformer or do I need to use an iron powder one? (resonant?)

5. Is the output near 50R?

 

[k7elp60]

If you read the writing below the schematic, it says L1 is 17 turns and the tap is 1 turn from the capacitor end. I don't know where you are getting 25uH for it. L2 needs to be some form of iron core from the schematic having the two parallel lines between the primary and secondary.

 

[Warpspeed]

Its completely ambiguous, it says one turn from the capacitor end, which is the "hot" end.

Problem is, if you tap it right up at the capacitor end, the base is going to load the tuned circuit and destroy the Q. Highest possible Q is a prime requirement not only for decent frequency stability, but also phase noise (frequency jitter).

The author is right about keeping the amplitude high, well above the noise floor. But probably even more important is the highest achievable Q by having a very lightly loaded tuned circuit.

Best material for the tuned circuit would be powdered iron grades 6 or 7, as these have the lowest temperature coefficient, and have low losses over the HF band for highest Q.

The 25uH choke could be ferrite, but personally I would prefer the same more stable grade of powdered iron as used for the tuned circuit for frequency stability reasons.

A ferrite toroid, maybe 50 grade should work well in the output. Number of turns depends on core size, operating frequency, and impedance matching requirements. Best found experimentally.

It might be best to use as few a secondary turns for least loading on the oscillator. But obviously this may reduce the output below what is required.

Its all far from easy, and many published "low noise" oscillators are far from that and pretty disappointing, when finally tested with a spectrum analyser.

In my own search for the holy grail, the most fruitful approach seems to be to use a DDS synthesiser driving a PLL, which becomes in effect a very narrow band tracking filter.

You build a super high Q tuned circuit as part of a fet oscillator.
Then phase lock that to your DDS source with a narrow bandwidth loop.
The fet VCO need not be ultra long term frequency stable, but it does need to be a very pure amplitude stabilised sine wave.

That creates the "nice clean" low noise output, but with the excellent frequency stability of the DDS, without all the DDS wideband crap.

Harmonics should be minimal, but are easily removed with the usual low pass filter on the output.